WO2024141740A1 - Device and method for analysing a raised inspection pattern of a wall of a glass container - Google Patents

Abstract

The invention relates to a device (1) for analysing light passing through a raised pattern of an inspection sector of a wall (3, 4) of a glass container (2), the device (1) comprising: a light source (5) for illuminating the entire inspection sector by transmission, the light source being able to emit an overall beam comprising multiple specific beams, the specific beams each having a different angular spectrum (α), and a different polarisation property; an image acquisition device (6) able to detect the specific polarisation property of at least one specific beam, in order to acquire n primary images, with n ≥ 2, each primary image (IP1, ..., IPn) being produced by means of a combination of different specific beams received from the overall beam by the acquisition device and selected according to the polarisation property of the specific beams; and a unit for processing the n primary images in order to analyse a raised pattern of a wall (3, 4) of the container.

Description

Description

Title of the invention: Device and method for analyzing an inspection relief of a wall of a glass container

[0001] The present invention relates to the technical field of inspection of glass containers.

[0002] More particularly, the invention relates to a device and a method for optical analysis of a glass container, with a view to analyzing reliefs present within an inspection sector of the container.

[0003] Typically, a glass container comprises side walls and a bottom wall. The side and/or bottom walls generally have reliefs.

[0004] A first category of reliefs comprises reliefs deliberately formed by molding on the surface of the glass of the container, during their forming in the finishing molds of the manufacturing machine. Such reliefs are for example projecting from the surface, and are called moldings. Such moldings are for example information to be decoded, such as a code relating to a number integrating two digits with seven segments or a distribution of pearls, generally organized in a ring, and the presence and position of each of which encode the number of mold.

Other known reliefs also falling into the first category are formed by removing material from the surface of the glass of the container, and are called engravings. Such engravings are for example bar codes, for example Datamatrix codes engraved by laser on the container leaving the oven, as described in documents EP 2114840 Bl, or JP H09128578.

Other known reliefs falling within the first category are for example inscriptions intended for the end user of the container, such as decorative or informative inscriptions linked to the contents of the container.

[0007] It is desired that the engravings and moldings be well rendered. A well-rendered engraving or molding is such that the formed surface has taken exactly the shape of the desired relief. In the case of moldings, this means example that the glass surface has taken the shape given by the mold. In the case of laser engraving, this means, for example, that the dots formed have the desired depth.

[0008] A second category of reliefs consists of unplanned or accidental reliefs, which are in reality manufacturing defects. Such accidental reliefs generally appear in the bottom wall of the container, and are generally characterized by a small projection of glass having a point shape going towards the interior of the glass container. An example of this type of defect called "punch pull" or "spike", located in the bottom wall of the final container, comes from adhesion of the glass to the punch forming the blank in a blank mold. Such a blank is made at the first stage of container forming in a press-and-blow process machine, which is the preferred process for making wide-mouthed containers such as glass canning jars.

[0009] Throughout the remainder of this text, the word “relief” refers to a relief belonging to the first or second category presented above.

[0010] The detection of reliefs is of great importance in the glass manufacturing process. Indeed, in the case of reliefs belonging to the first category, it is desired to know the information linked to the codes, and these must be able to be visible during an inspection. The information collected can in fact be used for traceability purposes.

[0011]In the case of decorative reliefs, it is necessary to be able to ensure that the reliefs are correctly produced.

[0012] For reliefs falling into the second category, in other words defects, these must be able to be identified as precisely as possible.

[0013] To this end, the detection of reliefs is currently carried out using their light refraction properties. Indeed, the refractive properties between a side and/or bottom wall of the container devoid of relief, and a side and/or bottom wall with relief are different. [0014] For this, a device is used comprising a light source intended to be projected onto the part of the container to be inspected, and a camera making it possible to create an image. The reliefs materialize on such an image by appearing darker than the rest of the image. By means of dedicated algorithms, for example segmentation algorithms, the pixel values constituting the acquired image are compared with the neighboring pixels of this image.

[0015] In order to allow accurate detection of the reliefs, that is to say a detection corresponding well to the relief present on the container, the image acquired by the camera must present optimum contrast of the reliefs.

[0016] Indeed, if the contrast is too low, then minor reliefs may potentially not be detected, running the risk of allowing non-compliant containers to pass due to the presence of a relief defect, or else to render illegible reliefs deliberately formed on the side or bottom wall of the container. If the contrast is too great, then dark areas corresponding to so-called parasitic reliefs, that is to say not corresponding to relief which should be detected, appear.

[0017] To deal with such a problem, it has been proposed to produce images using a device which creates contrast on the reliefs, and allows an adjustable contrast level. In practice, from test samples, the contrast adjustment (or modulation) device is adjusted, thus allowing calibration of the contrast according to the inspection parameters or the type of container to be detected. Note that the contrast adjustment once made is frozen for the entire production in progress. It can then be modified as required by modifying the adjustment device.

[0018]It is therefore necessary to be able to easily adjust the contrast of the image acquired during a relief inspection.

[0019] Document US 4,606,634 proposes a device making it possible to reinforce the contrast of appearance of reliefs in an image while adapting the device to the physical and optical characteristics of the object and thus adjusting said contrast. More particularly, such a document describes a light source presenting a diffuse emitting surface whose spatial intensity distribution has a predetermined extent and shape. Such extent and shape is controllable by a mask. In the direction of travel of the light, a converging lens, the container and a camera are arranged.

[0020] The converging lens allows the illumination of the entire sector to be inspected with, at any point in the sector, beams having an extent or width of angular spectrum, that is to say an angle between the extreme rays of said beam , constant. For the sake of simplification, throughout the remainder of this text, the expression “angular spectrum” refers to “extent of angular spectrum” or “width of angular spectrum”.

By varying the extent and shape of the mask, it was found that the angular spectrum changes. In this way, it can be seen that in the image formed in the camera, raised areas having a slope greater than a value depending on the aperture of the mask go out, and therefore appear black. The smaller the opening of the mask, the more the faint reliefs are blackened. By strong relief we mean a relief whose local slope is steep compared to a surface of slight curvature. In other words, the contrast obtained for reliefs in the images depends directly on the aperture of the mask. The contrast obtained for reliefs in the images also depends on the focal length and aperture of the camera lens, but for the sake of simplification these parameters are not discussed below.

[0022] Industrially, such a principle is used in container inspection devices, the mask being a diaphragm adjustable by an operator.

[0023] More particularly, when calibrating the inspection device, the opening of the diaphragm is adjusted according to the type of relief to be detected. The more pronounced the reliefs (for example in the case of strongly rendered engravings), the more open the diaphragm will be. In such a situation, faint reliefs are not legible.

[0024]In order to allow the reading of less marked reliefs, the aperture is reduced, but noise appears on the image, causing areas to appear dark while the corresponding areas on the container do not contain reliefs to be detected.

[0025]Also, a diaphragm adjustment compromise must be found in order to determine the correct diaphragm opening.

Such a compromise is difficult to determine, since it varies depending on the inspection conditions, for example the source, the type of container to be inspected, the thermal conditions, etc.

[0027] Furthermore, in the case of a series of containers of the same production having reliefs to be detected, for example moldings encoding information, we are in the presence of a mixture of containers carrying slightly moldings. marked and strongly marked moldings. In such a situation, it is very difficult to find a compromise adjustment of the diaphragm.

[0028] Furthermore, for a given setting, the risk of decoding errors on containers bearing weakly marked moldings nevertheless remains possible because part of the code will not be detected and/or for containers bearing strongly marked moldings , parasitic shadows can impact the analysis of the images and the final result may then be erroneous.

[0029] Adjusting the diaphragm is also a long process, requiring the intervention of an operator, and is therefore very restrictive in an industrial approach, that is to say for inspection in large series of containers. .

[0030] Finally, adjusting the diaphragm requires stopping production, since during such an operation, no detection can be made, which has a negative impact on the general availability of the production line of glass container.

The invention aims to respond to the aforementioned drawbacks.

[0032] A first object is to propose a device for analyzing through light a relief of an inspection sector of a wall of a glass container making it possible to improve the rate of detection of the reliefs to be detected . [0033] A second object is to propose such a device making it possible to adjust the source in a non-mechanical manner, and adapted to any type of glass container.

