WO2025204016A1 - Information processing system, feature point extraction method, and control program - Google Patents

Abstract

This information processing system comprises: a comparison unit 513 for comparing first feature point information for a web in first inspection data with second feature point information for the web in second inspection data, and extracting feature points that are the same in the first inspection data and the second inspection data; an analysis unit 514 for generating relationship information indicating the relationship between a coordinate position with respect to the entire web and the difference value of coordinates in the first inspection data and the second inspection data for each of the plurality of feature points extracted by the comparison unit; and an output unit 516 for outputting the relationship information.

Description

Information processing system, feature point extraction method, and control program

The present invention relates to an information processing system, a feature point extraction method, and a control program.

Liquid crystal display devices are increasingly being used in large-screen televisions and large monitors, and as a result, there is a demand for wider films to be used on the display surfaces of these devices. For example, there is a demand for wider films of 2000 mm or more. Furthermore, in order to anticipate substrate loss (film loss) and reduce transportation costs, there is a demand for the production of long film rolls with winding lengths of 1000 m or more, and even 3000 m or more.

When a defect or other quality issue occurs in a post-processing step that produces a product based on a web such as film, it is necessary to determine whether the defect occurred in the upstream process, i.e., during the production of the web and was originally present on the web, or whether it arose in the downstream process. If it is not possible to clearly determine that the defect occurred in the downstream process, improvements may be required in the upstream process. In attempting to respond to requests for improvement, the shipping standards for the upstream process may become stricter than necessary, resulting in over-specification. Also, in order to prevent quality issues before they occur, it is sometimes necessary to set overly strict shipping standards for web production based on estimates, even though it is not clear how this will affect the downstream process. Such cases also constitute over-specification. Over-specification leads to reduced yields, increased costs, is undesirable from an ecological perspective, and is undesirable for both the upstream and downstream process companies.

It is necessary to determine whether the defect was originally present on the web or whether it occurred during a post-processing step, and in doing so, it is necessary to match the coordinate system on the web from the previous process with the coordinate system from the subsequent process.

Patent Document 1 below discloses a technology for synchronizing the inspection results of a first and second inspection device in an optical film inspection system. In this inspection system, the inspection results of both devices are synchronized by using an encoder signal that indicates the distance between the first inspection device and the second inspection device, which is placed downstream in the transport direction, and the amount of film movement.

JP 2016-161576 A

The technology in Patent Document 1 synchronizes inspection results based on the mutual distance and movement distance between the inspection devices, and is unable to achieve synchronization when the film expands or contracts. Furthermore, synchronization is also unable to be achieved when two inspection devices are wound up with film in a roll and then have offline processing performed to unwind the roll.

The present invention was made in consideration of the above circumstances, and aims to efficiently collect information that is useful for process improvement and setting shipping standards both when manufacturing webs and in the manufacturing process where post-processing is performed using these webs.

The above-mentioned object of the present invention is achieved by the following means.

(1) an acquisition unit that acquires first inspection data in a first manufacturing process in which a web is manufactured or a manufactured web is post-processed, and second inspection data in a second manufacturing process in which post-processing using the web is performed after the first manufacturing process;
a comparison unit that compares first feature point information of the web in the first inspection data with second feature point information of the web in the second inspection data and extracts identical feature points between the first and second inspection data;
an analysis unit that generates relationship information indicating a relationship between a difference value between the coordinates of each of the plurality of feature points extracted by the comparison unit in the first inspection data and the second inspection data and a coordinate position on the entire web;
an output unit that outputs the relationship information.

(2) The information processing system described in (1) above, wherein the relationship information is a graph showing the relationship between coordinate difference values and overall coordinate positions.

(3) An information processing system according to (1) or (2) above, wherein the relationship information is statistical information indicating the relationship between coordinate difference values and overall coordinate positions.

(4) Further comprising a receiving unit that receives a correction amount of the coordinates of the first inspection data and/or the second inspection data,
The information processing system described in (1) above, wherein the comparison unit corrects the coordinates of the feature points with the correction amount received by the reception unit, then compares the first feature point information of the web in the first inspection data with the second feature point information of the web in the second inspection data, and extracts identical feature points in the first and second inspection data.

(5) the receiving unit receives the correction amount after the output unit displays the relationship information;
In response to the reception unit receiving the correction amount, the comparison unit extracts the same feature points again, the analysis unit generates the relationship information, and
The information processing system according to (4), wherein the output unit displays the relationship information again.

(6) The relationship information is a scatter diagram and a regression equation that represent the relationship between the coordinate difference value and the overall coordinate position,
The information processing system according to (1) above, further comprising an extraction unit that classifies the feature points in the scatter diagram according to a distance of the feature points to the regression equation.

(7) The relationship information includes a correlation coefficient indicating the relationship between the coordinate difference value and the overall coordinate position, or a variance of the difference value,
The information processing system described in (1) above, wherein the analysis unit generates an alert when the absolute value of the correlation coefficient is less than or equal to a first predetermined threshold value or when the variance is greater than or equal to a second predetermined threshold value, and the output unit notifies the alert.

(8) A method for extracting feature points of a web, comprising:
A step (a) of acquiring first inspection data in a first manufacturing process of manufacturing a web or post-processing a manufactured web;
(b) acquiring second inspection data in a second manufacturing process that uses the web and is performed after the first manufacturing process;
(c) comparing first feature information of the web in the first inspection data with second feature information of the web in the second inspection data;
(d) extracting identical feature points between the first and second inspection data based on the comparison result of the step (c);
a step (e) of generating relationship information indicating a relationship between a difference value between the coordinates of each of the plurality of feature points extracted in the step (d) in the first inspection data and the second inspection data and a coordinate position on the entire web;
and (f) outputting the relationship information generated in the step (e).

(9) The feature point extraction method described in (7) above, wherein the relationship information is a graph showing the relationship between coordinate difference values and overall coordinate positions.

(10) A control program for causing a computer to execute the extraction method described in (7) or (8) above.

The information processing system of the present invention includes an acquisition unit that acquires first inspection data from a first manufacturing process in which a web is manufactured or a manufactured web is post-processed, and second inspection data from a second manufacturing process in which post-processing using the web is performed after the first manufacturing process. The information processing system also includes a comparison unit that compares first feature point information of the web in the first inspection data with second feature point information of the web in the second inspection data and extracts identical feature points in the first and second inspection data, an analysis unit that generates relationship information indicating the relationship between the coordinate difference values in the first inspection data and the second inspection data for each of the multiple feature points extracted by the comparison unit and their coordinate positions across the entire web, and an output unit that outputs the relationship information. This makes it possible to efficiently collect information useful for process improvement and shipping specification setting both when manufacturing a web and in the manufacturing process in which post-processing is performed using the web.

Advantages and features provided by one or more embodiments of the present invention will be more fully understood from the following detailed description and the accompanying drawings, which are for purposes of illustration only and are not intended to be limiting.
FIG. 1 is a schematic diagram showing a film roll production line. 2 is a top view of the stretching process, drying process, and trimming process of the production line in FIG. 1. FIG. 1 is a schematic diagram illustrating a configuration of an inspection device. FIG. 1 is a schematic diagram illustrating a configuration of an inspection device. FIG. 1 is a schematic diagram illustrating a configuration of an inspection device. FIG. 1 is a schematic diagram illustrating an application example of an information processing system according to an embodiment of the present invention. 10 is a table for explaining extracted first to third type feature points. 1 is a block diagram showing a schematic configuration of an information processing system. 4 is an example of various data stored in a storage unit. 10 is an example of an inspection data DB stored in a storage unit. 10 is a flowchart showing a process for generating first inspection data performed in a first manufacturing process. 10 is a flowchart showing a process for generating second inspection data performed in a second manufacturing process. 10 is a flowchart showing a feature point extraction process executed in the information processing system. 10 is an example of an operation screen for accepting a scaling ratio. FIG. 10 is a schematic diagram for explaining the extraction processing of feature points. 10 is a subroutine flowchart showing a comparison process 1 in step S35. 10 is an example of a probability density function calculated by kernel density estimation, which shows the positions and intensities of feature points. 10 is an example of a corresponding point list. 10 is an example of a distribution map as relationship information. 10 is an example of a distribution map as relationship information. 10 is an example of a distribution map as relationship information. 10 is an example of a distribution map as relationship information. 10 is an example of a distribution map as relationship information. 10 is an example of an operation screen for accepting a correction amount. 10 is a subroutine flowchart showing a comparison process 2 in step S50. 10 is an example of a distribution map after applying a correction amount. 10 shows an example of a feature point extraction condition. FIG. 10 is a schematic diagram for explaining classification according to the distance from the regression equation. 10 is a flowchart illustrating a feature point extraction process according to the second embodiment. 10A and 10B are diagrams illustrating an example of a distribution map and an alert display.

