JP7428870B2 - Tire wear degree estimation device, tire wear degree learning device, tire wear degree estimation method, learned model generation method and program - Google Patents

Description

The present invention relates to a tire wear degree estimation device, a tire wear degree learning device, a tire wear degree estimation method, a learned model generation method, and a program.

As an example of a technique for estimating the degree of tire wear, Patent Document 1 describes a recognition method and a recognition device that can recognize tire types, wear conditions, etc. from tire images using a convolutional neural network.

Further, Patent Document 2 describes a technique for calculating the degree of wear of a part from an approximation line to surface shape data including information indicating the surface shape of the worn part and a worn portion included in the surface shape data.

JP2019-35626A JP2017-187418A

Since tire tread patterns vary depending on the type, the technology described in Patent Document 1 requires learning a huge amount of data in order to accurately estimate the degree of wear for various types of tires. be. Note that the technique described in Patent Document 2 does not use a machine learning model in the first place.

The present invention has been made in view of the above circumstances, and one of its objects is a tire wear degree estimation device that can accurately estimate the wear degree of various types of tires even if the amount of data used for learning is small. The present invention provides a tire wear degree learning device, a tire wear degree estimation method, a learned model generation method, and a program.

In order to solve the above problems, a tire wear degree estimating device according to the present invention includes a tread surface image acquisition means for acquiring a tread surface image in which a tread surface of a tire is shown, and a tread surface image acquisition means for acquiring a tread surface image in which a tread surface of a tire is shown; tread surface area of interest specifying means for specifying a tread surface area of interest; target input image generating means for generating a target input image that is an image of the tread surface area of interest with a predetermined number of pixels; and machine learning that has already learned the target input image. and estimating means for estimating the degree of wear of the tire based on the output when input to the model.

In one aspect of the present invention, the tread surface slip sign is shown at a given position within the tread surface region of interest.

Further, in one aspect of the present invention, the invention further includes a side image acquisition unit that acquires a side image that is an image showing a side surface of the tire, and the tread surface area of interest specifying unit is configured to locate a side slip sign that is reflected in the side surface image. The region of interest on the tread surface is specified based on the above.

Further, in one aspect of the present invention, the tread surface region of interest specifying means identifies the tread surface region of interest of the predetermined number of pixels, and the target input image generating means determines the tread surface region of interest from the tread surface image. The target input image is generated by extracting the target input image.

Alternatively, the target input image generation means generates the target input image by enlarging or reducing an image obtained by extracting the tread surface region of interest from the tread surface image to an image of the predetermined number of pixels.

Further, in one aspect of the present invention, the tread surface image includes a tread pattern of the tire.

Moreover, in one aspect of the present invention, the tread surface image is a depth image or a three-dimensional image.

Further, the tire wear degree learning device according to the present invention is a tire wear degree learning device that executes learning of a machine learning model used for estimating the degree of tire wear. A learning input image acquisition means for acquiring a learning input image of a predetermined number of pixels showing the tread surface of a tire, and an output when the learning input image is input to the machine learning model. A learning means for performing learning.

In one aspect of the present invention, the learning input image includes a tread pattern of the tire.

Further, in one aspect of the present invention, the learning input image is a depth image or a three-dimensional image.

Further, the tire wear degree estimation method according to the present invention includes the steps of: acquiring a tread surface image in which the tread surface of the tire is shown; and identifying a tread surface region of interest in which the tread surface slip sign is shown in the tread surface image; a step of generating a target input image that is an image of the tread surface region of interest with a predetermined number of pixels; and estimating the degree of wear of the tire based on an output when the target input image is input to a trained machine learning model. The method includes the steps of:

Further, a method for generating a trained model according to the present invention is a method for generating a trained model for executing learning of a machine learning model used for estimating the degree of wear of a tire. A step of acquiring a learning input image of a predetermined number of pixels that shows the tread surface of the tire, and learning of the machine learning model using the output when the learning input image is input to the machine learning model. and performing steps.