[0034] A third object is to propose such a device that can be implemented at reduced cost.

[0035] A fourth object is to propose a method of analyzing through light a relief of an inspection sector using the device presented above.

[0036] According to a first aspect, it is proposed a device for analyzing through light a relief of an inspection sector of a wall of a glass container, the device comprising a light source for illuminating in transmission the entire inspection sector, the light source being capable of emitting a global beam comprising several specific beams, the specific beams each having a different angular spectrum, and a different polarization property, an image acquisition device capable of detecting the specific polarization property of at least one specific beam, in order to acquire n primary images, with n > 2, each primary image being produced by means of a combination of different specific beams received from the overall beam by the acquisition device and selected according to their polarization property, a unit for processing n primary images to analyze a relief of a wall of a glass container.

[0037] Various complementary characteristics can be provided alone or in combination:

- the light source comprises an emitting surface composed of p light zones with p > 2, comprising a central zone, and at least one peripheral zone, the light zones each emitting a specific beam, each beam having its own and distinct polarization property of the other beams, an optical projection lens of the emitting surface, so that each point of the inspected sector is crossed by a light beam whose extent of the angular spectrum depends on the extent of the emitting surface emitting said beam, the device acquisition integrates a photoelectric image sensor capable of detect the polarization property of specific beams coming from the light source.

- the polarization property is the direction of polarization or a phase shift between the two orthogonal components of the electric field vector of the light waves composing the overall beam

- the acquisition device is capable of acquiring primary images so that during the acquisition of the n primary images, the acquisition device constitutes n-1 primary images by excluding the light having the polarization property of one of the associated light zones, all or part of the light flux emitted by each of the light zones contributes to the formation of at least one of the n primary images;

- the processing unit is capable of calculating by linear combinations the values of the pixels of several primary images, a first secondary image having pixels having a value corresponding to the image generated solely by the light coming from Ml contiguous central light zones taken among the p light zones, with Ml between 1 and p-1, at least one second secondary image presenting pixels having a value corresponding to the image generated by the light coming from M2 contiguous central light zones, with M2 taken from among the p light zones, M2 being between Ml+1 and p, the angular spectrum of the beams contributing to the second secondary image being broadened relative to the restricted angular spectrum of the beams contributing to the first secondary image, due to the fact that the extent of the M2 contiguous central light zones combined is greater than that of the Ml contiguous central light zones combined, analyze the n secondary images in order to determine the presence and or morphology of a relief carried by a wall of the container.

- the emitting surface is equipped with at least p-1 source polarizing filters having a linear polarization direction, each linear source polarizing filter being assigned to a peripheral zone, the p-1 filters having a polarization direction different from each other ; - the p-1 filters have a direction of polarization different from each other, chosen from the directions 0° 45° 90° and 135°;

- the emitting surface comprises a transparent or translucent diffuser, backlit uniformly by a primary light, a screen of liquid crystal cells is placed on the diffuser, the diffuser being controlled to present, depending on p peripheral zones, different birefringence properties in order to produce, for at least p-1 light zones, light with a polarization property different from each other;

- the acquisition device comprises a polarimetric camera;

- the lens is convergent, and is placed at a distance from the emitting surface substantially equal to the focal length;

- the lens is taken from one of the following lenses: plane convex lens, bi-convex lens, Fresnel lens

[0038] According to a second aspect, there is provided a method for analyzing through light a relief of an inspection sector of a wall of a glass container, the method comprising: an illumination step consisting of illuminating the inspection sector in transmission, the light source emitting a global beam comprising several specific beams, the specific beams each presenting a different angular spectrum, and a different polarization property, a step of acquiring n primary images , with n > 2, by means for each primary image of a combination of different specific beams received from the overall beam by the detection device and selected according to their polarization property, a step of processing the primary images to analyze a relief of a wall of a glass container.

[0039] Various complementary characteristics can be provided alone or in combination:

- during the illumination step, the light emitted by the source passes through the inspected sector of the wall of the container with a light source comprising, in the direction of travel of the light, an emitting surface, an optical projection lens of the emitting surface, so that that each point of the inspected sector receives a light beam having an angular spectrum, the emitting surface composed of p concentric annular light zones, with p > 2, the light zones each emitting a specific beam, each specific beam having its own polarization property and distinct from each other;

- the acquisition step comprises, by means of at least one acquisition device, the acquisition of n primary images, with n greater than or equal to two, such that at least each of n-1 primary images is made up of excluding the light having the property of polarization of one of the associated zones all or part of the light flux of each of the light zones contributes to the formation of at least one of the n primary images;

- the processing step comprises a sub-step of calculating by linear combinations the pixel values of the n primary images of a first secondary image presenting a pixel value corresponding to the image generated solely by the light coming from Ml light zones contiguous central areas, with Ml between 1 and n-1, and at least one second secondary image whose pixel value corresponds to the image generated by the light coming from M2 contiguous central light zones, with M2 between M+ l etn, the extent of the M2 contiguous central light zones combined is greater than that of the Ml contiguous central zones combined, a determination sub-step integrating the analysis of at least two secondary images, to determine the presence and/or the morphology of a relief carried by the inspection sector, the relief having inclined edges and horizontal regions, the inclined edges of the relief appearing in the secondary images differently contrasted compared to the horizontal regions of the inspected sector;

- during the illumination step, the different polarization property for the p light zones is a zero linear polarization property, or a linear polarization according to a polarization direction specific to the light zone, or a phase shift between the two orthogonal components of the electric field of the light waves composing the overall beam;

- during the illumination step, the number of light zones is three or four, the light zones emitting linearly polarized light in different directions, during the acquisition step, at least n primary images are acquired, and at least n-1 of these primary images are obtained by filtering the light through n-1 linear acquisition polarization filters each having their polarization direction orthogonal to the polarization orientation of each of the n-1 respective polarized light zones, by means of at least two cameras or a polarimetric camera;

- the luminous zones emit linearly polarized light in different directions chosen from the following orientations relative to the direction of propagation of the light: 0°, 45°, 90°, and 135°;

- a phase of examining each secondary image to obtain for each a relief analysis result integrating a read sign or a recognized defect and a confidence index;

- a choice phase in which a unique final reading or recognition result is the most certain result chosen from said relief analysis results;

- the determination sub-step comprises a phase of segmentation of each secondary image to obtain for each at least one object of interest, a phase of association of objects of interest corresponding between secondary images by at least partial superposition, a phase of analysis of the associated objects of interest by combining photometric or morphological characteristics of said objects of interest, in order to obtain at least one sign reading or defect recognition result corresponding to each associated object;

- the determination sub-step includes a phase of constructing a tertiary image integrating an analysis of the secondary images for constitute from pixels chosen in several secondary images, in a tertiary image, an object of interest representing a reconstituted sign, a sign reading phase consisting of analyzing the reconstituted sign in order to read it.

- the examination phase, the analysis phase or the sign reading phase is carried out by means of a classifier resulting from supervised learning, such as a Bayesian classifier or a neural network, trained to read signs signs or recognition of defects;

- in the processing step, the analysis of the primary images comprises the following sub-steps a presentation sub-step consisting of simultaneously presenting n primary images as input to an image classifier trained for reading or recognition of defects in primary images or a sub-step consisting of calculating n secondary images to present them as input to an image classifier trained to read or recognize defects in secondary images, a sub-step of classification by the previously trained image classifier.