Embodiments of the present invention will now be described with reference to the accompanying drawings. However, the scope of the present invention is not limited to the disclosed embodiments. In the description of the drawings, identical elements will be given the same reference numerals, and duplicate explanations will be omitted. Furthermore, the dimensional proportions in the drawings have been exaggerated for the sake of explanation and may differ from the actual proportions.

In this embodiment, the term "web" refers to a sheet-like material, including resin film and metal film. In the following, the web refers to a long resin film, and is described as a film roll and the material that is processed. Processing also includes the process of applying a coating liquid and the process of layering another film-like material.

First, the production line for film rolls 80 will be described with reference to Figures 1 and 2. Figure 1 is a schematic diagram showing a film roll production line. The production line shown in Figures 1 and 2 illustrates film production using a solution casting method, but is not limited to this and film production using a melt extrusion method is also possible.

As shown in FIG. 1, the film roll manufacturing apparatus 1000 has a casting section 01, a first drying section 02, a stretching section 03, a trimming section 05, a second drying section 06, and a winding section 07 (also called a winding device).

In the film roll manufacturing apparatus 1000, an inspection device 90 is positioned between multiple processes. The inspection device 90 monitors the surface of the film F8 being transported on the production line, photographs the film surface, and analyzes the resulting frame images to detect the transport status of the film F8. The configuration of the inspection device 90 will be described later (see Figure 3A, etc.). In Figure 1, the white arrow indicates the transport direction of the film, etc.

The casting unit 01 has a mirror-finished, strip-shaped metal casting belt (hereinafter referred to as the belt) 01a, an endless support that runs endlessly (in the direction of the arrow in the figure), and a die 01b that casts a dope, which is made by dissolving a resin in a solvent, onto the belt 01a. To stabilize the dope film that flows out of the die 01b, a decompression chamber (not shown) may be provided upstream of the die 01b in the belt transport direction, and a pressurization chamber (not shown) may be provided downstream.

The peeling roll 01d peels off the cast film 01c formed by casting onto the belt 01a, producing an unstretched film F8.

The first drying section 02 (first drying process) has a drying box 02a with a dry air intake 02b and an exhaust 02c, and a set of upper and lower transport rolls 02d, each consisting of multiple sets, that transport the film F8.

In the first drying section 02, it is possible to adjust the amount of solvent contained in the unstretched film F8 before it enters the stretching section 03 (stretching process). This adjustment of the solvent amount is performed by changing the drying temperature and conveying speed in the first drying section 02 using the control device 08.

(Stretching section 03, trimming section 05)
Figure 2 is a schematic top view of the stretching process (also referred to as the tenter process), drying process, and trimming process of the production line in Figure 1. In Figures 2, 3A, etc., the vertical direction is the Z direction, the transport direction of film F8 is the Y direction, and the direction perpendicular to the transport direction, the width direction of film F8, is the X direction. The Y direction or transport direction is also referred to as MD (Machine Direction). The X direction or width direction is also referred to as TD (Transverse Direction) or left-right direction. The Y direction is also referred to as the longitudinal direction, and the width direction is also referred to as the width direction in relation to the longitudinal direction.

The stretching unit 03 has an MD stretching unit 03a and a TD stretching unit 03b. The stretching unit 03 stretches the unstretched film F8 transported from the first drying unit 02. As shown in Figure 2, the stretching unit 03 grips both side edges of the heated film F8 with multiple pairs of left and right gripping units 301. The stretching unit 03 is equipped with gripping units 301, chains, drive units, closers, openers, etc. (Components other than gripping units 301 are not shown).

In the upstream MD stretching section 03a (MD stretching process), the film accelerates as it is transported downstream, gradually increasing its movement speed in the transport direction (indicated by the black arrow in Figure 2). As a result, in the MD stretching process, the unstretched film F8 is transported while being stretched in the MD direction.

In the next stage, TD stretching section 03b (TD stretching process), film F8 and the like are transported by gripping sections 301. The gripping sections 301 are, for example, clips, and the left and right gripping sections 301 are connected, for example, to an endless chain. The chain, wound around a sprocket, is rotated by a drive section, causing the gripping sections 301 to move in the transport direction and the like. A closer is located at the most upstream side of the TD stretching process in stretching section 03, and an opener is located at the most downstream side. As the gripping sections 301 reach the closer position, the open gripping sections 301 are sequentially closed, and as they reach the opener position, they are sequentially opened. The closed gripping sections 301 grip the side edges of film F8 and transport it. In this TD stretching section 03b (TD stretching process), the pair of left and right gripping sections 301 gripping film F8 and the like move in the transport direction and gradually move outward in the width direction (black arrows). As a result, during the TD stretching process, the distance (spacing) between the pair of gripping sections 301 gradually increases in the width direction, and the film F8 and other films gripped by these gripping sections are transported while being stretched in the TD direction.

In addition, this stretching process may also include an oblique stretching process other than the TD stretching process. The TD stretching process stretches the film in a direction parallel to the width direction. That is, in the TD stretching process, film F8 expands at an angle of 0° with the width direction. On the other hand, in the oblique stretching process, in order to adjust the optical properties of the film, both ends of the heated film are held by multiple gripping devices and the film is stretched obliquely at an angle between 40 and 50° with the width direction (referred to as "oblique stretching"). In this case, the cooling process may also include a heat-relaxing process in which the stretched film is continuously transported while being heat-treated to relieve stress.

The trimming unit 05 (trim process) has a cutting unit 05a and a recovery unit 05b. The cutting unit 05a includes two slitters 5110L, 5110R, one on the left and one on the right, and multiple rollers 5120. The slitters 5110L, 5110R are, for example, circular blades or dish-shaped blades rotatably supported on a shaft, and are rotated by a drive unit (not shown). The slitters 5110L, 5110R may rotate in the same direction as the transport direction of the film F8 at the cutting position, at a speed roughly matching that of the film F8, or may rotate in the opposite direction. The two slitters 5110L, 5110R separate the film F8 from the cutting position onwards into a central portion of the film F8 and trimmed films F801, F802 (also called selvages) on both sides.

The central portion of film F8 is the area that will become the product, and its width, for example, ranges from 1,000 mm to 2,500 mm. The widths (also called trim widths) of trimmed films F801 and F802 are both several tens to several hundred mm (for example, 170 mm). The central portion of film F8 is transported downstream in the transport direction and supplied to the subsequent winding process. Meanwhile, trimmed films F801 and F802 are turned 90 degrees downward by roller 5120 and transported, and are collected by collection unit 05b downstream. Collection unit 05b has a rotary cutter and a suction duct (both not shown), and the transported trimmed films F801 and F802 are cut by rotary cutter 522 into small chips of a few millimeters in size, which are collected in the dust box of the downstream vacuum cleaner. If the inspection device 90 is placed before the trimming process, the trim width may be input so that coordinate conversion according to the trim width is performed from the photographed data in pre-processing (step S32 in FIG. 11, described below).

The central portion of the film F8 is transported to the downstream winding section 07. A knurling process may be performed between the trimming section 05 and the winding section 07. In the knurling process, knurls are formed on both ends of the trimmed film F8.

The second drying unit 06 (second drying process) has the same basic configuration as the first drying unit 02, so a detailed explanation will be omitted.

(Winding section 07)
The winding section 07 includes a winder 07a that winds up the film F8, both ends of which have been knurled in the trimming section 05, an entrained air amount control device 07b, a contact or non-contact linear encoder 07c for detecting the running speed of the stretched film F8, a winding shaft rotation speed measuring device 07d, a tension control device 07e, and a thickness measuring device 6f.

As shown in Figure 1, the casting unit 01 dissolves the raw resin in a solvent, and then extrudes the resulting dope, which is prepared by adding various additives such as plasticizers, UV absorbers, anti-degradants, slip agents, and release promoters as needed, onto the endless belt 01a through a die 01b. The cast film formed on the belt 01a is then cast onto the endless support, and after the solvent has been removed to a certain extent, it is peeled off from the belt. It is then passed through the drying unit and stretching unit 03 by various conveying means, where both ends are trimmed and knurled as needed, before being wound up around a take-up shaft in the winding unit 07 to produce an optical film.

The material of the film F8 produced as shown in Figures 1 and 2 is not particularly limited, but typical examples include polycarbonate resin, polysulfone resin, acrylic resin, polyolefin resin, cyclic olefin resin, polyether resin, polyester resin, polyamide resin, polysulfide resin, unsaturated polyester resin, epoxy resin, melamine resin, phenolic resin, diallyl phthalate resin, polyimide resin, urethane resin, polyvinyl acetate resin, polyvinyl alcohol resin, styrene resin, cellulose acetate resin, and vinyl chloride resin. Furthermore, for example, the width of film F8 is preferably 1000 mm to 3200 mm, taking into account productivity, quality, etc. The thickness is preferably 15 μm to 500 μm, taking into account quality, handling, etc.