Further, the program according to the present invention includes a procedure for acquiring a tread surface image in which the tread surface of a tire is captured, a procedure for specifying a tread surface area of interest in which a tread surface slip sign is captured in the tread surface image, and a procedure for identifying a tread surface region of interest in which a tread surface slip sign is captured in the tread surface image, and A computer executes a step of generating a target input image that is an image of the area of interest, and a step of estimating the degree of wear of the tire based on the output when the target input image is input to a trained machine learning model. let

In addition, another program according to the present invention allows a computer that executes learning of a machine learning model used for estimating the degree of wear of a tire to detect a tire tread surface in which a tread surface slip sign is reflected at a given position. A procedure for acquiring a learning input image with a predetermined number of pixels, and a procedure for executing learning of the machine learning model using an output when the learning input image is input to the machine learning model.

1 is a diagram illustrating an example of the configuration of an information processing device according to an embodiment of the present invention. FIG. 3 is a diagram schematically showing an example of a left side surface of a vehicle. 3 is a diagram schematically showing an example of the top surface of the vehicle shown in FIG. 2. FIG. It is a figure which shows an example of a side image. It is a figure which shows an example of a tread surface image. FIG. 3 is a diagram showing an example of a target input image. FIG. 1 is a functional block diagram showing an example of functions of an information processing device according to an embodiment of the present invention. FIG. 3 is a flow diagram showing an example of the flow of learning processing performed by the information processing device according to an embodiment of the present invention. FIG. 2 is a flow diagram showing an example of the flow of estimation processing performed by the information processing device according to an embodiment of the present invention. It is a figure which shows an example of a tread surface image. It is a figure which shows an example of a side image.

Hereinafter, one embodiment of the present invention will be described with reference to the drawings.

FIG. 1 is a diagram showing an example of the configuration of an information processing device 10 according to an embodiment of the present invention. The information processing device 10 according to this embodiment is a computer such as a personal computer. As shown in FIG. 1, the information processing device 10 includes, for example, a processor 12, a storage section 14, a communication section 16, a display section 18, an operation section 20, a first photographing section 22a, and a second photographing section 22b.

The processor 12 is a program-controlled device such as a CPU that operates according to a program installed in the information processing device 10, for example.

The storage unit 14 is a storage element such as ROM or RAM, a hard disk drive, or the like. The storage unit 14 stores programs and the like executed by the processor 12.

The communication unit 16 is, for example, a communication interface such as a network board.

The display unit 18 is a display device such as a liquid crystal display, and displays various images according to instructions from the processor 12.

The operation unit 20 is a user interface such as a keyboard or a mouse, and receives user operation input and outputs a signal indicating the content to the processor 12.

The first photographing unit 22a is a photographing device such as a digital camera, and outputs an image generated by photographing to the processor 12.

The second photographing unit 22b is a photographing device such as a stereo camera capable of photographing a three-dimensional image or a depth camera capable of photographing a depth image, and outputs a three-dimensional image and a depth image generated by photographing to the processor 12. .

Note that the information processing device 10 may include an optical disk drive for reading optical disks such as a DVD-ROM or a Blu-ray (registered trademark) disk, a USB (Universal Serial Bus) port, and the like.

FIG. 2 is a diagram schematically showing an example of the left side surface of a vehicle 30 equipped with tires whose degree of wear is estimated. FIG. 3 is a diagram schematically showing an example of the top surface of the vehicle 30 shown in FIG. 2. As shown in FIG. The vehicle 30 is, for example, parked in a parking lot.

As shown in FIGS. 2 and 3, in this embodiment, the first photographing section 22a is arranged at a position where the left side surface of the left rear wheel 32 of the vehicle 30 can be photographed. Further, the second photographing section 22b is arranged at a position where the tread surface of the left rear wheel 32 of the vehicle 30 can be photographed.

FIG. 4 is a diagram illustrating an example of a side image 34, which is an image of the left side surface of the tire, photographed by the first photographing section 22a. As shown in FIG. 4, the side image 34 shows a side slip sign (hereinafter referred to as a side slip sign 36) stamped on the side surface (shoulder portion) of the tire. As shown in FIG. 4, side slip signs 36 are stamped on the side surface of the tire at equal intervals along the outer circumference of the tire. Generally, four to nine side slip signs 36 are stamped on the side surface of the tire at equal intervals along the outer circumference of the tire. The tire shown in FIG. 4 has six side slip signs 36 stamped on it.

In this embodiment, it is confirmed whether the side slip sign 36 is included in the side surface area of interest 38 in the side image 34 . Here, the side surface area of interest 38 is an area corresponding to the angle of view (photographing range) of the second photographing section 22b, and includes the position of the side surface area of interest 38 in the side image 34, and the shape and size of the side surface area of interest 38. is predetermined.

