Tire inspection device and inspection method, identification model generation method and generation device, and program

Abstract

To improve detection accuracy of a defect generated in an inner structure of a tire.SOLUTION: An image data acquisition unit 11a acquires image data of an X-ray image of an inspection object tire. A defect detection unit 11v detects a defect in an inner structure of the inspection object tire by inputting a portion of the image data of the inspection object tire to identification models M1, M2 generated by machine learning using image data of a defective region being a region including the defect in the inner structure of the tire and image data of a normal region being a region different from the defective region as teacher data.SELECTED DRAWING: Figure 5

Description

[Technical Field]

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

[0002] Patent Document 1 below discloses an inspection method for inspecting the internal structure of a tire. Some tires have steel carcass cords. In this type of tire, a defect called an across cord, in which adjacent carcass cords cross, can occur. Patent Document 1 discloses an inspection method for detecting this across cord. In the inspection disclosed in Patent Document 1, while the tire is rotated in the circumferential direction, X-rays are irradiated onto the inner surface of the tire from inside the tire, and an X-ray image is acquired by an imaging device arranged outside the tire. Then, based on the area of ​​the carcass cord image and whether or not the carcass cord image branches, it is determined whether or not the above-mentioned defect called an across cord has occurred. [Prior art documents] [Patent documents]

[0003] [Patent Document 1] Japanese Patent Application Laid-Open No. 2008-309644 DISCLOSURE OF THE INVENTION [Problem to be solved by the invention]

[0004] The inspection method disclosed in Patent Document 1 may result in overdetection, where a normal carcass cord is determined to be defective, and therefore there is a need for improvement in detection accuracy. [Means for solving the problem]

[0005] (1) A tire inspection device proposed in this disclosure includes an image data acquisition means for acquiring image data of an X-ray image of a tire to be inspected, and a defect detection means for inputting a part or all of the image data of the tire to be inspected into a discrimination model generated by machine learning using training data of image data of a defective area, which is an area containing defects in the internal structure of the tire, and image data of a normal area, which is an area different from the defective area. This inspection device can improve the accuracy of tire inspection.

[0006] (2) The tire inspection device described in (1) may further include a defect area candidate extraction means for extracting image data of a defect area candidate, which is a candidate for an area containing a defect in the internal structure, from the image data of the tire to be inspected. The defect detection means may input the image data of the defect area candidate as part of the image data of the tire to be inspected into the identification model and determine whether the defect area candidate contains a defect. This inspection device can further improve tire inspection accuracy.

[0007] (3) The internal structure of the tire may have a plurality of linear members aligned in one direction, and the defect in the internal structure of the tire may include an intersection of the linear members. In the tire inspection device described in (2), the defect area candidate extraction means may extract image data of the defect area candidate based on at least one of the area and shape of each linear member appearing in the image of the tire to be inspected.

[0008] (4) The defects in the internal structure of the tire may include a first type and a second type. In the tire inspection device described in (1), the identification model may be generated using image data of the defect area including the first type of defect, image data of the defect area including the second type of defect, and image data of the normal area as the training data. The defect detection means may input the part or all of the image data of the inspection target tire to the identification model and identify the type of defect in the internal structure of the inspection target tire.

[0009] (5) The tire to be inspected may include a belt and a carcass, and the carcass may have a carcass center portion that is covered by the belt and a carcass side portion that is not covered by the belt. The tire inspection device described in (1) may further include a carcass center portion image extraction means that extracts image data of the carcass center portion from the tire image data, and a carcass side portion image extraction means that extracts image data of the carcass side portion from the tire image data. The discrimination model may include a first discrimination model for detecting defects in the carcass center portion and a second discrimination model for detecting defects in the carcass side portion. The defect detection means may input part or all of the image data of the carcass center portion to the first discrimination model, and input part or all of the image data of the carcass side portion to the second discrimination model.

[0010] (6) The tire to be inspected may have an internal structure including a carcass layer, a belt layer, and a finishing layer. In the tire inspection device described in (1), the defect detection means may detect defects in the carcass layer, the belt layer, or the finishing layer.

[0011] (7) A tire inspection method proposed in the present disclosure includes an image data acquisition step of acquiring image data of an X-ray image of an inspection target tire, and a defect detection step of inputting a part or all of the image data of the inspection target tire to an identification model generated by machine learning using training data including image data of a defect area, which is an area containing a defect in the internal structure of the tire, and image data of a normal area, which is an area different from the defect area, to detect defects in the internal structure of the inspection target tire. This inspection method can improve the accuracy of tire inspection.

[0012] (8) A program proposed in the present disclosure inputs part or all of the image data of an X-ray image of a tire to be inspected into an image data acquisition means for acquiring image data of the tire to be inspected, and into a discrimination model generated by machine learning using, as training data, image data of a defective area, which is an area containing defects in the internal structure of the tire, and image data of a normal area, which is an area different from the defective area, and makes a computer function as defect detection means for detecting defects in the internal structure of the tire to be inspected. This program can improve the accuracy of tire inspection.

[0013] (9) The method for generating an identification model proposed in the present disclosure includes an image data acquisition step of acquiring image data of an X-ray image of a tire, a defect area extraction step of extracting, from the image data, image data of a defect area that is an area containing defects in the internal structure of the tire, a normal area extraction step of extracting, from the image data, image data of a normal area that is an area different from the defect area, and a model generation step of generating an identification model for detecting defects in the internal structure of a tire to be inspected using the image data of the defect area and the image data of the normal area as training data. Utilizing the identification model generated by this method can improve the accuracy of tire inspection.

[0014] (10) In the generation method described in (9), the defect area extraction step may include a candidate extraction step of extracting image data of defect area candidates, which are areas that may contain defects in the internal structure of the tire, from the tire image data, and a judgment result receiving step of receiving a worker's judgment on whether or not a defect is depicted in the defect area candidate. In the model generation step, image data of the defect area candidate that the worker judges to depict a defect may be used as image data of the defect area. Using an identification model generated in this way can further improve tire inspection accuracy.

[0015] (11) The tire includes a belt and a carcass, and the carcass has a carcass center portion that is covered by the belt and a carcass side portion that is not covered by the belt. In the generation method described in (9), the defect area extraction step may extract image data of the defect area from image data of the carcass center portion and extract image data of the defect area from image data of the carcass side portion, and the model generation step may generate a first discrimination model using the image data of the defect area extracted from the image data of the carcass center portion and generate a second discrimination model using the image data of the defect area extracted from the image data of the carcass side portion.

[0016] (12) The device for generating a discriminant model proposed in this disclosure includes an image data acquisition means for irradiating a tire with X-rays and acquiring image data of the tire, a defect area extraction means for extracting, from the image data, image data of a defect area that is an area containing defects in the internal structure of the tire, a normal area extraction means for extracting, from the image data, image data of a normal area that is an area different from the defect area, and a model generation means for generating a discriminant model for detecting defects in the internal structure of an inspected tire using the image data of the defect area and the image data of the normal area as training data. Utilizing the discriminant model generated by this device can improve the accuracy of tire inspection.

[0017] (13) A program proposed in the present disclosure causes a computer to function as image data acquisition means for acquiring image data of an X-ray image of a tire, defect area extraction means for extracting, from the image data, image data of a defective area that is an area containing defects in the internal structure of the tire, normal area extraction means for extracting, from the image data, image data of a normal area that is an area different from the defective area, and model generation means for generating an identification model for detecting defects in the internal structure of a tire to be inspected using the image data of the defective area and the image data of the normal area as training data. Utilizing the identification model with this program can improve the accuracy of tire inspection. [Brief explanation of the drawings]

[0018] [Figure 1] FIG. 1 is a block diagram showing hardware of a tire inspection device proposed in the present disclosure. [Figure 2] 3 is a diagram for explaining the arrangement of an X-ray irradiation unit and an imaging unit, and the internal structure of a tire. FIG. [Figure 3] FIG. 1 is a diagram showing an example of an X-ray image. [Figure 4A] FIG. 10 is a diagram showing an example of a carcass cord including a defect. [Figure 4B] FIG. 10 is a diagram showing an example of a carcass cord including a defect. [Figure 4C] FIG. 10 is a diagram showing an example of a carcass cord including a defect. [Figure 5] FIG. 2 is a block diagram showing functions of a control unit according to the first embodiment. [Figure 6A] 10A and 10B are diagrams for explaining processing by a carcass center portion image extraction unit. [Figure 6B] 10A and 10B are diagrams for explaining processing by a carcass center portion image extraction unit. [Figure 6C] 10A and 10B are diagrams for explaining processing by a carcass center portion image extraction unit. [Figure 6D] 10A and 10B are diagrams for explaining processing by a carcass side image extraction unit. [Figure 7A] FIG. 10 is a diagram for explaining the processing of a candidate extraction unit. [Figure 7B] FIG. 10 is a diagram for explaining the processing of a candidate extraction unit. [Figure 7C] FIG. 10 is a diagram for explaining the processing of a candidate extraction unit. [Figure 8A] 10 is a flowchart illustrating an example of processing executed in generating a discrimination model. [Figure 8B] 10 is a flowchart illustrating an example of processing executed in generating a discrimination model. [Figure 9] 1 is a flowchart illustrating an example of a process performed in tire inspection. [Figure 10] FIG. 10 is a block diagram showing functions of a control unit according to a second embodiment. [Figure 11] 11 is a diagram showing an example of an image extracted by the belt image extraction unit shown in FIG. 10. FIG. DETAILED DESCRIPTION OF THE INVENTION

[0019] The following describes a tire inspection device and inspection method, as well as a method and device for generating an identification model proposed in this disclosure.

