Tire inspection device and inspection method, method and device for generating identification model, and program

Abstract

To highly accurately detect the presence / absence of a defect in an inspection object tire.SOLUTION: An identification model M is stored in a storage unit 12. The identification model M is the model generated with machine learning using a data set including an image of a defect region being a region where a defect in a tire is expressed in a shearography image and tire information related to a material of the tire as teacher data. A tire inspection device 10 comprises: an image acquisition unit 11a which acquires the shearography image; a tire information acquisition unit 11b which acquires the tire information related to the material of the inspection object tire; and a defect determination unit 11v which inputs a portion of the shearography image and the tire information to the identification model M and detects a defect in the inspection object tire.SELECTED DRAWING: Figure 8

Description

Technical Field

[0001] The present disclosure relates to a tire inspection apparatus and inspection method, a method and apparatus for generating an identification model, and a program.

Background Art

[0002] The following Patent Document 1 discloses an apparatus for detecting defects inside a tire by shearography. In shearography, a laser beam is irradiated onto the tire surface, and a speckle image formed by the interference of the reflected light from the tire surface is acquired by a camera. Image subtraction processing is performed on two speckle images captured in two states with different temperatures and pressures, and the difference between the speckle images is acquired as a measurement result.

[0003] For example, in Patent Document 1, the tire is placed inside a pressure chamber. When the pressure inside the chamber is reduced, the tire surface deforms. If there are defects such as air bubbles in the rubber part of the tire, the air bubbles expand and the tire surface bulges locally. Image subtraction processing is performed on the speckle image captured before the pressure in the chamber is reduced and the speckle image captured after the pressure is reduced. Then, a measurement image (hereinafter referred to as a "shearography image") showing the change in the tire surface (local bulging) due to the pressure reduction is obtained. In this shearography image, interference fringes appear in the bulging part.

Prior Art Documents

Patent Documents

[0004]

Patent Document 1

Summary of the Invention

Problems to be Solved by the Invention

[0005] In a shared tomography image, defects inside the tire appear as stripe patterns. There is a correlation between the size of the stripe pattern (the size of the bulge appearing on the tire surface) and the actual size of the defect. However, the size of the stripe pattern also depends on the rigidity of the tire. For example, when the rigidity of the rubber part is high, even if the actual defect (e.g., air bubble) is large, the local bulge on the tire surface becomes small, and the stripe pattern appearing in the shared tomography image also becomes small. Therefore, it is difficult to perform highly accurate defect determination based only on the size of the stripe pattern.

Means for Solving the Problem

[0006] (1) In the storage means of the tire inspection device proposed in the present disclosure, an identification model is stored. The identification model is a model generated by machine learning using, as teacher data, a data set including an image of a defect region, which is a region where a defect inside the tire appears in a shared tomography image that is an image of the tire obtained by shared tomography, and tire information related to the material of the tire. The tire inspection device includes an image acquisition means for acquiring a shared tomography image of a tire to be inspected obtained by shared tomography, a tire information acquisition means for acquiring tire information related to the material of the tire to be inspected, and a defect determination means for inputting a part or all of the shared tomography image of the tire to be inspected and the tire information of the tire to be inspected into the identification model to detect a defect inside the tire to be inspected.

[0007] The material of the tire affects the rigidity of the tire. In the tire inspection device of (1), the tire information related to the material of the tire to be inspected is also input into the identification model. Therefore, it becomes possible to detect the presence or absence of a defect in the tire to be inspected with high accuracy.

[0008] (2) The tire includes a crown portion and a sidewall portion. In the tire inspection apparatus described in (1), the data set for generating the identification model may include, in addition to the image of the defect region and the tire information, part information representing the part of the tire including the defect region. The defect determination means may input a part of the shareography image of the tire to be inspected, the tire information of the tire to be inspected, and part information representing the part of the tire to be inspected including the part of the shareography image into the identification model.

[0009] Since the structures of the crown portion and the sidewall portion are different, the rigidity and the form of the defects generated in these two parts may be different. According to the tire inspection apparatus of (2), since the part information is included in the data input to the identification model, the accuracy of defect detection can be improved.

[0010] (3) The tire inspection apparatus described in (1) may further include a defect region candidate extraction means for extracting an image of a defect region candidate, which is a candidate for a region including a defect inside the tire to be inspected, from the shareography image of the tire to be inspected. The defect determination means may input the image of the defect region candidate into the identification model and determine whether a defect appears in the defect region candidate. According to this, the detection accuracy of the defect can be further improved.

[0011] (4) In the tire inspection apparatus described in (3), the defect region candidate extraction means may search for a pattern registered in advance in the shareography image of the tire to be inspected, and extract a region including the pattern found by the search as the defect region candidate.

[0012] (5) The tire inspection method proposed in this disclosure includes an image acquisition step of acquiring a shareography image, which is an image of the tire to be inspected obtained by shareography, and a tire information acquisition step of acquiring tire information related to the material of the tire to be inspected. Further, the identification model is a model generated by machine learning using, as teacher data, a dataset including an image of a defect region, which is a region where a defect inside the tire appears in the shareography image, and tire information related to the material of the tire. The tire inspection method includes a defect determination step of inputting a part or all of the shareography image of the tire to be inspected and the tire information of the tire to be inspected into the identification model to detect a defect inside the tire to be inspected.