The length of the optical film F8 in the film roll 80 wound around the winding shaft 82 (see Figure 1; the winding shaft is also called the winding core) is preferably 2000 m to 8000 m, taking into consideration productivity, winding quality, etc. The winding length is a value calculated from the speed and time. The winding speed is, for example, 100 m/min.

The inspection device 90 photographs the film, etc., and generates image data. In the example shown in Figure 1, multiple inspection devices 90 are placed before and after each process, including the stretching process, on the film production line. In Figure 1, the inspection devices 90 are positioned to photograph the surface of the film F8 immediately after the casting process (casting section 01), immediately after the drying process (first drying section 02), and immediately after the trimming process (trimming section 05). The configuration of the inspection device 90 will be described below with reference to Figures 3A to 3C.

(Inspection device 90)
The following types of devices are available for detecting irregularities on the surface of a transparent body such as film F8 as an object to be inspected, as well as bubbles, cracks, distortions in the internal structure, and the like present inside the transparent body.
(1) A transmission type inspection device that detects defects in an object to be inspected by irradiating the object with light and receiving the light that has passed through the object to be inspected.
(2) A reflection type inspection device that detects defects in an object by receiving light reflected from the object.

Furthermore, for both the transmissive and reflective types, there are bright-field inspection devices that receive non-scattered light from the surface and dark-field inspection devices that receive scattered light, depending on the relative positions of the camera's optical axis, light source, and object being inspected. In a bright-field inspection device, if there is no defect, there is no scattering of light, so light from the light source enters the light detection means without being blocked. If there is a defect, the light is blocked by the defect and does not enter the light detection means. Therefore, the defect is observed as a dark dot or streak against a bright background. In contrast, in a dark-field inspection device, if there is no defect, there is no scattering of light, so light does not enter the light detection means. However, if there is a defect, the light is scattered by the defect and enters the light detection means. Therefore, the defect is observed as a bright dot or streak against a dark background. Either type of inspection device may be used as the inspection device 90 in this embodiment. It is preferable, but not limited to, that the first inspection data and the second inspection data be acquired by the same type of inspection device.

(Reflection type inspection device)
FIG. 3A is a schematic diagram showing the configuration of a reflective inspection device 90 as viewed from the width direction (X direction). FIG. 3B is a schematic diagram showing the configuration of the inspection device 90 as viewed from the transport direction (Y direction). The inspection device 90 includes a light source 91, a camera 92 as an optical sensor, an image analysis unit 93 as a data processing device, and a memory unit 94. The inspection device 90 optically inspects feature points (hereinafter simply referred to as defects) that occur on the film F8 during transport. In the inspection device 90, the camera 92 optically inspects the film F8 in the film roll 80 and generates image data as inspection data. The image data includes not only still images but also video data consisting of a time-series of consecutive still images. Furthermore, feature points may be extracted directly from signal data without converting the data into image data. The number of cameras 92, the angle of view, and the distance to the film surface are set so that the entire width of the film F8 is the inspection area (capture range). The number of cameras is determined so that multiple cameras can be arranged in the width direction when a single camera cannot adequately capture the entire width of the film. FIG. 3B illustrates an example in which two cameras 92 are arranged in the width direction (X direction). The image analysis unit 93 may combine multiple images obtained by continuous shooting with one camera 92 to generate a single image data piece containing the entire film surface of the film roll 80, or may store multiple image data pieces in the storage unit 94 in association with the shooting times. Similar image data pieces obtained by multiple cameras 92 arranged in the width direction may also be combined. The image analysis unit 93 can determine the longitudinal position of the film F8 by referencing the stored transport speed (winding speed or unwinding speed) and the shooting times associated with the image data. In the following description, it is assumed that multiple image data pieces obtained by continuous shooting are stored for one film roll 80 in association with the shooting times. The image analysis unit 93 generates defect information by analyzing the image data. The inspection device 90 inspects the long film F8 for defects that occur during the manufacturing process, such as while the film F8 is being wound.

The light source 91 irradiates the inspection area of the film F8 with light. The light source 91 irradiates light uniformly across the width of the rolled film F8 (a direction perpendicular to the longitudinal direction of the film F8 and parallel to the film surface). Here, "uniform" means that the illuminance on the film F8 is approximately the same across the width of the film F8 (e.g., the difference between the maximum and minimum values is less than a specified value).

Camera 92 is an optical sensor that optically reads the inspection area of film F8. Camera 92 is equipped with imaging elements such as a CCD (Charge Coupled Device) or CMOS (Complementary Metal Oxide Semiconductor), lenses, etc. Camera 92 is an area sensor that generates two-dimensional image data from the output signals of each imaging element. Camera 92 detects diffused light that is irradiated by light source 91 and reflected in the inspection area of film F8. Here, either a color camera or a black and white camera (monochrome camera) may be used as camera 92.

The one or more cameras 92 have a shooting range that spans the entire width of the film F8, and in one shooting session, the entire width of the film F8 is simultaneously read. The cameras 92 may be those that detect light in the visible light range or those that detect light in the infrared range.

Furthermore, it is desirable that the contrast between the signal values corresponding to the illuminated areas on film F8 illuminated by light source 91 and the signal values corresponding to the non-illuminated areas not illuminated by light source 91 in the output signal from camera 92 be equal to or greater than a predetermined value. In other words, it is desirable that only the areas on film F8 illuminated by light from light source 91 (illuminated areas) appear bright.

Contrast is expressed as the difference or ratio between two values to be processed (here, the signal value corresponding to the irradiated area and the signal value corresponding to the non-irradiated area), and the greater the difference between the two values, the greater the contrast. To increase the contrast between the irradiated and non-irradiated areas, it is desirable to use a light source 91 that is powerful and highly directional.

Here, "strong" means that when the illuminance at an irradiation distance of 50 mm is E50, the illuminance E50 is 50,000 lx or more. Furthermore, "highly directional" means that when the illuminance at an irradiation distance of 50 mm is E50 and the illuminance at an irradiation distance of 100 mm is E100, the relationship (E50 - E100)/E50<0.5 is satisfied.

The image analysis unit 93 is composed of a CPU, RAM, etc., and reads various processing programs stored in the memory unit 94, loads them into RAM, and performs various processes in cooperation with these programs.

The storage unit 94 is composed of an HDD, SSD (Solid State Drive), etc., and stores various processing programs, data required to execute those programs, etc. The storage unit 94 also stores captured image data (inspection data) linked to the time of capture. The storage unit 94 also stores the winding speed (e.g., 100 m/min) of the film roll manufacturing apparatus 1000, or the unwinding conditions (e.g., 30 m/min) of film F8 in the product manufacturing apparatus 2000. These winding speeds and unwinding conditions may be included in the inspection list of the inspection DB (see table T3 in Figure 8).

The image analysis unit 93 performs data processing on the output signal of the camera 92 (optical sensor) to detect characteristic points (position and intensity) of defects, etc. on the film F8. The data processing includes image processing of the image data obtained from the output signal of the camera 92, defect determination processing to determine defects based on the data after image processing, and quantitative evaluation processing to quantitatively evaluate defects based on the data after image processing.

(Relative positions of camera 92 and light source 91)
The camera 92 may be positioned so as to receive specularly reflected light emitted from the light source 91 (in the case of a bright-field inspection method that receives non-scattered light).

Furthermore, the camera 92 may be positioned to avoid receiving specularly reflected light from the light source 91 (in the case of a dark-field inspection method that receives scattered light). In other words, it is preferable to position the camera 92 in a position that receives diffused light reflected from the object being inspected.

(Transmission type inspection device)
3C shows an example of a transmission type inspection device 90. In this way, a transmission type inspection device 90 in which the light source 91 is disposed opposite the camera 92 with the film F8 sandwiched therebetween may be employed.

FIG. 4 is a schematic diagram showing an application example of an information processing system 50 according to this embodiment. As shown in FIG. 4, the information processing system 50 communicates with terminal devices 70 and the like in factories A and B via a network. The network is a communication line such as a data communication network. Some networks may use a wired LAN, a wireless LAN, or the like (for example, a LAN conforming to the IEEE 802.11 standard).

The terminal device 70 is, for example, a PC (personal computer). For example, the terminal device 70 is a PC used by an employee of a manufacturing company that operates Factory A and Factory B.

Factory A is equipped with the above-mentioned film roll manufacturing apparatus 1000. Factory A is operated or managed by, for example, a film manufacturer. Factory A carries out the first manufacturing process of manufacturing a film roll 80. The film surface of the film roll 80 is inspected by an inspection device 90.