[0020] 1 is a block diagram showing the hardware of a tire inspection device 10 proposed in the present disclosure. As shown in Fig. 1, the tire inspection device 10 has a control unit 11, a display unit 13, an operation unit 14, an imaging unit 15, an X-ray irradiation unit 16, and a support unit 17. The control unit 11 and other elements of the tire inspection device 10 function as a generation device for an identification model used by the tire inspection device 10.

[0021] Fig. 2 is a diagram for explaining the arrangement of the X-ray irradiation unit 16 and the imaging unit 15, as well as the internal structure of a tire, shown in Fig. 1. Reference numeral 90 in Fig. 2 denotes both the tire to be inspected and the tire used to generate an identification model.

[0022] [Internal structure of the tire] 2, a tire 90 has a carcass layer 91, a belt layer 92, and a finishing layer 93. Each of these three layers 91, 92, and 93 is composed of a plurality of linear members (cords) aligned in one direction.

[0023] 2, the carcass layer 91 has a plurality of carcass cords 91a arranged in the circumferential direction of the tire 90. Each carcass cord 91a extends from one sidewall portion 90R of the tire 90 to the other sidewall portion 90L.

[0024] 2, the belt layer 92 is disposed inside the tread portion 90T and covers a central portion 91A of the carcass layer 91. The tire 90 may have a plurality of belt layers 92. Each belt layer 92 has a plurality of belt cords 92a arranged in the circumferential direction of the tire 90. Each belt cord 92a is disposed obliquely with respect to the circumferential direction of the tire 90.

[0025] 2, the finishing layer 93 is disposed on the edges of the sidewall portions 90R and 90L, and covers the beads 94. The finishing layer 93 is composed of a plurality of finishing cords 93a arranged in the circumferential direction of the tire 90.

[0026] The linear members (carcass cord 91a, belt cord 92a, finishing cord 93a) that make up the three layers 91, 92, and 93 are made of metal (specifically, steel).

[0027] [Inspection equipment hardware] The X-ray irradiation unit 16 is an X-ray tube that radiates X-rays. As shown in Fig. 2, the X-ray irradiation unit 16 is arranged, for example, inside the tire 90, and irradiates the tire 90 with X-rays. The imaging unit 15 is arranged outside the tire 90, and outputs image data of the X-rays that have passed through the tire 90. Because the X-rays pass through the rubber part of the tire 90, an X-ray image that shows the internal structure of the tire 90 is obtained.

[0028] 2, the imaging unit 15 is disposed, for example, on the upper side (radial outer side) of the tire 90. In addition to or instead of the imaging unit 15, the tire inspection device 10 may have imaging units disposed diagonally upward and to the right and to the left with respect to the tire 90. Furthermore, the tire inspection device 10 may have imaging units disposed on the right and left sides of the sidewall portions 90R and 90L of the tire 90, respectively.

[0029] The tire 90 is supported by a support unit 17 (FIG. 1) so as to be rotatable. The support unit 17 rotates the tire 90 at a predetermined speed during tire inspection and when generating image data (teacher data) for machine learning. The imaging unit 15 continuously images the tire 90 at a frequency according to the rotational speed of the tire 90 and outputs image data for one circumference of the tire 90. The imaging unit 15 is, for example, a line sensor camera, but may also be an area camera.

[0030] The control unit 11 has arithmetic devices 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 includes a main storage device configured with a RAM (Random Access Memory) and an auxiliary storage device configured with a ROM (Read Only Memory), an SSD (Solid State Drive), an HDD (Hard Disk Drive), etc. The storage unit 12 stores programs executed by arithmetic devices such as the CPU, image data output by the imaging unit 15, generated classification models, etc.

[0031] The display unit 13 is a display device such as a liquid crystal display, and displays various images according to instructions from the control unit 11.

[0032] The operation unit 14 is a user interface such as a keyboard or a mouse, and receives operation inputs from the user and outputs signals indicating the contents of the inputs to the control unit 11.

[0033] FIG. 3 is a diagram showing an example of an X-ray image acquired by the imaging unit 15. In FIG. 3, the vertical direction (Y direction) of the image corresponds to the circumferential direction of the tire 90. The carcass layer 91 and the belt layer 92 appear in the central portion of the image (the region indicated by reference numeral 91A in the figure). As described above, the belt layer 92 has a plurality of belt cords 92a arranged in the circumferential direction of the tire 90. Each belt cord 92a is disposed obliquely with respect to the circumferential direction of the tire 90. Furthermore, a plurality of carcass cords 91a are arranged in the circumferential direction of the tire 90. Each carcass cord 91a extends in the horizontal direction in this image. The carcass layer 91 has a side portion 91B that is not covered by the belt layer 92. In this specification, the central portion 91A of the carcass layer 91 that is covered by the belt layer 92 in the carcass layer 91 is referred to as the "carcass central portion," and the side portion 91B of the carcass layer 91 that is not covered by the belt layer 92 is referred to as the "carcass side portion."

[0034] [Defects in the tire's internal structure] The tire inspection device 10 is a device that inspects, for example, carcass cords 91a. The multiple carcass cords 91a are arranged at equal intervals in the circumferential direction of the tire 90. However, some carcass cords 91a may be arranged diagonally, resulting in a defect where the carcass cords 91a cross over adjacent carcass cords 91a. Such a defect is called an across-cord.

[0035] For example, as shown in FIG. 4A, a defect may occur in which two carcass cords 91a intersect while making line contact. Also, as shown in FIG. 4B, two carcass cords 91a may intersect in an X-shape. Also, as shown in FIG. 4C, a defect may occur in which one carcass cord 91a is curved in a U-shape and intersects with the adjacent carcass cord 91a at two positions. The tire inspection device 10 detects such defects by processing described below. Hereinafter, such defects will be referred to as "cord defects."

[0036] [Processing of the control unit in the first embodiment] The functions of the control unit 11 are described below. FIG. 5 is a block diagram showing the functions of the control unit 11. The control unit 11 has, as its functions, an image data acquisition unit 11a, a preprocessing unit 11b, a training data generation unit 11e, a model generation unit 11j, a defect area candidate extraction unit 11u, and a defect detection unit 11v. These functions are realized by the control unit 11 executing a program stored in the storage unit 12. The control unit 11 also has discrimination models M1 and M2. The discrimination models M1 and M2 are stored in the storage unit 12.

[0037] The image data acquisition unit 11a, the preprocessing unit 11b, the teacher data generation unit 11e, and the model generation unit 11j generate (learn) the discrimination models M1 and M2. The image data acquisition unit 11a, the preprocessing unit 11b, the defect area candidate extraction unit 11u, and the defect detection unit 11v inspect the internal structure of the tire 90 using the discrimination models M1 and M2.

[0038] The control unit 11 may be configured with multiple personal computers. Alternatively, the control unit 11 may be configured with one or multiple personal computers and one or multiple server computers. In this case, some of the functions of the control unit 11 (e.g., the defect area candidate extraction unit 11u and the defect detection unit 11v) may be executed by a personal computer, and other functions of the control unit 11 (e.g., the model generation unit 11j) may be executed by another personal computer or server computer.

[0039] The types of intersections of the carcass cords 91a that should be determined as cord defects are not the same in the carcass central portion 91A and the carcass side portions 91B. For example, there are intersections that should be determined as cord defects when they occur in the carcass central portion 91A, but should be determined as normal when they occur in the carcass side portions 91B. Also, there are types of intersections that are likely to occur in the carcass central portion 91A but unlikely to occur in the carcass side portions 91B. Therefore, the tire inspection device 10 has a first identification model M1 used to inspect the carcass central portion 91A and a second identification model M2 used to inspect the carcass side portions 91B.