[0013] The material of a tire affects the rigidity of the tire. (5) In the tire inspection method, tire information related to the material of the tire to be inspected is also input into the identification model. Therefore, it becomes possible to detect the presence or absence of a defect in the tire to be inspected with high accuracy.

[0014] (6) The program proposed in this disclosure causes a computer to function as an image acquisition means for acquiring a shareography image, which is an image of the tire to be inspected obtained by shareography, and a tire information acquisition means for acquiring tire information related to the material of the tire to be inspected. Further, the program also causes the computer to function as a defect determination means. The defect determination means inputs a part or all of the shareography image of the tire to be inspected and the tire information of the tire to be inspected into an identification model generated by machine learning using, as teacher data, a dataset including an image of a defect region, which is a region where a defect inside the tire appears in the shareography image, and tire information related to the material of the tire, to detect a defect inside the tire to be inspected.

[0015] (7) The method for generating an identification model proposed in this disclosure includes an image acquisition step of acquiring a shared-lography image, which is an image of a tire obtained by shared-lography, a defective region extraction step of extracting an image of a defective region, which is a region including a defect inside the tire, from the shared-lography image, a tire information acquisition step of acquiring tire information related to the material of the tire, and a model generation step of generating an identification model for detecting a defect inside a tire to be inspected using, as teacher data, a data set including the image of the defective region and the tire information. In the method of (7), tire information related to the material of the tire is also used for generating the identification model. Therefore, according to an inspection using this identification model, the presence or absence of a defect in a tire to be inspected can be detected with high accuracy.

[0016] (8) The tire includes a crown portion and a sidewall portion. The data set may include, in addition to the image of the defective region and the tire information, part information representing the part of the tire including the defective region. Since the structures of the crown portion and the sidewall portion are different, the rigidity and the form of the defect generated may be different in these two parts. In the method of (8), part information representing the part of the tire including the defective region is also used for generating the identification model. Therefore, according to an inspection using this identification model, the accuracy of defect detection can be further improved.

[0017] (9) The apparatus for generating an identification model proposed in this disclosure includes an image acquisition means for acquiring a shared-lography image of a tire obtained by shared-lography, a defective region extraction means for extracting an image of a defective region, which is a region including a defect inside the tire, from the shared-lography image acquired in the image acquisition step, a tire information acquisition means for acquiring tire information related to the material of the tire, and a model generation means for generating an identification model for detecting a defect inside a tire to be inspected using, as teacher data, a data set including the image of the defective region and the tire information.

[0018] (10) The program proposed in the present disclosure causes a computer to function as an image acquisition means for acquiring a shareography image of a tire acquired by shareography, a defect area extraction means for extracting an image of a defect area, which is an area including a defect inside the tire, from the shareography image, a tire information acquisition means for acquiring tire information related to the material of the tire, and a model generation means for generating an identification model for detecting a defect inside a tire to be inspected using a data set including the image of the defect area and the tire information as teacher data.

Brief Description of the Drawings

[0019]

Figure 1

Figure 2

Figure 3

Figure 4

Figure 5

Figure 6

Figure 7

Figure 8

Embodiments for Carrying Out the Invention

[0020] Hereinafter, a tire inspection apparatus and inspection method proposed in the present disclosure, and a method and apparatus for generating an identification model will be described.

[0021] FIG. 1 is a block diagram showing the hardware of the tire inspection apparatus 10 proposed in the present disclosure. As shown in FIG. 1, the tire inspection apparatus 10 includes a control unit 11, a display unit 13, an input unit 14, a share lithography apparatus 15, and a decompressor 17. Elements such as the control unit 11 included in the tire inspection apparatus 10 function not only as the tire inspection apparatus 10 but also as a generation apparatus for an identification model used by the tire inspection apparatus 10.

[0022] As shown in FIG. 1, the share lithography apparatus 15 includes a laser irradiation unit 15A, an interferometer 15B, and a light receiving unit 15C. FIGS. 2 and 3 are diagrams for explaining the share lithography method. FIG. 2 shows an outline of the share lithography apparatus, and FIG. 3 shows an example of an image obtained by the share lithography apparatus.

[0023] The laser irradiation unit 15A includes a laser diode and irradiates the surface S of the tire 90 with laser light. The irradiated laser light is reflected by the surface of the tire 90 and enters the interferometer 15B as shown in FIG. 2. As the interferometer 15B, for example, a Michelson interferometer can be used. In FIG. 2, reference numeral 15a is a half mirror, and reference numerals 15b and 15c are mirrors. The interferometer 15B may be a type of interferometer different from the Michelson interferometer. Reference numeral 15C is a light receiving unit. The light receiving unit 15C includes a CCD or a CMOS image sensor. The image captured by the light receiving unit 15C is supplied to the control unit 11.

[0024] The laser light reflected by the surface S of the tire 90 and incident on the interferometer 15B recombines after passing through two optical paths L1 and L2 and reaches the image plane (light receiving unit 15C). In the example of FIG. 2, the optical path L1 is the path of the light reflected by the half mirror 15a and reflected by one of the mirrors 15b, and the optical path L2 is the path of the light transmitted through the half mirror 15a and reflected by the other mirror 15c. The laser light that has passed through the two optical paths L1 and L2 interferes at the light receiving unit 15C (image plane). The surface S of the tire 90 is microscopically a rough surface, and the optical path difference generated in the interferometer 15B changes irregularly due to the unevenness of the surface S of the tire 90. Therefore, a speckle image in which a speckle pattern appears is captured by the light receiving unit 15C.