Factory B is equipped with product manufacturing equipment 2000. Factory B is operated or managed, for example, by a coating manufacturer (hereinafter also referred to as a user company or user). There are multiple user companies operating each factory B. Factory B manufactures products using film rolls 80 shipped and transported from factory A. Factory B performs the first or second manufacturing process, which involves unwinding film (film F8, described below) from film roll 80 and coating, or superimposing or adhering it to other films for post-processing.

In this embodiment, as a typical example, the process of manufacturing a film (web) is called the first manufacturing process, and the post-processing carried out at factory B using this manufactured film is called the second manufacturing process. For example, the post-processing process involves a stretching process, or a stretching process and a coating process in which a functional layer is applied to the surface. In the second manufacturing process, the film surface of the film roll 80 is also inspected by inspection device 90.

As a non-typical example of this embodiment, if the film manufactured in factory B undergoes multiple post-processing steps, the first post-processing step may be referred to as the first manufacturing process, and the later post-processing step may be referred to as the second manufacturing process. Also, the film roll 80 manufactured in factory A may be unwound and post-processed in a later process. In this case, the process of manufacturing the film roll 80 may be referred to as the first manufacturing process, and the later post-processing step performed in the same factory A after unwounding the film roll 80 may be referred to as the second manufacturing process. The second manufacturing process performed in factory A includes stretching and drying steps.

(Outline of feature point extraction process)
The feature point extraction process will be described in detail later, but an overview of the feature point extraction process will be described below with reference to Fig. 4. In the following, as a typical example, the process of manufacturing a film (web) will be described as the first manufacturing process, and the post-processing performed at factory B using this manufactured film will be described as the second manufacturing process.

In the first manufacturing process at Factory A, during product inspection, the film is optically inspected by inspection device 90, and inspection data (also referred to as failure data or defect data) is generated. Specifically, inspection data (hereinafter referred to as first inspection data) containing feature point information (feature points, positions, and intensity) is generated by analyzing image data obtained by photographing the film's surface. Information processing system 50 acquires the first inspection data from terminal device 70 at Factory A (Step S1).

Here, feature points are defects on the film, and are generated by analyzing image data. Image analysis can be performed using known techniques to extract, as feature points, pixels in image data captured of the film surface whose pixel values deviate by a specified amount from the average value of the surrounding pixels (the difference is greater than a specified amount). Alternatively, feature points can be calculated using the "image processing for feature point generation" method described below. Tens to thousands of feature points are often generated from one or more image data captured of a single film roll 80 (total length of several km). Defects include both defects that could result in product defects and minor defects that do not result in product defects. Feature points include defects related to poor adhesion when films are bonded together (by ultrasonic welding, etc.), axial irregularities, etc. Feature point information includes size and position (x, y coordinates). Feature point information can also be obtained by grouping (clustering) multiple nearby feature points together. The image processing for feature point generation will be described later.

The film roll 80 manufactured in Factory A is transported to Factory B. In Factory B's second manufacturing process, during product inspection, the film roll 80 is optically inspected by inspection device 90, and inspection data is generated. Image data obtained by photographing the film surface is analyzed to generate inspection data containing feature points (hereinafter referred to as second inspection data). The information processing system 50 acquires the second inspection data from the terminal device 70 in Factory B (Step S2). It is preferable that the inspection device 90 in Factory A (first manufacturing process) and the inspection device 90 in Factory B (second manufacturing process) are the same, i.e., have the same measurement system and the same measurement conditions, but this is not limited to this. Factories A and B may have different required performance, quality, and product standards (hereinafter referred to as product standards, etc.), and an inspection device 90 with an appropriate measurement system and measurement conditions may be used depending on the respective product standards, etc.

The information processing system 50 performs a feature point extraction process by comparing the first and second inspection data for the same film roll 80 (step S3). Specifically, the information processing system 50 performs a feature point extraction process by comparing feature points at corresponding positions on the film surface in the first and second inspection data. In this specification, detecting feature points from image data is referred to as "generating feature points." Comparing the first and second inspection data and classifying the feature points into one of the following first to third type feature points is referred to as "extracting feature points."

Figure 5 is a table explaining the first to third type feature points extracted by the feature point extraction process. Multiple feature points generated from the inspection data of one film roll 80 may be classified into each of the first to third type feature points. For example, some of the hundreds of feature points may be extracted as first type feature points, some as second type feature points, and the rest as third type feature points.

(First type feature point)
The first type feature points are feature points that are present in the first inspection data but not present in the second inspection data. The first type feature points are feature points that disappear in the second manufacturing process (e.g., the coating process). These first type feature points are feature points that do not need to be managed in the first manufacturing process. In this case, the manufacturing conditions that cause the first type feature points to occur may be subject to relaxed standards in the first manufacturing process.

(Second type feature point)
The second type feature points are feature points that are not present in the first inspection data but are present in the second inspection data. The second type feature points are feature points that newly appear in the second manufacturing process. Because these second type feature points are feature points that originate in the second manufacturing process, they can be used to improve the second manufacturing process.

(Third-class feature point)
The third type feature points are feature points that exist in both the first inspection data and the second inspection data. These third type feature points are feature points that originate from the first manufacturing process and require management. Because these third type feature points originate from the first manufacturing process, they can be used to improve the first manufacturing process.

The information processing system 50 may also provide feedback on the feature point extraction results to users such as employees of Factory A and Factory B (Step S4). For example, the extraction results may be sent to terminal device 70 in response to access from the terminal device 70 of the user of the manufacturer of the target film roll 80 and the terminal device 70 of the user to whom the film roll 80 was delivered. This completes the overview of the feature point extraction process. More detailed content of the process will be described later.

(Information Processing System 50)
The information processing system 50 will be described below with reference to Fig. 6 to Fig. 8. Fig. 6 is a block diagram showing a schematic configuration of the information processing system 50. The information processing system 50 is, for example, a server. As shown in Fig. 6, the information processing system 50 includes a control unit 51, a storage unit 52, and a communication unit 53.

(Control unit 51)
The control unit 51 has a CPU and memories such as RAM, ROM, etc. The CPU is a control circuit configured with a multi-core processor or the like that controls each of the above-mentioned units and executes various arithmetic processing in accordance with a program, and each function of the information processing system 50 is realized by the CPU executing the corresponding program.

The control unit 51 functions as an acquisition unit 511 and a reception unit 512 by working in cooperation with the communication unit 53. The control unit 51 also functions as a comparison unit 513, an analysis unit 514, an extraction unit 515, and an output unit 516. The acquisition unit 511 acquires the first and second inspection data obtained by inspection in the first and second manufacturing processes. The reception unit 512 receives input of the amount of correction from the user. The reception unit 512 may also receive input of the expansion/contraction ratio from the user. Here, the amount of correction includes the X direction, Y direction, θ direction, r direction, and front/back correction. In the following, the correction amount will be described as the X direction, Y direction, and θ direction. In the following, the received expansion/contraction ratio and correction amount will be collectively referred to as the correction amount, etc. The comparison unit 513 extracts feature points from each of the first and second inspection data, and performs a comparison process using the received correction amount, etc. The comparison unit 513 searches for corresponding feature points between the first and second test data through a comparison process. In the comparison process, the comparison unit 513 generates feature point descriptors (descriptor 2, described below) for the two test data (image data), performs feature point matching between the test data using the feature point descriptors, and outputs the comparison results (a corresponding point list shown in FIG. 16, described below). The analysis unit 514 calculates an index indicating the relationship (hereinafter referred to as relationship information) using the corresponding point list. The relationship information includes graphs and statistical information such as averages, regression equations, variances, standard deviations, and correlation coefficients. Graphs include scatter plots, histograms, Pareto charts, bubble charts, and the like. Regression equations include regression lines and polynomial regressions of second or higher order. In addition to simple regression, regression methods such as multiple regression and logistic regression may be used for the regression equation (regression model). The main extraction unit 515 uses the comparison results to extract (classify) first to third type feature points. The output unit 516 transmits the feature point extraction results to the terminal device 70 or displays them on a display unit (not shown) in response to a request from the terminal device 70, etc.

(Storage unit 52)
The memory unit 52 is a large-capacity auxiliary storage device that stores various programs including an operating system and various data. For example, a hard disk, a solid-state drive, a flash memory, a ROM, etc. are used as the storage. The memory unit 52 stores a user list, a lot list, an inspection data DB, etc. Of these, the user list and lot list are managed and registered by an administrator who accesses the terminal device 70. For example, this administrator is a person in charge of the relevant department at the manufacturer that operates Factory A.

(User list)
7 shows an example of various data stored in the storage unit 52. Table T1 shown in FIG. 7 is an example of a user list. The user list stores user IDs, user names, contact information, etc. Each user is assigned access rights to a search data DB (inspection database), and is granted access rights to various data (inspection data, extracted data, etc.) related to the film roll 80 (identified by lot ID) that the user is involved in.