[0040] Other characteristics and advantages of the invention will appear more clearly and concretely on reading the following description of embodiments, which is made with reference to the appended drawings in which:

[0041] [Fig. 1] Figure 1 illustrates a schematic representation of an example of a relief analysis device;

[0042] [Fig. 2] Figure 2 schematically represents a perspective view of an exemplary embodiment of a photoelectric sensor of a polarimetric camera;

[0043] [Fig. 3] Figure 3 schematically represents a top view of the photoelectric sensor of Figure 2, with a detail on a group of photoelectric cells; [0044] [Fig. 4] Figure 4 schematically illustrates a top view of an acquisition device of the analysis device, the acquisition device having several cameras;

[0045] [Fig. 5] Figure 5 schematically illustrates a top view of an acquisition device of the analysis device, the acquisition device presenting a single camera equipped with several sensors;

[0046] [Fig. 6] Figure 6 schematically illustrates a light source provided with two polarizing source filters of circular shape;

[0047] [Fig. 7] Figure 7 schematically illustrates a light source provided with four circular source polarizing filters;

[0048] [Fig. 8] Figure 8 schematically illustrates a light source provided with four polarizing source filters of rectangular shape;

[0049] [Fig. 9] Figure 9 illustrates a block diagram of an analysis method;

[0050] [Fig. 10] Figure 10 illustrates a block diagram of a processing step of the method of Figure 9;

[0051][Fig. 11] Figure 11 schematically illustrates an implementation of the method

[0052] [Fig. 12] Figure 12 schematically illustrates an implementation of the method, with detail on a processing step according to a first variant;

[0053] [Fig. 13] Figure 13 schematically illustrates an implementation of the method, with detail on a processing step according to another variant;

[0054] [Fig. 14] Figure 14 schematically illustrates an example of four primary images and four secondary images of a container bottom wall;

[0055] [Fig. 15] Figure 15 schematically illustrates an example of a point in an inspected sector, a specific beam being generated by a first light zone;

[0056] [Fig. 16] Figure 16 schematically illustrates three emission configurations for the analysis of an inspection sector. [0057] Referring to Figure 1 illustrating an analysis device 1 making it possible to analyze in passing light an inspection sector of a glass container 2, so as to allow the analysis, it is that is to say the characterization of a relief on the wall of the container 2.

[0058] In the embodiment described, the container 2 is a bottle, but in other embodiments, the container 2 is a pot, or a bottle.

[0059] In the embodiment described, the inspection sector is located on a bottom wall 3 of the container 2. In other embodiments, the inspection sector is located on a side wall 4 of the container 2. .

[0060] In the embodiment currently described, the inspection sector occupies the entirety of the bottom wall 3, but in other embodiments not shown, the inspection sector occupies a portion of the wall of bottom 3 or side wall 4.

[0061] In the embodiment shown, the analysis device 1 comprises a light source 5, and an acquisition device 6.

[0062] The light source 5 emits for example a global beam comprising several specific beams each presenting a different angular spectrum α, the angular spectrum being able for example to be visualized in FIGS. 15 and 16.

In other words, the overall beam integrates specific beams each having a difference between the extreme angles of incidence.

[0064] In this way, the light source 1 uniformly illuminates the inspection sector.

[0065] In order to allow the emission of specific beams, the light source 1 comprises, in the direction of travel of the light, an emitting surface 7 having a predetermined shape and dimension.

Advantageously, the emitting surface 7 is oriented to emit light towards the inspection sector, the aim being to allow lighting going from the outside towards the inside of the container 2. [0067]Advantageously, and more specifically, the emitting surface 7 is composed of p light zones Zl, Zp, with p > 2, the light zones Zl, Zp each emitting a specific beam, each specific beam having its own polarization property and distinct from other specific bundles. We describe later in this document how the p light zones are created.

[0068] In Figure 15, an example of the analysis device 1 can be seen with an inspection point, at the location of which a relief is represented. The first central light zone Zl emits a specific beam (grayed) having an angular spectrum a having a value al, the second light zone Z2 being assumed to be off or unused for an acquired image. When the first Zl and the second light zone Z2 are both lit, they emit a specific beam having an angular spectrum a having a value a2. As it can be observed a2 is greater than al, due to the size of the source presents a beam emitted by the first light zone Zl and second light zone Z2.

[0069] In the case of a multiplicity of observation points on the wall of the container 2, as illustrated in Figure 16, each of the points in the inspected sector receives a specific beam. As can be noticed, for a given source, each specific beam presents an identical angular spectrum. The zones Zl, Z2 each have a different emission surface Cl, C2, C3, the angular spectrum of each specific beam is different. In this exemplary embodiment, two primary images IPI, IP2 are acquired at the same time each with the angular spectra al and a3+a4 respectively. It is then possible to calculate secondary images by combining the primary images to obtain, for example, images IS1 and IS2 corresponding to angular spectra of respective extents al and a2 = a3+a4.

[0070] We define a polarization property of light as being a state of polarization of light. In other words, a polarization property of light is characterized by the angle of the linear polarization direction, or by a phase shift value between two orthogonal components of the field electrical light waves or even, if said phase shift is worth a quarter of a wavelength, by the direction of circular polarization.

[0071] In order to be able to project the light rays to infinity, the analysis device 1 comprises a converging lens 8 placed at a distance from the light source 5 close to the focal length. Thanks to such a characteristic, each point of the inspected area receives a light beam whose extent of the angular spectrum a depends on the surface area of the emitting surface 7.

Advantageously, the converging lens 8 is a convex lens, or a bi-convex lens or a Fresnel lens.

[0073] Advantageously, the converging lens 8 is made in such a way that the material does not modify the state of polarization of the light. The converging lens 8 is advantageously made of glass, for example optical glass.

The analysis device 1 comprises an image acquisition device 6 for detecting the specific polarization property of at least one specific beam. The combination of specific beams received from the global beam allows the production of at least two images.

[0075]Also, in view of what proceeds, we can see that the analysis device 1 offers the possibility of producing a plurality of primary images IPI, ..., IPn, illuminated with different polarizations. Such characteristics thus make possible the simultaneous formation of different secondary images IS1, ..., ISn under different contrasts, as can be seen in Figure 14.

The analysis device 1 comprises a processing unit 9, advantageously intended to analyze the presence of reliefs.

[0077] Advantageously, such an analysis is carried out using primary images IPI, IPn from the acquisition device 6.

[0078]Advantageously, such an analysis consists in particular of constructing secondary images IS1, ..., ISn by a calculation whose methods are specified later in this description. Such secondary images IS1, ISn are then advantageously exploited to enable the detection of reliefs.

[0079] The processing unit 9 allows in particular, from primary images IPI, ..., IPn, which are acquired by the acquisition device 6, the construction of secondary images. The exploitation of such secondary images allows in certain embodiments the detection of reliefs, to detect relief defects, or to read inscriptions on the wall 3, 4 of the container 2.

We now describe the light source 5 more particularly, referring to Figures 1 and 6 to 8.

[0081]Advantageously, the emitting surface 7 is composed of p contiguous light zones Zl, ..., Zp, that is to say nested within each other.

[0082] In the embodiments shown in Figures 6 and 7, the light zones Zl, ..., Zp have a circular shape, and are concentric. Such a shape is most frequently encountered, because it gives the beams conical envelope angular spectra with a uniform effect on the contrasts in the inspection sector. .

[0083] In other implementations, for example that shown in Figure 8, the luminous zones are rectangles having their center of gravity all combined, and their sides respectively parallel to each other. In this way, it is possible to adapt to container shapes 2 having a section having a shape other than circular, or having an irregular distribution of the glass thickness, particularly on the bottom of the container.

[0084] Advantageously, the emitting surface 7 is equipped with source polarizing filters 10, the number of which corresponds to p-1. In this way one of the light zones Zl, ..., Zp is devoid of polarizing filter of source 10, and emits unpolarized light, which is considered in the rest of the text as light having a “zero” polarization property.

[0085] Advantageously, the source polarizing filters 10 are linear polarization filters, in other words the filtered light waves have as polarization property the direction of the electric field, called the direction of linear polarization.

[0086] In other embodiments not shown, the source polarizing filters 10 are circular polarization filters, in other words the filtered light waves have as polarization property the direction of rotation of the electric field called circular polarization direction.

[0087] Advantageously, the emitting surface 7 includes a transparent or translucent diffuser, backlit uniformly by a primary light. Such a diffuser is not visible in the figures. It is located upstream of the source polarizing filters 10 in the direction of travel of the light.

[0088] In another embodiment, in the direction of travel of the light, are arranged on said diffuser, in place of the source polarizing filters 10, a single linear polarization filter then a screen of liquid crystal cells ( not visible in the figures) called LC (liquid crystal) screen.