[0025] The tire 90 is disposed inside the chamber. For example, with the inside of the chamber maintained at atmospheric pressure, a speckle image of the surface S of the tire 90 is captured. Thereafter, the decompressor 17 decompresses the inside of the chamber. If there are defects such as air bubbles (solid line P1 in FIG. 2) inside the tire 90, the air bubbles P1 expand due to the decompression inside the chamber (see the broken line P1 in FIG. 2), and the surface S of the tire 90 bulges locally (see the broken line S in FIG. 2). Then, after such bulging occurs, a speckle image of this surface S is acquired. Then, a speckle pattern also appears in this speckle image, similar to before the decompression. The brightness and darkness of the speckles appearing in the speckle image after decompression are the same as those in the speckle image before decompression at positions where the change in the optical path length due to the local bulging of the surface S is an integral multiple of the wavelength of the laser light. On the other hand, at positions where the change in the optical path length is a half-integral multiple of the wavelength of the laser light, the brightness and darkness of the speckles after decompression are inverted from those before decompression.

[0026] Therefore, by calculating the difference (the difference in pixel values of each pixel) between the two speckle images, the speckle image before decompression and the speckle image after decompression, and making it into an image, as shown in FIG. 3, the bulging portion Sd of the surface S of the tire 90 appears. In this specification, an image obtained from the difference in speckle images is referred to as a "shearography image".

[0027] As shown in FIG. 4, the tire 90 has a crown portion 91 and left and right sidewall portions 93. A rubber portion 90a is formed on the crown portion 91 and the sidewall portions 93. A tread is formed on the surface of the rubber portion 90a (the surface of the crown portion 91). The tire 90 has a belt 94, a carcass 95, a finishing member 96, etc. inside the rubber portion 90a. The carcass 95 is formed from the left sidewall portion 93 to the right sidewall portion 93 of the tire 90. The belt 94 is disposed inside the crown portion 91 and covers the central portion of the carcass 95. The finishing member 96 is disposed at the edge of the sidewall portion 93 and covers the bead 97. The belt 94, the carcass 95, and the finishing member 96 are formed of metal. The belt 94 and the central portion of the carcass 95 are located in the crown portion 91. The carcass 95 and the finishing member 96 are located in the sidewall portion 93.

[0028] The tire 90 is supported by a support device (not shown) that rotates it in the circumferential direction so that the shareography device 15 can obtain the overall shareography image of the tire 90. The support device rotates the tire 90 at a predetermined speed during the generation of the shareography image (teacher data) for machine learning and during the inspection of the tire 90. The light receiving portion 15C continuously captures the speckle image of the surface of the tire 90 at a frequency corresponding to the rotation speed of the tire 90 and outputs the image data for one rotation of the tire 90. Further, the shareography device 15 may be relatively movable with respect to the tire 90 in a direction along the axis of the tire 90 so that the shareography images of the crown portion 91 and the sidewall portion 93 of the tire 90 can be obtained.

[0029] The control unit 11 (see FIG. 2) has arithmetic units such as a CPU (Central Processing Unit) and a GPU (Graphics Processing Unit). The control unit 11 also has a storage unit 12. The storage unit 12 has a RAM (Random access memory), a ROM (Read only memory), etc. The storage unit 12 may include a storage device capable of both reading and writing, such as an SSD (Solid State Drive) or an HDD (hard disk drive). The storage unit 12 stores programs executed in arithmetic units such as the CPU, image data captured by the shared lithography device 15, and identification models generated by the processes described later.

[0030] The display unit 13 is a display device such as a liquid crystal display, and displays various images according to the instructions of the control unit 11.

[0031] The input unit 14 is a user interface such as a keyboard or a mouse, receives user operation inputs, and inputs signals indicating the contents to the control unit 11.

[0032] [Processing performed by the control unit] Hereinafter, the functions of the control unit 11 will be described. FIG. 5 is a block diagram showing the functions of the control unit 11. As its functions, the control unit 11 has a shared lithography image acquisition unit 11a, a tire information acquisition unit 11b, a teacher data generation unit 11e, a model generation unit 11j, a defect area candidate extraction unit 11u, and a defect determination unit 11v. These functions are realized by the control unit 11 executing programs stored in the storage unit 12. The control unit 11 also has an identification model M. The identification model M is stored in the storage unit 12.

[0033] The identification model M is generated (learned) by the shareography image acquisition unit 11a, the tire information acquisition unit 11b, the teacher data generation unit 11e, and the model generation unit 11j. Further, an inspection of the interior of the tire 90 using the identification model M is executed by the shareography image acquisition unit 11a, the tire information acquisition unit 11b, the defect area candidate extraction unit 11u, and the defect determination unit 11v.

[0034] The control unit 11 may be composed of a plurality of personal computers. Alternatively, the control unit 11 may be composed of one or more personal computers and one or more server computers. In this case, some functions of the control unit 11 (for example, the defect area candidate extraction unit 11u and the defect determination unit 11v) may be executed by a personal computer, and other functions of the control unit 11 (for example, the model generation unit 11j) may be executed by another personal computer or a server computer.