(Lot list)
Table T2 in Fig. 7 is an example of a lot list. The lot list records a lot ID assigned to each film roll, a product name (also called a type), a delivery destination user ID (orderer), multiple manufacturing conditions, size (width, length, thickness), manufacturing date, etc.

(Test data DB)
FIG. 8 is an example of an inspection DB (database) stored in the storage unit 52. The inspection data DB stores data related to the inspection of various film rolls 80, such as the first and second inspection data and feature point extraction results shown in FIG. 8. As described above, the first inspection data is data obtained from the inspection in the first manufacturing process. The second inspection data is data obtained from the inspection in the second manufacturing process. The feature point extraction results are data generated by the information processing system 50 using the first and second inspection data.

Table T3 shown in Figure 8 is an example of an inspection list registered in the inspection data DB. The inspection list stores the inspection ID, lot ID, inspection device ID, inspection data, inspection date and time, etc.

Table T4 in Figure 8 shows an example of the contents of inspection data (inspection ID: i0101) in the inspection list. The inspection data contains feature point IDs, which are automatically assigned consecutive numbers to each feature point, and feature point descriptors 1 and 2 (hereinafter simply referred to as descriptor 1, etc.) for each feature point ID.

Descriptor 1 is information about a feature point alone, recording its XY coordinate position and intensity. Intensity is the rank of the feature point, which will be described later. Intensity information may also include information about the feature point's size (diameter, area) and brightness. The XY coordinate position is based on the origin of the film surface (for example, the left edge of the leading edge). X is the coordinate in the width direction of the film, and can range from 0 to 3000 mm, for example, depending on the film size (see Table T2). Y is the coordinate in the length direction of the film, and can range from 0 to 10000 m, for example, depending on the film size. Descriptor 1 is generated by the image analysis unit 93 of the inspection device 90.

Descriptor 2 is peripheral information, and is vector or array information that represents surrounding information such as relationships with other feature points. For example, SIFT features may be used as descriptors, or the probability density function of feature points calculated by kernel density estimation may be used as descriptors. Descriptor 2 is mainly generated by the comparison unit 513.

Table T5 in Figure 8 is an example of extraction result data (hereinafter simply referred to as extracted data). The extracted data records the inspection IDs of the original first and second inspection data, as well as the extraction results for each feature point. The extraction results (Types 1 to 3) are classified as shown in Figure 5 above. Integrated feature point IDs are automatically assigned consecutive numbers, and integrated feature points are generated corresponding to feature points found in either or both of the first and second inspection data. The number of integrated feature point IDs is greater than or equal to the number of first inspection and second inspection feature point IDs. The extracted data is updated as correction amounts, etc. are input. This is because descriptors 1 and 2 are updated by accepting correction amounts, etc., as described below (steps S402 and S502 in Figures 14 and 15), and the updated descriptors 1 and 2 are compared.

It should be noted that multiple first and second inspection data may be generated for one lot ID by multiple inspection devices. For example, in the second manufacturing process, the film roll 80 in its original wound state may be inspected (photographed), and multiple second inspection data may be generated by inspections in several downstream processes. In this case, the information processing system 50 may generate multiple feature point extraction result data for one piece of first inspection data in a one-to-many relationship with multiple pieces of second inspection data. The user may also be able to select which second inspection data to associate with (for example, buttons b1 and b2 in Figure 12, described below).

(Communication unit 53)
The communication unit 53 also serves as an interface for network connection with an external device such as a PC.

(Generation of first and second test data)
The process of generating the first and second inspection data performed in the first and second manufacturing steps will be described below with reference to Figures 9 and 10. Figure 9 is a flowchart showing the process of generating the first inspection data performed in the first manufacturing step.

(First test data generation process)
(Step S11)
In the typical example described above, in the first manufacturing step, the film roll 80 is manufactured by the film roll manufacturing apparatus 1000 .

(Step S12)
The inspection device 90 photographs the film and stores the image data. The inspection device 90 is the same as that described with reference to FIG. 3A and other figures.

(Step S13)
The image analysis unit 93 performs image processing, which will be described below, on the image data to generate a plurality of feature points.

(Image processing for generating feature points)
The image analysis unit 93 acquires two-dimensional image data generated by the camera 92 and stored in the storage unit 94 .

The image analysis unit 93 performs data processing on the image data (examination data) acquired from the camera 92.

The image analysis unit 93 divides the image data into multiple regions. For example, the image analysis unit 93 divides the image data into n regions (e.g., several to several tens) in the width direction (hereinafter referred to as regions a1 to an).

Next, the image analysis unit 93 acquires image data for one area a1 and performs mathematical processing on the image data for area a1. Appropriate mathematical processing is provided depending on the type of defect to be detected (gauge band, vertical wrinkles, diagonal wrinkles, uneven bright spots, light leaks, etc.).

Mathematical processing includes preprocessing, enhancement processing, signal processing, image feature extraction, etc.

Pretreatment includes the following:
- Image cropping,
- Low-pass filter, high-pass filter, Gaussian filter, median filter, bilateral filter,
-Morphological transformation, color transformation (L*a*b*, sRGB, HSV, HSL), contrast adjustment, noise removal, restoration of blurred and shaken images, mask processing, Hough transform, projection transformation, etc.

Examples of enhancement processing include the Sobel filter, Scharr filter, Laplacian filter, Gabor filter, and Canny algorithm.

The signal processing includes the following:
- Basic statistics (maximum, minimum, average, median, standard deviation, variance, quartile), square root of sum of squares, difference, sum, product, ratio, distance matrix calculation, differential and integral calculus, threshold processing (binarization, adaptive binarization, etc.),
-Fourier transform, wavelet transform, peak detection (peak value, number of peaks, half-width, etc.), etc.

Examples of image feature extraction include template matching and SIFT features.

Next, the image analysis unit 93 performs threshold processing on the values (feature amounts) obtained by mathematical processing of the image data of area a1. Threshold processing is a process that determines whether or not the defect is the target of detection based on a predetermined defect determination threshold, and also determines the rank (intensity) of the defect.

In threshold processing, determining the presence and type of defect corresponds to "defect determination processing." Also, in threshold processing, classifying defects into multiple ranks according to the threshold corresponds to "quantitative evaluation processing."

For example, defects are classified into multiple ranks for a parameter (feature) that takes a value between 1 and 100. For example, ranks are classified according to the size (diameter or area) of the defect. Ranks classified by size may also be further subdivided according to the parameter value.

The image analysis unit 93 performs similar processing on areas other than area a1.

After processing each of the regions a1-an, the image analysis unit 93 integrates the results for each of the regions a1-an, and data processing ends. Specifically, the image analysis unit 93 generates data that associates the rank of the detected defects with their location (x and y coordinates) for each region (each position in the width direction of the film F8).

After data processing, the image analysis unit 93 stores the results of the data processing in the storage unit 94. The image analysis unit 93 performs this type of data processing on each of the multiple image data obtained in the inspection of one film roll 80, and obtains the processing results. By aggregating these processing results, inspection data such as that shown in Table T4 in Figure 8 is generated.

(Step S14)
Terminal device 70 in the first manufacturing process sends inspection data including information on multiple feature points obtained through the processing up to step S13 to information processing system 50. Acquisition unit 511 of information processing system 50 stores the acquired inspection data in the inspection data DB of storage unit 52 as first inspection data.

(Generation process of second test data)
FIG. 10 is a flowchart showing the process of generating second inspection data performed in the second manufacturing process.

(Step S21)
In the second manufacturing process, after the first manufacturing process, the product manufacturing apparatus 2000 performs post-processing using the film roll 80 to manufacture a product using the film F8.

(Step S22)
The inspection device 90 photographs the surface of the film F8 before post-processing, or during or after post-processing, and stores the image data.

(Step S23)
The image analysis unit 93 stores the generated inspection data including the feature point information of the plurality of feature points in the storage unit 94 by the same process as in step S13.

(Step S24)
Terminal device 70 in the second manufacturing process sends inspection data including information about multiple feature points obtained through the processing up to step S23 to information processing system 50. Acquisition unit 511 of information processing system 50 stores the acquired inspection data as second inspection data in the inspection data DB of storage unit 52. The second inspection data is described by feature point IDs and feature point descriptors 1 and 2, similar to the first inspection data shown in Table T4.

(Feature point extraction process)
The feature point extraction process executed by the information processing system 50 will be described below with reference to Figs. 11 to 18. Fig. 11 is a flowchart showing the feature point extraction process. Fig. 12 is an example of an operation screen displayed on the terminal device 70. Fig. 13 is a schematic diagram for explaining the feature point extraction process.