When it is confirmed that the side slip sign 36 is captured in the side surface area of interest 38 in the side image 34, the second imaging unit 22b creates a depth image or a three-dimensional image showing the tread surface of the left rear wheel 32 of the vehicle 30. A tread surface image 40 illustrated in FIG. 5, which is an image, is photographed.

The tread surface image 40 includes a tread surface slip sign (hereinafter referred to as a tread surface slip sign 42). The tread surface slip sign 42 is provided in the main groove 44 near the side slip sign 36. In this embodiment, as described above, since the side slip sign 36 is shown in the side surface area of interest 38, the tread surface slip sign 42 is shown in the tread surface image 40.

The tread surface image 40 may include a tread pattern. For example, the periodic structure of the tread pattern may be shown in the tread surface image 40. Note that in the tread surface image 40 shown in FIG. 5, the description of the tread pattern is omitted.

In this embodiment, based on the position of the side slip sign 36 shown in the side image 34, a tread surface area of interest 46 in which the tread surface slip sign 42 and the surface of the tread surface are shown in the tread surface image 40 is specified. The tread surface region of interest 46 is a strip-shaped region (in the example of FIG. 5, a horizontally long strip-shaped region).

The shape and size of the tread surface region of interest 46 within the tread surface image 40 may be determined in advance. Further, the aspect ratio of the tread surface region of interest 46 may be determined in advance. Furthermore, the length of the tread surface region of interest 46 in the longitudinal direction may be predetermined.

The position of the side slip sign 36 within the tire reflected in the side image 34 and the position of the tread surface region of interest 46 within the tread surface image 40 are in one-to-one correspondence. For example, a reference line passing through the center of the left rear wheel 32 reflected in the side image 34 (in the example of FIG. 4, a line in the right direction) is connected to the center and the side slip sign 36 disposed within the side surface area of interest 38. Let the angle between the line and the line be a. Further, the length from the upper side of the tread surface image 40 to the center of the tread surface region of interest 46 is defined as b. In this case, the length b is uniquely specified based on the angle a.

Here, for example, the relationship between the angle a and the length b is such that the length from the upper side of the tread surface image 40 to the center of the tread surface slip sign 42 when the above-mentioned angle is a is the above-mentioned length b. may be set in advance. For example, the relationship between the angle a and the length b is such that the length from the upper side of the tread surface image 40 to the center of the central tread surface slip sign 42 when the above-mentioned angle is a is the above-mentioned length b. may be set in advance. Then, the length b may be uniquely specified based on the relationship and the angle a. In this way, the tread surface slip sign 42 may be shown at a given position within the tread surface region of interest 46.

In this embodiment, a target input image 48 illustrated in FIG. 6, which is an image of a tread surface region of interest 46 having a predetermined number of pixels, is generated. Here, for example, by extracting a tread surface region of interest 46 of a predetermined number of pixels from the tread surface image 40, a target input image 48 of a predetermined number of pixels may be generated. Alternatively, the target input image 48 with a predetermined number of pixels may be generated by enlarging or reducing an image in which the tread surface region of interest 46 is extracted from the tread surface image 40 to an image with a predetermined number of pixels.

In this embodiment, the wear degree of the tire of the left rear wheel 32 is estimated based on the output when the target input image 48 is input into a trained machine learning model.

Hereinafter, learning of the machine learning model implemented in the information processing device 10 and estimation of the degree of tire wear by the information processing device 10 will be further described.

FIG. 7 is a functional block diagram showing an example of functions implemented in the information processing device 10 according to the present embodiment. Note that the information processing apparatus 10 according to the present embodiment does not need to implement all of the functions shown in FIG. 7, and functions other than the functions shown in FIG. 7 may be implemented.

As shown in FIG. 7, the information processing device 10 functionally includes, for example, a machine learning model 50, a learning data storage unit 52, a learning data acquisition unit 54, a learning unit 56, a side image acquisition unit 58, and a side view area of interest. It includes a specifying section 60, a side slip sign position specifying section 62, a tread surface image obtaining section 64, a tread surface region of interest specifying section 66, a target input image generating section 68, and an estimating section 70.