[0089]Advantageously such an LC screen is controlled or controlled to present, according to p-1 peripheral zones, different birefringence properties in order to produce, for at least p-1 light zones Zl, ..., Zp, a light with a polarization property different from each other, in particular a different phase shift between the two orthogonal components of the electric field of the light waves composing the overall beam.

[0090] The advantage of such a characteristic is that the shape and size of the light zones Zl, ..., Zp can be controlled electrically to be easily adapted to the container models 2 when adjusting the analysis device 1 .

[0091] We now describe the acquisition device 6 more specifically with reference for example to Figures 2 to 5.

[0092] Advantageously, the acquisition device 6 comprises a photoelectric sensor 12. Such a photoelectric sensor 12 comprises a plurality of photoelectric cells 13, that is to say two-dimensional elements making it possible to transform a light intensity into a signal electric. Such photoelectric cells 13 are organized in line for a linear sensor, and in a two-dimensional matrix in the case of a matrix sensor.

[0093] In the embodiments shown, the acquisition device 6 is capable of detecting a polarization property of the specific light beams, that is to say of selecting the specific light beams according to their property or characteristic of polarization: linear or circular polarization, angle of orientation of the polarization of such a specific beam in the case of a specific linearly polarized beam.

[0094]Advantageously, the acquisition device 6 comprises at least one analysis polarization filter, called acquisition polarizing filter 14, advantageously placed in front of the photoelectric sensor 12. In this way, it is possible to adapt the manner of which the photoelectric sensor 12 is illuminated.

[0095] Generally speaking, such a photoelectric sensor 12 is capable of delivering digital images of a field of vision determined by an objective 15 of the acquisition device 6, for example a CMOS (Complementary Metal- Oxide-Semiconductor) or CCD (Charge-Coupled Device).

[0096] In the embodiment shown, the acquisition device 6 comprises a polarimetric camera 11. Advantageously and as shown in Figure 2, the polarimetric camera 11 comprises a photoelectric sensor 12, equipped with individual polarization filters, or individual analyzers 14 in front of each photoelectric cell 13, thus forming a polarimetric sensor 18.

[0097] In Figures 2 and 3, which represent an example of such a polarimetric sensor 18, we can see four different individual polarizing acquisition filters 14, also called individual analyzers 14, arranged in front of photoelectric cells 13 d a group 19 of four connected photoelectric cells 13.

[0098] Such a group 19 comprises two photoelectric cells 13 in rows, two photoelectric cells 13 in columns, the group 19 forming a square of photoelectric cells 13. The individual analyzers 14 are advantageously linear polarization filters, and each have a different polarization direction.

[0099] Advantageously, in the polarimetric sensor 19, a photoelectric cell 12 is equivalent to a pixel of an image. Due to the very small size of each photoelectric cell 12, it is possible to simultaneously acquire four light beams each having different polarization properties.

[0100] The use of such a polarimetric sensor 18 has in particular the consequences that the primary images IPI, ..., IPn thus acquired have a definition, or quality lower than that acquired by means of a photoelectric sensor 12 of same size. However, such degradation is sufficiently low not to be problematic in the context of an application for the relief analysis of a container 2.

[0101] More generally, the number of different individual analyzers 14 makes it possible to determine the number of different specific beams which can be acquired by the acquisition unit 6.

[0102]Also, from a single image acquisition of a wall 3, 4 of a container 2, four primary images IPI, ..., IPn can be obtained, each of them being constructed in based on the signal coming from the photoelectric cells 13 covered by the same individual analyzer 14, that is to say coming from a set of pixels equipped with an individual analyzer 14 of the same direction, polarization orientation.

[0103] Together, the acquisition polarization filters 14 or individual analyzers 14, of a polarimetric camera 11, having the same orientation play, for each primary image, exactly the same role as the acquisition polarization filters 14, used during an acquisition with several sensors.

[0104] The use of a polarimetric camera 11 has the advantage of avoiding the use of several conventional cameras provided with separate acquisition filters 14, which makes it possible to limit the size, cost and complexity of implementation of the analysis device 1. [0105]Advantageously, in the case of a source polarizing filter 10 having circular polarization in a given direction over a given zone Zl, Zp, a quarter-wave delay plate (not shown), presenting a rapid axis at 45° from the direction of linear polarization of one of the individual analyzers 14, which allows the extinction of said light zone Zl, ..., Zp given for one of the primary images IPI,. .., IPn acquired by the polarimetric sensor 19.

[0106] Advantageously, such a polarimetric sensor 18 is equipped with individual linear analyzers 14 oriented in complementary directions, for example 0°, 45°, 90° and 135°.

[0107] In the example of Figure 2, the photoelectric sensor 11 additionally comprises an array of micro-lenses 16, each of which is associated with a photoelectric cell 13. Such micro-lenses 16 make it possible in particular to compensate for the fact that the efficiency depends on the photons constituting the beam reaching a sensitive part of each pixel, and the non-sensitive areas between the pixels.

[0108] In another embodiment shown in Figure 4, the acquisition device 6 comprises a plurality of photoelectric sensors 12, each arranged on cameras 20, black and white or color, but not polarimetric. Each of the cameras 20 is provided with a polarizing acquisition filter 14 having a different polarization property, common to all of its image sensor, placed in front, inside or behind the lens.

[0109] In such embodiments, the cameras 20 cannot be placed strictly at the same location, it is necessary to apply a processing algorithm known as a repositioning algorithm to allow the superposition between them of the acquired primary images IPI, ..., IPn by the cameras of the same inspection sector. Such an algorithm is advantageously applied before processing with a view to analyzing the reliefs of the container 2. [0110] In other embodiments not shown, the acquisition device 6 comprises a camera 20 with a single photoelectric sensor 12, in front of which is placed a movable polarizing acquisition filter 14.

[0111]An example of such a movable analyzer filter 14 is a rotating filter placed on a rotating barrel mounted on a rotating barrel, placed in front of the lens of the camera 20. Each of the primary images IPI, ..., IPn is then acquired by moving the filter between each shot. Such an embodiment finds an advantageous application in the laboratory, but does not find application in high-speed industrial relief detection, that is to say for large series.

[0112] In other embodiments shown for example in Figure 5, a camera 20 is used having a plurality of photoelectric sensors 12, each being provided with an acquisition polarization filter 14, but only one objective. Such a black and white camera 20 is currently more expensive than a polarimetric camera 11.

[0113] In an exemplary embodiment not shown, use is made of so-called tri-CCD cameras, each sensor being assigned a polarizing acquisition filter 14. Although allowing the acquisition of several primary IPI images, .. ., IPn with different polarizations, such a tri-CCD camera is however currently more expensive than polarimetric cameras making their use less frequent.

[0114] More generally, the acquisition device 6 comprises polarizing acquisition filters 14, and if necessary a quarter-wave delay plate, in the case of circularly polarized light. The acquisition polarizers 14 have polarization properties which are determined as a function of the light source 5, so as to constitute lighting configurations. Such lighting configurations are then detected and transformed into primary image IPI, ..., IPn.

[0115] Advantageously, the acquisition device 6 makes it possible to acquire n primary images IPI, ..., IPn. Among these images n primary images IPI, ..., IPn, n-1 primary images IPI, ..., IPn are acquired by excluding the light having the polarization property of one of the associated light zones Zl, Zp. In this way, there are at least n-1 primary images formed when at least one light zone Zl, ..., Zp is extinguished.

[0116] Advantageously, for the case where a source polarizing filter 10 is a linear filter, such exclusion of a light zone Zl, ..., Zp is carried out by providing a polarizing acquisition filter 14 having a direction of polarization orthogonal to the polarization direction of the source polarizing filter 10.

[0117] Advantageously, for the case where a source polarizing filter 10 is a circularly polarizing filter, such exclusion of a luminous zone Zl, ..., Zp is for example carried out by providing after the container 2 a quarter delay blade wave and at least one polarizing acquisition filter 14 with a linear polarization direction. Such a configuration is not shown in the figures.

[0118] Furthermore, the acquisition device is configured so that each of the light zones Zl, ..., Zp emits a luminous flux, all or part of said luminous flux contributing to the formation of at least one of the n images primary IPI, ..., IPn.

