WO2001055705A1 - Installation and method for detecting light reflecting faults - Google Patents

Abstract

The invention concerns an installation for detecting light reflecting faults in the wall of flasks (40) moving past, in a bottling or a production line, said detection being carried out in a test zone, comprising: a system of measuring light sources (1); at least an observation camera (3) for detecting luminous objects, whereof the optical axis forms an angle between 65 DEG and 95 DEG relative to the lighting direction of the sources (1), and an angle between 5 DEG and 35 DEG relative to the forward displacement direction of the flasks (40); a data processing system; an assembly of sealed and pressurised housings for protecting the optical and electrical elements from being splashed with water, outlets being provided in front of the external optical elements to protect them with a clean air stream derived from the pressurised housings.

Description

 INSTALLATION AND METHOD FOR DETECTION OF GLAZING

Subject of the invention

The present invention relates to an original installation for detecting glaze type defects in the wall of scrolling, transparent or translucent containers.

The invention also relates to the method implemented in the installation.

Technological background and state of the art

In beverage bottling lines, there is an increasing desire to use recyclable plastic bottles or flasks. However, these bottles, when they are reused, or even sometimes when they are new, can present a type of defect called glaze. An glaze is an alteration or break in the thickness of the wall of the container, oblong, glassy in appearance, detectable with the naked eye or by optical means. The longitudinal extension of these glazes can reach several millimeters while their transverse extension can range from a few micrometers to 1 or 2 mm. The bottles affected by this type of defect must imperatively be carefully identified and eliminated from the bottling line, automatically. Indeed, this defect is particularly damaging in the event of re-use with carbonated drinks.

In the prior art, it is known to use a method which consists in illuminating the bottle scrolling from different angles and recording the image of the area to be inspected from different angles as well. A conventional processing of the images thus collected, by means of digital filters, then makes it possible to distinguish the glazes among the numerous sources of reflections emanating from the bottle and recorded by the camera on the same image. The specific choice of filters is determined by a learning procedure for each type of bottle to be inspected. The method described above generally allows, at the cost of heavy computer processing, to distinguish glazes on clean and new bottles. If it is a question of distinguishing the glazes on used bottles, possibly covered with droplets and steam, residues from the preliminary cleaning step on the bottling line, said process leads to an error rate such that it becomes unusable in practice. Indeed, on the one hand, “real” defects, the glazes, can escape detection, which leads to the reuse of defective bottles. On the other hand, “false” defects can be detected and lead to the qualitative rejection of non-defective bottles. In conclusion, this inspection method is linked to significant additional manufacturing costs, both in terms of the IT to be implemented and the level of detection error rate.

Aims of the invention

The invention aims to obtain a new method for the detection of glazes, not having the drawbacks of the state of the art, and in particular which makes it possible to considerably increase the brightness of the glazes only on the recorded images. The contrast between the signal due to a glaze and the spurious signals must be such that it becomes easy to distinguish the desired glaze, even on a damaged bottle covered with droplets and steam. The method of the invention will apply, a fortiori, to clean and / or new bottles. A complementary object of the invention is to achieve a very low rate of false rejections, that is to say less than 0.5 per thousand, while the standard imposed by the methods of the prior art hardly reaches 1 percent. A particular object of the invention is a possible use, without degrading performance, on used bottles, covered ^* with droplets or mist.

Another object of the invention is to obtain a cost price less than or equal to that corresponding to the methods of the prior art.

Finally, the method of the invention aims to be independent, to a very large extent, of the computer processing used, therefore without learning for each type of bottle.

Main characteristic elements of the invention The present invention relates to an installation for the detection of glaze type defects in the wall of bottles subjected to a moving movement, in a bottling or manufacturing line, said detection being carried out in a test area, including

A battery of measurement light sources, activated sequentially during the passage of said vial to be tested, each source producing a parallel beam delimiting a surface to be tested on said vial, called an illuminated area, the beams of the different sources of the battery being adjacent, - • at least one observation camera making it possible to detect luminous objects, the optical axis of which forms on the one hand an angle of between 65 ° and 95 ° relative to the direction of illumination of the sources, and on the other hand an angle between 5 ° and 35 ° relative to the mean direction of movement of the bottles, said at least one camera being provided with a programmable reading detector so that an acquisition is made in a window delimiting said illuminated area and following said bottle in its movement;

An information processing system making it possible to differentiate said light objects detected by said at least one camera and to reject said objects which do not have the appearance or the properties of a glaze, in particular a characteristic elongation;

• a set of sealed and pressurized boxes intended to protect optical and electronic elements from water splashes, leaks being organized in front of external optical elements in order to protect these by a clean air flow coming from pressurized boxes.