The machine learning model 50 is mainly implemented using the processor 12 and the storage unit 14. The learning data storage section 52 is mainly implemented using the storage section 14 . The learning data acquisition unit 54, the learning unit 56, the side surface area of interest identification unit 60, the side slip sign position identification unit 62, the tread surface area of interest identification unit 66, the target input image generation unit 68, and the estimation unit 70 are mainly implemented with the processor 12. be done. The side image acquisition section 58 is mainly implemented with the processor 12 and the first photographing section 22a. The tread surface image acquisition unit 64 is mainly implemented with the processor 12 and the second imaging unit 22b.

The above functions may be implemented by having the processor 12 execute a program that is installed in the information processing device 10, which is a computer, and includes commands corresponding to the above functions. This program may be supplied to the information processing device 10 via a computer-readable information storage medium such as an optical disk, a magnetic disk, a magnetic tape, a magneto-optical disk, or a flash memory, or via the Internet. .

The functions of the machine learning model 50, the learning data storage unit 52, the learning data acquisition unit 54, and the learning unit 56 in the information processing device 10 are tire wear models that execute learning of the machine learning model 50 used to estimate the degree of tire wear. This corresponds to the function of a degree learning device.

Further, in the information processing device 10, a machine learning model 50, a side image acquisition section 58, a side surface region of interest specifying section 60, a side slip sign position specifying section 62, a tread surface image obtaining section 64, a tread surface region of interest specifying section 66, a target input The functions of the image generation section 68 and the estimation section 70 correspond to the function of a tire wear degree estimating device that estimates the degree of tire wear.

In this embodiment, the machine learning model 50 is a machine learning model in which machine learning such as Adaboost, random forest, neural network, support vector machine (SVM), and nearest neighbor classifier is implemented.

The machine learning model 50 according to the present embodiment may be, for example, a determination model that estimates whether the degree of tire wear is normal or abnormal. In this case, the machine learning model 50 may output, for example, 1-bit data indicating whether the degree of tire wear is normal or abnormal, depending on the input.

Further, the machine learning model 50 may be a regression model that outputs data indicating the degree of tire wear. Here, for example, the degree of tire wear may be an index indicating the extent to which the tire is worn, which increases as the amount of remaining groove on the tire decreases. Further, the data indicating the degree of wear of the tire may be data indicating the amount of remaining groove of the tire itself. In this case, the machine learning model 50 may output, for example, an estimated value of the degree of tire wear according to the input.

In this embodiment, the learning data storage unit 52 stores, for example, learning data used for learning the machine learning model 50. The learning data includes a learning input image that is an image of a predetermined number of pixels showing the tread surface of a tire, and given teacher data indicating the degree of wear of the tire shown in the learning input image.

The learning input image is, for example, an image of a predetermined number of pixels showing the tread surface of a tire and showing the tread surface slip sign 42 at a given position, similar to the target input image 48 illustrated in FIG. 6 . Here, the number of pixels of the learning input image is the same as the number of pixels of the target input image 48 described above.

Here, for example, the learning input image is an image in which the center of any tread surface slip sign 42 (for example, the center tread surface slip sign 42) is placed on a line extending in the left-right direction passing through the center of the learning input image. It may be.

Further, for example, the learning input image may be generated based on an image similar to the tread surface image 40 illustrated in FIG. 5 . For example, the learning input image may be generated by extracting a region of a predetermined number of pixels corresponding to the tread surface region of interest 46 from an image similar to the tread surface image 40. Further, for example, a learning input image having a predetermined number of pixels may be generated by enlarging or reducing an image of a region corresponding to the tread surface region of interest 46 extracted from an image similar to the tread surface image 40.

Here, for example, the learning input image may be generated using a method similar to the method for generating the target input image 48 described above. Further, for example, the learning input image may be generated based on an image similar to the tread surface image 40 by the operator performing an image editing operation while visually viewing the image.

The learning input image may be a depth image or a three-dimensional image. For example, when the target input image 48 is a depth image, the learning data storage unit 52 stores learning data including the learning input image that is a depth image. Further, when the target input image 48 is a three-dimensional image, the learning data storage unit 52 stores learning data including the learning input image that is a three-dimensional image.

Note that the second photographing unit 22b may be a photographing device such as a digital camera, and the tread surface image 40 and the target input image 48 may be two-dimensional images. In this case, the learning data storage unit 52 stores learning data including a learning input image that is a two-dimensional image.