[0035] [Shareography Image Acquisition Unit] The shareography image acquisition unit 11a acquires a shareography image (Figure 3) calculated from the speckle image captured by the light receiving unit 15C. The shareography image acquisition unit 11a acquires a shareography image for each of the crown portion 91 and the sidewall portion 93. Further, during the learning (generation) of the identification model M, the shareography image acquisition unit 11a acquires shareography images for a plurality of tires 90 having different material compositions. Thereby, an identification model M can be generated that includes the material composition of the tire, in other words, the rigidity of the tire, as an element for defect determination.

[0036] The shareography image acquisition unit 11a may acquire a shareography image generated by another image processing device and stored in the storage unit 12. Alternatively, the shareography image acquisition unit 11a may acquire the two above-described speckle images (speckle images before and after decompression in the chamber) from the light receiving unit 15C and generate a shareography image from the difference between the speckle images.

[0037] During the inspection of the tire 90, the sharedography image acquisition unit 11a acquires a sharedography image for one full rotation of the tire 90. On the other hand, during the learning (generation) of the identification model M, the sharedography image acquisition unit 11a acquires a sharedography image of the learning tire 90 (an image for generating teacher data). In this case, the sharedography image does not necessarily have to be for one full rotation of the tire 90.

[0038] [Tire information acquisition unit] The tire information acquisition unit 11b acquires tire information related to the material of the tire 90. More specifically, the tire information is information for specifying the material of the rubber portion 90a of the tire 90. The tire information may be information that directly specifies the material of the rubber portion 90a of the tire 90. For example, the tire information may be the mass ratio of natural rubber and synthetic rubber, or the mass ratio of each of one or more compounding agents (carbon, silica, oil). Further, the tire information may be information that indirectly specifies the material of the rubber portion 90a. For example, when the model number for specifying the type of the tire 90 and the material of the rubber portion 90a are in one-to-one correspondence, the model number may be used as the tire information.

[0039] [Process for generating an identification model] A process for generating (learning) the identification model M using the sharedography image will be described. The teacher data generation unit 11e and the model generation unit 11j shown in FIG. 5 execute a process for generating the identification model M. As shown in FIG. 5, the teacher data generation unit 11e has a defect area extraction unit 11f. This defect area extraction unit 11f has a candidate extraction unit 11g and a determination result reception unit 11h.

[0040] The candidate extraction unit 11g extracts an area (see Im2 in Fig. 6) in the rubber part 90a where a defect may have occurred from a shared lithography image for learning (see Im1 in Fig. 6). Defects in the rubber part 90a include, for example, air bubbles P1 in the rubber part 90a (see Fig. 2), separation (peeling of the rubber part 90a from the belt 94, etc.), and contamination of foreign matter into the rubber part 90a. Hereinafter, these defects are referred to as "rubber part defects". Also, an area where such a rubber part defect may have occurred is referred to as a "defect area candidate".

[0041] The process by the candidate extraction unit 11g is executed as follows, for example. As described with reference to Fig. 3, in the shared lithography image, a pattern Sd corresponding to the shape and size of the defect in the rubber part 90a appears. The candidate extraction unit 11g searches for each of a plurality of patterns (patterns in which rubber part defects appear) registered in advance in the storage unit 12 in the shared lithography image. For this search, for example, evaluation of similarity (matching by feature vectors) or pattern matching can be used.

[0042] As a method for evaluating similarity, the candidate extraction unit 11g calculates, for example, a feature vector (feature quantity) for a partial area (referred to as an evaluation target area) in the shared lithography image, and calculates the similarity between the feature vector and a predetermined feature vector. Here, as the feature vector, for example, the spatial frequency of the pattern can be used. The candidate extraction unit 11g may scan the entire shared lithography image for this evaluation target area and evaluate the similarity (similarity of spatial frequency) for each evaluation target area. Then, the candidate extraction unit 11g may extract an evaluation target area with a similarity greater than a threshold as a defect area candidate.

[0043] Further, as pattern matching, the candidate extraction unit 11g may search the entire shared-lithography image for a plurality of patterns (patterns in which rubber part defects appear) registered in advance in the storage unit 12. At this time, the Euclidean distance may be used as a value representing the degree of similarity to the patterns registered in the storage unit 12. And when a pattern that matches the registered pattern is found in the shared-lithography image, the circumscribed rectangle of the found pattern may be extracted as a defect region candidate.

[0044] As yet another method, the candidate extraction unit 11g may perform a two-dimensional Fourier transform over the entire shared-lithography image. And a frequency band different from the frequencies corresponding to a plurality of patterns (patterns in which rubber part defects appear) registered in advance in the storage unit 12 may be cut out from the shared-lithography image. Thereby, a portion where rubber part defects may appear appears. The circumscribed rectangle of that portion may be extracted as a defect region candidate.

[0045] The determination result reception unit 11h displays the image of the defect region candidate extracted by the candidate extraction unit 11g on the display unit 13. And the determination result reception unit 11h receives the determination result of the operator as to whether the displayed image actually represents a rubber part defect. The operator can input the determination result through the input unit 14. When a plurality of defect region candidates are stored in the storage unit 12, the determination result reception unit 11h may display the plurality of defect region candidates on the display unit 13 in order and receive the determination result of the operator for each defect region candidate.

[0046] In addition to the size of the defect area candidate, the operator determines whether the extracted defect area candidate represents a rubber part defect while considering the material composition of the learning tire 90. For example, when the material of the tire 90 contains many compounding agents that increase the hardness of the rubber part 90a, even if the size of the defect area candidate is small, the defect actually occurring in the rubber part 90a is likely to be large. Therefore, in such a case, the operator may determine that this defect area candidate represents a rubber part defect. Conversely, when the amount of the compounding agent that increases the hardness of the rubber part 90a is small, even if the size of the defect area candidate is large, the defect actually occurring in the rubber part 90a may be small. In such a case, the operator may determine that this defect area candidate does not represent a rubber part defect.