The information processing system 50 starts the processing from step S31 onward in response to a start instruction from the user via the operation screen on the terminal device 70, or when the second inspection data is registered in the inspection data DB in the storage unit 52 and a pair of first and second inspection data is obtained. Figure 12 is an example of the operation screen 701 displayed on the terminal device 70. After selecting a lot, the user selects the first and second inspection data from the multiple inspection data linked to that lot using buttons b1 and b2. The second inspection data is inspection data obtained in a process downstream of the first inspection data. In addition, in this embodiment, it is described assuming that a stretching process occurs between the positions where the first inspection data and the second inspection data were collected.

(Step S31)
The acquisition unit 511 acquires the same lot, i.e., a pair of first and second inspection data, from the inspection data DB. The same lot is linked by the lot ID assigned to the inspection ID. The same lot is linked by the lot ID assigned to the inspection ID. Note that a branch ID or a new lot ID may be assigned to the lot ID in one of the processes, and the lot ID may not correspond one-to-one, but may correspond one-to-n or n-to-1. For example, a 3,000-meter film lot may be divided into three 1,000-meter pieces in a subsequent process in a pre-process. In this case, a branch ID of the lot ID or a new lot ID is assigned in the subsequent process. Conversely, two or three film rolls may be spliced in a pre-process to form a single film for use in a subsequent process. In this case, the pre-process has two or three lot IDs, and the subsequent process has one lot ID. In either case, the correspondence between the lot IDs in each process is described in the lot list (see FIG. 7 ).

(Steps S32 and S33)
The comparison unit 513 performs preprocessing on the first inspection data under a first condition, and performs preprocessing on the second inspection data under a second condition.

As shown in FIG. 13, as preprocessing for the second condition, the comparison unit 513 performs preprocessing to reverse the Y coordinate (up and down) of the second inspection data to match any differences between winding (first manufacturing process) and unwinding (second manufacturing process). Furthermore, in the second manufacturing process, the comparison unit 513 performs preprocessing to reverse the X coordinate (left and right) of the second inspection data (or first inspection data) depending on whether the camera 92's shooting area is set to the front or back of the film F8. Furthermore, in the second manufacturing process, if the film F8 expands or contracts due to the heating temperature and tension settings in post-processing, and the expansion rate can be estimated in advance, the comparison unit 513 performs coordinate conversion for the first inspection data or second inspection data using the expansion rate received by the reception unit 512.

Furthermore, the comparison unit 513 performs at least one of the following noise removal processes on the first and second test data as the noise removal processes included in the first and second conditions.
(1) Removal of low-rank feature points.
(2) Remove extremely small feature points.
(3) Eliminate consecutive hits.
(4) Eliminate concentrated impacts in the width direction. This occurs especially at the leading and trailing edges of film F8.

(Step S34)
The accepting unit 512 accepts the expansion/contraction ratio from the user. For example, the accepting unit 512 accepts input of the expansion/contraction ratio through an operation screen 701 displayed on the terminal device 70 as shown in FIG. 12. The user presses buttons b3 and b4 within a predetermined range (for example, 50 to 300%).
Any numerical value can be entered using the numeric keypad. The reception unit 512 receives input of the expansion/contraction rate in the Y direction (longitudinal direction) through user input using button b3. The reception unit 512 also receives input of the expansion/contraction rate in the X direction (widthwise direction) through user input using button b4. The operation screen 701 in FIG. 12 shows a state in which 140% has been entered for each value. If the numerical value is greater than 100%, it is assumed that the film in the second inspection data has stretched more than the film in the first inspection data.

In Figure 12, the stretching unit 03 performs TD stretching in the X direction and MD stretching in the Y direction, and therefore an example is shown in which the stretching ratios for both the X direction and the Y direction are accepted accordingly. This is not limiting; for example, if the stretching unit 03 performs a diagonal stretching process, the accepting unit 512 may accept the stretching ratio for the diagonal stretching. In this case, the accepting unit 512 accepts input of a stretching angle value within a predetermined range (40 to 50°).

(Step S35)
The comparison unit 513 searches for feature points in one test data that are identical to or correspond to feature points in the other test data by comparison process 1, and matches the feature points with each other. Fig. 14 is a subroutine flowchart showing the processing of step S35.

(Step S401)
The comparison unit 513 shifts the position of one of the feature points in the first and second inspection data according to the expansion/contraction ratio received in step S34. In other words, the comparison unit 513 converts the coordinate system. In the following description, the comparison unit 513 shifts the feature points in the first inspection data according to the expansion/contraction ratio. For example, in the input example shown in FIG. 12 , the comparison unit 513 converts the Y coordinate of each feature point in the first inspection data into a value multiplied by the expansion/contraction ratio, with the origin (0) as the base. Here, the origin is the leading position (start of winding) of film F8 during production of the film roll 80 (the end when unwound). The comparison unit 513 also converts the X coordinate of each feature point in the first inspection data into a value multiplied by the expansion/contraction ratio, with the center point as the base. In other words, the first inspection data is transformed so that it expands on both sides of the center point. Here, the center point is the center position of the film in the width direction in the first inspection data. As another example, in the case of oblique stretching, the comparison unit 513 calculates the stretch rate in the X direction and the stretch rate in the Y direction corresponding to the position in the width direction of the film in the first inspection data in accordance with the input stretch angle and stretch rate, and applies these. As yet another example, when converting the second inspection data side, the comparison unit 513 can perform coordinate conversion using the same processing as when converting the first inspection data side, except for dividing by the stretch rate.

(S402)
The comparison unit 513 generates a descriptor 2 for each feature point of the first and second test data. For example, the comparison unit 513 uses a SIFT feature amount as the descriptor, or a probability density function of the feature points calculated by kernel density estimation as the descriptor.

(Steps S403 and S404)
The comparison unit 513 compares the feature points of the first and second test data to search for the most similar feature points. The comparison unit 513 evaluates the similarity between the feature points using Descriptor 1 and Descriptor 2, and regards the most similar feature points as corresponding feature points.

For example, when comparing feature points between the first and second test data, the comparison unit 513 searches for feature points in the first test data that correspond to feature points of interest in the second test data. In this case, the comparison unit 513 determines that feature points in the first test data that match the X and Y coordinates of descriptor 1 of the feature point in the second test data or that have the closest intensity are the same point. Alternatively, the comparison unit 513 determines that feature points with the closest distance (Euclidean distance) between the X and Y coordinates of descriptor 1 are the same point (corresponding points). The comparison unit 513 excludes feature points that are farther apart than a set threshold from the determination of corresponding points. Note that in this embodiment, the threshold in the Y direction is set to a value that is two to three orders of magnitude larger than the threshold in the X direction. For example, the threshold in the X direction is several millimeters, and the threshold in the Y direction is several meters. The threshold in the Y direction is two to three orders of magnitude larger than the threshold in the X direction because there is a greater amount of change.

Furthermore, the comparison unit 513 may use descriptor 2 in addition to descriptor 1 to extract feature points that are most similar based on the distance between vectors in high-dimensional vector space. In this case, the comparison unit 513 may use a probability density function calculated by kernel density estimation as described above as descriptor 2. Figure 15 is an example of a probability density function calculated by kernel density estimation that indicates the position and intensity (density) of feature points. In Figure 15, the vertical and horizontal axes are XY coordinates, and it is shown that the higher the concentration, the higher the density.

The comparison unit 513 determines that one feature point corresponds to only one other feature point. For example, for a feature point in the second test data, the feature point in the first test data whose descriptor vectors are closest to the feature point in the second test data is registered as the corresponding point in the corresponding point list.

Figure 16 is an example of a corresponding point list stored in the storage unit 52. The corresponding point list associates each feature point in the second inspection data with the most similar feature point in the first inspection data. The corresponding point list also describes the X and Y coordinates of the feature points in the second inspection data, the Euclidean distance between the associated feature points, the difference dx in the X coordinate, and the difference dy in the Y coordinate. A ratio may also be used as the difference.

(Step S405)
The analysis unit 514 generates correlation information between feature points that exist in both the first and second test data. That is, the analysis unit 514 generates one of the following four types of scatter diagrams as data indicating correlation information for feature points that have corresponding feature points in the corresponding point list. Furthermore, the analysis unit 514 may perform principal component analysis on the following four relationships and generate a scatter diagram projected onto a coordinate system that maximizes the variance of the data.
(1) Horizontal axis: X coordinate, vertical axis: dx (X-dx scatter diagram),
(2) X-coordinate on the horizontal axis, dy on the vertical axis (X-dy scatter diagram),
(3) Horizontal axis: Y coordinate, vertical axis: dy (Y-dy scatter diagram),
(4) Horizontal axis is Y coordinate, vertical axis is dx (Y-dx scatter diagram).