Furthermore, the learning input image may include a tread pattern. For example, the periodic structure of the tread pattern may be shown in the learning input image.

In this embodiment, the learning data acquisition unit 54 acquires learning data stored in the learning data storage unit 52, for example.

In this embodiment, the learning unit 56 inputs a learning input image to the machine learning model 50, for example. In this embodiment, the learning unit 56 executes learning of the machine learning model 50 using, for example, the output when the learning input image is input to the machine learning model 50. Here, for example, a difference between an output when a learning input image included in the learning data is input to the machine learning model 50 and teacher data included in the learning data may be specified. Then, supervised learning may be performed in which the values of the parameters of the machine learning model 50 are updated based on the identified difference.

In this embodiment, learning of the machine learning model 50 is performed using a plurality of learning data. Then, using the trained machine learning model 50 (trained model), the degree of wear of the tire reflected in the tread surface image 40 photographed by the second photographing section 22b is estimated.

In this embodiment, the side image acquisition unit 58 acquires, for example, a side image 34 that is an image showing the side surface of the tire.

In this embodiment, the side surface area of interest specifying unit 60 specifies, for example, the side surface area of interest 38 in the side image 34 .

In this embodiment, the side slip sign position specifying unit 62 specifies, for example, the position of the side slip sign 36 shown within the side surface area of interest 38.

In this embodiment, the tread surface image acquisition unit 64 acquires, for example, a tread surface image 40 that is an image showing the tread surface of a tire. Here, as described above, the tread surface image 40 may be a depth image or a three-dimensional image. Further, the tread surface image 40 may be a two-dimensional image.

In this embodiment, the tread surface region of interest specifying unit 66 specifies, for example, a tread surface region of interest 46 in which the tread surface slip sign 42 is captured in the tread surface image 40 . The tread surface region of interest identifying unit 66 identifies the tread surface region of interest 46 based on the position of the side slip sign 36 shown in the side image 34, for example. Here, the tread surface slip sign 42 may be captured at a given position within the tread surface region of interest 46.

In this embodiment, the target input image generation unit 68 generates the target input image 48, which is, for example, an image of the tread surface region of interest 46 having a predetermined number of pixels.

Here, for example, the tread surface region of interest specifying unit 66 may specify the tread surface region of interest 46 having a predetermined number of pixels. Then, the target input image generation unit 68 may generate the target input image 48 with a predetermined number of pixels by, for example, extracting the tread surface region of interest 46 with a predetermined number of pixels from the tread surface image 40.

Alternatively, the target input image generation unit 68 generates the target input image 48 with a predetermined number of pixels by enlarging or reducing the image of the tread surface region of interest 46 extracted from the tread surface image 40 to an image with a predetermined number of pixels. Good too.

In this embodiment, the estimating unit 70 estimates the degree of wear of the tires appearing in the target input image 48 based on the output when the target input image 48 is input to a trained machine learning model 50 (trained model), for example. do.

Here, when the machine learning model 50 is a judgment model, the machine learning model 50 may output 1-bit data indicating whether the degree of tire wear is normal or abnormal, depending on the input. . Further, when the machine learning model 50 is a regression model, the machine learning model 50 may output an estimated value of the degree of wear indicating the degree to which the tire is worn, depending on the input. Furthermore, the machine learning model 50 may output an estimated value of the amount of remaining groove in the tire according to the input.

Here, an example of the flow of the learning process performed by the information processing apparatus 10 according to the present embodiment will be described with reference to the flow diagram illustrated in FIG. 8. Here, it is assumed that a plurality of pieces of learning data are stored in the learning data storage section 52, for example.

First, the learning data acquisition unit 54 acquires one piece of learning data that has not been subjected to the process shown in S102 from among the plurality of pieces of learning data stored in the learning data storage unit 52 (S101).

Then, the learning unit 56 executes learning of the machine learning model 50 using the output when the learning input image included in the learning data acquired in the process shown in S101 is input to the machine learning model 50 (S102). . Here, for example, the values of the parameters of the machine learning model 50 may be updated based on the difference between the output and the teacher data included in the learning data acquired in the process shown in S101.

