WO2025084668A1 - Apparatus and method for collecting defect image on basis of multi-processing for high-speed and wide-area reflective film - Google Patents

Abstract

An apparatus and a method for collecting a defect image on basis of multi-processing for a high-speed and wide-area reflective film are disclosed. The apparatus for collecting a defect image according to an embodiment of the present invention may include: a multiprocessing-based on-memory queuing module which performs a memory sharing process of loading image data generated in association with a reflective film to a shared memory and a queuing process of loading replicated image data obtained by copying the image data to a queue; and a web GUI module for Blackbox-based data collection, which performs a streaming and saving process of streaming the image data obtained from the shared memory and storing defect image data determined to be a defect among the image data and the replicated image data obtained from the queue, and a defect streaming process of checking for defects in the defect image data.

Description

Multiprocessing-based defect image acquisition device and method for high-speed and wide-area reflective films

The present invention relates to a defect image collection device and method based on multi-processing for a high-speed and wide-area reflective film, which separates a process of generating image data by photographing a reflective film and a process of streaming the image data, shares the image data between the two processes through a shared memory, and stores product surface image data having a defect in a localized area through a queuing process that limits and stores streaming for only a specific N-second period.

Registration No. 10-2541166 (2023.06.02) "AI Vision Inspection System Using Robots"

Registration No. 10-2541166 discloses an AI vision inspection system using a robot that can quickly and accurately analyze and inspect whether a product is defective or not by using a video taken of the front of the product by a vision camera mounted on the robot.

Registration No. 10-2491420 (2023.01.18) "Vision inspection method for vision inspection using dynamic plate based on artificial intelligence learning"

Registration No. 10-2491420 discloses a vision inspection method that performs a vision inspection based on data learned in advance through artificial intelligence using real-time shooting information on an object to be inspected using multiple cameras.

Publication No. 10-2023-0083134 (2023.06.09) "Vision inspection device and its control method"

Publication number 10-2023-0083134 discloses a vision inspection device and a control method thereof, which perform vision inspection accurately and more quickly to increase process efficiency.

A vision system can be a system that observes and classifies objects on behalf of human eyes.

Vision systems are commonly used in manufacturing and production industries to detect or classify defects in manufactured products.

Since the products targeted for defect detection in vision systems are clearly visible to the naked eye, small in size, and produced at slow speeds, many vision systems detect defects using vision technologies such as contrast, shading correction, and line defect extraction.

However, if the product being produced is a reflective film, it is not possible to isolate the external environment and the production environment like with conventional vision systems.

In addition, reflective films cannot properly utilize existing vision technologies due to reasons such as light reflection, wide shooting area, and fast production speed.

With regard to the above light reflection, in the case of a white reflective film, whiteout due to light reflection occurs when the surface brightness is 20,000 Lux or higher, and in the case of a black reflective film, blackout may occur when the surface brightness is 20,000 Lux or lower.

In relation to the above wide shooting area, the size of the reflective film can be at least 2.4 times (approximately 1.2 m) larger than the shooting target size of the existing vision system (0.1 to 0.5 m). Accordingly, there may be limitations in terms of speed when analyzing all pixels of the image using the existing vision technology.

Artificial intelligence technology can automatically perform exception handling for defect detection for light reflection and the external environment through learning when learning a defect detection model, and can analyze a large number of high-resolution images in a short period of time by utilizing special equipment called GPU (Graphics Processing Unit).

In order to learn an AI-based defect detection model, a large amount of training defect image data is required. For example, the training data (MNIST Numeric) required for stable classification of handwritten digits may require approximately 60,000.

However, since the defective images that are the target of data collection are captured in high resolution, storing them when defects are found after continuous shooting may require a lot of load on the storage device.

For example, if a defect is found after only 1 hour of work, it would require approximately 300Gb (4K image, 30fps, 60 minutes) of storage space, which could take a lot of storage time.

On the other hand, in order to reduce the load on the storage device, there is a limitation in that the act of having to manually rewind the equipment to place the defect within the camera's shooting range when a defect is discovered and monitor whether the defect is within the shooting range causes a lot of manpower consumption.

Therefore, there is an urgent need for an improved model to collect efficient bad image data for high-speed and wide-area reflective films by overcoming the limitations of storage space, storage time, and manpower consumption.

An embodiment of the present invention aims to provide a multi-processing-based defective image collection device and method for a high-speed and wide-area reflective film, thereby improving the existing data collection process by proposing a method for collecting image data in consideration of the production process of the reflective film.

In addition, the embodiment of the present invention aims to expand the range of existing collectable data by proposing a method for collecting high-resolution, high-speed image data for artificial intelligence learning.

In addition, an embodiment of the present invention aims to improve the efficiency of data collection by proposing a method for collecting a large amount of image data for artificial intelligence learning.

In addition, an embodiment of the present invention aims to alleviate worker fatigue in collecting data by switching to black box-based data collection.

According to one embodiment of the present invention, a multi-processing-based defective image collection device for a high-speed and wide-area reflective film may include a multi-processing-based On-memory Queuing module that performs a Memory Sharing Process for loading image data generated in connection with a reflective film into a shared memory, and a Queuing Process for loading duplicate image data copied from the image data into a Queue; and a Blackbox-based data collection Web GUI module that performs a Streaming & Saving Process for streaming image data acquired from the shared memory and storing defective image data determined to be defective among the image data and duplicate image data acquired from the Queue, and a Defect Streaming Process for confirming a defect in the defective image data.

In addition, a method for collecting defective images based on multi-processing for a high-speed and wide-area reflective film according to an embodiment of the present invention may be configured to include a step of performing a Memory Sharing Process for loading image data generated in association with a reflective film into a shared memory in a multi-processing-based On-memory Queuing module; a step of performing a Queuing Process for loading duplicate image data copied from the image data into a Queue in the multi-processing-based On-memory Queuing module; a step of performing a Streaming & Saving Process for streaming image data acquired from the shared memory in a Blackbox-based data collection Web GUI module and saving defective image data determined to be defective among the image data and duplicate image data acquired from the Queue; and a step of performing a Defect Streaming Process for confirming a defect in the defective image data in the Blackbox-based data collection Web GUI module.

According to one embodiment of the present invention, by proposing a method for collecting image data by considering the production process of a reflective film, a multi-processing-based defective image collection device and method for a high-speed and wide-area reflective film can be provided, which improves the existing data collection process.

In addition, according to one embodiment of the present invention, by proposing a method for collecting high-resolution, high-speed image data for artificial intelligence learning, the range of existing collectable data can be expanded.