[0119] More specifically, the acquisition device has for example polarizing acquisition filters 14 chosen and/or oriented as a function of the polarizing filters of the source 10, so that for each of the primary images acquired, at least one of the zones lights Zl, ..., Zp is lit at least once. In other words, there is no primary image formed in a lighting configuration for which all of the light zones Zl, ..., Zp are extinguished.

[0120] Different examples of implementation of the source and the acquisition device 6 are described below for obtaining the primary images IPI, ..., IPn using the polarimetric camera 11.

[0121] In the analysis below, the fact that a small part of the light intensity is absorbed by the polarizing filter is not taken into account, including for the case where the filter polarizing the emitted light and the polarizing acquisition filter 14 have the same direction of polarization.

[0122] In the analysis below, it is also not taken into account that there cannot be total extinction of the light intensity even in the case where the source polarizing filter 10 and the polarizing filter d acquisition 14 are orthogonal to each other.

[0123] In the embodiment shown in Figure 6, the light source 5, via the emitting surface 7, has a first central light zone Zl and a second light zone Z2, peripheral to the first light zone Zl.

[0124] According to such an embodiment, the central zone Zl is devoid of a source polarizing filter 10, the peripheral zone Z2 is provided with a linear source polarizing filter 10 having an orientation at 45°. The acquisition device 6 is provided with a linear polarizing acquisition filter 14 having an orientation at -45°.

[0125] In such a lighting configuration, a first primary image IPI can be obtained, formed by the photoelectric cells 13 devoid of polarizing acquisition filter 14, and a second primary image IP2, formed by the coated photoelectric cells 13. a polarizing acquisition filter at -45°.

[0126] In such a lighting and observation configuration, the first primary image IPI is formed thanks to the light intensity of all the light zones Zl, Z2. The primary image IP2 is formed via the light intensity coming from the central light zone Zl, the light zone Z2 then being considered to be off, since the source polarizing filter 10 and the acquisition polarizing filter 14 have polarization directions orthogonal to each other, respectively 45° and - 45°.

[0127] In such a lighting configuration, shown in Figure 6, it is possible for example to use a polarimetric camera, so as to obtain a first primary image IPI, formed by the photoelectric cells 13 each provided with a filter individual acquisition polarizer 14 to 45°, thanks to the light intensity of all the light zones Zl, Z2. A second primary image IP2 is obtained, formed by the photoelectric cells 13 coated with a polarizing acquisition filter at -45°, such a primary image IP2 being formed via the light intensity coming from the zone central light Zl, the light zone Z2 then being considered to be off.

Referring for example to the embodiment shown in Figure 7, the light source 5, via the emitting surface 7, has a first light zone Z1, a second light zone Z2, a third light zone Z3, and a fourth light zone Z4. The first light zone Z1 is central, while the second light zone Z2, the third light zone Z3, and the fourth light zone Z4 are peripheral light zones. In such an example, the peripheral zones are circular and concentric.

[0129] According to such an embodiment, the first light zone Z1 does not have a polarizing source filter 10, the second light zone Z2 is provided with a linear source polarizing filter 10 having an orientation at 0°, the third zone Z3 light zone is equipped with a 10 to 45° source polarizing filter, the fourth Z4 light zone is equipped with a 10 to 90° source polarizing filter. The acquisition device 6 is provided with a polarizing acquisition filter 14, linear having an orientation at 45°, with a polarizing acquisition filter 14 linear having an orientation at 90°, with a polarizing filter of linear acquisition 14 having an orientation at -45°, of a polarizing acquisition filter 14 linear at 0°.

[0130] In such a lighting configuration, the first primary image IPI is formed thanks to the light intensity passing through a linear polarizing acquisition filter 14 having an orientation at 45°, therefore resulting from half the light intensity of the first light zone Zl, half the intensity of the second light zone Z2, all of the intensity of the third light zone Z3, and half the intensity of the fourth light zone Z4 in accordance with Malus' law, no zone being considered extinguished during the formation of the first primary image.

[0131] In such a lighting configuration, the second primary image IP2 is formed thanks to the light intensity passing through a linear acquisition polarizing filter 14 having an orientation at 90°, therefore results from half the intensity of the first light zone Zl, half the light intensity of the third light zone Z3, and the entire intensity of the fourth light zone Z4, the light zone Z2 being considered extinguished for the formation of this second primary image IP2.

[0132] In such a lighting configuration, the third primary image IP3 is formed thanks to the light intensity passing through a linear polarizing acquisition filter 14 having an orientation at -45°, therefore resulting from half of the intensity of the first light zone Zl, half the intensity of the second light zone Z2, and half the intensity of the fourth light zone Z4, the light zone Z3 being considered off for the formation of this third image primary IP3.

[0133] In such a lighting configuration, the fourth primary image IP4 is formed thanks to the light intensity passing through a linear polarizing acquisition filter 14 having an orientation at 0°, therefore results from half the intensity of the first light zone Zl, of the entire intensity of the second light zone Z2, and half of the intensity of the third light zone Z3, the light zone Z4 being considered extinguished for the formation of this fourth image primary IP4.

[0134] We now describe the processing unit 9 and its operation.

[0135] The processing unit 9 has the particular function of processing the primary images IPI, ..., IPn recovered in a single acquisition, for example using the acquisition device 6, so as to be able to analyze the reliefs on a wall 3, 4 of the container 2.

[0136] According to certain embodiments, the primary images IPI, ..., IPn are analyzed by the processing unit 9, directly giving a result of processing, that is to say without it being necessary to calculate combinations of the primary images IPI, ..., IPn.

[0137] According to other embodiments, the primary images IPI, ..., IPn are combined so as to obtain secondary images IS1, ..., ISn. Such secondary images IS1, ..., ISn are images equivalent to images which would be acquired, using an analysis device using a diaphragm or opaque mask according to the prior art, one after the other. others with different aperture or mask aperture values for each acquisition.

[0138] More precisely, in theory according to the prior art for a similar result, it would be necessary to successively acquire n images of each container. Furthermore, the result would not be completely equivalent. Indeed, the articles move, it would be necessary to adopt either a very expensive solution with several acquisition stations, or to use a single station and take several images hoping that the inspection sector remains in the field of observation and lighting of the system. The implementation of one in addition to the other of the arrangements described above would impose not only a slowing down of the rate, but an imperfect step of matching the acquired images.

[0139] In reality, according to the prior art, only a single image is obtained with a single acquisition and therefore a single contrast adjustment. Therefore it cannot be possible to benefit from the advantages of the analysis device 1 subject of the present text, which makes it possible to obtain by a single simultaneous acquisition n secondary images IS1, ..., ISn, of different contrasts for the reliefs.

[0140] Indeed, as explained above, in the devices of the prior art, a single image is acquired with a given diaphragm opening, therefore a given angular spectrum extent. In the present analysis device 1, several images can be acquired simultaneously with the equivalent of several different apertures, therefore several extents of angular spectrum a, as can be observed in the example shown in Figure 16.

[0141] Advantageously, the processing unit 9 makes it possible to calculate by linear combinations the values of the pixels of several primary images IPI, .., IPn of the secondary images IS1, ..., ISn. [0142] Such linear combinations depend, by Malus' law, on the relative orientations of the linear polarization directions of the source polarizing filters 10 and the acquisition polarizing filters, or individual analyzers 14.

[0143] Other linear combinations are used when the source polarizing filters 10 are circular polarizers and the acquisition polarizing filters 14 are preceded by quarter-wave delay plates.

[0144] In the embodiment shown in Figure 6, the linear combinations are very simple. In fact, the secondary image IS1 is equal to the first primary image IPI which is formed, through a polarizing filter of acquisition of direction -45°, thanks to the light intensity of all the light zones Zl, Z2. Also, the secondary image IS2 is equal to the second primary image IP2 which is formed, through a polarizing filter for acquisition of direction +45°, via the light intensity coming from the central luminous zone Zl with only light from Zl. The secondary image IS1 is therefore the equivalent of an image acquisition with a source of relatively large extent Z1+Z2, while the secondary image IS2 is therefore the equivalent of an image acquisition with a source of relatively large extent Zl. small.

[0145] Advantageously, the processing unit 9 produces a first secondary image IS1 presenting pixels having a value corresponding to the image generated solely by the light coming from Ml contiguous central light zones Zc taken from the p light zones Zl,. .., Zp, with Ml between 1 and p-1.