[0040] [Image data acquisition section] The image data acquisition unit 11a acquires image data of an X-ray image of the tire 90 from the imaging unit 15. Alternatively, the image data acquisition unit 11a may acquire image data of an X-ray image of the tire 90 stored in the memory unit 12. During inspection, the image data acquisition unit 11a acquires image data of the tire 90 to be inspected. During learning (generation) of the identification models M1 and M2, the image data acquisition unit 11a acquires image data for learning (image data for generating teacher data). During learning, the image data acquisition unit 11a may acquire image data of multiple tires. Each image data may be image data for one circumference of the tire.

[0041] [Preprocessing section] As shown in FIG. 5, the pre-processing unit 11b has a carcass center image extraction unit 11c and a carcass side image extraction unit 11d.

[0042] [Carcass center image extraction section] The carcass center portion 91A is covered with the belt layer 92, and in the area of ​​the carcass center portion 91A, a carcass cord 91a and a belt cord 92a appear as shown in Fig. 3. The carcass center portion image extraction unit 11c extracts an image of the carcass cord 91a of the carcass center portion 91A from the image data of the tire 90 acquired by the image data acquisition unit 11a. This extraction can be performed, for example, by the following process.

[0043] The average density value in the region of the carcass center portion 91A is darker than the average density value in the region of the carcass side portion 91B. Therefore, the carcass center portion image extraction unit 11c generates an image in which the density values ​​of multiple pixels arranged in the vertical direction of the image (the circumferential direction of the tire 90) are smoothed, as shown in FIG. 6A, for example. In this image, the average density value of multiple pixels arranged in the vertical direction is assigned to the multiple pixels. In this image, the boundary between the region of the carcass center portion 91A and the region of the carcass side portion 91B (edges E1 and E2 in the example of FIG. 6A) appears. Then, the carcass center portion image extraction unit 11c performs edge detection on this image and extracts the region between the right edge E1 and the left edge E2 from the image data of the tire acquired by the image data acquisition unit 11a. As a result, image data of the region of the carcass center portion 91A is extracted, as shown in FIG. 6B.

[0044] In the image of the area of ​​the carcass center portion 91A, both the carcass cord 91a and the belt cord 92a appear, as shown in Fig. 6B. The carcass center portion image extraction unit 11c executes processing to clarify the carcass cord 91a in the carcass center portion 91A.

[0045] Each belt cord 92a is disposed obliquely relative to the circumferential direction of the tire 90, and the multiple belt cords 92a are arranged at equal intervals in the circumferential direction of the tire 90. Therefore, in the image of the tire 90, the belt cords 92a appear periodically as oblique lines. Therefore, the carcass center portion image extraction unit 11c performs, for example, a two-dimensional Fourier transform process on the image data of the extracted carcass center portion 91A (an image including the carcass layer 91 and the belt layer 92). The carcass center portion image extraction unit 11c then removes image data of the frequency components corresponding to the belt cords 92a. By doing so, an image in which the carcass cords 91a in the carcass center portion 91A are clearly shown can be obtained, as shown in FIG. 6C. The carcass center portion image extraction unit 11c may perform a binarization process on the image obtained in this manner. This binarization process may be a dynamic binarization process. Hereinafter, the image extracted by the carcass center portion image extraction unit 11c (FIG. 6C) will be referred to as the "carcass center portion image."

[0046] [Carcass side image extraction section] The carcass side image extraction unit 11d extracts image data of the carcass cord 91a of the carcass side portion 91B from the tire image data acquired by the image data acquisition unit 11a. This extraction can utilize part of the processing performed by the carcass center image extraction unit 11c. That is, the carcass side image extraction unit 11d generates an image (see FIG. 6A) in which the density values ​​of multiple pixels aligned in the vertical direction of the image (the circumferential direction of the tire 90) are smoothed, and performs edge detection on the generated image. The carcass side image extraction unit 11d then extracts the region to the right of the right edge E1 (see FIG. 6A) and the region to the left of the left edge E2 (see FIG. 6A). This results in image data of the left and right carcass side portions 91B, as shown in FIG. 6D. The carcass side image extraction unit 11d may perform binarization processing on this image data. This binarization processing may also be dynamic binarization processing. Hereinafter, the image extracted by the carcass side image extractor 11d (FIG. 6D) will be referred to as a "carcass side image."

[0047] [Generate a discrimination model] The following describes the process for generating the discrimination models M1 and M2 using the learning image data (carcass center image and carcass side image) extracted by the process of the preprocessing unit 11b. The training data generation unit 11e and the model generation unit 11j shown in Figure 5 execute the process for generating the discrimination models M1 and M2.

[0048] [Teacher data generation section] 5, the teacher data generating unit 11e has a defective area extracting unit 11f and a normal area extracting unit 11i. The defective area extracting unit 11f has a candidate extracting unit 11g and a determination result receiving unit 11h.

[0049] [Candidate extraction section] The candidate extraction unit 11g extracts areas where a cord defect (e.g., the defects shown in FIGS. 4A to 4C) may occur from each of the learning carcass center image and the learning carcass side image. Hereinafter, an area where a cord defect may occur is referred to as a "defect area candidate." The candidate extraction unit 11g extracts image data of the defect area candidate based on at least one of the area and shape of the carcass cord 91a appearing in the carcass center image and the carcass side image. This process is performed, for example, as follows.

[0050] FIG. 7A is an example of an enlarged image of the carcass center portion. In this figure, ten carcass cords 91a are shown, but some of the carcass cords 91a (carcass cords within the area indicated by the two-dot chain line L) overlap. When multiple carcass cords 91a are lined up normally, each of the multiple carcass cords 91a appears in the binarized image as a connected area in which pixels with a density value of 1 are connected. However, when multiple carcass cords 91a intersect, the entire multiple carcass cords 91a appear as a single connected area. For example, in the example of FIG. 7A, three carcass cords 91a within the area indicated by the two-dot chain line L form a single connected area. Therefore, the area of ​​the intersecting carcass cords 91a (the number of pixels in the connected area) is larger than the area of ​​the other carcass cords 91a.

[0051] Therefore, the candidate extraction unit 11g extracts a defect area candidate based on the area of ​​the carcass cord 91a appearing in the binarized carcass center image. For example, the candidate extraction unit 11g calculates the average value of the area of ​​each carcass cord 91a and detects an image of the carcass cord 91a having an area that is a predetermined magnification (e.g., 1.5 times) or more of the average value. The candidate extraction unit 11g extracts an area including the detected image of the carcass cord 91a as a defect area candidate. For example, as shown by the two-dot chain line L in Figures 7A and 7B, a circumscribed rectangle of the carcass cord 91a having an area that is a predetermined magnification or more of the average value is extracted as a defect area candidate.

[0052] The candidate extraction unit 11g may extract a defect area candidate based on the shape of the carcass cord 91a appearing in the carcass center portion image. For example, the candidate extraction unit 11g may search for a branch shape of the carcass cord 91a in the carcass center portion image. If a branch shape is found, the candidate extraction unit 11g may extract an area including the branch shape as a defect area candidate.

[0053] For example, the candidate extraction unit 11g may perform a thinning process on the region extracted based on area (the image exemplified in FIG. 7B). This thinning process gradually reduces the line width of the image of each carcass cord 91a, and each carcass cord 91a is represented as a line having a width of, for example, one pixel (see FIG. 7C). The candidate extraction unit 11g may then search for a branching shape in the thinned image of the carcass cord 91a. The search for a branching shape can be performed, for example, by scanning the image of the thinned carcass cord 91a in the horizontal direction of the image. If a branching shape is present, the candidate extraction unit 11g may extract a circumscribing rectangle of a connected region (the region shown in FIG. 7B) including the branching shape as a defect region candidate. By performing such a process, for example, when two adjacent carcass cords 91a are close to each other but do not intersect, it is possible to prevent the image extracted based on area from being immediately extracted as a defect region candidate.

[0054] As described herein, the candidate extraction unit 11g extracts image data of the candidate defective area based on, for example, the area and shape (presence or absence of a branching shape) of the carcass cord 91a appearing in the carcass center portion image. Alternatively, the candidate extraction unit 11g may extract image data of the candidate defective area based only on the area of ​​the carcass cord 91a appearing in the carcass center portion image. As yet another example, the candidate extraction unit 11g may extract image data of the candidate defective area based only on the shape (presence or absence of a branching shape) of the carcass cord 91a appearing in the carcass center portion image.