In addition, according to one embodiment of the present invention, by proposing a method for collecting a large amount of image data for artificial intelligence learning, the efficiency of data collection can be improved.

In addition, according to one embodiment of the present invention, by switching to black box-based data collection, worker fatigue in collecting data can be alleviated.

FIG. 1 is a block diagram illustrating a configuration of a multi-processing-based defective image collection device for a high-speed and wide-area reflective film according to one embodiment of the present invention.

Figure 2 is a diagram showing the entire operation process of the defective image collection device.

Figure 3 is a diagram to explain the difference in image processing methods depending on whether parallel processing is used.

Figure 4 is a diagram explaining the structure of a queue.

Figure 5 is a diagram to explain the difference in operation between processes according to the data life cycle.

Figure 6 is a diagram for explaining an example of a Bayer pattern.

Figure 7 is a diagram explaining the image storage process depending on the presence or absence of the Image Lazy Convert Process.

Figure 8 is a diagram explaining the process of collecting defective image data from the worker's perspective according to whether the product is used or not (blackbox/whitebox).

Figure 9 is a diagram showing the results of implementing a GUI process for real-time streaming and data storage.

FIG. 10 is a flowchart illustrating a multi-processing-based defective image collection method for a high-speed and wide-area reflective film according to one embodiment of the present invention.

Hereinafter, embodiments will be described in detail with reference to the attached drawings. However, since various modifications may be made to the embodiments, the scope of the patent application rights is not limited or restricted by these embodiments. It should be understood that all modifications, equivalents, or substitutes to the embodiments are included in the scope of the rights.

The terms used in the examples are for the purpose of description only and should not be construed as limiting. The singular expression includes the plural expression unless the context clearly indicates otherwise. In this specification, the terms "comprises" or "has" and the like are intended to specify the presence of a feature, number, step, operation, component, part or combination thereof described in the specification, but should be understood to not exclude in advance the possibility of the presence or addition of one or more other features, numbers, steps, operations, components, parts or combinations thereof.

Unless otherwise defined, all terms used herein, including technical or scientific terms, have the same meaning as commonly understood by one of ordinary skill in the art to which the embodiments belong. Terms defined in commonly used dictionaries, such as those defined in common dictionaries, should be interpreted as having a meaning consistent with the meaning they have in the context of the relevant art, and shall not be interpreted in an idealized or overly formal sense, unless expressly defined in this application.

In addition, when describing with reference to the attached drawings, the same components will be given the same reference numerals regardless of the drawing numbers, and redundant descriptions thereof will be omitted. When describing an embodiment, if it is determined that a detailed description of a related known technology may unnecessarily obscure the gist of the embodiment, the detailed description thereof will be omitted.

FIG. 1 is a block diagram illustrating a configuration of a multi-processing-based defective image collection device for a high-speed and wide-area reflective film according to one embodiment of the present invention.

Referring to FIG. 1, a multi-processing-based defective image collection device for a high-speed and wide-area reflective film according to one embodiment of the present invention (hereinafter, abbreviated as 'defective image collection device', 100) may be configured to include a multi-processing-based On-memory Queuing module (110) and a Blackbox-based Web GUI module for data collection (120).

First, the multiprocessing-based On-memory Queuing module (110) performs a Memory Sharing Process that loads image data generated in connection with a reflective film into a shared memory. That is, when a reflective film is photographed by a vision camera and image data, which is vision data, is generated, the multiprocessing-based On-memory Queuing module (110) can play a role in storing the generated image data in a shared memory.

The multiprocessing-based On-memory Queuing module (110) can load image data for a reflective film produced at high speed into shared data in real time as the image data is continuously generated.

The camera generating image data may be an area camera that captures the exterior of the reflective film in horizontal/vertical pixel units and generates precise image data for each pixel.

When generating image data, the camera can generate the image data by photographing a reflective film under conditions in which a certain amount of light is irradiated from a vision light.

In addition, the multiprocessing-based On-memory Queuing module (110) performs a Queuing Process that loads duplicate image data, which is the copied image data, into the Queue. That is, the multiprocessing-based On-memory Queuing module (110) can play a role in linking with the generation of image data, copying the generated image data in real time to create duplicate image data, and sequentially loading the copied image data into the Queue.

In loading duplicate image data into a queue, a multiprocessing-based on-memory queuing module (110) copies the image data generated from the present to a previously set period of time into the duplicate image data and loads it into the queue, but can remove the duplicate image data loaded first from the queue in a first-in, first-out manner as time passes.

That is, the multiprocessing-based On-memory Queuing module (110) accumulates and loads duplicate image data up to the storage capacity of the queue, and when the storage capacity is exceeded, the earliest loaded duplicate image data is removed, thereby securing storage space for newly loaded duplicate image data.

Image data loaded into the above shared memory can be acquired and streamed later by the Blackbox-based data collection Web GUI module (120).

In the above streaming, the multiprocessing-based On-memory Queuing module (110) can perform an Image Lazy Convert Process for converting the image data and the duplicated image data generated and duplicated in BayerRG8 format into an RGB format for the streaming, as the camera that photographs the reflective film is set to the BayerRG8 shooting mode that uses the Bayer pattern.

As the camera captures a reflective film and generates/duplicates image data and duplicate image data in BayerRG8 format, the multiprocessing-based On-memory Queuing module (110) can perform an Image Lazy Convert Process to convert the image data/duplicate image data from BayerRG8 format to RGB format, thereby making the image data/duplicate image data into a format suitable for streaming.

According to an embodiment, the bad image collection device (100) can multi-process the loading of the image data into the shared memory and the streaming of the image data acquired from the shared memory separately and simultaneously.

That is, the bad image collection device (100) can independently acquire image data from the shared memory and stream it while loading the generated image data into the shared memory through multi-processing.

Since the loading of the image data into the shared memory and the streaming of the image data acquired from the shared memory are multi-processed separately and simultaneously, the multi-processing-based On-memory Queuing module (110) can cause the camera to photograph the reflective film immediately after loading the image data into the shared memory without waiting for the image data to stream, thereby generating new image data.

The Web GUI module (120) for Blackbox-based data collection performs a Streaming & Saving Process that streams image data acquired from the shared memory and saves defective image data determined to be defective among the image data and duplicate image data acquired from the Queue. That is, the Web GUI module (120) for Blackbox-based data collection can play a role in determining whether the acquired image data/duplicate image data is defective and identifying and storing defective image data determined to be defective.