Advantageously, an optical device detects the passage of each bottle in the test area in order to produce and control the sequential ignition of the sources.

Preferably, said optical device emits a beam whose wavelength is different from that of the beams of the measurement sources, the interference between the two types of beams being eliminated by means of optical filters.

Each measurement source preferably comprises a laser diode, a collimating lens and a diaphragm. In a particularly advantageous manner, the detector of the camera (s) is a detector of the CMOS type and the detector of the camera (s) is (are) inclined with respect to the optical axis of an angle necessary for the correction of the focusing corresponding to the axial displacement of said bottles.

The invention also relates to a method for the detection of glaze type defects in the wall of bottles subjected to a moving movement, in a bottling or manufacturing line, said detection being carried out in a test area, characterized in that :

• a battery of measurement light sources is switched on, activated sequentially during the passage of said bottle to be tested, each source producing a parallel beam delimiting a surface to be tested on said bottle, called an illuminated area, the beams of the different sources of the battery being adjacent ;

• luminous objects are detected, by means of at least one observation camera, the optical axis of which forms on the one hand an angle of between 65 ° and 95 ° relative to the direction of illumination of the sources, and on the other hand, an angle of between 5 ° and 35 ° relative to the mean direction of movement of the bottles, said at least one camera being provided with a programmable reading detector so that an acquisition is made in a window delimiting said illuminated area and following said bottle in its movement;

• we differentiate, by means of an information processing system, said light objects detected by said at least one camera and we reject said objects which do not have the appearance or properties of a glaze, in particular an elongation feature ; • it protects, thanks to a set of sealed and pressurized boxes, optical and electronic elements of water splashes, leaks being organized in front of external optical elements in order to protect them by a clean air flow coming from the boxes pressurized. Advantageously, the information processing process forming part of the method according to the invention, for the differentiation of the detected light objects, including the glazes, comprises the following steps:

• a thresholding is carried out so as to transform a grayscale image into a binary image which is faster to process subsequently; • on the thresholded image, objects are detected, which are defined as a set of pixels of the same value, adjacent in an octagonal grid;

• among said objects, those which do not reach a critical size defined by calibration and those touching the edges of the window are eliminated;

• for the selected objects, a characterization is made on the basis of the surface equal to the number of pixels contained, the center of gravity, the moments of inertia, preferably the main moments of inertia and the ellipse of inertia, preferably its orientation and its aspect ratio;

• the glazes are identified, preferably on the basis of the characteristics of said inertia ellipse obtained after calibration; • errors are detected, preferably by the fact that the images containing too many objects having a total surface greater than a determined level during the calibration of the system generate a error code.

Brief description of the drawings

Figures l.a (elevation view) and l.b (plan view) schematically represent the geometry of detection of glazes according to the method of the invention.

Figure 2 shows schematically the installation for the implementation of the method according to

1 invention. Figure 3 shows schematically the acquisition area and the types of faults.

Figure 4 shows schematically the source block with the measurement and position beams respectively.