[0055] The candidate extraction unit 11g performs the same processing on the carcass side image as on the carcass center image. That is, the candidate extraction unit 11g may extract a defect area candidate from the carcass side image based on the area of ​​the carcass cord 91a appearing in the binarized carcass side image. The candidate extraction unit 11g may also extract a defect area candidate from the carcass side image based on the shape of the carcass cord 91a appearing in the carcass side image.

[0056] [Judgment result reception section] The judgment result receiving unit 11h displays the image of the defect area candidate extracted by the candidate extracting unit 11g on the display unit 13. Then, the judgment result receiving unit 11h receives the worker's judgment result as to whether the displayed image actually contains a code defect. The worker can input the judgment result through the operation unit 14. When multiple defect area candidates are stored in the memory unit 12, the judgment result receiving unit 11h displays the multiple defect area candidates in order on the display unit 13 and receives the worker's judgment result for each defect area candidate.

[0057] The determination result receiving unit 11h assigns an identification label to each defect area candidate based on the input determination result, and stores the result in the storage unit 12. Specifically, if the input determination result indicates that a cord defect appears in the defect area candidate, the determination result receiving unit 11h assigns the identification label "defect" to the defect area candidate. Then, the determination result receiving unit 11h stores the image data of the defect area candidate as a defect area in the storage unit 12. On the other hand, if the input determination result indicates that a cord defect does not appear in the defect area candidate, the determination result receiving unit 11h assigns the identification label "normal" to the defect area candidate, and stores the image data of the defect area candidate as a normal area in the storage unit 12. The defect area and normal area are stored in the storage unit 12 together with information on the location where they existed (carcass center portion 91A or carcass side portion 91B).

[0058] If a code defect appears in the defect area candidate, the worker may input the type of the code defect. Examples of types of code defects include those shown in FIGS. 4A to 4C. For example, if a shape similar to the code defect shown in FIG. 4A appears in the defect area candidate, the determination result receiving unit 11h assigns an identification label "Type 1 defect" to this defect area candidate. The determination result receiving unit 11h then stores image data of this defect area candidate in the storage unit 12 as a defect area of ​​the "Type 1 defect." Similarly, if a code defect with a shape similar to the code defect shown in FIG. 4B appears in the defect area candidate, the determination result receiving unit 11h assigns an identification label "Type 2 defect" to this defect area candidate and stores image data of this defect area candidate in the storage unit 12 as a defect area of ​​the "Type 2 defect." The same applies when a code defect with a shape similar to the code defect shown in FIG. 4C appears in the defect area candidate. The number of defect types is not limited to this. The number of defect types may be two or more than three.

[0059] [Normal area extraction part] The normal area extraction unit 11i extracts, as a normal area, a portion of an area that was not extracted as a defective area candidate in the processing by the candidate extraction unit 11g. The normal area extraction unit 11i assigns an identification label "normal" to the image extracted as a normal area, and stores image data of this normal area in the storage unit 12. The normal area extraction unit 11i extracts normal areas from each of the carcass center image and the carcass side image. The image data of the normal areas extracted in this manner is stored in the storage unit 12 together with information on the area where they existed (carcass center portion 91A or carcass side portion 91B).

[0060] The normal region extraction process can be performed, for example, as follows: The normal region extraction unit 11i randomly extracts an image having a predetermined number of pixels and a predetermined aspect ratio from the carcass center image. Then, if the randomly extracted image does not have any overlapping portion with the image extracted as the defective region candidate, the normal region extraction unit 11i may determine this extracted image as the normal region. Here, the predetermined number of pixels and the predetermined aspect ratio may be, for example, the average number of pixels and the average aspect ratio of the defective region candidate.

[0061] The teacher data generation unit 11e converts each of the image data of the plurality of defect areas 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 will be referred to as the "model input pixel number," and this aspect ratio will be referred to as the "model input aspect ratio.") The teacher data generation unit 11e also converts the image data of the plurality of normal areas obtained by the above-described process into the model input pixel number and model input aspect ratio. In other words, the teacher data generation unit 11e unifies the dimensions of the teacher data.

[0062] [Model generation section] The model generation unit 11j inputs the image data and identification labels (i.e., correct labels) of the defective region as training data into the pre-learning discrimination model. The model generation unit 11j also inputs the image data and identification labels (i.e., correct labels) of the normal region as training data into the same discrimination model. In this way, the model generation unit 11j generates discrimination models M1 and M2 for detecting defects in the internal structure of the tire 90.

[0063] For example, neural networks may be used as the discrimination models M1 and M2. Convolutional neural networks (CNNs) may also be used as the discrimination models M1 and M2. Alternatively, support vector machines (SVMs) or random forests may also be used as the discrimination models M1 and M2.

[0064] As described above, the training data generating unit 11e extracts defect areas and normal areas from each of the carcass center image and the carcass side image. The model generating unit 11j generates a first identification model M1 for detecting a cord defect in the carcass center portion 91A using the defect areas and normal areas extracted from the carcass center image. The model generating unit 11j also generates a second identification model M2 for detecting a cord defect in the carcass side portion 91B using the defect areas and normal areas extracted from the carcass side image.

[0065] As described above, the types of intersections of the carcass cords 91a that should be determined as cord defects are different between the carcass central portion 91A and the carcass side portions 91B. For example, there are intersections that should be determined as cord defects when they occur in the carcass central portion 91A, but should be determined as normal when they occur in the carcass side portions 91B. Also, there are types of intersections that are likely to occur in the carcass central portion 91A but unlikely to occur in the carcass side portions 91B. In this embodiment, two mutually different identification models M1 and M2 are generated for each of the two portions of the carcass layer 91 (the carcass central portion 91A and the carcass side portions 91B). This improves the accuracy of cord defect detection using the identification models M1 and M2.

[0066] [Tire inspection] The following describes the process for inspecting a tire using the carcass center image and carcass side image of the tire to be inspected. The defect area candidate extraction unit 11u and defect detection unit 11v perform the inspection process.

[0067] [Defect area candidate extraction section] The defect area candidate extraction unit 11u extracts defect area candidates (see FIG. 7B) from the image data of the tire 90 to be inspected extracted by the preprocessing unit 11b, that is, from each of the carcass center image and the carcass side image.

[0068] The processing performed by the defect area candidate extraction unit 11u may be the same as that performed by the candidate extraction unit 11g of the training data generation unit 11e. That is, the defect area candidate extraction unit 11u extracts image data of defect area candidates based on the area of ​​the carcass cord 91a appearing in each of the carcass center image and the carcass side image. Instead of or in addition to the area of ​​the carcass cord 91a, the defect area candidate extraction unit 11u may extract image data of defect area candidates based on the shape of the carcass cord 91a appearing in the carcass center image and the carcass side image. For example, the defect area candidate extraction unit 11u may search for the branching shape of the carcass cord 91a in the carcass center image and the carcass side image, and extract an area including the branching shape as a defect area candidate.

[0069] [Defect detection section] The defect detection unit 11v inputs the image data of the defect area candidates extracted by the defect area candidate extraction unit 11u to the identification models M1 and M2. The defect detection unit 11v inputs the defect area candidates extracted from the carcass center image to the first identification model M1 and classifies the defect area candidates into one of multiple classes. The defect detection unit 11v also inputs the defect area candidates extracted from the carcass side image to the second identification model M2 and classifies the defect area candidates into one of multiple classes.

[0070] Here, the multiple classes are, for example, a class in which a cord defect appears in the defect area candidate (a class to which the identification label "defect" is assigned), or a class in which a cord defect does not appear in the defect area candidate (a class to which the identification label "normal" is assigned). Through such processing by the defect detection unit 11v, defects in the internal structure of the tire 90 to be inspected can be detected. The output of the defect detection unit 11v is, for example, the probability that the defect area candidate falls into each class. Alternatively, the output of the defect detection unit 11v may be the identification label of the class to which the defect area candidate has the highest probability of falling.

[0071] As described above, when generating the identification models M1 and M2, image data of defect areas to which identification labels indicating the types of defects (e.g., "Type 1 defect" or "Type 2 defect") have been assigned may be used as training data. In this case, the multiple classes classified by the defect detection unit 11v may include multiple classes corresponding to the types of code defects appearing in the defect area candidates. For example, the multiple classes may include a class corresponding to Type 1 code defects (see FIG. 4A) (a class assigned with the identification label "Type 1 defect") and a class corresponding to Type 2 code defects (see FIG. 4B) (a class assigned with the identification label "Type 2 defect"). This processing by the defect detection unit 11v enables not only the presence or absence of defects but also the type of defects to be detected.