In addition, the Web GUI module (120) for Blackbox-based data collection performs a Defect Streaming Process for checking for defects in the above-mentioned defective image data. In other words, the Web GUI module (120) for Blackbox-based data collection can play a role in allowing the operator to re-check for defects in the defective image data identified and stored as defective image data.

In addition, the Web GUI module (120) for Blackbox-based data collection can further perform Other Processes for checking the data collection status related to the frequency at which the defective image data is stored and the local storage capacity related to the remaining capacity in the local storage.

That is, the Web GUI module (120) for Blackbox-based data collection can measure the frequency with which image data/duplicate image data is judged to be defective and provide the worker with the data collection status, and can also provide the worker with local storage capacity regarding the remaining capacity in the local storage where the defective image data is stored so that the worker can check it.

The Web GUI module (120) for Blackbox-based data collection can obtain image data loaded into shared memory and stream it to the operator.

In streaming image data, the Blackbox-based data collection Web GUI module (120) can store the streamed image data of t-1 and the duplicate image data of t (where t is the creation time) acquired from the Queue in the main memory.

That is, the Web GUI module (120) for Blackbox-based data collection can store image data/duplicate image data with different creation times in the main memory.

Thereafter, the Web GUI module (120) for Blackbox-based data collection can determine whether data having different creation times and separated life cycles are defective. That is, the Web GUI module (120) for Blackbox-based data collection can determine whether image data of t-1 and duplicate image data of t are defective.

The Web GUI module (120) for Blackbox-based data collection can store the defective image data determined to be defective in a local storage and remove the remaining data not determined to be defective from the main memory.

That is, the Web GUI module (120) for Blackbox-based data collection can secure free space for the main memory by storing data judged as defective in local storage and removing data judged as normal.

In addition, the Web GUI module (120) for Blackbox-based data collection can temporarily stop streaming of the image data at the observation point when the image data being streamed is observed as defective by the operator. That is, the Web GUI module (120) for Blackbox-based data collection can stop streaming when the operator finds a defect in the image data during streaming and generates a set event, so that the image data in which a defect is found can be distinguished.

Afterwards, the Web GUI module (120) for Blackbox-based data collection can determine whether the paused image data is defective.

The Web GUI module (120) for collecting Blackbox-based data can store the paused image data as defective image data together with a defective information summary including the defective type and defective location if it is determined to be defective. That is, the Web GUI module (120) for collecting Blackbox-based data can store the image data as defective image data together with the defective information summary if a defective is confirmed as a result of the determination, and can create a defective information summary describing the defective type and defective location of the confirmed defective data, identify the corresponding image data as defective image data, and store the image data in a local storage together with the defective information summary.

According to one embodiment of the present invention, by proposing a method for collecting image data by considering the production process of a reflective film, a multi-processing-based defective image collection device and method for a high-speed and wide-area reflective film can be provided, which improves the existing data collection process.

In addition, according to one embodiment of the present invention, by proposing a method for collecting high-resolution, high-speed image data for artificial intelligence learning, the range of existing collectable data can be expanded.

In addition, according to one embodiment of the present invention, by proposing a method for collecting a large amount of image data for artificial intelligence learning, the efficiency of data collection can be improved.

In addition, according to one embodiment of the present invention, by switching to black box-based data collection, worker fatigue in collecting data can be alleviated.

The defective image collection device (100) of the present invention may be related to collecting defective image data based on multi-processing for high-speed and wide-area reflective films.

The bad image collection device (100) can be configured to include a multiprocessing-based On-memory Queuing module for preventing unnecessary storage space waste and securing storage speed, and a Blackbox-based data collection Web GUI (Graphic User Interface) module for preventing manpower consumption and efficient data collection.

The multiprocessing-based On-memory Queuing module can perform Memory Sharing Process for real-time streaming, Queuing Process to store only the most recent N seconds of streaming in memory, and Image Lazy Convert Process to secure image storage efficiency.

The Web GUI module for Blackbox-based data collection can perform Streaming & Saving Process, Defect Streaming Process, and Other Process.

Figure 2 is a diagram showing the entire operation process of the defective image collection device.

In step 201, the bad image collection device (100) can initialize the camera as a Memory Sharing Process.

In step 202, the bad image collection device (100) can specify a queue size as a queuing process.

In step 203, the defective image collection device (100) can take pictures of the reflective film as a Memory Sharing Process with the camera initial settings and queue size specified.

In step 203, the defective image collection device (100) can generate image data in BayerRG8 format and duplicate image data that is a copy of the image data through camera shooting.

In addition, in step 216, the bad image collection device (100) may perform, as Other Process, in parallel, the verification of the local storage capacity in which the image data generated according to the camera shooting in step 203 is stored.

In step 204, the bad image collection device (100) can load image data generated by camera shooting into shared memory as a Memory Sharing Process.

In step 205, the bad image collection device (100) can load the duplicate image data, which is the copied image data, into the queue as a queuing process.

In step 206, the bad image collection device (100) can obtain image data in shared memory as a Memory Sharing Process.

In step 207, the bad image collection device (100) can acquire duplicate image data within the queue as a queuing process.

In step 208, the defective image collection device (100) can determine whether there is a defect in the image data/duplicate image data.

If there are no defects (No direction in step 208), in step 209, the defective image collection device (100) can convert the image data/duplicate image data to RGB (Bayer to RGB) as an Image Lazy Convert Process.

In step 210, the bad image collection device (100) can stream image data in RGB format as a Streaming & Saving Process.

In step 211, the bad image collection device (100) can save image and video copy image data in RGB format as a Streaming & Saving Process.

At step 212, the bad image collection device (100) can refresh the storage of images and videos as a Streaming & Saving Process.

If there is a defect (Yes direction in step 208), in step 213, the defect image collection device (100) can select the defect type and location as a Streaming & Saving Process.

After selecting the defect type and location, the defect image collection device (100) can refresh the storage of images and videos by performing step 212.

After refreshing the storage of images and videos, the defective image collection device (100) can perform step 214 to check previously captured defects as a Defect Streaming Process, or perform step 215 to check the data collection status as an Other Process.

The multiprocessing-based On-memory Queuing module can perform Memory Sharing Process for real-time streaming.

The Memory Sharing Process can stream image data for the reflective film to the Web GUI module.

If the camera shooting operation and the operation of streaming the image to the Web GUI model are performed in one process, a bottleneck phenomenon may occur, so the multiprocessing-based On-memory Queuing module can separate the two operations through multiprocessing.