[0047] The structures of the crown part and the sidewall part are different. For example, for example, a belt 94 is located in the crown part 91, whereas the belt 94 is not located in the sidewall part 93. Therefore, the defect in which the rubber part 90a peels off from the belt 94 appears only in the crown part 91. The operator may determine whether this defect area candidate is a rubber part defect while taking into account the shape and the part of the pattern appearing in the defect area candidate. That is, the operator may determine whether the extracted defect area candidate corresponds to a rubber part defect while considering the size of the defect area candidate, the material composition of the learning tire 90, and the part where the defect area candidate is found.

[0048] Based on the input determination results, the defect area extraction unit 11f assigns identification labels to each defect area candidate. Specifically, when a determination result indicating that a rubber part defect appears in a defect area candidate is input, the defect area extraction unit 11f assigns the identification label "defect" to the defect area candidate. (Hereinafter, a defect area candidate determined to have a rubber part defect is referred to as a "defect area".) Then, the defect area extraction unit 11f uses the image data of the defect area and the tire information acquired by the tire information acquisition unit 11b as one data set, and stores this data set and the identification label (defect) in the storage unit 12 in a corresponding manner. As another example, the defect area extraction unit 11f may use the image data of the defect area, the tire information, and the part information indicating the part where the defect area candidate is found (for example, the crown part 91 or the sidewall part 93) as one data set, and store this data set and the identification label (defect) in the storage unit 12 in a corresponding manner.

[0049] On the other hand, when a determination result indicating that a rubber part defect does not appear in a defect area candidate is input, the defect area extraction unit 11f assigns the identification label "normal" to the defect area candidate. (Hereinafter, a defect area candidate determined to have no rubber part defect is referred to as a "normal area".) Also in this case, the defect area extraction unit 11f uses the image data of the defect area candidate (here, the normal area) and the tire information acquired by the tire information acquisition unit 11b as one data set, and stores this data set and the identification label (normal) in the storage unit 12 in a corresponding manner. As another example, the defect area extraction unit 11f may use the image data of the defect area candidate (here, the normal area), the tire information, and the part information indicating the part where the defect area candidate is found (for example, the crown part 91 or the sidewall part 93) as one data set, and store this data set and the identification label (normal) in the storage unit 12 in a corresponding manner.

[0050] As shown in FIG. 5, the teacher data generation unit 11e may include a normal region extraction unit 11i. For example, the normal region extraction unit 11i extracts a part of the region that was not extracted as a defect region candidate in the process of the candidate extraction unit 11g as a normal region. The normal region extraction unit 11i assigns the identification label "normal" to the image extracted as the normal region. The normal region extraction unit 11i uses the image data of the normal region and the tire information acquired by the tire information acquisition unit 11b as one data set, and stores this data set and the identification label (normal) in the storage unit 12 in association with each other. In another example, the normal region extraction unit 11i may use the image data of the normal region, the tire information, and the part information indicating the part where the defect region candidate was found (for example, the crown part 91 or the sidewall part 93) as one data set, and store this data set and the identification label (normal) in the storage unit 12 in association with each other.

[0051] The extraction process of the normal region can be performed as follows, for example. The normal region extraction unit 11i randomly extracts an image having a predetermined number of pixels and a predetermined aspect ratio from the shared lithography image. When the randomly extracted image does not have an overlapping part with the image extracted as the defect region candidate, the normal region extraction unit 11i may use this extracted image as the normal region. Here, the predetermined number of pixels and the predetermined aspect ratio may be, for example, the average of the number of pixels and the average of the aspect ratios of the defect region candidates.

[0052] The teacher data generation unit 11e converts each of the image data of the plurality of defect regions obtained by the above-described process into image data having a predetermined number of pixels and a predetermined aspect ratio. (Hereinafter, this number of pixels is referred to as the "model input pixel number", and this aspect ratio is referred to as the "model input aspect ratio".) Also, the teacher data generation unit 11e converts the image data of the plurality of normal regions obtained by the above-described process into the model input pixel number and the model input aspect ratio. That is, the teacher data generation unit 11e unifies the dimensions of the teacher data.

[0053] The model generation unit 11j inputs a data set including the image data of the defective region into the discrimination model before learning as teacher data. By doing so, the model generation unit 11j generates a discrimination model M for detecting defects in the internal structure of the tire 90. In addition to the data set including the image data of the defective region, the model generation unit 11j may input a data set including the image data of the normal region into the discrimination model before learning as teacher data.

[0054] As the discrimination model M, for example, a neural network may be used. As the discrimination model M, a convolutional neural network (CNN) may be used. Alternatively, a support vector machine (SVM), a random forest, or the like may be used as the discrimination model M.

[0055] [Processing for tire inspection] A process for performing a tire inspection using the shareography image of the tire to be inspected will be described. The defective region candidate extraction unit 11u and the defect determination unit 11v execute the process for inspection.