FIG. 17A is an example of a scatter plot corresponding to (1) above. FIG. 17B is an example of a scatter plot corresponding to (2) above. FIG. 17C is an example of a scatter plot corresponding to (3) above. FIG. 17D is an example of a scatter plot corresponding to (4) above. FIG. 18 is another example of a scatter plot corresponding to (1) above, showing bowing-like expansion and contraction during the stretching process. Common to these figures, the horizontal axis indicates the X coordinate (or Y coordinate) of the feature point in the second inspection data. Furthermore, the analysis unit 514 may calculate statistical information (e.g., regression equations, correlation coefficients, etc.) as correlation information instead of or in addition to the scatter plot. In FIGS. 17A-17D and 18, the regression equations for the scatter data are also shown. Furthermore, a histogram, Pareto chart, or bubble chart may be used instead of the scatter plot.

(Step S36)
The output unit 516 displays and outputs the correlation information or outputs an alert. For example, it displays and outputs the correlation information on the terminal device 70. The output unit 516 displays the correlation information on the operation screen of the terminal device 70. The output unit 516 may also notify the user of the alert by audio information such as an alarm sound or tactile information such as vibrating the mobile terminal. FIG. 19 shows an example of an operation screen 702 for accepting correction amounts, which displays a scatter plot as correlation information. This operation screen 702 is displayed following the operation screen 701. For example, the scatter plot as correlation information generated in step S405 is displayed in the preview area b10 of the operation screen 702 shown in FIG. 19. Note that the preview area b10 may also display a correlation coefficient. The user can determine whether the correspondence is appropriate by referring to the correlation information displayed in the preview area b10 as shown in FIG. 19. By referring to a graph such as a scatter plot, the user can determine whether the correspondence between the feature points of the two test data is satisfactory by reviewing the correction amounts.

As a variation, the display range may be configured so that the user can set the display range (preview X-axis) in the longitudinal or lateral directions, and the preview displayed within that range. In this case, the amount of correction may be applied to a section of the display range rather than the entire range, or a separate setting may be configured so that the setting of a section is applied individually to each section.

As another variation, only the correlation coefficient may be displayed, or statistical information such as the correlation coefficient may be displayed in a window in the preview area b10.

Furthermore, as a further modification, the following measures may be taken. For example, when a correction amount is input and the feature points recalculated using this correction amount are displayed in the preview area b1, the plot before the correction amount is applied (result data) and the plot after the correction amount is applied may be displayed side by side, or may be displayed superimposed by using different colors. This allows the user to compare the result data before and after correction and understand the validity of the correction amount they input.

(Step S37)
If the user determines by looking at the scatter diagram that the correlation is poor, the user inputs a correction amount to see if it can be improved. In response to this, the control unit 51 advances the process to step S50 in response to the reception unit 512 receiving the correction amount from the user (YES).

The types of correction amounts accepted by the accepting unit 512 include the following: Note that inversion correction may also be included as a correction amount. Inversion correction is a correction in which the X coordinate or Y coordinate is inverted relative to a reference (the center of the width or the end of the length).
(a) Tilt correction amount (rotation) (θ direction),
(b) X coordinate correction amount (X direction),
(c) Y coordinate correction amount (Y direction).

(a) By accepting the tilt correction amount, the coordinates of one of the two test data are transformed. This tilt correction amount corresponds to the expansion/contraction ratio. For example, in Figures 17 and 18, applying the tilt correction amount has the same effect as rotating the entire scatter plot. The input of (a) may also be accepted as (a11) an X correction amount at the X coordinate correction position, or (a12) a Y correction amount at the Y coordinate correction position. (b) For the X coordinate correction amount, the comparison unit 513 shifts the X coordinate of the entire test data in accordance with the input correction amount. For example, the comparison unit 513 shifts the X coordinate of the first test data in accordance with the input correction amount. In this case, in Figure 18, the same effect occurs as if the entire scatter plot were shifted up or down. The same applies to (c) the Y coordinate correction amount. In this case, the Y coordinate of the entire test data in accordance with the input correction amount. For example, in the example of Figure 17, the entire data is shifted up or down in accordance with the input Y coordinate correction amount.

The input example shown in Figure 19 is an example of (a12). The user selects the Y-dy scatter plot with button b11 on the operation screen 702 in Figure 19, and in response to the selection, a Y-dy scatter plot like the one shown in Figure 17 is displayed in the preview area b10. The user enters -6.0 m in the input area b12 and 0 m in b13. By operating the recalculate button b14, a recalculation is performed based on this input correction amount, and the display in the preview area b10 is updated (redisplayed). In this case, the center of the regression equation is fixed, and the leftmost side (b13: Y = 0 m) moves by -6.0 m. In other words, the entire coordinate system rotates counterclockwise so that the left end of the regression equation moves down by 6.0 m.

(Step S50)
FIG. 20 is a subroutine flowchart showing the comparison process 2 in step S50.

(Step S501)
The comparison unit 513 moves the position of one of the feature points of the first and second test data in accordance with the correction amount (hereinafter referred to as the input correction amount) received in step S37. That is, the comparison unit 513 converts the coordinate system. In the following description, the comparison unit 513 is assumed to move the feature point of the first test data in accordance with the input correction amount.

For example, in the input example of Figure 19, the comparison unit 513 shifts the Y coordinate of the feature point whose Y coordinate is at the origin (0 m) using the input correction amount (-6.0 m). For Y coordinates other than the origin, the Y coordinate of the feature point is shifted by a value obtained by multiplying the input correction amount (-6.0 m) by the ratio of the distance to the center position ((yc-y)/yc). Here, yc is the coordinate of the center position. yc is half the value of the distance from the origin to the end of the data. The end is the last Y coordinate data of the first inspection data, and corresponds to the length of the film on the roll (the size of table T2 in Figure 7). Figure 21 shows an example in which a correction amount is applied. The same is true for the X coordinate; the coordinate of the feature point is shifted according to the user's input correction amount. In other words, by shifting the coordinate of the feature point through these processes, the entire system rotates counterclockwise (see arrow) around the axis of rotation on the regression equation, as shown in Figure 21.

(Steps S502 to S505)
The processing here is the same as steps S402 to S405 in FIG. 14. The comparison unit 513 generates new feature point descriptor information using the coordinates converted in step S501, performs a matching process between feature points, associates corresponding feature points, and stores the associated feature points in a corresponding point list. The analysis unit 514 references the updated corresponding point list and generates correlation information between feature points present in both the first and second test data. The correlation information includes at least one of a scatter plot and statistical information, as described above. This completes the processing in FIG. 20, and the processing returns to the processing in FIG. 11. For example, a scatter plot like the one in FIG. 21 is displayed in the preview area b10 of the operation screen 702 in FIG. 19.

(Step S51)
The output unit 516 again displays the relevance information. The processing here is the same as that in step S36. For example, the preview area b10 in FIG. 19 is the feature point after the coordinates have been moved in accordance with the received correction amount, and the generated relevance information is displayed in the preview area b10.

(Step S37: NO)
When the user determines that the correlation is correct to a certain extent by referring to the correlation information such as a scatter diagram, or when the user determines that correction (re-correction) of the correction amount is unnecessary (NO), the user presses the register button b16 on the operation screen 701. In response to the operation of the register button b16, the control unit 51 advances the process to step S38.

(Step S38)
FIG. 22 shows an example of the feature point extraction conditions that are determined and recorded in the inspection DB of the storage unit 52.
The control unit 51 records the expansion/contraction ratio and correction amount received in steps S34 and S37 in association with the lot and the combination of the first and second inspection data. Thereafter, when using film rolls of the same type as the lot, the control unit 51 can read out the feature point extraction conditions to eliminate the need for adjustments, and by comparing the same conditions with the previous lot, it is possible to understand the differences between the lots. Note that feature point extraction may be performed and recorded using intermediate correction amounts before they are finalized. For example, feature points may be extracted (classified) according to the correction amount each time a correction amount is input.

Furthermore, the extraction unit 515 classifies third-type feature points, which are matched feature points in the first and second inspection data, into two types: third-type a (rank A) and third-type b (rank B) based on their distance from the regression equation. Figure 23 is a schematic diagram illustrating classification based on distance from the regression equation. The extraction unit 515 classifies feature points whose distance from the regression equation is less than a predetermined threshold into third-type a, and feature points whose distance exceeds the predetermined threshold into third-type b. This classification is performed using the Y-dy scatter diagram (3) described above, but other scatter diagrams (e.g., X-dx) may also be used. Furthermore, classification may be performed based on both the distance from the regression equation of the Y-dy scatter diagram and the distance from the X-dx scatter diagram, and the two may be compared and presented. For example, if both are classified as rank A, the scatter diagram is classified as A1 (rank S), and if only one of them is ranked A, the scatter diagram is classified as A2.