Additionally, the multiprocessing-based On-memory Queuing module can utilize shared memory to share images between two processes performing different operations.

The difference between parallel processing of camera shooting and image streaming is explained below through Figure 3.

Figure 3 is a diagram to explain the difference in image processing methods depending on whether parallel processing is used.

Figure 3 (a) is a schematic diagram explaining the occurrence of a bottleneck phenomenon.

In step 301, the existing module can perform camera initialization.

At step 302, the existing module can take a camera shot on the reflective film according to the camera initial settings.

In step 303, the existing module can perform image streaming through the GUI module.

During this time, camera shooting may freeze between image streaming.

Once image streaming is complete, the original module can resume camera shooting.

Figure 3 (b) is a schematic diagram explaining parallel processing through multi-processing.

At step 311, the multiprocessing-based On-memory Queuing module can perform camera initialization.

At step 312, the multiprocessing-based On-memory Queuing module can take pictures of the reflective film according to the camera initial settings.

In step 313, the multiprocessing-based On-memory Queuing module can load images generated by camera shooting into shared memory and induce continuous shooting.

The multiprocessing-based On-memory Queuing module can perform multiprocessing that processes steps 314 to 315 in parallel, independently of steps 311 to 313.

At step 314, the multiprocessing-based On-memory Queuing module can acquire and post-process the image.

In step 315, the multiprocessing-based On-memory Queuing module can perform image streaming and induce continuous streaming through the GUI module.

The multiprocessing-based On-memory Queuing module can perform initial camera settings to capture clear images of the reflective film, which is the subject of the shooting, that can be identified with the naked eye.

Afterwards, the multiprocessing-based On-memory Queuing module can repeat the process of loading images captured by the camera into shared memory.

The multiprocessing-based On-memory Queuing module can repeat the process of acquiring images stored in shared memory through the GUI module, post-processing them, and streaming them.

The multiprocessing-based On-memory Queuing module can perform a Queuing Process to store only the most recent N seconds of streaming in memory.

Storing the entire image data that was captured over a long period of time without meaning to store bad image data located in a local area can be very inefficient in terms of storage space/speed.

Queuing Process can be a space/speed efficient process to overcome these limitations.

A queue can have a first-in, first-out structure, where old data can be discarded and new data can be inserted when the maximum size is reached.

In the Queuing Process, by utilizing the Queue, only images captured for a fixed amount of time can be stored and previously captured meaningless image data can be deleted.

Figure 4 is a diagram explaining the structure of a queue.

As in Fig. 4, when the length of the Queue is '7', the multiprocessing-based On-memory Queuing module can store seven image data of Data 1 to 7 in the Queue. At this time, the multiprocessing-based On-memory Queuing module can store Data 7 in the Queue while removing Data 0, which is stored first in the Queue, in a first-in, first-out manner.

If Data 8 is newly generated, the multiprocessing-based On-memory Queuing module can remove Data 1, which is stored first in the Queue, in a first-in-first-out manner and store the new Data 8 in the Queue, so that up to 7 image data can always be maintained.

The multiprocessing-based On-memory Queuing module can specify the maximum size of the Queue to be loaded into memory.

The maximum size of a queue can be calculated according to [Equation 1].

[Formula 1]

Max Queue Size = Image Resolution(4096 * 768) * FPS * Seconds

Seconds = Human Detect Time + Idle Time

Here, Image Resolution is the resolution of the image being captured, and FPS can be the number of shots per second of the camera.

Seconds is the shooting time, which is the sum of the time from when a defect is captured until it is discovered by a worker (Human Detect Time) and the time it takes for a worker to save it after discovery (Idle Time). It can be a hyperparameter value depending on the worker's work efficiency.

For example, the maximum size of the Queue to store images with a resolution of 4096 * 768 streamed at 30FPS for the last 10 seconds can be calculated as (4096*768*30*10) according to Equation 1.

The multiprocessing-based On-memory Queuing module can generate and preprocess image data by shooting a reflective film through a camera after specifying the maximum size of the queue, and input duplicate image data that is a copy of the preprocessed image data into the queue.

The multiprocessing-based On-memory Queuing module can copy preprocessed image data because the life cycle of the data is the same when using the same memory space as the image data being used in the Memory Sharing Process.

Figure 5 explains why the life cycles are the same.

Figure 5 is a diagram to explain the difference in operation between processes according to the data life cycle.

Figure 5 (a) is a diagram for explaining two processes that share data of the same life cycle.

In step 501, a multiprocessing-based On-memory Queuing module can generate image 1 by taking a picture of a reflective film with a camera.

At step 502, the multiprocessing-based On-memory Queuing module can load image 1 into shared memory as a Memory Sharing Process.

In step 503, the multiprocessing-based On-memory Queuing module can stream images through the GUI module.

Additionally, in step 504, the multiprocessing-based On-memory Queuing module can load image 1 into the Queue as a Queuing Process.

At step 505, the multiprocessing-based On-memory Queuing module can stream or store the image1 loaded into the Queue into the Main Memory.

After being stored in the main memory, the multiprocessing-based On-memory Queuing module can perform step 506 to remove the image as a Memory Sharing Process, or perform step 507 to store the image when a defect is found through the GUI module.

Figure 5 (b) is a diagram for explaining two processes for handling data whose life cycle is separated through image copying.

At step 511, the multiprocessing-based On-memory Queuing module can generate image 1 by taking a picture of the reflective film with a camera.

At step 512, the multiprocessing-based On-memory Queuing module can load image 1 into shared memory as a Memory Sharing Process.

At step 513, the multiprocessing-based On-memory Queuing module can stream images through the GUI module.

Additionally, in step 514, the multiprocessing-based On-memory Queuing module can copy image 1 as a Queuing Process. The copied image 1 can be output as image 2.

At step 515, the multiprocessing-based On-memory Queuing module can load image 2 into the Queue as a Queuing Process.

At step 516, the multiprocessing-based On-memory Queuing module can stream or store the image1 and image2 loaded into the Queue into the Main Memory.

After being stored in the main memory, the multiprocessing-based On-memory Queuing module can perform step 517 to remove the image as a Memory Sharing Process, or perform step 518 to store the image when a defect is found through the GUI module.

The multiprocessing-based On-memory Queuing module can stop the process if a defect is found during the process of repeatedly inputting images into the queue and store the image data captured over the last N seconds.

The multiprocessing-based On-memory Queuing module can perform the Image Lazy Convert Process.