[0072] The defect detection unit 11v displays the output, for example, together with an image of the defect area candidate on the display unit 13 or stores it in the memory unit 12. When multiple defect area candidates for one tire 90 are stored in the memory unit 12, the defect detection unit 11v may input all of the multiple defect area candidates to the identification models M1 and M2 in order and display each of the outputs on the display unit 13 or store them in the memory unit 12.

[0073] [Learning Flow] 8A and 8B are diagrams showing an example of processing executed by the control unit 11 to generate identification models M1 and M2.

[0074] The control unit 11 drives the X-ray irradiation unit 16 to irradiate X-rays onto the tire 90 for generating learning data. The image data acquisition unit 11a acquires an X-ray image of the tire 90 from the imaging unit 15 (S101). The image data acquisition unit 11a may acquire an X-ray image stored in the storage unit 12. The carcass center image extraction unit 11c extracts a carcass center image (see FIG. 6C) from the image data acquired in S101. Furthermore, the carcass side image extraction unit 11d extracts a carcass side image (see FIG. 6D) from the image data acquired in S101 (S102).

[0075] Next, the training data generating unit 11e executes a process of generating training data (S103). Specifically, as shown in FIG. 8B, the defect area extracting unit 11f (candidate extracting unit 11g) extracts areas where a cord defect may occur (defect area candidates, FIG. 7B) from the carcass center image extracted in S102 (S201). The defect area extracting unit 11f (determination result receiving unit 11h) displays the extracted defect area candidates on the display unit 13. Then, the defect area extracting unit 11f (determination result receiving unit 11h) receives the operator's determination result as to whether the displayed defect area candidates include a cord defect, and assigns an identification label (correct answer label) to the defect area candidate according to the determination result (S202). The identification label is, for example, "normal" or "defect." Image data assigned with the identification label "defect" is stored in the storage unit 12 as a defective area, and image data assigned with the identification label "normal" is stored in the storage unit 12 as a normal area. As the identification label, a label indicating the type of the code defect (for example, "type 1 defect" or "type 2 defect") may be given.

[0076] The normal area extraction unit 11i extracts a part of the area that was not extracted (selected) as a defect area candidate in S201 from the carcass center image as a normal area, and assigns an identification label "normal" to the image data of the normal area (S203). This image data assigned the identification label "normal" is stored in the storage unit 12 as a normal area.

[0077] Next, the training data generating unit 11e performs the same processes as S201 to S203 on the carcass side portion 91B. Specifically, the defect area extracting unit 11f (candidate extracting unit 11g) extracts areas (defect area candidates) where a cord defect may occur from the carcass side portion image extracted in S102 (S204). The defect area extracting unit 11f (determination result receiving unit 11h) displays the defect area candidates on the display unit 13. Then, the defect area extracting unit 11f (determination result receiving unit 11h) receives the operator's determination result as to whether the displayed defect area candidates include a cord defect, and assigns an identification label (correct label) to the defect area candidate according to the determination result (S205). The identification label is, for example, "normal" or "defect." Image data assigned with the identification label "defect" is stored in the storage unit 12 as a defective area, and image data assigned with the identification label "normal" is stored in the storage unit 12 as a normal area. Instead of "defect," a label indicating the type of the code defect (for example, "type 1 defect" or "type 2 defect") may be assigned as the identification label.

[0078] The normal area extraction unit 11i extracts a part of the area that was not extracted (selected) as a defect area candidate in S204 from the carcass side image as a normal area, and assigns an identification label "normal" to the image data of the normal area (S206). This image data assigned the identification label "normal" is stored in the storage unit 12 as a normal area.

[0079] The training data generation unit 11e determines whether a predetermined number of sets of image data (predetermined number of sets of training data) have been prepared for each of the multiple classes through the processes of S201 to S206 (S207). Here, the multiple classes include, for example, a class in which the defect area candidate has a cord defect (a class assigned the identification label "defect") and a class in which the defect area candidate does not have a cord defect (a class assigned the identification label "normal"). In this case, the training data generation unit 11e determines whether the number of defect areas extracted from the carcass center image, normal areas extracted from the carcass center image, defect areas extracted from the carcass side image, and normal areas extracted from the carcass side image has reached a predetermined number. Here, the predetermined number is the number deemed necessary to generate the identification models M1 and M2. If the number of image data for each area has not reached the predetermined number, the training data generation unit 11e returns to the process of S201 and executes the subsequent processes. The plurality of classes into which the defect area candidates are classified may include a plurality of classes each corresponding to a type of code defect, such as a "first type defect" and a "second type defect."

[0080] The teacher data generation unit 11e converts the number of pixels and aspect ratio of each image data (defective area and normal area) stored in the memory unit 12 into a predetermined number of pixels (the above-mentioned "model input pixel number") and a predetermined aspect ratio (the above-mentioned "model input aspect ratio"), respectively (S208).

[0081] The order of the processes performed by the training data generating unit 11e is not limited to the example shown in Fig. 8B. For example, the process of extracting a defect area and a normal area from the carcass side image (S204 to S206) may be performed before the process of extracting a defect area and a normal area from the carcass side image (S201 to S203). Furthermore, the defect area extracting unit 11f may extract multiple defect area candidates from the carcass center image (or the carcass side image). The defect area extracting unit 11f may then display the multiple defect area candidates on the display unit 13 at once.

[0082] 8A, the model generation unit 11j inputs the image data (teaching data) extracted in S201 and S202 into the pre-learning discrimination model M1, and generates (trains) the discrimination model M1 for detecting cord defects in the carcass center portion 91A (S104). The model generation unit 11j also inputs the image data (teaching data) extracted in S204 and S206 into the pre-learning discrimination model M2, and generates (trains) the discrimination model M2 for detecting cord defects in the carcass side portion 91B (S105). The above is an example of the processing executed by the control unit 11 to generate the discrimination models M1 and M2.

[0083] [Inspection flow] Next, an example of processing executed by the control unit 11 to detect defects in the internal structure of a tire will be described with reference to FIG.

[0084] The control unit 11 drives the X-ray irradiation unit 16 to irradiate the tire 90, which is the object of inspection, with X-rays. The image data acquisition unit 11a acquires image data of the X-ray image of the tire 90 from the imaging unit 15 (S301). The pre-processing unit 11b (carcass center image extraction unit 11c) extracts a carcass center image (see FIG. 6C) from the image data acquired in S301 (S302). In addition, the pre-processing unit 11b (carcass side image extraction unit 11d) extracts a carcass side image (S302).

[0085] The defect area candidate extraction unit 11u extracts areas where a cord defect may occur (defect area candidates) from each of the carcass center image extracted in S302 and the carcass side image extracted in S302 (S303).The defect area candidate extraction unit 11u converts the pixel count and aspect ratio of the extracted defect area candidates into a model input pixel count and a model input aspect ratio (S304).

[0086] The defect detection unit 11v inputs image data of the defect area candidates extracted from the carcass center portion image to the first identification model M1 (S305). Then, the defect detection unit 11v classifies the defect area candidates into multiple classes (S306). The multiple classes are, for example, a class in which a cord defect appears in the defect area candidate (a class assigned the identification label "defect"), or a class in which a cord defect does not appear in the defect area candidate (a class assigned the identification label "normal"). The multiple classes classified by the defect detection unit 11v may include multiple classes corresponding to different types of cord defects. This process by the defect detection unit 11v can detect not only the presence or absence of a defect in the internal structure of the tire 90 being inspected, but also the type of defect. The defect detection unit 11v determines whether all of the defect area candidates extracted from the carcass center portion image in S303 have been classified (S307). If there are still any unclassified defect area candidates remaining, the defect detection section 11v returns to S305 and executes the subsequent processes for the unclassified defect area candidates.

[0087] Next, the defect detection unit 11v performs the same processes as S305 to S307 on the defect area candidates extracted from the carcass side image. That is, the defect detection unit 11v inputs image data of the defect area candidates extracted from the carcass side image to the second identification model M2 (S308). Then, the defect detection unit 11v classifies the defect area candidates into a plurality of classes (S309). The defect detection unit 11v determines whether all of the defect area candidates extracted from the carcass side image in S303 have been classified (S310). If there are still defect area candidates that have not been classified, the defect detection unit 11v returns to S308 and performs the subsequent processes on the unclassified defect area candidates.

[0088] If it is determined in S310 that classification of all defect area candidates extracted from the carcass side image has been completed, the control unit 11 ends the process. In other words, the control unit 11 ends inspection of the internal structure of the tire 90.