[0146] Advantageously, the processing unit 9 produces at least a second secondary image IS2 presenting pixels having a value corresponding to the image generated by the light coming from M2 contiguous central light zones Zc, with the M2 central light zones taken among the p light zones Zl, ..., Zp, M2 being between Ml+1 and p, the angular spectrum of the beams contributing to the second secondary image being broadened relative to the restricted angular spectrum of the beams contributing to the first image secondary, due to the fact that the extent of the M2 contiguous central light zones combined is greater than that of the Ml contiguous central light zones combined.

[0147] Advantageously, the processing unit 9 determines the presence of a relief, and/or the morphology of a relief carried by the inspection sector, for example of a side wall 3 or a wall of bottom 4 of container 2.

[0148] Thanks to such a calculation, in the case where a relief has inclined edges, such reliefs appear in the secondary images IS1, ..., ISn with a different gray level compared to the regions extending horizontally in the inspected area.

[0149] The reliefs are thus differently contrasted in the different secondary images IS1, ..., ISn, depending on the relationship between the angular spectrum a of the beam consisting of the specific beams corresponding to each secondary image, and the slope of the edges of the relief. In this way, a diaphragm analysis system is reproduced. It is thus possible to detect and/or analyze reliefs without it being necessary to use a diaphragm.

[0150] We now describe a method for analyzing relief in through light of a glass container 2, with reference to Figures 9 to 14.

[0151]As can be observed in Figures 9 and 10, the method comprises computer steps 300, in particular computer calculation steps which are controlled, controlled, implemented for example by the processing unit 9. The processing unit 9 is for example made up of a computer system not shown in the figures, such as a computer, or a terminal.

[0152] Generally speaking, such a computer system comprises for example one or more microprocessors, one or more electronic memory units and one or more display interfaces (screen, projector, holographic display, etc.), input interfaces (keyboard , mouse, touchpad, touch screen, etc.), and/or communication (USB, Ethernet®, Wi-Fi®, Bluetooth®, Zigbee®, etc.).

[0153] Advantageously, the computer system is connected to a computer network sharing data with one or more other computers of the network, or with other networks, for example by an Internet or Ethernet® protocol.

[0154] The through-light analysis method comprises a step of illumination 100 in transmission of the inspection sector, the light source emitting a global beam comprising several specific beams, the specific beams each presenting a different angular spectrum a, and a different polarization property.

[0155] During an acquisition step 200, n primary images are acquired, with n > 2, by means for each primary image of a combination of different specific beams received from the overall beam by the acquisition device 6 and selected based on their polarization property.

[0156] During a processing step 300, the n primary images IPI, ...IPn are processed to analyze a relief of a wall 3, 4 of a container 2.

[0157] Advantageously, during the illumination step 100, the light passed through by the light source 5 crosses the inspected sector of a wall 3, 4 of the container 2 via a light source 5 comprising in the direction of travel of the light, an emitting surface 7, a converging lens 8 for optical projection of the emitting surface, so that each point of the inspected sector receives a light beam having an angular spectrum a.

[0158] Advantageously, the emitting surface 7 comprises p peripheral light zones Zl, ..., Zp, with p > 2, the light zones Zl, ..., Zp each emit a specific beam, each specific beam having a property of polarization specific and distinct from each other.

[0159] Advantageously, during the acquisition step 200, by means of at least one acquisition device 6 of n primary images IPI, ..., IPn, with n greater than or equal to two, at least each of the n-1 primary images IPI, ..., IPn is constituted in such a way that all or part of the light zones Zl, ..., Zp contributes to the formation of at least one of the n primary images IPI, .. ., IPn, while excluding light having the polarization property of one of the associated zones. Such exclusion is obtained by extinction of one of the zones of light Zl, Zp in relation to the polarization property of the emitted light and the nature of a corresponding analyzer filter.

[0160]Advantageously, during the illumination step 100, the different polarization property for the p light zones Zl, ..., Zp is a property of zero polarization. Such a property of zero polarization is for example obtained by unpolarized light, or linear polarization according to a polarization direction specific to the light zone Zl, ..., Zp, or a phase shift between the two orthogonal components of the electric field of the light waves composing the overall beam. In the case of a phase shift to produce circular polarization, the phase shift is advantageously plus or minus a quarter of a wavelength.

[0161] Advantageously, during the illumination step 100, at least one light zone Zl, ..., Zp emits linearly polarized light in a polarization direction, at least one primary image is acquired through it as a filter acquisition polarizer 14 a polarization analyzer orthogonal to this direction of polarization which extinguishes or attenuates the light of this luminous zone Zl, ..., Zp.

[0162]Advantageously, during the illumination step 100, the number of light zones Zl, ..., Zp is between two and four, the light zones Zl, ..., Zp emitting light linearly polarized according to different directions.

[0163] Advantageously, during the acquisition step 200, at least n primary images IPI, ..., IPn are acquired, and at least n-1 of these primary images IPI, ..., IPn are obtained by filtering the light through n-1 linear analyzers each having their polarization analysis direction orthogonal to the polarization orientation of each of the respective n-1 polarized light zones, by means of at least two cameras or one polarimetric camera 11.

[0164] Advantageously, the light zones Zl, ..., Zp emit linearly polarized light in different directions chosen from the following orientations relative to a normal to the direction of propagation of the light 0°, 45°, 90° , and 135°. [0165] The analysis method comprises a processing step 300, implemented for example by the processing unit 9.

[0166] According to the first implementations shown in Figures 11 and 12, such a processing step 300 allows the construction of secondary images IS1, ..., ISn and includes for example a sub-step 310 of calculation by linear combinations of the pixel values of n primary images IPI, ..., IPn of a first secondary image IS1, ..., ISn presenting a pixel value corresponding to the image generated solely by the light coming from Ml contiguous central light zones, with Ml between 1 and n-1, and at least one second secondary image IS1, ..., ISn whose pixel value corresponds to the image generated by the light coming from M2 contiguous central light zones, with M2 between M+l and n, the extent of the M2 contiguous central light zones combined is greater than that of the Ml contiguous central zones combined.

[0167] Advantageously, the processing step 300 comprises a determination sub-step 320 integrating the analysis of at least two secondary images IS1,..., ISn, to determine the presence and/or morphology of a relief carried by the inspection sector, the relief having inclined edges and horizontal regions, the inclined edges of the relief appearing in the secondary images IS1, ..., ISn differently contrasted compared to the horizontal regions of the inspected sector.

[0168] We now describe more specifically alternative embodiments of the determination step 320. Figure 10 summarizes the different alternative embodiments.

[0169] According to a first alternative embodiment, shown in Figure 11, during the processing step 300, the processing unit 9 processes and analyzes the secondary images IS1, ..., ISn separately from each other .

[0170]In such a first variant, each secondary image IS1, ..., ISn is analyzed, for example during an examination phase 321 to enable an analysis result Ri to be obtained, namely fault detection or reading a relief, and a confidence index, forming for example a list of reading results RI .. Rn.

[0171] The examination phase 321 comprises for example for each secondary image IS1, ..., ISn: a segmentation step to obtain regions of interest, a step of determining the morphological and geometric characteristics of the regions of interest , then on the basis of the determined characteristics, a step of classification of the regions of interest to give a result Ri and a confidence index in the result.

[0172] In such a first variant, during a choice phase 322, a final analysis result R is selected from the list of analysis results Ri, as being the analysis result Ri with the index highest confidence. The result R is therefore the most certain. A unique final reading or recognition result R is thus determined.

[0173] According to a second variant not shown, the secondary images IS1, ..., ISn are analyzed together.

[0174] More particularly, in such a second variant, the determination sub-step 320 comprises a segmentation phase 321' of secondary images IS1, ..., ISn, which consists for example of dividing the image into regions or segments, that is to say to assign to the pixels of a secondary image IS1, ..., ISn a membership in a region of interest or object of interest. Such a segmentation phase 321' makes it possible to determine and obtain one or more objects of interest in a secondary image IS1,..., ISn, for example by filtering, thresholding, or contour tracking operations.