To stream high-resolution RGB images, inputting them into shared memory as RGB images can reduce memory efficiency.

Additionally, inputting RGB images into the Queue can also reduce memory efficiency.

Image Lazy Convert Process can avoid limitations that reduce memory efficiency.

The multiprocessing-based On-memory Queuing module can set the image format to BayerRG8 during camera initialization.

Figure 6 is a diagram for explaining an example of a Bayer pattern.

As shown in Fig. 6, BayerRG8 can be an image expression pattern that can express only one color per pixel, with 50% green, 25% blue, and 25% red, taking into account human visual characteristics.

Through this, the multiprocessing-based On-memory Queuing module can secure about three times more memory space than RGB, which expresses three colors in one pixel.

The multiprocessing-based On-memory Queuing module can acquire data corresponding to BayerRG8 through camera shooting and perform the above-described Memory Sharing Process and Queuing Process.

The multiprocessing-based On-memory Queuing module can convert the image data into RGB image format when streaming it.

Figure 7 explains the image storage process depending on whether Image Lazy Convert Process is present or not.

Figure 7 is a diagram explaining the image storage process depending on the presence or absence of the Image Lazy Convert Process.

Figure 7 (a) is a schematic diagram when the image conversion process is not applied.

In step 701, the multiprocessing-based On-memory Queuing module can perform camera initialization.

In step 702, the multiprocessing-based On-memory Queuing module can take a camera shot on a reflective film according to the camera initial settings. At this time, the generated image data can be in RGB format.

In step 703, the multiprocessing-based On-memory Queuing module can load image data (RGB) generated by camera shooting into shared memory.

In step 704, the multiprocessing-based On-memory Queuing module can obtain image data (RGB) from shared memory.

In step 705, the multiprocessing-based On-memory Queuing module can perform image streaming of image data (RGB) through the GUI module.

Figure 7 (b) is a schematic diagram when applying the image conversion process.

In step 711, the multiprocessing-based On-memory Queuing module can perform camera initialization. At this time, the multiprocessing-based On-memory Queuing module can set the BayerRG8 shooting mode using the Bayer pattern.

In step 712, the multiprocessing-based On-memory Queuing module can take pictures of the reflective film according to the camera initial settings. At this time, the generated image data can be in BayerRG8 format.

At step 713, the multiprocessing-based On-memory Queuing module can load image data (BayerRG8) generated by camera shooting into shared memory.

At step 714, the multiprocessing-based On-memory Queuing module can obtain image data (BayerRG8) from shared memory.

At step 716, the multiprocessing-based On-memory Queuing module can perform image conversion of converting BayerRG8 to RGB as an Image Lazy Convert Process.

In step 716, the multiprocessing-based On-memory Queuing module can perform image streaming of image data (RGB) through the GUI module.

A Web GUI module for Blackbox-based data collection can be designed and implemented to prevent manpower consumption and ensure efficient data collection.

Previously, separate technical guidance was required for workers who had no experience with artificial intelligence to use GUI-based software provided by industrial camera vendors.

Additionally, previously, a separate file had to be created or the data name had to be changed to record information such as defect type, shooting time, and defect location.

In addition, previously, it was a significant burden on AI experts (in charge of learning) because the recorded video had to be separated into frames for conversion to image data.

The Web GUI module for Blackbox-based data collection can solve these limitations and make data collection easier by simply providing workers with only the functions they need.

Figure 8 is a diagram explaining the process of collecting defective image data from the worker's perspective according to whether the product is used or not (blackbox/whitebox).

Fig. 8 (a) is a diagram for explaining a conventional method for collecting defective image data.

In a conventional bad image data collection method at step 801, camera software can be run.

In a conventional bad image data collection method at step 802, camera initial settings can be made.

In a conventional bad image data collection method at step 803, image data can be generated by performing a production task.

In step 804, in a conventional defective image data collection method, it is possible to determine whether a defect is found in the generated image data.

In the conventional defective image data collection method at step 805, if a defect is found, the equipment can be rewinded so that the defect is visible in the shooting range.

In a conventional bad image data collection method at step 806, a camera can be used to take pictures of the reflective film.

In a conventional bad image data collection method at step 807, the camera can be stopped after shooting.

In the conventional method of collecting defective image data at step 808, after changing the file name of the defective type, defective location, shooting time, etc., the process can return to step 803.

Fig. 8 (b) is a diagram for explaining a method for collecting defective image data of the present invention.

In step 811, the Web GUI module for Blackbox-based data collection can run data collection software.

At step 812, the Web GUI module for Blackbox-based data collection can perform production work to generate image data.

In step 813, the Web GUI module for Blackbox-based data collection can determine whether a defect is found in the generated image data. At this time, the Web GUI module for Blackbox-based data collection can perform internal operations based on the Blackbox.

In step 814, the multiprocessing-based On-memory Queuing module can return to step 812 after changing the file name such as the defect type, defect location, and shooting time.

The Web GUI module for Blackbox-based data collection implemented on the web can perform Streaming & Saving Process, Defect Streaming Process, and Other Process.

Streaming & Saving Process can play a role in showing the product currently being shot and saving the shot in real time.

The Defect Streaming Process can be used to determine whether previously captured and recently saved defective image data by a worker is distinct enough to be seen by the naked eye.

Other Processes may consist of a process for visualizing the remaining capacity in local storage for user convenience and a process for visualizing the accumulated status of bad image data.

In the Streaming & Saving Process, the Web GUI module for Blackbox-based data collection can show the product currently being photographed and save the photographed copy when a defect is found.

The Web GUI module for Blackbox-based data collection first allows the operator to observe the product currently being photographed as high-resolution image data.

The image data being observed may be image data entered into shared memory in the Memory Sharing Process.

The image data will be converted to RGB format through the Image Lazy Convert Process and may be erased within the shared memory for the next image streaming.

The Web GUI module for Blackbox-based data collection allows operators to save defects observed during the streaming process as defect image data.

The Web GUI module for Blackbox-based data collection can perform pause, select defective type, select defective location, save, and perform an initialization function to re-select if a defective type or defective location is incorrectly selected, thereby saving defective image data.

The Web GUI module for Blackbox-based data collection can stop the Queuing Process when a pause is performed, preventing newly captured image data from being saved.

At the time the pause is performed, all data existing in the Queue is converted into video footage at a preset FPS after going through the Image Lazy Convert Process. This is to prevent a significant amount of time from being consumed due to a bottleneck when converting bad image data into video footage at the time of saving.