[0089] [Processing of the control unit in the second embodiment] Furthermore, the tire inspection device 10 described above detects cord defects occurring in the carcass layer 91. However, cord defects may also occur in the belt layer 92 or the finishing layer 93. For example, two adjacent belt cords 92a may cross each other, or two adjacent finishing cords 93a may cross each other. The tire inspection device may also detect cord defects occurring in the belt layer 92 or the finishing layer 93. When detecting cord defects in the finishing layer 93, the tire inspection device 10 may have an imaging unit facing the sidewall portions 90L and 90R of the tire 90, in addition to the imaging unit 15 located radially outward from the tire 90.

[0090] Fig. 10 is a block diagram showing the functions of the control unit 111 of the inspection device of this type. Below, an example of detecting a cord defect occurring in the belt layer 92 and a cord defect occurring in the carcass layer 91 will be described with reference to Fig. 10.

[0091] The control unit 111 includes an image data acquisition unit 11a, a preprocessing unit 111b, a training data generation unit 111e, a model generation unit 111j, a defect area candidate extraction unit 111u, and a defect detection unit 111v. The preprocessing unit 111b includes a belt image extraction unit 111p and a carcass image extraction unit 111q.

[0092] Like the image data acquisition unit 11a of the control unit 11 described above, the image data acquisition unit 11a acquires image data of an X-ray image of a tire through the imaging unit 15. During inspection, the image data acquisition unit 11a acquires image data of the tire 90 to be inspected. During learning, the image data acquisition unit 11a acquires image data of the tire 90 for generating image data for learning (teacher data).

[0093] The belt image extraction unit 111p extracts image data of the belt cord 92a from the tire image data acquired by the image data acquisition unit 11a. This extraction can be performed, for example, by the following process.

[0094] The belt layer 92 covers the carcass center portion 91A, and the average density value of the region where the belt layer 92 appears is higher than the average density value of the region where only the carcass layer 91 appears (region of the carcass side portion 91B). Therefore, similar to the above-described carcass center portion image extraction unit 11c, the belt image extraction unit 111p generates an image (see FIG. 6A) by smoothing the density values ​​of multiple pixels aligned in the vertical direction of the image (circumferential direction of the tire 90). In this image, the boundary (edges E1 and E2 in the example of FIG. 6A) between the region of the belt layer 92 (region of the carcass center portion 91A) and the region of the carcass side portion 91B appears. Then, the belt image extraction unit 111p performs edge detection on this image and extracts the region between the right edge E1 and the left edge E2 from the image data acquired by the image data acquisition unit 11a. As a result, image data of the belt layer 92 and the carcass center portion 91A (see FIG. 6B) is extracted.

[0095] The belt image extraction unit 111p performs processing to clarify the belt cord 92a. For example, the belt image extraction unit 111p performs two-dimensional Fourier transform processing on the extracted image (an image including the carcass cord 91a and the belt cord 92a). Then, the belt image extraction unit 111p removes image data of frequency components corresponding to the carcass cord 91a. By doing so, an image in which the belt cord 92a is clearly shown is obtained, as shown in FIG. 11. The belt image extraction unit 111p may perform binarization processing on the image obtained in this manner. This processing may be dynamic binarization processing. Hereinafter, the image obtained by these processings of the belt image extraction unit 111p will be referred to as a "belt image."

[0096] The tire 90 may have multiple belt layers 92 in which the angles of the belt cords 92a are different. For example, the tire 90 may have a first belt layer 92 made up of belt cords 92a arranged at a first angle (+45 degrees with respect to the circumferential direction of the tire 90) and a second belt layer 92 made up of belt cords 92a arranged at a second angle (-45 degrees with respect to the circumferential direction of the tire 90). In this case, the belt image extraction unit 111p removes the carcass cords 91a and the belt cords 92a of the second belt layer 92 by two-dimensional Fourier transform, and generates a belt image (see FIG. 11) in which the belt cords 92a of the first belt layer 92 are clearly shown. The belt image extraction unit 111p also removes the carcass cords 91a and the belt cords 92a of the first belt layer 92 by two-dimensional Fourier transform, and generates a belt image in which the belt cords 92a of the second belt layer 92 are clearly shown.

[0097] The carcass image extraction unit 111q performs a two-dimensional Fourier transform process on the image acquired by the image data acquisition unit 11a to remove frequency components corresponding to the belt cord 92a. The carcass image extraction unit 111q also performs a binarization process to generate an image in which the carcass cord 91a is clearly visible. Hereinafter, the image obtained by this process of the carcass image extraction unit 111q will be referred to as a "carcass image." Hereinafter, the carcass image may be, for example, an image including both the carcass center portion 91A and the carcass side portion 91B described above.

[0098] [Generate a discrimination model] The following describes the process for generating the discrimination models M3 and M4 using the training belt image and carcass image. A training data generation unit 111e and a model generation unit 111j execute the process for generating the discrimination models M3 and M4. The training data generation unit 111e includes a candidate extraction unit 111g and a determination result reception unit 111h.

[0099] [Candidate extraction section] The candidate extraction unit 111g extracts areas (defective area candidates) where a cord defect may occur from the belt image. When multiple belt cords 92a intersect, the area of ​​the intersecting belt cords 92a (the number of pixels in the connected area) is larger than the area of ​​each of the other belt cords 92a. Therefore, the candidate extraction unit 111g extracts image data of defective area candidates based on, for example, the area of ​​the belt cords 92a appearing in the belt image. For example, the candidate extraction unit 111g detects an image of a belt cord 92a whose area is equal to or greater than the average area of ​​each belt cord 92a by a predetermined magnification (for example, 1.5 times), and designates the circumscribed rectangle of the belt cord 92a as a defective area candidate.

[0100] The candidate extractor 111g may extract image data of candidate defective areas based on the shape of the belt cord 92a that appears in the belt image. For example, the candidate extractor 111g may search for a branching shape of the belt cord 92a in the belt image and extract an area that includes the branching shape as a candidate defective area.

[0101] The candidate extraction unit 111g also extracts defect area candidates from the carcass image. This process may be similar to the process of extracting defect area candidates from the belt image. That is, the candidate extraction unit 111g detects an image of a carcass cord 91a having an area that is a predetermined magnification (e.g., 1.5 times) or more of the average area of ​​each carcass cord 91a, and sets the circumscribed rectangle of the carcass cord 91a as a defect area candidate. The candidate extraction unit 111g may search for the branch shape of the carcass cord 91a in the carcass image and extract an area including the branch shape as a defect area candidate.

[0102] [Judgment result reception section] The judgment result receiving unit 111h displays the defect area candidates extracted from the carcass image and the defect area candidates extracted from the belt image on the display unit 13. The judgment result receiving unit 11h also receives the worker's judgment results as to whether or not each of these defect area candidates contains a defect. The judgment result receiving unit 11h then assigns an identification label ("defect" or "normal") according to the judgment result to the image data of the defect area candidate. The image data assigned with the identification label "defect" is stored in the memory unit 12 as a defective area. The image data assigned with the identification label "normal" is stored in the memory unit 12 as a normal area.

[0103] If a cord defect appears in the defect area candidate, the worker may input the type of cord defect as an identification label using the operation unit 14. For example, if a shape similar to the cord defect shown in FIG. 4A appears in the defect area candidate, the determination result receiving unit 11h assigns an identification label of "Type 1 defect" to this defect area candidate. Similarly, for other types of cord defects, the determination result receiving unit 11h assigns an identification label such as "Type 2 defect" or "Type 3 defect" to each defect area candidate. The number of defect types is not limited to this. The number of defect types may be two or more than three. Types of cord defects may be specified separately for defects occurring in the carcass cord 91a and defects occurring in the belt cord 92a.

[0104] [Normal area extraction part] The normal region extraction unit 111i extracts, as normal regions, portions of regions not selected as defective region candidates in the processing by the candidate extraction unit 111g from each of the carcass image and the belt image. For example, the normal region extraction unit 111i randomly extracts images having a predetermined number of pixels and a predetermined aspect ratio from each of the carcass image and the belt image. Then, if the randomly extracted image does not have any overlapping portions with the defective region candidates, the normal region extraction unit 111i determines the randomly extracted image as a normal region. The normal region extraction unit 111i assigns an identification label "normal" to the extracted image and stores it in the storage unit 12.

[0105] [Model generation section] The model generation unit 111j uses the image data (image data of defective areas and image data of normal areas) generated by the training data generation unit 111e as training data to generate identification models M3 and M4 for detecting defects in the internal structure of the tire.