[0175]Advantageously, during an association phase 322', each of the secondary images IS1, ..., ISn of the same container 2 are associated, or matched with each other. Thus, the objects of interest of the different secondary images IS1, ..., ISn correspond to the same relief to be analyzed, but with different contrasts

[0176] Advantageously, the objects of interest are associated during the association phase 322' if they are at least partially or completely superimposed. Thus, in the case of using a polarimetric camera 11, the objects of interest are completely superimposed thanks to such an acquisition means. It is thus possible to obtain several analysis results of a relief of the container 2 represented by an object of interest in each secondary image, IS1, ..., ISn with a different contrast.

[0177]Advantageously, during an analysis phase 323', morphological and geometric characteristics of the associated regions are taken into account in order to obtain at least one sign reading or defect recognition result corresponding to each relief of the container 2 represented by an object of interest in each secondary image.

[0178] According to a third alternative embodiment shown in Figure 12, the secondary images IS1, ..., ISn are segmented together. For example after their association or matching according to criteria of superposition and belonging to the same relief of the container 2, the segmentation is done by constructing a common region of interest which can be recorded in the form of a tertiary image (not shown in the figures). Such an operation is carried out during a phase of constructing a tertiary image (321").

[0179] The construction of such an object of common interest takes advantage of the fact that the secondary images IS1, ..., ISn have a different contrast. In fact, portions of faint relief are not distinguishable in a secondary image IS1, ..., ISn with low contrast, but distinguishable in a secondary image IS1, ..., ISn with high contrast.

[0180] For example in Figure 14, the top bar of the number 3 in image IS3 is incomplete while it is complete in image IS1. On the other hand, it would be possible in IS1 to have parasitic shadows attached to the code, which are not part of the sign to be deciphered. We can therefore use rules for combining the pixel values taken in IS1 and IS3 in order, during the segmentation phase, to correctly construct a sign corresponding to a complete mold number code in the object of common interest.

[0181] Other variants than those described above are possible for carrying out the determination substep 320. [0182] In variants not shown n secondary images IS1, ISn, are analyzed together during the analysis, comprising a step of matching, or superposition of the secondary images IS1, ..., ISn, to constitute an image multidimensional, which is a form of tertiary image, then a step of segmenting the multidimensional image, then classifying the objects of interest of the multidimensional image, uned

[0183] In variants not shown, photometric and/or geometric characteristics of the segmented objects, common or specific to each secondary image, are measured after segmentation and before the classification steps. In such variants, dassifiers having these measured characteristics as input data are used to classify segments and obtain as a result the recognition of a fault or the reading of a code or sign. Such variants allow the use of any classifier including programmed decision tree type dassifiers, or supervised learning dassifiers. As supervised dassifiers, dassifiers based on measured characteristics, or image dassifiers are used, for example.

[0184] In certain embodiments, for example, a supervised learning image classifier capable of recognizing objects in images, or image portions is used as a classifier, as does for example a neural network. convolutional (CNN).

[0185] By “image classifier”, we mean a classifier which, having as input data one or more images, gives an analysis result, namely a fault detection or the reading of a relief. Such an image classifier is trained directly from a training set consisting of images and expected results.

[0186] According to a variant of these embodiments, the image classifier is trained on a training set consisting of groups of n primary images IPI, ..., IPn obtained during an acquisition step 200 on containers, each group being associated with an expected sign reading or defect recognition result for these groups of images. [0187] According to a second implementation, the processing step 300 is carried out directly by the use of an image dassiifier.

[0188] In a variant of the second implementation, visible in Figure 13, following the path in solid lines, the processing step 300, for the analysis of the primary images IPI, ..., IPn comprises by example the following sub-steps:

- a presentation sub-step 330 consisting in particular of simultaneously presenting n primary images IPI, ..., IPn as input to an image dassiifier trained to read or recognize defects in primary images,

- a classification sub-step 340 by the previously trained image dassiifier.

[0189]Advantageously, the image dassifier is trained on a training set consisting of groups of n secondary images IS1, ..., ISn obtained from an acquisition step 200 on primary image containers IPI, ..., IPn followed by a sub-step 310 of calculating the n secondary images IS1, ..., ISn, each group of n secondary images IS1, ..., ISn being associated with an expected sign reading result or fault recognition for these groups of images.

[0190] In another variant of the second implementation, visible in Figure 13, following the path in dotted lines, the processing step 300, for the analysis of the primary images IPI, ..., IPn comprises the following sub-steps:

- a sub-step 310 for calculating the n secondary images IS1, ..., ISn, to present them simultaneously as input to an image dassiifier trained to read or recognize defects in secondary images,

- a classification sub-step 340 by the previously trained image dassiifier.

[0191]Thus in these latter embodiments, the processing step 300 does not necessarily implement secondary images IS1, ..., ISn, that is to say, the primary images IPI, ... , IPn are possibly analyzed directly. [0192] More particularly, such a supervised learning image classifier, for example a CNN, is capable of classifying an image region, for example a rectangular portion of an image, directly without segmentation and without programmed measurements of photometric characteristics. or geometric.

[0193] Advantageously, only portions of predefined images, that is to say portions almost certainly containing a sign or code to be decoded, are processed. In such a case, the image classifier by supervised learning is trained with as training set sets of associated primary images or portions of associated primary images, corresponding to the same relief of a container 2, said sets being associated with an expected result.

[0194] The analysis device 1, and the analysis method presented above offer a large number of advantages, in particular:

- possibility of instantly having, from a single image acquisition, the equivalent of several images with contrast levels of different reliefs in said images

- a high code reading rate for mass production including containers with disparate relief renderings depending on production conditions

- ease of implementation during a container manufacturing process 2, without impacting the overall efficiency of the container manufacturing line, adaptation to any shape of glass container

- a moderate cost due to the possibility of using a single polarimetric camera 11.