Even at the time of selecting the bad type and bad location, video conversion can continue to be performed in the background.

The converted video footage can be stored on the server and used as a video to be played in the Defect Streaming Process.

The Web GUI module for Blackbox-based data collection can receive defect types from workers.

The Web GUI module for Blackbox-based data collection can add and define new types of faults that have not been previously defined.

The Web GUI module for Blackbox-based data collection can operate in a manner that adds defect types each time they are selected when multiple defects exist within a single product.

The Web GUI module for Blackbox-based data collection can select the defective location as left, center, right, or global, taking into account the product characteristics, such as wide width in the case of reflective films, depending on the characteristics of the product.

The Web GUI module for Blackbox-based data collection can perform storage once both the fault type and fault location are selected.

The Web GUI module for Blackbox-based data collection can store previously converted video footage and images corresponding to each frame on the server.

Video footage and images stored on the server can be used as training data at a later point in time when training an artificial intelligence model.

The Web GUI module for Blackbox-based data collection can record the type of defect, location of the defect, storage time, etc. in the file name for the convenience of data labeling (ex: point defect, line defect_left_230809_01_15frame.png).

The Web GUI module for Blackbox-based data collection can finally refresh the entire screen and repeat the Streaming & Saving Process.

Figure 9 is a diagram showing the results of implementing a GUI process for real-time streaming and data storage.

Figure 9 shows an example of a Streaming & Saving Process in which streaming image data is paused when a defect is found by a worker, and the paused image data is saved as defective image data, including a summary of defect information (defect type, defect location).

Defect Streaming Process can be performed to reconfirm whether defective image data saved through Streaming & Saving Process was captured correctly.

The Web GUI module for Blackbox-based data collection can perform speed setting, enlargement, and data deletion as a Defect Streaming Process.

The Web GUI module for Blackbox-based data collection can set the playback speed when playing converted video clips.

For example, if the playback speed of a video clip is too fast for an operator to visually recognize a defect, the Web GUI module for Blackbox-based data collection can set a playback speed setting that reduces the playback speed of the video clip.

The Web GUI module for Blackbox-based data collection can support speed settings such as x0.25, x0.5, x1, x2, etc.

The Web GUI module for Blackbox-based data collection can support a zoom function to enable operators to accurately check small defects.

The magnification function may be essential for observing small defects in products manufactured with a width of 1m or more through a monitor.

The Web GUI module for Blackbox-based data collection can delete defective image data when a part suspected to be defective by an operator is determined to be normal.

The Web GUI module for Blackbox-based data collection can eliminate the inconvenience of having to contact an AI expert (learning manager) separately whenever a situation arises where a defect needs to be restored to normal through the deletion function.

Other Processes are processes that may be performed primarily by production managers or AI experts rather than workers, such as checking the status of collecting bad image data and checking local storage capacity.

To introduce AI-based vision services into production facilities, production managers must check the status of collection of defective image data.

In order to learn a stable defect detection AI model, a large amount of learning data is required, and the number of learning data required can be determined by considering the defect occurrence rate and defect type through compromise between production managers and AI experts.

Production managers need to observe the training of defect detection AI models to determine how insufficient the training data is and what types of defects require more training data.

The Web GUI module for Blackbox-based data collection can provide production managers with statistical figures for measuring the defect rate within production facilities by supporting the collection status of defective image data, and can allow AI experts to observe changes in AI model performance according to the number of collected defective image data.

Production managers and AI experts need to check the local storage capacity of the computers that collect bad image data.

The bad image data being collected can be a collection of high-resolution, high-FPS, long-time captured image data.

The local storage capacity requirement for a single bad event can be approximately 5 Gb per minute.

If these defects occur 10 times a day for a week (excluding holidays), the amount of defective image data generated could be approximately 250 Gb.

To store 250Gb of bad image data, it takes a lot of time, so instead of HDD, which is a high-capacity, low-speed storage device, SSD, which is a low-capacity, high-speed storage device, can be used for local storage.

In case of insufficient management, the storage space of the SSD will become full and no more data can be collected until the bad image data is moved to the backup storage.

The Web GUI module for Blackbox-based data collection can support checking of local storage capacity, thereby allowing production managers and AI experts to separately discuss data collection/backup storage upload plans according to local storage capacity.

According to the present invention, by proposing a vision data collection method that takes into account the product production process, the existing vision data collection process can be improved.

In addition, the present invention proposes a method for collecting high-resolution, high-speed image data for artificial intelligence learning, thereby expanding the range of vision data that can be collected.

In addition, the present invention can improve the efficiency of collecting vision data by proposing a method for collecting high-capacity artificial intelligence learning data images.

In addition, the present invention can alleviate worker data collection fatigue by switching to black box-based data collection.

Below, FIG. 10 describes in detail the work flow of the defective image collection device (100) according to embodiments of the present invention.

FIG. 10 is a flowchart illustrating a multi-processing-based defective image collection method for a high-speed and wide-area reflective film according to one embodiment of the present invention.

The multi-processing-based defective image collection method for a high-speed and wide-area reflective film according to the present embodiment can be performed by a defective image collection device (100).

First, a Memory Sharing Process is performed (1010) to load image data generated in connection with a reflective film into a shared memory in a multiprocessing-based On-memory Queuing module of a defective image collection device (100). Step (1010) may be a process of storing the generated image data in a shared memory by a multiprocessing-based On-memory Queuing module when a reflective film is photographed by a vision camera and image data, which is vision data, is generated.

The multiprocessing-based on-memory queuing module can load image data for high-speed reflective films into shared data in real time as the image data is continuously generated.

The camera generating image data may be an area camera that captures the exterior of the reflective film in horizontal/vertical pixel units and generates precise image data for each pixel.

When generating image data, the camera can generate the image data by photographing a reflective film under conditions in which a certain amount of light is irradiated from a vision light.

In addition, in the multiprocessing-based On-memory Queuing module of the bad image collection device (100), a Queuing Process is performed (1020) to load duplicate image data copied from the image data into the Queue. Step (1020) may be a process of sequentially loading the generated image data into duplicate image data by copying the generated image data in real time in conjunction with the generation of the image data by the multiprocessing-based On-memory Queuing module into the Queue.

In loading duplicate image data into a queue, a multiprocessing-based on-memory queuing module copies the image data generated from the present to a previously set period of time into the duplicate image data and loads it into the queue, but can remove the duplicate image data loaded earliest from the queue in a first-in, first-out manner over time.