[0056] The defective region candidate extraction unit 11u extracts defective region candidates (Im2, see FIG. 6) from the image data of the tire 90 to be inspected, which is acquired by the shareography image acquisition unit 11a. The process of the defective region candidate extraction unit 11u may be the same as, for example, the process of the candidate extraction unit 11g of the teacher data generation unit 11e described above. That is, the defective region candidate extraction unit 11u searches for each of a plurality of patterns registered in advance in the storage unit 12 in the shareography image.

[0057] The defect determination unit 11v inputs the image data of the defect area candidates extracted by the defect area candidate extraction unit 11u into the identification model M, and detects defects inside the tire 90 to be inspected. For example, the defect determination unit 11v inputs the image data of the defect area candidates into the identification model M, and classifies this defect area candidate into one of a plurality of classes. Here, the plurality of classes are, for example, a class in which a rubber part defect appears in the defect area candidate (a class to which the identification label "defect" is assigned), or a class in which a rubber part defect does not appear in the defect area candidate (a class to which the identification label "normal" is assigned). By this process of the defect determination unit 11v, defects in the internal structure of the tire 90 to be inspected can be detected. The output of the defect determination unit 11v is, for example, the probability that the defect area candidate corresponds to each class. Alternatively, the output of the defect determination unit 11v may be the identification label of the class with the highest probability that the defect area candidate corresponds to.

[0058] The defect determination unit 11v may display the determination result (the return value from the identification model M) on the display unit 13 together with the image of the defect area candidate input to the identification model M and the part information indicating the part (crown part 91 or sidewall part 93) where the defect area candidate was found, or store them in the storage unit 12. When a plurality of defect area candidates are found for one tire 90, the defect determination unit 11v may input all of the plurality of defect area candidates into the identification model M in order, and display each of the outputs on the display unit 13 or store them in the storage unit 12. When there is a candidate classified into the "defect" class among the plurality of defect area candidates, the defect determination unit 11v may determine that the tire 90 is defective.

[0059] In contrast, the identification model M may be generated to determine even minute defects (defects that should be tolerated) as "defects". In that case, when the number of such minute defects exceeds the threshold set for each of the sites (the crown portion 91 and the sidewall portion 93) of the tire 90, it may be determined that the site is defective. For example, when the number of defective regions found in the shareography image of the crown portion 91 is greater than the threshold, the defect determination unit 11v may determine that there is a defect in the crown portion 91. Conversely, when the number of defective regions found in the shareography image of the sidewall portion 93 is greater than the threshold, the defect determination unit 11v may determine that there is a defect in the sidewall portion 93.

[0060] [Flow of the process for generating the identification model] FIG. 7 is a diagram showing an example of the process executed by the control unit 11 for generating the identification model M. The control unit 11 executes the process shown in FIG. 7 for a plurality of tires 90 having different material compositions, and generates one identification model M using the data obtained from these plurality of tires 90.

[0061] The control unit 11 acquires a shareography image and acquires tire information for specifying the material of the tire (S101). This process is the process executed by the shareography image acquisition unit 11a and the tire information acquisition unit 11b described above.

[0062] Next, the teacher data generation unit 11e generates teacher data using the shareography image and the tire information. Specifically, the candidate extraction unit 11g extracts regions (defect region candidates) where rubber part defects may occur from the shareography image acquired in S101 (S102). The determination result reception unit 11h displays the extracted defect region candidates on the display unit 13. Then, the determination result reception unit 11h receives the operator's determination result as to whether or not the displayed defect region candidates include rubber part defects, and assigns an identification label (defect or normal) corresponding to the determination result to the defect region candidates (S103).

[0063] The defect area extraction unit 11f associates the extracted image data with an identification label (normal or defective), and stores them in the storage unit 12 (S104). More specifically, for the image data (defect area) with the identification label "defect" assigned, the defect area extraction unit 11f regards this image data, the tire information obtained in S101, and the part information indicating the part where the area of this image data was found as one data set, associates this data set with the identification label (defect), and stores it in the storage unit 12. Also, for the image data (normal area) with the identification label "normal" assigned, the defect area extraction unit 11f regards this image data, the tire information obtained in S101, and the part information indicating the part where the area of this image data was found as one data set, associates this data set with the identification label (normal), and stores it in the storage unit 12.

[0064] The normal area extraction unit 11i extracts a part of the area that was not extracted (selected) as a defect area candidate in S102 from the shared lithography image as a normal area (S105). The normal area extraction unit 11i regards the image data of the normal area, the tire information, and the part information as one data set, associates this data set with the identification label (normal), and stores it in the storage unit 12 (S106).

[0065] Next, the teacher data generation unit 11e determines whether a predetermined number of data sets have been prepared for each of the plurality of classes by the processes of S104 and S106 (S107). Here, the plurality of classes are, for example, a class including a rubber part defect (a class with the identification label "defect" assigned) and a class not including a rubber part defect (a class with the identification label "normal" assigned). The predetermined number is the number recognized as necessary to generate the identification model M. If the number of data sets has not reached the predetermined number for each part (the crown part 91 and the sidewall part 93), the teacher data generation unit 11e returns to the process of S102 and executes the subsequent processes.

[0066] The teacher data generation unit 11e converts the number of pixels and the aspect ratio of each image data (defective area and normal area) stored in the storage unit 12 into the model input number of pixels and the model input aspect ratio described above (S108).

[0067] The model generation unit 11j inputs the data set stored in the storage unit 12 in S104 and S106 into the identification model M for which learning has not yet been completed, and generates (learns) an identification model M for detecting rubber part defects (S109). The above is an example of the process executed by the control unit 11 for generating the identification model M.