Claims

Claims 1. Device (1) for analyzing through light a relief of an inspection sector of a wall (3, 4) of a glass container (2), the device (1) comprising: - a light source (5) for illuminating the entire inspection sector in transmission, the light source (5) being capable of emitting a global beam comprising several specific beams, the specific beams each having a different angular spectrum range ( a), and a different polarization property, - an image acquisition device (6) capable of detecting the specific polarization property of at least one specific beam, in order to acquire n primary images (IPI, ..., IPn), with n > 2, each primary image (IPI, ..., IPn) being produced by means of a combination of different specific beams received from the global beam by the acquisition device and selected according to their polarization property, - a unit for processing n primary images (IPI, ..., IPn) to analyze a relief of a wall (3, 4) of a glass container (2). 2. Device (1) according to the preceding claim, characterized in that - the light source (5) comprises: o an emitting surface (7) composed of p light zones (Zl, ..., Zp), with p > 2, comprising a central zone, and at least one peripheral zone, the light zones (Zl, ..., Zp) each emitting a specific beam, each beam having its own polarization property distinct from the other beams, o an optical projection lens (8) of the emitting surface, so that each point of the inspected sector is crossed by a light beam whose extent of the angular spectrum (a) depends on the extent of the emitting surface (7) emitting said beam, - the acquisition device (6) integrates a photoelectric image sensor capable of detecting the polarization property of specific beams coming from the light source. 3. Device (1) according to claim 1 or 2, characterized in that the polarization property is a direction of polarization or a phase shift between the two orthogonal components of the electric field vector of the light waves composing the overall beam. 4. Device (1) according to any one of the preceding claims, characterized in that the acquisition device (6) is capable of acquiring primary images (IPI, ..., IPn) so that i. during the acquisition of n primary images (IPI, ..., IPn), the acquisition device (6) constitutes n-1 primary images (IPI, ..., IPn-1) by excluding the light having the polarization property of one of the associated light zones (Zl, ..., Zp), ii. all or part of the light flux emitted by each of the light zones (Zl, ..., Zp) contributes to the formation of at least one of the n primary images (IPI, ..., IPn). 5. Device (1) according to the preceding claim, characterized in that the processing unit is capable of calculating by linear combinations the values of the pixels of several primary images (IPI, ..., IPn): a. a first secondary image presenting pixels having a value corresponding to the image generated solely by the light coming from Ml contiguous central light zones taken from the p light zones (Zl, ..., Zp), with Ml between 1 and p -1, b. at least one second secondary image presenting pixels having a value corresponding to the image generated by the light coming from M2 contiguous central light zones, with M2 taken from the p light zones (Zl, ..., Zp), M2 being included between Ml+1 and p, the angular spectrum (a) of the beams contributing to the second secondary image being broadened relative to the restricted angular spectrum (a) of the beams contributing to the first secondary image, due to the fact that the extent of the M2 zones contiguous central luminous zones combined is greater than that of the Ml contiguous central luminous zones combined, vs. Analyze the n secondary images (IS1, ISn) in order to determine the presence and/or morphology of a relief carried by a wall (3, 4) of the container (2). 6. Device (1) according to claim 2 or claim 3, characterized in that the emitting surface (7) is equipped with at least p-1 source polarizing filters (10) having a direction of linear polarization, each polarizing filter of source (10) being assigned to a peripheral zone, the p-1 source polarizing filters (10) having a polarization direction different from each other. 7. Device (1) according to the preceding claim, characterized in that the p-1 source filters (10) have a direction of polarization different from each other, chosen from the directions 0° 45° 90° and 135°. 8. Device (1) according to any one of claims 2 to 7, characterized in that the emitting surface (7) comprises a transparent or translucent diffuser, uniformly backlit by a primary light, a screen of liquid crystal cells being arranged on the diffuser, the diffuser being controlled to present, according to p peripheral zones, different birefringence properties in order to produce, for at least p-1 light zones (Zl, ..., Zp), light with a polarization property different from each other. 9. Device (1) according to any one of the preceding claims, characterized in that the acquisition device (6) comprises a polarimetric camera (11). 10. Device (1) according to any one of the preceding claims, characterized in that the lens (8) is convergent, and is arranged at a distance from the emitting surface (7) substantially equal to the focal distance. 11. Device (1) according to the preceding claim, characterized in that the lens (8) is taken from one of the following lenses: - plane convex lens, - bi-convex lens, - Fresnel lens. 12. Method for analyzing through light a relief of an inspection sector of a wall (3, 4) of a glass container (2), the method comprising: - an illumination step (100) consisting of illuminating the inspection sector in transmission, the light source (5) emitting a global beam comprising several specific beams, the specific beams each having a different angular spectrum range (a) , and a different polarization property, - an acquisition step (200) of n primary images (IPI, ..., IPn), with n > 2, using for each primary image (IPI, ..., IPn) a combination of specific beams different received from the overall beam by the detection device and selected according to their polarization property. - a step (300) of processing primary images (IPI, ..., IPn) to analyze a relief of a wall (3, 4) of a glass container (2). 13. Method according to the preceding claim characterized in that: - during the illumination step (100), where the light emitted by the source passes through the inspected sector of the wall (3, 4) of the container (2) with a light source (5) comprising in the direction path of the light, an emitting surface (7), an optical projection lens (8) of the emitting surface (7), so that each point of the inspected sector receives a light beam having an angular spectrum (a) , o the emitting surface (7) being composed of p concentric annular light zones (Zl, ..., Zp), with p > 2, the light zones (Zl, ..., Zp) each emitting a specific beam, each specific beam having its own polarization property distinct from each other. 14. Method according to claim 12 or 13, characterized in that the acquisition step (200) comprises, by means of at least one acquisition device (6), the acquisition of n primary images (IPI, . .., IPn), with n greater than or equal to two, such that o at least each of n-1 primary images (IPI, IPn-1) is constituted by excluding the light having the polarization property of one of the associated light zones (Zl, Zp), o all or part of the luminous flux of each of the light zones (Zl, ..., Zp) contributes to the formation of at least one of the n primary images (IPI, ..., IPn). 15. Method according to claim 13 or 14, characterized in that the processing step (300) comprises: o a sub-step (310) of calculating by linear combinations of the pixel values of the n primary images (IPI, .. ., IPn): ■ a first secondary image presenting a pixel value corresponding to the image generated solely by the light coming from Ml contiguous central light zones, with Ml between 1 and n-1, ■ and at least a second secondary image whose pixel value corresponds to the image generated by the light coming from M2 contiguous central light zones, with M2 between M+l andn, the extent of the M2 contiguous central light zones combined is greater than that of the Ml contiguous central zones combined, o a determination sub-step (320) integrating the analysis of at least two secondary images (IS1,..., ISn), to determine the presence and/or the morphology of a relief carried by the inspection sector, the relief having inclined edges and horizontal regions, the inclined edges of the relief appearing in the secondary images (IS1, ..., ISn) differently contrasted compared to the regions horizontals of the inspected sector. 16. Method according to the preceding claim, characterized in that during the illumination step (100), the different polarization property for the p light zones (Zl, ..., Zp) is a polarization property zero linear, or a linear polarization according to a polarization direction specific to the light zone (Zl, ..., Zp), or a phase shift between the two components orthogonal to the electric field of the light waves composing the overall beam. 17. Method according to the preceding claim, characterized in that - during the illumination step (100), at least one light zone (Zl, ..., Zp) emits linearly polarized light in a source polarization direction, and - during the acquisition step (200), at least one primary image (IPI, ..., IPn) is acquired through an acquisition filter (14) having a polarization direction orthogonal to the direction of source polarization which extinguishes or attenuates the light of the luminous zone (Zl, ..., Zp). 18. Method according to any one of claims 12 to 17, characterized in that - during the illumination step (100), the number of light zones (Zl, ..., Zp) is three or four, the light zones (Zl, ..., Zp) emitting polarized light linearly in different directions, - during the acquisition step (200), at least n primary images (IPI, ..., IPn) are acquired, and at least n-1 of these primary images (IPI, ..., IPn) are obtained by filtering the light through n-1 linear acquisition polarization filters (14) each having their polarization direction orthogonal to the polarization orientation of each of the n-1 light zones (Zl, ..., Zp ) respective polarized, by means of at least two cameras (20) or a polarimetric camera (11). 19. Method according to any one of claims 12 to 18, characterized in that the light zones (Zl, ..., Zp) emit linearly polarized light in different directions chosen from the following orientations relative to the direction of light propagation: 0°, 45°, 90°, and 135°. 20. Method according to any one of claims 15 to 19, characterized in that the determination sub-step (320) comprises: - an examination phase (321) of each secondary image (IS1, ...ISn) to obtain for each a result of analysis of a relief integrating: o a sign read or a recognized defect and o a confidence index - a choice phase (322) in which a unique final reading or recognition result is the most certain result chosen from said relief analysis results. 21. Method according to any one of claims 15 to 19, characterized in that the determination sub-step (320) comprises: - a segmentation phase (321') of each secondary image to obtain at least one object of interest for each, - an association phase (3229 objects of interest corresponding between secondary images (IS1, ..., ISn) by at least partial superposition, - an analysis phase (323') of the associated objects of interest by combining photometric or morphological characteristics of said objects of interest, in order to obtain at least one sign reading or defect recognition result corresponding to each object partner. 22. Method according to any one of claims 15 to 21, characterized in that the determination sub-step (320) comprises: - a phase of construction of a tertiary image (321") integrating an analysis of the secondary images (IS1, ..., ISn) to constitute from pixels chosen in several secondary images, in a tertiary image, an object of interest representing a reconstituted sign, - a sign reading phase (322") consisting of analyzing the reconstituted sign in order to read it. 23. Method according to one of claims 20 to 22, characterized in that the examination phase (321), the analysis phase (323') or the sign reading phase (322") is carried out by means a dassifier resulting from supervised learning, such as a Bayesian dassifier or a neural network, trained to read signs or recognize faults. 24. Method according to any one of claims 12 to 14 or 16 to 19, characterized in that in the processing step (300), the analysis of the primary images (IPI, ..., IPn) comprises the sub -following steps : - a presentation sub-step (330) consisting of simultaneously presenting n primary images (IPI, IPn) as input to an image classifier (340) trained to read or recognize defects in primary images - or a sub-step (310) consisting of calculating n secondary images (IS1, ISn) to present them as input to an image classifier (340) trained to read or recognize defects in secondary images, - a classification sub-step (340) by the image classifier (340) previously trained.