That is, the multiprocessing-based On-memory Queuing module accumulates and loads duplicate image data up to the storage capacity of the Queue, and when the storage capacity is exceeded, the earliest loaded duplicate image data is removed to secure storage space for newly loaded duplicate image data.

Image data loaded into the above shared memory can be acquired and streamed later by a Web GUI module for Blackbox-based data collection.

In the above streaming, the multiprocessing-based On-memory Queuing module can perform an Image Lazy Convert Process for converting the image data and the duplicated image data generated and replicated in BayerRG8 format into an RGB format for the streaming, as the camera that photographs the reflective film is set to the BayerRG8 shooting mode that uses the Bayer pattern.

As the camera captures a reflective film and generates/duplicates image data and duplicate image data in BayerRG8 format, the multiprocessing-based On-memory Queuing module can perform an Image Lazy Convert Process to convert the image data/duplicate image data from BayerRG8 format to RGB format, thereby making the image data/duplicate image data into a format suitable for streaming.

According to an embodiment, the bad image collection device (100) can multi-process the loading of the image data into the shared memory and the streaming of the image data acquired from the shared memory separately and simultaneously.

That is, the bad image collection device (100) can independently acquire image data from the shared memory and stream it while loading the generated image data into the shared memory through multi-processing.

Since the loading of the image data into the shared memory and the streaming of the image data acquired from the shared memory are multi-processed separately and simultaneously, the multi-processing-based on-memory queuing module can cause the camera to photograph the reflective film to generate new image data immediately after loading the image data into the shared memory without waiting for the image data to stream.

Continuing, in the Web GUI module for Blackbox-based data collection of the defective image collection device (100), the image data acquired from the shared memory is streamed, and the Streaming & Saving Process for storing defective image data determined to be defective among the image data and duplicate image data acquired from the Queue is performed (1030). Step (1030) may be a process for determining whether the acquired image data/duplicate image data is defective, and identifying and storing the defective image data determined to be defective by the Web GUI module for Blackbox-based data collection.

In addition, in the Web GUI module for collecting data based on the Blackbox of the defective image collection device (100), a Defect Streaming Process is performed (1040) to confirm defects in the defective image data. Step (1040) may be a process for re-confirming defects from an operator's perspective in the defective image data identified and stored as defective image data by the Web GUI module for collecting data based on the Blackbox.

In addition, the Web GUI module for Blackbox-based data collection can further perform Other Processes for checking the data collection status related to the frequency at which the above-mentioned bad image data is stored and the local storage capacity related to the remaining capacity in the local storage.

That is, the Web GUI module for Blackbox-based data collection can measure the frequency with which image data/duplicate image data is judged to be defective and provide the worker with the status of data collection, and can also provide the worker with local storage capacity regarding the remaining capacity in the local storage where the defective image data is stored so that the worker can check it.

The Web GUI module for Blackbox-based data collection can acquire image data loaded into shared memory and stream it to the operator.

In streaming of image data, the Blackbox-based data collection Web GUI module can store the streamed image data of t-1 and the duplicate image data of t (where t is the creation time) acquired from the Queue in the main memory.

That is, the Web GUI module for Blackbox-based data collection can store image data/duplicate image data with different creation times in the main memory.

Afterwards, the Web GUI module for Blackbox-based data collection can determine whether data having different creation times and separated life cycles are defective. That is, the Web GUI module for Blackbox-based data collection can determine whether image data of t-1 and duplicate image data of t are defective.

The Web GUI module for Blackbox-based data collection can store the defective image data judged as defective in a local storage and remove the remaining data not judged as defective from the main memory.

That is, the Web GUI module for Blackbox-based data collection can secure free space for main memory by storing data judged as defective in local storage and removing data judged as normal.

In addition, the Web GUI module for Blackbox-based data collection can pause streaming of the image data at the observation point if the image data being streamed is observed as defective by the operator. That is, the Web GUI module for Blackbox-based data collection can stop streaming when the operator discovers a defect in the image data during streaming and generates a set event, so that the image data in which a defect is discovered can be distinguished.

Afterwards, the Web GUI module for Blackbox-based data collection can determine whether the paused image data is defective.

The Web GUI module for Blackbox-based data collection can store the paused image data as defective image data together with a defective information summary including the defective type and defective location if it is determined to be defective. That is, the Web GUI module for Blackbox-based data collection can, if a defective is confirmed as a result of the judgment, create a defective information summary describing the defective type and defective location of the confirmed defective data, identify the corresponding image data as defective image data, and store the image data in local storage together with the defective information summary.

According to one embodiment of the present invention, by proposing a method for collecting image data by considering the production process of a reflective film, a multi-processing-based defective image collection device and method for a high-speed and wide-area reflective film can be provided, which improves the existing data collection process.

In addition, according to one embodiment of the present invention, by proposing a method for collecting high-resolution, high-speed image data for artificial intelligence learning, the range of existing collectable data can be expanded.

In addition, according to one embodiment of the present invention, by proposing a method for collecting a large amount of image data for artificial intelligence learning, the efficiency of data collection can be improved.

In addition, according to one embodiment of the present invention, by switching to black box-based data collection, worker fatigue in collecting data can be alleviated.

The method according to the embodiment may be implemented in the form of program commands that can be executed through various computer means and recorded on a computer-readable medium. The computer-readable medium may include program commands, data files, data structures, etc., alone or in combination. The program commands recorded on the medium may be those specially designed and configured for the embodiment or may be those known to and available to those skilled in the art of computer software. Examples of the computer-readable recording medium include magnetic media such as hard disks, floppy disks, and magnetic tapes, optical media such as CD-ROMs and DVDs, magneto-optical media such as floptical disks, and hardware devices specially configured to store and execute program commands such as ROMs, RAMs, flash memories, etc. Examples of the program commands include not only machine language codes generated by a compiler but also high-level language codes that can be executed by a computer using an interpreter, etc. The hardware devices described above may be configured to operate as one or more software modules to perform the operations of the embodiment, and vice versa.

The software may include a computer program, code, instructions, or a combination of one or more of these, which may configure a processing device to perform a desired operation or may independently or collectively command the processing device. The software and/or data may be permanently or temporarily embodied in any type of machine, component, physical device, virtual equipment, computer storage medium or device, or transmitted signal waves, for interpretation by the processing device or for providing instructions or data to the processing device. The software may also be distributed over network-connected computer systems, and stored or executed in a distributed manner. The software and data may be stored on one or more computer-readable recording media.