[0106] The model generation unit 111j generates a first identification model M3 using training data obtained from the belt image, and generates a second identification model M4 using training data obtained from the carcass image.

[0107] For example, neural networks may be used as the discriminant models M3 and M4. Convolutional neural networks (CNNs) may be used as the discriminant models M3 and M4. Alternatively, support vector machines (SVMs) or random forests may be used as the discriminant models M3 and M4.

[0108] [Tire inspection] The following describes the process for inspecting a tire using a belt image and a carcass image of the tire to be inspected. The defect area candidate extraction unit 111u and the defect detection unit 111v perform the inspection process.

[0109] [Defect area candidate extraction section] The defect area candidate extraction unit 111u extracts defect area candidates from the image data (that is, the carcass image and the belt image) of the tire 90 to be inspected, which is obtained by the pre-processing unit 11b.

[0110] Like the candidate extractor 111g, the candidate defect area extractor 111u extracts image data of candidate defect areas from the belt image based on, for example, the area of ​​the belt cord 92a appearing in the belt image. Alternatively, the candidate defect area extractor 111u may search for a branching shape of the belt cord 92a in the belt image and extract an area including the branching shape as a candidate defect area.

[0111] The defect area candidate extraction unit 111u also extracts defect area candidates from the carcass image. For example, the defect area candidate extraction unit 111u extracts image data of defect area candidates from the carcass image based on the area of ​​the carcass cord 91a appearing in the carcass image. The defect area candidate extraction unit 111u may also search for the branch shape of the carcass cord 91a in the carcass image and extract an area including the branch shape as a defect area candidate.

[0112] [Defect detection section] The defect detection unit 111v inputs defect area candidates obtained from the belt image into a first identification model M3 and classifies the defect area candidates into multiple classes. The defect detection unit 111v also inputs defect area candidates extracted from the carcass image into a second identification model M4 and classifies the defect area candidates into multiple classes.

[0113] Here, the multiple classes are, for example, a class in which a cord defect appears in the defect area candidate (a class to which an identification label "defect" is assigned), or a class in which a cord defect does not appear in the defect area candidate (a class to which an identification label "normal" is assigned). By such processing by the defect detection unit 111v, defects in the belt layer 92 and defects in the carcass layer 91 can be detected. The output of the defect detection unit 111v is, for example, the probability that the defect area candidate falls into each class. Alternatively, the output of the defect detection unit 111v may be the identification label of the class to which the defect area candidate has the highest probability of falling.

[0114] As described above, when generating the identification models M3 and M4, image data to which identification labels indicating the types of defects (e.g., "Type 1 defect" or "Type 2 defect") are assigned may be used as training data. In this case, the multiple classes classified by the defect detection unit 111v may be multiple classes corresponding to the types of code defects appearing in the defect area candidates.

[0115] The defect detection unit 11v displays the output from the identification models M3 and M4, for example, together with an image of the defect area candidate on the display unit 13 or stores it in the memory unit 12. If multiple defect area candidates for one tire 90 are stored in the memory unit 12, the defect detection unit 111v may input all of the multiple defect area candidates to the identification models M3 and M4 in order and display each of the outputs on the display unit 13 or store them in the memory unit 12.

[0116] [Variations] The tire inspection device proposed in the present disclosure is not limited to the first and second embodiments described above.

[0117] For example, in the tire inspection device described above, defective area candidates are extracted from an image of the tire 90 to be inspected, and only the image data of the defective area candidates is input to the identification model. Alternatively, for example, the image of the tire 90 to be inspected (such as a carcass center image, a carcass side image, a carcass image, and a belt image) may be divided into multiple parts, and all of these parts may be input to the identification models M1, M2, M3, and M4.

[0118] In the first embodiment, the image data of the defect area candidates extracted from the carcass center image and the image data of the defect area candidates extracted from the carcass side image are input to two identification models M1 and M2, respectively. Alternatively, these two types of image data may be input to a common identification model. In the second embodiment, the image data of the defect area candidates extracted from the carcass image and the image data of the defect area candidates extracted from the belt image may be input to a common identification model. [Explanation of symbols]

[0119] 10: tire inspection device, 11·111: control unit, 11a·111a: image data acquisition unit, 11b·111b: pre-processing unit, 11c: carcass center image extraction unit, 11d: carcass side image extraction unit, 11e·111e: training data generation unit, 11f·111f: defect area extraction unit, 11g·111g: candidate extraction unit, 11h·111h: judgment result reception unit, 11i·111i: normal area extraction unit, 11j·111j: model generation unit, 11u·111u: defect area candidate extraction unit, 11v·11 1v: defect detection unit, 12: memory unit, 13: display unit, 14: operation unit, 15: imaging unit, 16: X-ray irradiation unit, 17: support unit, 90: tire, 90L·00R: sidewall unit, 90T: tread unit, 91: carcass layer, 91A: carcass center unit, 91B: carcass side unit, 91a: carcass cord, 92: belt layer, 92a: belt cord, 93: finishing layer, 93a: finishing cord, 94: bead, 111p: belt image extraction unit, 111q: carcass image extraction unit.

Claims

1. A tire inspection device for inspecting a tire including a belt layer including a belt cord and a carcass layer including a carcass cord, the carcass layer having a carcass center portion that is a portion covered by the belt layer, image data acquisition means for acquiring image data of an X-ray image of the tire to be inspected; a pre-processing means including a carcass image extraction means for removing a frequency component corresponding to the belt cord from the image data of the tire to be inspected by a two-dimensional Fourier transform; and a belt image extraction means for removing a frequency component corresponding to the carcass cord from the image data of the tire to be inspected by a two-dimensional Fourier transform. a part or all of the image data obtained by the carcass image extraction means is input into a carcass defect identification model generated by machine learning using training data including image data of a defective region, which is a region including a defect in the carcass layer, from which frequency components corresponding to the belt cord have been removed, and image data of a normal region, which is a region different from the defective region, from which frequency components corresponding to the belt cord have been removed by the two-dimensional Fourier transform, to detect defects in the carcass layer; a defect detection means for detecting a defect in the belt layer by inputting a part or all of the image data obtained by the belt image extraction means into a belt defect identification model generated by machine learning using training data including image data of a defective region, which is a region including a defect in the belt layer, from which a frequency component corresponding to the carcass cord has been removed, and image data of a normal region, which is a region different from the defective region, from which a frequency component corresponding to the carcass cord has been removed by the two-dimensional Fourier transform; A tire inspection device having:

2. The method further includes a defect area candidate extraction means for extracting image data of a carcass defect area candidate, which is a candidate for an area including a defect in the carcass layer, from the image data obtained by the carcass image extraction means, and extracting image data of a belt defect area candidate, which is a candidate for an area including a defect in the belt layer, from the image data obtained by the belt image extraction means, The defect detection means inputs image data of the carcass defect area candidate as part of the image data obtained by the carcass image extraction means into the carcass defect identification model, and determines whether or not a defect is included in the carcass defect area candidate, and inputs image data of the belt defect area candidate as part of the image data obtained by the belt image extraction means into the belt defect identification model, and determines whether or not a defect is included in the belt defect area candidate.

2. A tire inspection device according to claim 1.

3. The defect area candidate extraction means extracts image data of the carcass defect area candidate based on at least one of the area and shape of the carcass cord appearing in the image obtained by the carcass image extraction means, and extracts image data of the belt defect area candidate based on at least one of the area and shape of the belt cord appearing in the image obtained by the belt image extraction means.

3. A tire inspection device according to claim 2.

4. the carcass layer defects include a first carcass defect type and a second carcass defect type; the carcass defect identification model is generated using, as the training data, image data of the defect area including a defect of the first carcass defect type, image data of the defect area including a defect of the second carcass defect type, and image data of the normal area; The defect detection means inputs the part or all of the image data obtained by the carcass image extraction means into the carcass defect identification model, and identifies the type of defect in the carcass layer.

2. A tire inspection device according to claim 1.

5. the carcass layer has a carcass side portion that is a portion not covered by the belt layer, the tire inspection device further comprises a carcass center image extraction means for extracting image data of the carcass center from the image data of the inspection target tire obtained by the image data acquisition means, and a carcass side image extraction means for extracting image data of the carcass side from the image data of the inspection target tire obtained by the image data acquisition means, the carcass defect identification model includes a first carcass defect identification model for detecting defects in the carcass center portion and a second carcass defect identification model for detecting defects in the carcass side portion; The defect detection means inputs a part or all of the image data of the carcass center portion into the first carcass defect identification model, and inputs a part or all of the image data of the carcass side portion into the second carcass defect identification model.