The distance between the feature point and the regression equation is measured using the distance on the vertical axis (dy), but it may also be measured using Euclidean distance. The predetermined threshold uses a preset value, but may also be set using statistical information such as standard deviation (σ). The extraction unit 515 classifies feature points that fall within the range of +1σ to -1σ from the regression equation into a third type a, and other feature points into a third type b. The extraction unit 515 may record the separation of the third types a and b in a corresponding point list. Feature points classified as the third type a with rank A have a higher degree of confidence that they are the same feature point.

(Step S39)
The extraction unit 515 classifies feature points from the correspondence list, the first test data, and the second test data. Specifically, the extraction unit 515 classifies, in the correspondence list, feature points of the second test data that are not associated with feature points of the first test data as second-type feature points. Furthermore, the extraction unit 515 classifies, in the correspondence list, feature points of the first test data that are not included in the correspondence list (feature points that are not associated with feature points of the second test data) as first-type feature points (see Table T5 in FIGS. 5 and 8 ).

(Step S40)
The output unit 516 registers the extraction results (extraction data) generated in step S38 in the test data DB, and transmits the extraction results to the terminal device 70. This completes the feature point extraction process shown in Fig. 11 (END).

As described above, the information processing system according to this embodiment includes a comparison unit that compares first feature point information of a web in the first inspection data with second feature point information of a web in the second inspection data and extracts identical feature points in the first and second inspection data. The information processing system also includes an analysis unit that generates relationship information indicating the relationship between the coordinate difference between the first inspection data and the second inspection data for each of the feature points extracted by the comparison unit and the coordinate position on the entire web, and an output unit that outputs the relationship information. This allows efficient collection of information useful for process improvement and shipping standard setting both during the production of film or web for film rolls and during post-processing processes using such film or web. In particular, by checking the displayed relationship information, the user can determine whether the feature points in the first and second inspection data are correctly matched. Furthermore, by accepting a correction amount input by the user, more accurate feature point matching (third-type feature points) can be achieved, which ultimately leads to more accurate classification of the first, second, and third-type feature points.

Second Embodiment
In the second embodiment, an alert is issued in accordance with the correlation coefficient in a scatter diagram. Fig. 24 is a flowchart showing the process of extracting feature points in the second embodiment.

In Figure 24, steps S71 to S76 and S77 to S91 correspond to steps S31 to S36 and S37 to S51, respectively, in the first embodiment shown in Figure 12. In the second embodiment, step S765 is performed.

(Step S765)
The analysis unit 514 calculates the correlation coefficient, and issues an alert if the correlation coefficient is equal to or less than a first predetermined threshold. For example, if the absolute value of the correlation coefficient is 0.5 or less, an alert is displayed on the terminal device 70. Alternatively, if the regression equation is a regression line and its slope is close to zero (i.e., flat), the analysis unit 514 calculates the variance of the difference value (dy or dx) of the feature points, and if the variance is equal to or greater than a second threshold, an alert is displayed.

Figure 25 shows an example of a distribution diagram and an alert display. Figure 25 (A) shows a case where the correlation coefficient is high or the variance of the difference dy is small, indicating normality. On the other hand, Figures 25 (b) and (c) show a case where the correlation coefficient is low or the variance of the difference dy is large, indicating abnormality. In such cases, the output unit 516 displays an alert on the display unit of the terminal device 70. This allows the user to understand that some kind of malfunction has occurred.

The information processing system 50 described above is the main configuration used to explain the features of the above embodiment, and is not limited to the above configuration, and various modifications can be made within the scope of the claims. Furthermore, configurations that are included in general information processing devices/systems are not excluded. For example, in this embodiment, it is assumed that there is a stretching process between the positions where the first test data and the second test data are collected, and although an example of accepting an expansion/contraction ratio is shown in Figures 11 and 12, the process of accepting the expansion/contraction ratio (step S34) may be omitted. For example, if there is no stretching process between the first and second test data, the acceptance of the expansion/contraction ratio is omitted. In this case, the comparison unit 513 performs the comparison process at step S35 using a fixed expansion/contraction ratio.

Furthermore, when a lot is selected, if there is a history of producing products under the same production conditions in the past, and the correction amounts, etc. set and confirmed by the user are recorded as feature point extraction conditions (see Figure 22), the correction amounts, etc. applied to the same product in the past can be read out and presented to the user. This reduces the number of trial and error steps required by the user, making it easier to set the appropriate correction amounts.

Furthermore, for example, the information processing system 50 may include an inspection device 90 arranged in the first manufacturing process and/or the second manufacturing process. The feature point generation function of the image analysis unit 93 of the inspection device 90 may also be performed by the control unit 51 of the information processing system 50. In this case, the inspection device 90 sends image data of an image of the film surface and the shooting conditions (information such as transport speed, camera direction, and angle of view) to the information processing system 50, and the feature point generation process is performed on the control unit 51 side.

The means and methods for performing various processes in the information processing system 50 according to the above-described embodiment can be realized by either a dedicated hardware circuit or a programmed computer.
The program may be provided by a computer-readable recording medium such as a Digital Versatile Disc-ROM, or may be provided online via a network such as the Internet. In this case, the program recorded on the computer-readable recording medium is usually transferred to and stored in a storage unit such as a hard disk. The program may also be provided as standalone application software, or may be incorporated into the software of a device as one of its functions.

Although embodiments of the present invention have been described and illustrated in detail, the disclosed embodiments are made for purposes of illustration and example only and are not intended to be limiting. The scope of the present invention should be interpreted by the language of the appended claims.

This application is based on a Japanese patent application (Patent Application No. 2024-056247) filed on March 29, 2024, the disclosure of which is incorporated herein by reference in its entirety.

50 Information processing system 51 Control unit 511 Acquisition unit 512 Reception unit 513 Comparison unit 514 Analysis unit 515 Extraction unit 516 Output unit 52 Storage unit 90 Inspection device 1000 Film roll manufacturing device 2000 Product manufacturing device

Claims (10)

an acquisition unit that acquires first inspection data in a first manufacturing process in which a web is manufactured or a manufactured web is post-processed, and second inspection data in a second manufacturing process in which post-processing using the web is performed after the first manufacturing process;
a comparison unit that compares first feature point information of the web in the first inspection data with second feature point information of the web in the second inspection data and extracts identical feature points between the first and second inspection data;
an analysis unit that generates relationship information indicating a relationship between a difference value between the coordinates of each of the plurality of feature points extracted by the comparison unit in the first inspection data and the second inspection data and a coordinate position on the entire web;
an output unit that outputs the relationship information. The information processing system of claim 1, wherein the relationship information is a graph showing the relationship between coordinate difference values and overall coordinate positions. The information processing system of claim 1 or claim 2, wherein the relationship information is statistical information indicating the relationship between coordinate difference values and overall coordinate positions. a receiving unit that receives a correction amount of the coordinates of the first inspection data and/or the second inspection data,
2. The information processing system according to claim 1, wherein the comparison unit corrects the coordinates of the feature points with the correction amount received by the reception unit, then compares the first feature point information of the web in the first inspection data with the second feature point information of the web in the second inspection data, and extracts identical feature points in the first and second inspection data. the receiving unit receives the correction amount after the output unit displays the relationship information;
In response to the reception unit receiving the correction amount, the comparison unit extracts the same feature points again, the analysis unit generates the relationship information, and
The information processing system according to claim 4 , wherein the output unit displays the relationship information again. the relationship information is a scatter diagram and a regression equation that represent the relationship between the coordinate difference values and the overall coordinate positions;
The information processing system according to claim 1 , further comprising an extraction unit that classifies the feature points in the scatter diagram according to a distance from the feature points to the regression equation. The relationship information includes a correlation coefficient indicating a relationship between a difference value of the coordinates and an overall coordinate position, or a variance of the difference value;
2. The information processing system according to claim 1, wherein the analysis unit generates an alert when an absolute value of the correlation coefficient is equal to or less than a first predetermined threshold value or when the variance is equal to or greater than a second predetermined threshold value, and the output unit issues an alert. A method for extracting feature points on a web, comprising:
A step (a) of acquiring first inspection data in a first manufacturing process of manufacturing a web or post-processing a manufactured web;
(b) acquiring second inspection data in a second manufacturing process that uses the web and is performed after the first manufacturing process;
(c) comparing first feature information of the web in the first inspection data with second feature information of the web in the second inspection data;
(d) extracting identical feature points between the first and second inspection data based on the comparison result of the step (c);
a step (e) of generating relationship information indicating a relationship between a difference value between the coordinates of each of the plurality of feature points extracted in the step (d) in the first inspection data and the second inspection data and a coordinate position on the entire web;
and (f) outputting the relationship information generated in the step (e). The feature point extraction method of claim 8, wherein the relationship information is a graph showing the relationship between coordinate difference values and overall coordinate positions. A control program for causing a computer to execute the extraction method described in claim 8 or claim 9.