[0068] Note that the process performed by the teacher data generation unit 11e is not limited to the example shown in FIG. 7. For example, the control unit 11 may not extract the normal area in S105 or store the data set including the image data of the normal area in the storage unit 12 in S106. Further, the control unit 11 may store only the data set including the image data of the defective area in the storage unit 12 in S104, input only the data set including the image data of the defective area into the identification model in S109, and generate (learn) the identification model. In this case, as the identification model M, a neural network or a convolutional neural network may be used. In the tire inspection process, the probability that the data set input to the identification model M corresponds to "defect" may be output.

[0069] [Flow of the process for inspecting a tire] Next, an example of the process executed by the control unit 11 for detecting rubber part defects of a tire will be described with reference to FIG. 8.

[0070] The control unit 11 acquires a shareography image and acquires tire information for specifying the material of the tire (S201). This process is the process executed by the shareography image acquisition unit 11a and the tire information acquisition unit 11b described above.

[0071] The defect area candidate extraction unit 11u extracts an area where a rubber part defect may have occurred (defect area candidate) from the share lithography image acquired in S201 (S202). In S202, the defect area candidate extraction unit 11u extracts all defect area candidates from the share lithography image. Further, the defect area candidate extraction unit 11u converts the number of pixels and the aspect ratio of the extracted defect area candidates into the model input number of pixels and the model input aspect ratio (S203).

[0072] The defect determination unit 11v inputs the image data of the defect area candidate, the tire information acquired in S101, and the part information indicating the part where the defect area candidate was found as a single data set into the discrimination model (S204). Then, the defect determination unit 11v classifies the defect area candidates into a plurality of classes (S205). The plurality of classes are, for example, a class including a rubber part defect (a class to which the identification label "defect" is assigned), or a class not including a rubber part defect (a class to which the identification label "normal" is assigned). The defect determination unit 11v determines whether there are still defect area candidates that have not been classified (S206). If there are still defect area candidates that have not been classified, the defect determination unit 11v returns to S204 and executes subsequent processing for the unclassified defect area candidates.

[0073] In S206, when it is determined that the classification of all defect area candidates extracted from the share lithography image has been completed, the control unit 11 ends the process. That is, the control unit 11 ends the inspection of the internal structure of the tire 90.

[0074] [Summary] The storage unit 12 of the tire inspection device 10 stores an identification model M. The identification model M is a model generated by machine learning using, as teacher data, a data set including an image of a defect area in a shared-lithography image and tire information related to the material of the tire. The tire inspection device 10 includes a shared-lithography image acquisition unit 11a that acquires a shared-lithography image of a tire 90 to be inspected, a tire information acquisition unit 11b that acquires tire information related to the material of the tire 90 to be inspected, and a defect determination unit 11b that inputs image data of a defect area candidate in the shared-lithography image of the tire 90 to be inspected and the tire information of the tire 90 to be inspected into the identification model M to detect a defect in the interior (rubber portion 90a) of the tire 90 to be inspected. The material of the tire affects the rigidity of the tire. In the tire inspection device 10 described above, since the tire information related to the material of the tire 90 to be inspected is also input into the identification model M, it becomes possible to detect the presence or absence of a defect in the tire 90 to be inspected with high accuracy.

[0075] The identification model M is generated by machine learning using, as teacher data, a data set including, in addition to the image of the defect area and the tire information, part information representing the part of the tire including the defect area. The defect determination unit 11v inputs the image data of the defect area candidate extracted from the shared-lithography image of the tire 90 to be inspected, the tire information of the tire 90 to be inspected, and the part information representing the part of the tire 90 to be inspected including the defect area candidate into the identification model M. Since the structures of the crown part and the sidewall part of the tire are different, there is a possibility that the rigidity and the form of the defect generated are different in these two parts. In the inspection by the tire inspection device 10, since the data input into the identification model M includes part information, the accuracy of defect detection can be improved.

[0076] In the method for generating the identification model M proposed in the present disclosure, a shareography image of a tire is acquired, an image of a defect area inside the tire is extracted from the shareography image, tire information related to the material of the tire is acquired, and an identification model is generated using, as teacher data, a data set including the image of the defect area and the tire information. In this method, tire information related to the material of the tire is also used in generating the identification model. Therefore, according to the inspection using this identification model, the presence or absence of a defect in the tire to be inspected can be detected with high accuracy. Further, in the method for generating the identification model M, the data set input to the identification model includes, in addition to the image of the defect area and the tire information, part information representing the part of the tire including the defect area. According to the inspection using this identification model M, the accuracy of defect detection can be further improved.

[0077] [Others] The present disclosure is not limited to the above-described tire inspection device 10 and the like proposed in the present disclosure.

[0078] For example, the model generation unit 11j may separately generate an identification model for detecting a rubber part defect in the crown part 91 and an identification model for detecting a rubber part defect in the sidewall part 93. In this case, in the tire inspection process, the defect area candidate extracted from the shareography image of the crown part 91 may be input to the identification model for the crown part 91. Further, the defect area candidate extracted from the shareography image of the sidewall part 93 may be input to the identification model for the sidewall part 93.

[0079] In yet another example, the control unit 11 (teacher data generation unit 11e) may not have the normal area extraction unit 11i.