2. A tire inspection device according to claim 1.

6. The inspection target tire has the carcass layer, the belt layer, and a finishing layer, The defect detection means detects defects in the finishing layer in addition to the carcass layer and the belt layer.

2. A tire inspection device according to claim 1.

7. A tire inspection method for inspecting a tire including a belt layer including a belt cord and a carcass layer including a carcass cord, the carcass layer having a carcass center portion that is a portion covered by the belt layer, an image data acquisition step of acquiring image data of an X-ray image of the tire to be inspected; a pre-processing step including a carcass image extraction step of removing frequency components corresponding to the belt cord from the image data of the tire to be inspected by two-dimensional Fourier transform; and a belt image extraction step of removing frequency components corresponding to the carcass cord from the image data of the tire to be inspected by two-dimensional Fourier transform; a carcass defect identification model generated by machine learning using training data including image data of a defect area, which is an area including a defect in the carcass layer, from which frequency components corresponding to the belt cord have been removed, and image data of a normal area, which is an area different from the defect area, from which frequency components corresponding to the belt cord have been removed by the two-dimensional Fourier transform; and a part or all of the image data obtained by the carcass image extraction step is input into the carcass defect identification model to detect defects in the carcass layer; a defect detection step of detecting a defect in the belt layer by inputting a part or all of the image data obtained in the belt image extraction step into a belt defect identification model generated by machine learning using training data including image data of a defective region, which is a region including a defect in the belt layer, from which frequency components corresponding to the carcass cord have been removed, and image data of a normal region, which is a region different from the defective region, from which frequency components corresponding to the carcass cord have been removed by the two-dimensional Fourier transform; A tire inspection method comprising:

8. A program that causes a computer to function as a tire inspection device that inspects a tire that includes a belt layer including belt cords and a carcass layer including carcass cords, and that has a carcass center portion that is a portion of the carcass layer covered by the belt layer, image data acquisition means for acquiring image data of an X-ray image of the tire to be inspected; a pre-processing means including a carcass image extraction means for removing a frequency component corresponding to the belt cord from the image data of the tire to be inspected by a two-dimensional Fourier transform; and a belt image extraction means for removing a frequency component corresponding to the carcass cord from the image data of the tire to be inspected by a two-dimensional Fourier transform; a part or all of the image data obtained by the carcass image extraction means is input into a carcass defect identification model generated by machine learning using training data including image data of a defective region, which is a region including a defect in the carcass layer, from which frequency components corresponding to the belt cord have been removed, and image data of a normal region, which is a region different from the defective region, from which frequency components corresponding to the belt cord have been removed by the two-dimensional Fourier transform, to detect defects in the carcass layer; a defect detection means for detecting a defect in the belt layer by inputting a part or all of the image data obtained by the belt image extraction means into a belt defect identification model generated by machine learning using training data including image data of a defective region, which is a region including a defect in the belt layer, from which frequency components corresponding to the carcass cord have been removed, and image data of a normal region, which is a region different from the defective region, from which frequency components corresponding to the carcass cord have been removed by the two-dimensional Fourier transform; A program that makes a computer function as a

9. 1. A method for generating an identification model used in a tire inspection device for inspecting a tire including a belt layer including a belt cord and a carcass layer including a carcass cord, the carcass layer having a carcass center portion that is a portion covered by the belt layer, an image data acquisition step of acquiring image data of an X-ray image of the tire; a pre-processing step including a carcass image extraction step of removing frequency components corresponding to the belt cord from the tire image data by two-dimensional Fourier transform, and a belt image extraction step of removing frequency components corresponding to the carcass cord from the image data of the tire to be inspected by two-dimensional Fourier transform; a defect area extraction step of extracting image data of a defect area, which is an area including a defect in the carcass layer, from the image data acquired in the carcass image extraction step, and extracting image data of a defect area, which is an area including a defect in the belt layer, from the image data acquired in the belt image extraction step; a normal region extraction step of extracting image data of a normal region, which is a region different from the defective region of the carcass layer, from the image data acquired by the carcass image extraction step, and extracting image data of a normal region, which is a region different from the defective region of the belt layer, from the image data acquired by the belt image extraction step; a model generation step of generating a carcass defect identification model for detecting defects in the carcass layer using image data of the defect area of ​​the carcass layer and image data of the normal area of ​​the carcass layer as training data, and generating a belt defect identification model for detecting defects in the belt layer using image data of the defect area of ​​the belt layer and image data of the normal area of ​​the belt layer as training data; A method for generating a discriminative model including:

10. The defect area extraction step includes: a candidate extraction step of extracting image data of a candidate defect area, which is an area that may contain a defect in the carcass layer, from the image data of the tire; and a determination result receiving step of receiving a determination by an operator as to whether or not a defect in the carcass layer is depicted in the candidate defect area. a candidate extraction step of extracting image data of a candidate defect area, which is an area that may contain a defect in the belt layer, from the image data of the tire; and a determination result receiving step of receiving a determination by an operator as to whether or not a defect in the belt layer is depicted in the candidate defect area, In the model generation step, image data of the defect area candidate determined by the operator to depict a defect in the carcass layer or the belt layer is used as image data of the defect area. The method for generating a discriminative model according to claim 9.

11. the carcass layer has a carcass side portion that is a portion not covered by the belt layer, In the defect area extraction step, image data of the defect area of ​​the carcass layer is extracted from image data of the carcass center portion, and image data of the defect area of ​​the carcass layer is extracted from image data of the carcass side portion, In the model generation step, a first carcass defect identification model is generated using image data of the defect area extracted from image data of the carcass center portion, and a second carcass defect identification model is generated using image data of the defect area extracted from image data of the carcass side portion. The method for generating a discriminative model according to claim 9.

12. 1. A device for generating an identification model used in a tire inspection device that inspects a tire including a belt layer including belt cords and a carcass layer including carcass cords, the carcass layer having a carcass center portion that is a portion covered by the belt layer, an image data acquisition means for irradiating a tire with X-rays and acquiring image data of the tire; a pre-processing means including a carcass image extraction means for removing a frequency component corresponding to the belt cord from the tire image data by a two-dimensional Fourier transform; and a belt image extraction means for removing a frequency component corresponding to the carcass cord from the tire image data by a two-dimensional Fourier transform; a defect area extraction means for extracting image data of a defect area, which is an area including a defect in the carcass layer, from the image data acquired by the carcass image extraction means, and for extracting image data of a defect area, which is an area including a defect in the belt layer, from the image data acquired by the belt image extraction means; a normal region extraction means for extracting image data of a normal region, which is a region different from the defective region of the carcass layer, from the image data acquired by the carcass image extraction means, and for extracting image data of a normal region, which is a region different from the defective region of the belt layer, from the image data acquired by the belt image extraction means; a model generating means for generating a carcass defect identification model for detecting defects in the carcass layer by using image data of the defective area of ​​the carcass layer and image data of the normal area of ​​the carcass layer as training data, and for generating a belt defect identification model for detecting defects in the belt layer by using image data of the defective area of ​​the belt layer and image data of the normal area of ​​the belt layer as training data; A device for generating a discriminative model including:

13. A program that causes a computer to function as a device for generating an identification model used in a tire inspection device that inspects a tire that includes a belt layer including belt cords and a carcass layer including carcass cords, the carcass layer having a carcass center portion that is a portion covered by the belt layer, image data acquisition means for acquiring image data of an X-ray image of a tire; a carcass image extraction means for removing a frequency component corresponding to the belt cord from the tire image data by two-dimensional Fourier transform; and a belt image extraction means for removing a frequency component corresponding to the carcass cord from the tire image data by two-dimensional Fourier transform. a defect area extraction means for extracting image data of a defect area, which is an area including a defect in the carcass layer, from the image data acquired by the carcass image extraction means, and for extracting image data of a defect area, which is an area including a defect in the belt layer, from the image data acquired by the belt image extraction means; a normal region extraction means for extracting image data of a normal region, which is a region different from the defective region of the carcass layer, from the image data acquired by the carcass image extraction means, and for extracting image data of a normal region, which is a region different from the defective region of the belt layer, from the image data acquired by the belt image extraction means; and a model generating means for generating a carcass defect identification model for detecting defects in the carcass layer by using image data of the defective area of ​​the carcass layer and image data of the normal area of ​​the carcass layer as training data, and for generating a belt defect identification model for detecting defects in the belt layer by using image data of the defective area of ​​the belt layer and image data of the normal area of ​​the belt layer as training data; A program that makes a computer function as a

Images