Detailed description of the invention

Definition and principle of the measurement The glaze is a rupture of the material of which the bottle is made throughout its thickness. The thin layer of air separating the two edges of the rupture allows total reflection of the incident light to be obtained at a critical angle inside the wall. This total reflection which takes place only for the critical angles returns the light to the specular angle with an intensity much higher than that of the light scattered by all the other reflective defects or surfaces from which we want to distinguish the glaze. As the glaze is not a perfect plane, the conditions for brilliant specular reflection are met in a cone with an angle sufficient to make the process practical, otherwise a source and a detector would be required for each angle. The schematic diagram is shown in profile view in FIG. 1 a. The collimated light coming from the source 1 touches the glaze 2 and is returned preferably in the specular direction, with respect to the glaze, where it is detected by a matrix detector 3. At the same time, the fog 4 (or any other disturbing element), diffuses the light in all the directions and preferably in the direction specular 5 in relation to the surface of the bottle. This diffusion will therefore be very weakly in the direction 2,3 since the glaze is perpendicular to the surface of the bottle. A judicious choice of the direction of lighting 1,2 and the direction of observation 2,3 makes it possible to greatly increase the contrast between the signal due to a glaze and the parasitic signal due to reflection in a droplet or due to diffusion on the mist or on the bottle's defects (molding seam, scratches, frosted surface).

An angle of the order of 70 ° between the source beam and the direction of observation makes it possible to clearly highlight the glazes (Figure l.b). The displacement of the bottles is then made perpendicular to the source beam.

The images detected by the matrix detector 3 generally include a wide variety of objects among which it is necessary to locate the only glazes. This location is greatly facilitated by the capture method intended to highlight objects giving rise to specular reflection. It is however necessary to subject the data collected to relatively simple and rapid processing. This treatment consists in locating the typical characteristics which distinguish the glazes of the parasitic images: the localization (window isolating the region of the image where the angular relations are favorable to specular reflection), the intensity (thresholding) and the geometry (elongation and orientation). Description of a typical installation. To detect any glaze on a scrolling bottle, a simple solution is to rotate the bottle on itself at least 360 ° during the measurement. In order to keep the mean angle of incidence constant during the movement of the bottle, a series of sources is preferably used which light up sequentially when the detection zone is in the right place. The precise instant when each source must be switched on and then off is determined, preferably by an optical position detector which determines when the bottle enters the field of each source. The system is thus made independent of the regularity of the movement of the bottles.

The detector is a "smart" camera, composed of a lens, an addressable matrix detector, for example of the CMOS type, a frame grabber, control electronics and a microprocessor. These various elements can be included in a single box resembling, in this case, a conventional camera, or be made up of the assembly of several boxes. The optical axis of the camera is placed at a very small angle relative to the direction of movement of the vials so that the angle of observation varies little between the extreme positions covered by the camera. If the distance between these two extreme positions is not sufficient for the bottle to have rotated at least 360 ° during the time from one extreme position to the other, we will use a second camera which will take the relay of the first.

Description of a preferred embodiment of the invention An example of general geometry Illumination and observation is given in FIG. 2. The bottles are transported here by a carousel 6 represented by two curves guiding the movement. The same bottle is represented in three successive positions PI, P2 and P3 (arrow 70). This bottle carries a glaze 2 which rotates with the bottle in the direction indicated by the arrows 71. The sources 1 are placed along a tangent to the displacement and are switched on and off sequentially, in the direction indicated by the arrow 72, so to always illuminate, by a parallel beam 11, an area located behind the neck of the bottle, area where the glazes must be detected. Figure 2 indicates the source which is in operation for the three represented positions of the bottle. We see that the bottle does a little more than a full turn on itself before leaving the area covered by the sources. The total displacement of the bottle having made a full turn on itself is too large to be covered by the field of a single camera. Two cameras 3 have therefore been represented, designated by C1 and C2 in FIG. 2, each covering half of the total displacement with an overlap in the center.

The detector of each camera 3 is mounted at an angle such that the bottles remain properly focused during their movement. The glazes are observable at an angle a around the ideal position. This angle is about 25 °. The distance between two neighboring sources 1 must be less than the displacement of the bottle corresponding to the angle of rotation a during which a glaze is detectable. The distance between the sources, the total length of the sequence and the overlap of the field of the two cameras are calculated according to the rotation and the displacement of the bottles so that a glaze located in any azimuth cannot escape observation. In this calculation, a safety factor between 1.5 and 2 is taken into account.

Data detection and processing

The cameras are advantageously provided with a CMOS type matrix detector. This type of detector can easily be programmed so that only part of the detector's pixels are read at each acquisition. In the example given, a reading window is programmed to follow the area illuminated by the source when the bottle is moved. This window covering only a fraction corresponding to about one fortieth of the detector, the reading and the acquisition can be done at a rate forty times greater than if one read the whole detector with each acquisition. Such a window is illustrated in Figure 3.