Then, the learning unit 56 checks whether the process shown in S102 has been executed for all of the plurality of learning data stored in the learning data storage unit 52 (S103).

If it is confirmed that the process shown in S102 has not been executed for all of the plurality of learning data stored in the learning data storage unit 52 (S103: N), the process returns to the process shown in S101.

If it is confirmed that the process shown in S102 has been executed for all of the plurality of learning data stored in the learning data storage unit 52 (S103: Y), the process shown in this processing example is ended.

Next, an example of the flow of the estimation process performed by the information processing apparatus 10 according to the present embodiment will be described with reference to the flow diagram illustrated in FIG. 9.

First, the side image acquisition section 58 instructs the first photographing section 22a to photograph the side image 34, and acquires the side image 34 photographed by the first photographing section 22a (S201).

Then, the side surface area of interest specifying unit 60 specifies the side surface area of interest 38 within the side image 34 (S202).

Then, the side slip sign position specifying unit 62 confirms whether the side slip sign 36 is reflected in the side surface area of interest 38 (S203).

If it is confirmed that the image is not captured (S203: N), the process shown in this process example is ended.

If it is confirmed that the side slip sign 36 is captured (S203: Y), the side slip sign position specifying unit 62 specifies the position of the side slip sign 36 that is captured in the side surface area of interest 38 (S204). Here, for example, the angle a shown in FIG. 4 may be specified.

Then, the tread surface image acquisition section 64 instructs the second photographing section 22b to photograph the tread surface image 40, and acquires the tread surface image 40 photographed by the second photographing section 22b (S205).

Then, the tread surface region of interest identifying section 66 identifies the tread surface region of interest 46 in the tread surface image 40 based on the position of the side slip sign 36 identified in the process shown in S204 (S206). Here, for example, the length b shown in FIG. 5 may be specified based on the above-mentioned angle a.

Then, the target input image generation unit 68 generates a target input image 48, which is an image of a predetermined number of pixels of the tread surface region of interest 46 specified in the process shown in S206 (S207).

Then, based on the output when the target input image 48 generated in the process shown in S207 is input into the trained machine learning model 50 (trained model), the estimation unit 70 calculates the size of the tire appearing in the target input image. The degree of wear is estimated (S208). Then, the processing shown in this processing example is ended.

In addition, in the above explanation, the degree of wear of the tire of the left rear wheel 32 included in the vehicle 30 was estimated, but the degree of wear of tires of tires other than the tire of the left rear wheel 32 included in the vehicle 30 can also be estimated by the same method. Of course it is possible.

Tire tread patterns vary depending on the type. Therefore, in order to accurately estimate the degree of wear using the machine learning model 50 for various types of tires using tread patterns as clues, it is necessary to learn a huge amount of data.

In this embodiment, the degree of tire wear is estimated using the target input image 48 near the tread surface slip sign 42 as input to the machine learning model 50, regardless of the type of tire. The tread surface slip sign 42 exists regardless of the type of tire. Furthermore, we believe that the differences in the characteristics of images near the tread surface slip sign 42 due to differences in the degree of tire wear are more noticeable than the differences in the characteristics of images near the tread surface slip sign 42 due to differences in tire types. It will be done.

Therefore, according to the present embodiment, the degree of wear of various types of tires can be accurately estimated even if the amount of data used for learning is small.

Further, according to the present embodiment, it is possible to estimate the amount of tire wear using a smartphone or the like that can take depth images and three-dimensional images without using a high-performance sensor.

Further, according to the present embodiment, by providing the first photographing section 22a and the second photographing section 22b at the parking position of the vehicle 30, it is possible to periodically estimate the amount of tire wear without human intervention. Can be done. Furthermore, by using this embodiment, it is possible to provide a service in which, for example, when the tires of the vehicle 30 reach a level of wear that requires replacement, the owner of the vehicle 30 is notified to that effect. becomes.

Note that the present invention is not limited to the above-described embodiments.

For example, in the present embodiment, the tread surface region of interest specifying unit 66 may specify, based on the tread surface image 40, a belt-shaped main groove region 80 in which the main groove 44 is shown, as illustrated in FIG. For example, the maximum depth among the depths of each pixel indicated by the tread surface image 40 may be specified. Then, a band-shaped region surrounding a pixel group whose difference from the maximum depth is less than or equal to a predetermined value may be specified as the main groove region 80. In the example of FIG. 10, six main groove regions 80 are identified: upper left, upper center, upper right, lower left, lower center, and lower right. In the example of FIG. 10, the main groove region 80 is a belt-shaped region extending in the vertical direction.