Although the embodiments have been described with limited drawings as described above, those skilled in the art can apply various technical modifications and variations based on the above. For example, even if the described techniques are performed in a different order than the described method, and/or the components of the described system, structure, device, circuit, etc. are combined or combined in a different form than the described method, or are replaced or substituted by other components or equivalents, appropriate results can be achieved.

Therefore, other implementations, other embodiments, and equivalents to the claims are also included in the scope of the claims described below.

Claims (15)

A multiprocessing-based On-memory Queuing module that performs a Memory Sharing Process that loads image data generated in connection with a reflective film into a shared memory, and a Queuing Process that loads duplicate image data copied from the image data into a Queue; and A Web GUI module for data collection based on a Blackbox, which performs a Streaming & Saving Process for streaming image data acquired from the shared memory, storing defective image data judged to be defective among the image data and duplicate image data acquired from the Queue, and a Defect Streaming Process for checking for defects in the defective image data. A multi-processing based defect image acquisition device for high-speed and wide-area reflective films, including: In the first paragraph, The above multiprocessing-based On-memory Queuing module, Copy the image data generated from the present to the previous set period as the duplicate image data and load it into the Queue, but remove the duplicate image data loaded first from the Queue in a first-in, first-out manner over time. A multiprocessing-based defect image acquisition device for high-speed and wide-area reflective films. In the first paragraph, Since the loading of the image data into the shared memory and the streaming of the image data obtained from the shared memory are multi-processed separately and simultaneously, The above multiprocessing-based On-memory Queuing module, Without waiting for the image data to stream, immediately after loading the image data into the shared memory, the camera photographs the reflective film to generate new image data. A multiprocessing-based defect image acquisition device for high-speed and wide-area reflective films. In the first paragraph, The above Blackbox-based data collection Web GUI module is: The image data of the above streaming t-1 and the duplicate image data of t (where t is the creation time) obtained from the above Queue are stored in the main memory, The data with different creation times and separate life cycles are judged to be defective. Store the defective image data judged as defective in the local storage, and remove the remaining data not judged as defective from the main memory. A multiprocessing-based defect image acquisition device for high-speed and wide-area reflective films. In the first paragraph, The above multiprocessing-based On-memory Queuing module, As the camera that photographs the above reflective film is set to the BayerRG8 shooting mode that uses the Bayer pattern, the image data generated and copied in the BayerRG8 format and the copied image data are converted into an Image Lazy Convert Process for the RGB format for the streaming. A multiprocessing-based defect image acquisition device for high-speed and wide-area reflective films. In the first paragraph, The above Blackbox-based data collection Web GUI module is: For the image data being streamed, if it is observed as defective by the operator, the streaming of the image data is paused at the observation point, Determine whether the above image data that has been paused is defective or not. If judged to be defective, the image data that was paused is saved as defective image data, along with a summary of defective information including the type of defect and the location of the defect. A multiprocessing-based defect image acquisition device for high-speed and wide-area reflective films. In the first paragraph, The above Blackbox-based data collection Web GUI module is: Other Processes are performed to further collect data related to the frequency at which the above-mentioned bad image data is stored, and to check the local storage capacity related to the remaining capacity in the local storage. A multiprocessing-based defect image acquisition device for high-speed and wide-area reflective films. In a multiprocessing-based On-memory Queuing module, a step of performing a Memory Sharing Process for loading image data generated in relation to a reflective film into shared memory; In the above multiprocessing-based On-memory Queuing module, a step of performing a Queuing Process for loading copied image data into a Queue; In a Web GUI module for Blackbox-based data collection, a step of performing a Streaming & Saving Process for streaming image data acquired from the shared memory and saving defective image data judged to be defective among the image data and duplicate image data acquired from the Queue; and In the Web GUI module for collecting the above Blackbox-based data, a step of performing a Defect Streaming Process for checking for defects in the above defective image data A multi-processing based defect image acquisition method for high-speed and wide-area reflective films, including: In Article 8, The steps for performing the above Queuing Process are: A step of copying the image data generated from the present to a previously set period of time as the duplicate image data and loading it into the Queue, but removing the duplicate image data loaded first in a first-in, first-out manner from the Queue as time passes. A multi-processing based defect image acquisition method for high-speed and wide-area reflective films, including: In Article 8, Since the loading of the image data into the shared memory and the streaming of the image data obtained from the shared memory are multi-processed separately and simultaneously, In the above multiprocessing-based On-memory Queuing module, A step of generating new image data by photographing the reflective film by the camera immediately after loading the image data into the shared memory without waiting for the image data to stream. A multi-processing based defect image acquisition method for high-speed and wide-area reflective films, which further includes. In Article 8, The steps for performing the above Streaming & Saving Process are: In the Web GUI module for collecting the Blackbox-based data, a step of storing the image data of the streaming t-1 and the duplicate image data of t (where t is the time of creation) acquired from the Queue in the main memory; In the Web GUI module for collecting data based on the Blackbox, a step for determining whether data with different creation times and separated life cycles are defective; and In the Web GUI module for collecting the above Blackbox-based data, a step of storing the defective image data judged as defective in a local storage and removing the remaining data not judged as defective from the main memory A multi-processing based defect image acquisition method for high-speed and wide-area reflective films, including: In Article 8, In the above multiprocessing-based On-memory Queuing module, a step of performing an Image Lazy Convert Process for converting the image data and the duplicate image data into images, wherein, as the camera photographing the reflective film is set to a BayerRG8 shooting mode using a Bayer pattern, the step of converting the image data and the duplicate image data, which are generated and duplicated in a BayerRG8 format, into an RGB format for the streaming. A multi-processing based defect image acquisition method for high-speed and wide-area reflective films, which further includes. In Article 8, In the Web GUI module for collecting the above Blackbox-based data, if the streaming image data is observed as defective by an operator, a step for pausing the streaming of the image data at the observation point; In the Web GUI module for collecting the above Blackbox-based data, a step for determining whether the paused image data is defective; and In the Web GUI module for collecting the above Blackbox-based data, if it is judged to be defective, a step for saving the paused image data as defective image data together with a summary of defective information including the defective type and defective location. A multi-processing based defect image acquisition method for high-speed and wide-area reflective films, which further includes. In Article 8, In the Web GUI module for collecting data based on the Blackbox, a step for performing Other Process for checking the data collection status related to the frequency with which the defective image data is stored and the local storage capacity related to the remaining capacity in the local storage A multi-processing based defect image acquisition method for high-speed and wide-area reflective films, which further includes. A computer-readable recording medium recording a program for executing the method of Article 8.

Images