The observation window is represented by the rectangle 7 encompassing the entire figure. This window moves to follow the bottle, so that it always includes the area illuminated by the active source. This area is represented by the closed curve 8. The frame grabber is programmed so that it reads only the active window and in this window only the pixels or groups of useful pixels. These are determined by the spatial resolution necessary to distinguish error-free glazes. In our case, we determined that the smallest detectable object corresponded to the Nyquist cutoff frequency projected on the bottle, i.e. 0.25 mm if we activate every other pixel of the detector. This resolution is more than enough to distinguish the smallest glazes (> 1.5 mm) from the water droplets detected as a point reflection of the order of 0.25 mm in diameter. The active pixels are represented in FIG. 3 by a network of points. Various objects corresponding to glazes 21, 22 or to a water droplet 23, 24 are shown in the same figure. Image processing comprises the following steps: - thresholding is carried out so as to transform a grayscale image into a binary image which is faster to process subsequently. During this thresholding, fixed pixel defects can be eliminated. The threshold is also chosen so ^'to eliminate weak diffuse images; on the “thresholded” image represented in FIG. 3, the “objects” are detected. These are defined as a set of pixels of the same value, adjacent in an octagonal grid. For example, FIG. 3 contains three objects 21, 22 and - 23. The signal 24 corresponding to the reflection of a droplet of water has been eliminated by this stage of the treatment. Among these objects are eliminated those which do not reach a critical size defined by calibration, and those touching the edges of the window; for the selected objects, the processing proceeds to a characterization: surface equal to the number of pixels contained, center of gravity, moments of inertia and main moments of inertia, ellipse of inertia (orientation, ratio of elongation); the glazes are identified essentially based on the characteristics of the inertia ellipse after calibration; errors are detected by the fact that the images containing too many objects having a total surface area greater than a determined level during the calibration of the system generate an error code. The principle of processing, which could also follow another logic, is to minimize the implementation time by taking into account the type of matrix detector used. Thus, in the illustrated case, we have already seen the gain factor forty linked to the reading of a movable window following the bottle. A factor of four is gained by reading only one pixel out of two. Thresholding and object detection then only process the few tens of pixels representing the selected objects. The characterization operations, although relatively complex, therefore only apply to an extremely limited number of data in each image, and most often to no data at all. The entire identification process thus takes between 3 and 4 ms, depending on the cleanliness of the bottles, and allows rates of the order of 15 bottles per second with current cameras and a system examining one bottle at a time. This rate could easily be doubled if necessary.

Other installation configurations. The arrangement illustrated in Figure 2. is particular to the example given. The number of cameras required depends on the ratio between the speed of rotation and movement of the bottles. One or two cameras 3 will therefore be required depending on whether this ratio is greater or less, and depending on the security factor chosen. The device described can also be adapted to a linear movement of the bottles. It is therefore not essential to pass the bottles through a carousel.

The illustrated example applies particularly to vertical glazes. It can be adapted without major modification to the detection of horizontal glazes by placing the sources above the bottles so as to reproduce the angular relationship between the source beam, the glaze and the direction of observation.

Synchronization of the sequential ignition of sources. The optical position detector which controls the sequential ignition of the sources is advantageously placed directly above the sources. One detector per source is used, which makes it possible to completely overcome irregularities in the movement of the bottles. such as speed variations. A simple embodiment is shown in Figure 4.

The source 1 produces the collimated measurement beam 11 which covers the area 8 to be measured. The source 10 advantageously mounted in the same source block 13, produces the position detection beam 12 detected by the position detector 30. When the bottle passes, it cuts the beam 12 and the reduction in the signal is detected by 30. It suffices then to adjust the delay between this information and the moment when the source 1 must be switched on to illuminate the zone 8, in order to obtain perfect synchronization of the sources 1 with the movement of the bottle. The fact of mounting the sources 1 and 10 in the same block 13 increases the long-term stability between the two beams. The wavelengths of the two sources 1 and 10 are chosen far enough to avoid any interference, by using filters placed in front of the cameras and the position detectors.