Then, the tread surface region of interest specifying unit 66 selects a portion of the tread surface of the tread surface that is within the tread surface region of interest 46 and is located on the extension line of the band of the main groove region 80 and has a smaller depth than the pixels within the main groove region 80. It may be specified as the tread surface slip sign area 82 where the slip sign 42 is shown.

Then, the tread surface region of interest identifying section 66 may correct the position of the tread surface region of interest 46 based on the tread surface slip sign region 82 . For example, the tread surface is focused so that the center of any tread surface slip sign area 82 (for example, the central tread surface slip sign area 82) is located on a line extending in the left-right direction passing through the center of the tread surface area of interest 46. The position of region 46 may be corrected.

Furthermore, the tread surface region of interest 46 may be specified by a method different from the method described above. Further, the target input image 48 may be generated based on the tread surface image 40 without using the side image 34.

Furthermore, in the present embodiment, the first photographing unit 22a may photograph a side image 84 illustrated in FIG. 11, which is an image of a portion corresponding to the side surface area of interest 38 illustrated in FIG. Then, based on the side image 84, the position of the tread surface region of interest 46 within the tread surface image 40 may be specified. In this case, the position of the side slip sign 36 in the side image 84 and the position of the tread surface region of interest 46 in the tread surface image 40 are in one-to-one correspondence. Therefore, based on the position of the side slip sign 36 in the side image 84, the position of the tread surface region of interest 46 in the tread surface image 40 can be uniquely specified.

Furthermore, for example, instead of an image of an actual tire tread surface, an image representing a tire tread surface generated through simulation may be used as the learning input image.

Further, the specific numerical values and character strings mentioned above and the specific numerical values and character strings in the drawings are merely examples, and the present invention is not limited to these numerical values and character strings.

Reference Signs List 10 information processing device, 12 processor, 14 storage unit, 16 communication unit, 18 display unit, 20 operation unit, 22a first imaging unit, 22b second imaging unit, 30 vehicle, 32 left rear wheel, 34 side image, 36 side Slip sign, 38 Side area of interest, 40 Tread surface image, 42 Tread surface slip sign, 44 Main groove, 46 Tread surface area of interest, 48 Target input image, 50 Machine learning model, 52 Learning data storage unit, 54 Learning data acquisition unit , 56 learning section, 58 side image acquisition section, 60 side surface area of interest identification section, 62 side slip sign position identification section, 64 tread surface image acquisition section, 66 tread surface area of interest identification section, 68 target input image generation section, 70 estimation part, 80 main groove region, 82 tread surface slip sign region, 84 side image.

Claims (11)