[0080] In the above description, a part of the shareography image (defect area candidate) was extracted, and the discrimination model M was generated using this defect area candidate. Also, even during the inspection of the tire, a part of the shareography image (defect area candidate) was extracted, and this defect area candidate was input into the discrimination model M, and tire defects were detected. Differently, without such a defect candidate area being extracted, the entire shareography image may be input into the discrimination model M before learning, and the discrimination model M may be generated. Also, even during the inspection of the tire, the entire shareography image may be input into the discrimination model M, and tire defects may be determined.

Explanation of Signs

[0081] 10: Tire inspection device, 11: Control unit, 11a: Shareography image acquisition unit, 11b: Tire information acquisition unit, 11e: Teacher data generation unit, 11f: Defect area extraction unit, 11g: Candidate extraction unit, 11h: Judgment result reception unit, 11i: Normal area extraction unit, 11j: Model generation unit, 11u: Defect area candidate extraction unit, 11v: Defect judgment unit, 12: Storage unit, 13: Display unit, 14: Input unit, 15: Shareography device, 15A: Laser irradiation unit, 15B: Interferometer, 15C: Light receiving unit, 15a: Half mirror, 15b·15c: Mirror, 17: Vacuum pump, 90: Tire, 90a: Rubber part, 91: Crown part, 93: Sidewall part, 94: Belt, 95: Carcass, 96: Finishing member, 97: Bead, M: Discrimination model.

Claims

1. A storage means storing an identification model generated by machine learning using, as teacher data, a data set including an image of a defect region, which is a region where a defect inside the tire appears in a share-lography image that is an image of the tire acquired by share-lography, and tire information related to the material of the tire; An image acquisition means for acquiring a share-lography image of a tire to be inspected acquired by share-lography; A tire information acquisition means for acquiring tire information related to the material of the tire to be inspected; A defect determination means for inputting a part or all of the share-lography image of the tire to be inspected and the tire information of the tire to be inspected into the identification model to detect a defect inside the tire to be inspected; A tire inspection device having the above.

2. The part of the tire includes a crown part and a sidewall part, The data set for generating the identification model includes, in addition to the image of the defect region and the tire information, part information representing the part of the tire including the defect region, The defect determination means inputs a part of the share-lography image of the tire to be inspected, the tire information of the tire to be inspected, and part information representing the part of the tire to be inspected including the part of the share-lography image into the identification model. The tire inspection device according to Claim 1.

3. The tire inspection device further includes a defect region candidate extraction means for extracting an image of a defect region candidate, which is a candidate for a region including a defect inside the tire to be inspected, from the share-lography image of the tire to be inspected, The defect determination means inputs the image of the defect region candidate into the identification model to determine whether a defect appears in the defect region candidate. The tire inspection device according to Claim 1.

4. The tire inspection device according to Claim 3, wherein the defect region candidate extraction means searches for a pattern registered in advance in the share-lography image of the tire to be inspected and extracts, as the defect region candidate, a region including the pattern found by the search.

5. An image acquisition step of acquiring a share-lography image, which is an image of a tire to be inspected acquired by share-lography; A tire information acquisition step of acquiring tire information related to the material of the tire to be inspected; A defect determination step of inputting a part or all of the shared tomography image of the tire to be inspected and the tire information of the tire to be inspected into an identification model generated by machine learning using, as teacher data, a data set including an image of a defect area that is an area where a defect inside the tire appears in the shared tomography image and tire information related to the material of the tire A tire inspection method having the above. **Claim 6** An image acquisition means for acquiring a shared tomography image which is an image of a tire to be inspected acquired by shared tomography, A tire information acquisition means for acquiring tire information related to the material of the tire to be inspected, and A defect determination means for inputting a part or all of the shared tomography image of the tire to be inspected and the tire information of the tire to be inspected into an identification model generated by machine learning using, as teacher data, a data set including an image of a defect area that is an area where a defect inside the tire appears in the shared tomography image and tire information related to the material of the tire, and detecting a defect inside the tire to be inspected A program for causing a computer to function as the above. **Claim 7** An image acquisition step of acquiring a shared tomography image which is an image of a tire acquired by shared tomography, A defect area extraction step of extracting, from the shared tomography image, an image of a defect area which is an area including a defect inside the tire, A tire information acquisition step of acquiring tire information related to the material of the tire, and A model generation step of generating an identification model for detecting a defect inside a tire to be inspected using, as teacher data, a data set including the image of the defect area and the tire information A method for generating an identification model including the above. **Claim 8** The tire includes a crown part and a sidewall part, The data set includes, in addition to the image of the defect area and the tire information, part information representing a part of the tire including the defect area The method for generating an identification model according to claim 7. **Claim 9** An image acquisition means for acquiring a shared tomography image of a tire acquired by shared tomography, A defect area extraction means for extracting, from the shared tomography image acquired by the image acquisition means, an image of a defect area which is an area including a defect inside the tire, A tire information acquisition means for acquiring tire information related to the material of the tire Model generation means for generating an identification model for detecting defects inside a tire to be inspected, using as teacher data a data set including an image of the defective area and the tire information An identification model generation device including the above **Claim 10** Image acquisition means for acquiring a shareography image of a tire obtained by shareography, Defective area extraction means for extracting an image of a defective area, which is an area including a defect inside the tire, from the shareography image, Tire information acquisition means for acquiring tire information related to the material of the tire, and Model generation means for generating an identification model for detecting defects inside a tire to be inspected, using as teacher data a data set including an image of the defective area and the tire information A program for causing a computer to function as the above

Images