Claims

1. Installation for the detection of type 2 glaze defects in the wall of vials 40 subjected to ^a scrolling movement, in a chain of traffic jam or manufacturing, said detection being carried out in a test area comprising: • a battery of measurement light sources 1, activated sequentially during the passage of said bottle to be tested, each source producing a parallel beam 11 delimiting a surface to be tested on said bottle, called illuminated area 8, the beams of the different sources of the battery being adjacent; At least one observation camera 3 making it possible to detect luminous objects, the optical axis of which makes on the one hand an angle comprised between 65 ° - and 95 ° with respect to the direction of illumination of the sources 1, and d 'on the other hand an angle between 5 ° and 35 ° relative to the average direction of movement of the bottles 40, said at least one camera 3 being provided with a programmable reading detector so that an acquisition is made in a window delimiting said illuminated zone and following said bottle in its movement; • an information processing system making it possible to differentiate said light objects detected by said at least one camera 3 and to reject said objects which do not have the appearance or properties of a glaze 2, in particular a characteristic elongation; • a set of sealed and pressurized boxes intended to protect optical and electronic elements from water splashes, leaks being organized in front of external optical elements in order to protect them by a flow of clean air coming from pressurized boxes. 2. Installation according to claim 1, characterized in that an optical device detects the passage of each bottle in the test area in order to produce and control the sequential ignition of the sources. 3. Installation according to claim 2, characterized in that said optical device emits a beam whose wavelength is different from that of the beams of the measurement sources, the interference between the two types of beams being eliminated by means of optical filters . 4. Installation according to any one of the preceding claims, characterized in that each measurement source 1 comprises a laser diode, a collimating lens and a diaphragm. 5. Installation according to any one of the preceding claims, characterized in that the detector of said at least one camera 3 is a detector of the CMOS type. 6. Installation according to any one of the preceding claims, characterized in that the detector of said at least one camera 3 is inclined relative to the optical axis by an angle necessary for the correction of the focus corresponding to the displacement axial of said bottles. 7. Method for the detection of glaze type 2 defects in the wall of bottles 40 subjected to a moving movement, in a bottling or manufacturing line, said detection being carried out in a test zone, characterized in that: we turn on a battery of measurement light sources 1, activated sequentially during the passage of said bottle to be tested, each source producing a parallel beam delimiting a surface to be tested on said bottle, called the illuminated area 8, the beams of different sources of the battery being adjacent; • luminous objects are detected, by means of at least one observation camera 3, the optical axis of which forms on the one hand an angle of between 65 ° and 95 ° relative to the direction of illumination of the sources, and on the other hand an angle of between 5 ° and 35 ° relative to the direction of movement of the bottles, said at least one camera being provided with a detector with programmable reading so that an acquisition is made in a window delimiting said illuminated zone and following said bottle in its movement; • we differentiate, by means of an information processing system, said light objects detected by said at least one camera and we reject said objects which do not have the appearance or properties of a glaze 2, in particular a characteristic elongation; • it protects, thanks to a set of sealed and pressurized boxes, optical and electronic elements from water splashes, leaks being organized in front of external optical elements in order to protect them by a clean air flow coming from- s pressurized housings. 8. Method according to claim 7, characterized in that the information processing process, for the differentiation of the detected light objects, including the glazes, comprises the following steps: • a thresholding is carried out so as to transform a grayscale image into a binary image which is faster to process subsequently; • on the thresholded image, objects are detected, which are defined as a set of pixels of the same value, adjacent in an octagonal grid; • among said objects, those which do not reach a critical size defined by calibration and those touching the edges of the window are eliminated; • for the selected objects, characterization is performed on the basis of the surface equal to the number of pixels contained, the center of gravity, the moments of inertia, preferably the main moments of inertia and the ellipse of inertia, preferably its orientation and its aspect ratio; • the glazes are identified, preferably on the basis of the characteristics of said inertia ellipse obtained after calibration; • errors are detected, preferably by the fact that the images containing too many objects having a total surface greater than a determined level during the calibration of the system generate an error code.