tread surface image acquisition means for acquiring a tread surface image that is a depth image or a three-dimensional image that includes depth information of each pixel that shows the tread surface of the tire;
In the tread surface image, the tread is a band-shaped region extending in the width direction of the tire, which is each one shallow part of the main groove of the tire, and in which a plurality of tread surface slip signs spaced apart in the width direction of the tire are captured. A surface area of interest is specified, and at least one of the plurality of tread surface slip signs is arranged in the tread surface area of interest on a line extending in the width direction of the tire passing through the center of the tread surface area of interest. , means for specifying a region of interest on a tread surface;
Target input image generation means for generating a target input image that is an image of the tread surface region of interest having a predetermined number of pixels;
Estimating means for estimating a degree of wear indicating the extent to which the tire is worn or the amount of remaining tread of the tire itself, based on an output when the target input image is input to a trained machine learning model;
A tire wear degree estimation device comprising: Further comprising a side image acquisition means for acquiring a side image that is an image showing the side surface of the tire,
The tread surface area of interest specifying means specifies the tread surface area of interest based on the position of the side slip sign shown in the side image.
The tire wear degree estimating device according to claim 1. The tread surface region of interest identifying means identifies the tread surface region of interest of the predetermined number of pixels,
The target input image generation means generates the target input image by extracting the tread surface region of interest from the tread surface image.
The tire wear degree estimating device according to claim 1 or 2, characterized in that: The target input image generation means generates the target input image by enlarging or reducing an image of the tread surface region of interest extracted from the tread surface image to an image of the predetermined number of pixels.
The tire wear degree estimating device according to claim 1 or 2, characterized in that: The tread surface image includes a tread pattern of the tire;
The tire wear degree estimating device according to any one of claims 1 to 4. A tire wear degree learning device that executes learning of a machine learning model used for estimating the degree of wear indicating the degree of wear of a tire or the amount of remaining tread of the tire itself, comprising:
a shallow portion of the main groove of the tire, each extending in the width direction of the tire with a predetermined number of pixels showing the tread surface of the tire, in which a plurality of tread surface slip signs spaced apart in the width direction of the tire are shown; A learning input image that is a depth image or a three-dimensional image containing information about the depth of each pixel in a strip shape is obtained, and the learning input image includes a line extending in the width direction of the tire passing through the center of the learning input image. Learning input image acquisition means, in which at least one of the plurality of tread surface slip signs is arranged;
learning means for executing learning of the machine learning model using an output when the learning input image is input to the machine learning model;
A tire wear degree learning device comprising: The learning input image includes a tread pattern of the tire.
7. The tire wear degree learning device according to claim 6. Obtaining a tread surface image that is a depth image or a three-dimensional image that includes depth information of each pixel that shows the tread surface of the tire;
In the tread surface image, each is a shallow portion of the main groove of the tire, and is a band-shaped region extending in the width direction of the tire, in which a plurality of tread surface slip signs spaced apart in the width direction of the tire are captured. A tread surface area of interest is specified, and at least one of the plurality of tread surface slip signs is arranged in the tread surface area of interest on a line extending in the width direction of the tire passing through the center of the tread surface area of interest. There are steps and
generating a target input image that is an image of the tread surface region of interest with a predetermined number of pixels;
estimating a degree of wear indicating the extent to which the tire is worn or the amount of remaining tread of the tire itself, based on an output when the target input image is input to a trained machine learning model;
A tire wear degree estimation method characterized by comprising: A method for generating a trained model that executes learning of a machine learning model used for estimating the degree of wear indicating the degree of wear of a tire or the amount of remaining tread of the tire itself, the method comprising:
a shallow portion of the main groove of the tire, each extending in the width direction of the tire with a predetermined number of pixels showing the tread surface of the tire, in which a plurality of tread surface slip signs spaced apart in the width direction of the tire are shown; A learning input image that is a depth image or a three-dimensional image containing information about the depth of each pixel in a strip shape is obtained, and the learning input image includes a line extending in the width direction of the tire passing through the center of the learning input image. a step on which at least one of the plurality of tread surface slip signs is arranged;
executing learning of the machine learning model using an output when inputting the learning input image to the machine learning model;
A method for generating a trained model, the method comprising: A procedure for acquiring a tread surface image that is a depth image or a three-dimensional image that includes depth information of each pixel, showing the tread surface of the tire;
In the tread surface image, each is a shallow portion of the main groove of the tire, and is a band-shaped region extending in the width direction of the tire, in which a plurality of tread surface slip signs spaced apart in the width direction of the tire are captured. A tread surface area of interest is specified, and at least one of the plurality of tread surface slip signs is arranged in the tread surface area of interest on a line extending in the width direction of the tire passing through the center of the tread surface area of interest. There are, procedures,
a step of generating a target input image that is an image of the tread surface region of interest having a predetermined number of pixels;
a step of estimating a degree of wear indicating the extent to which the tire is worn or the amount of remaining tread of the tire itself, based on the output when the target input image is input to a trained machine learning model;
A program that causes a computer to execute. A computer that executes learning of a machine learning model used for estimating the degree of wear indicating the degree of wear of the tire or the amount of remaining tread of the tire itself,
a shallow portion of the main groove of the tire, each extending in the width direction of the tire with a predetermined number of pixels showing the tread surface of the tire, in which a plurality of tread surface slip signs spaced apart in the width direction of the tire are shown; A learning input image that is a depth image or a three-dimensional image containing information about the depth of each pixel in a strip shape is obtained, and the learning input image includes a line extending in the width direction of the tire passing through the center of the learning input image. a step in which at least one of the plurality of tread surface slip signs is arranged;
a step of performing learning of the machine learning model using an output when the learning input image is input to the machine learning model;
A program characterized by executing.

Images