Tire footprint image generator

Abstract

To obtain footprint images of a tire more easily than conventional arts.SOLUTION: A CPU 11 (processor) executes a processing process, a setting process, and an extraction process. In the processing process, the CPU 11 processes a planar tread surface image from an exterior image of a tire 100 including a curved tread surface 110. In the setting process, the CPU 11 obtains an extraction range of the planar tread surface image. In the extraction process, the CPU 11 extracts part of the planar tread surface images based on the extraction range. The extracted image is output as a footprint image. In addition, in the setting process, the CPU 11 determines a contact patch width and a contact patch length of the tire 100 based on specification data of the tire 100 and vehicle weight data. Further, in the setting process, the CPU 11 determines the extraction range based on the contact patch width and the contact patch length.SELECTED DRAWING: Figure 2

Description

[[TECHNICAL FIELD]]

[0001] Disclosed herein is an apparatus for generating a tire footprint image.

[0002] In Patent Document 1, tire analysis is performed based on the tire footprint. A footprint refers to the contact patch of a tire. Therefore, a footprint image is an image of the contact patch of a tire.

[0003] For example, grooves are formed on the tire surface. In the contact patch of the tire, a pipe structure is formed by the tire grooves and the road surface. In this pipe structure, aerodynamic pipe resonance occurs, which causes noise to be generated from the tire. By acquiring a footprint image, the pipe structure such as pipe length can be analyzed. [[PRIOR ART DOCUMENTS]] [[PATENT DOCUMENTS]]

[0004] [[Patent Document 1]] Japanese Unexamined Patent Publication No. 2007-237751 [[SUMMARY OF THE INVENTION]] [[Problem to be Solved by the Invention]]

[0005] By the way, when acquiring a footprint, using a measuring instrument complicates the work process. For example, a worker mounts a tire on a vehicle. Then, the worker operates the vehicle so that the tire rests on the measuring instrument. The measuring instrument includes a plurality of pressure sensors. The measuring instrument obtains the pressure distribution in the contact patch of the tire. This pressure distribution is used as the footprint image.

[0006] Disclosed herein is an apparatus for generating a tire footprint image, which can acquire a tire footprint image more easily than conventional techniques. [[Means for Solving the Problem]]

[0007] This specification discloses a device for generating tire footprint images. This device comprises a processor. The processor performs processing, setting, and extraction processes. In the processing process, the processor processes a planar tread surface image from an external image of the tire, including a curved tread surface. In the setting process, the processor determines the extraction range of the planar tread surface image. In the extraction process, the processor extracts a portion of the planar tread surface image based on the extraction range. The extracted image is output as a footprint image. In the setting process, the processor determines the tire's contact width and contact length based on tire specifications and vehicle weight data. Furthermore, in the setting process, the processor determines the extraction range based on the contact width and contact length.

[0008] According to the above configuration, a footprint image is generated using the tire's appearance image, tire specifications, and the weight data of the vehicle on which the tire is mounted as input data.

[0009] Furthermore, in the above configuration, the processor performs the following processing during the processing stage: The processor determines the center line of the tire's external image. The processor then divides the tire's external image using the center line as a dividing line. Next, the processor performs a projection transformation on each of the divided images. Finally, the processor combines the divided images after the projection transformation to generate a planar tread surface image.

[0010] According to the above configuration, the curved tread surface is transformed into a planar surface. In other words, a projection transformation can generate an image similar to the tread surface at the contact surface.

[0011] Furthermore, in the above configuration, the processor performs the following processing during the setup process: The processor generates an extraction frame based on the grounding width and grounding length. The processor then performs radius machining on the extraction frame.

[0012] With the above configuration, an image closer to the actual footprint can be generated.

[0013] Furthermore, in the above configuration, during the projection transformation, the processor sets the vertices of the projection transformation. In doing so, the processor sets corresponding points on the tangents to the boundary edges of the curved tread surface included in the segmented image before the transformation.

[0014] The above configuration enables projection transformation according to the curvature of the tread surface.

[0015] Furthermore, in the above configuration, the processor implements a contact length neural network and a contact width neural network. The contact length neural network uses tire specification data and vehicle weight data as input layers and the contact length as its output layer. The contact width neural network uses tire specification data and vehicle weight data as input layers and the contact width as its output layer.

[0016] With the above configuration, it is possible to train the ground contact length neural network and the ground contact width neural network using conventionally accumulated data related to footprint images as training data. [Effects of the Invention]

[0017] The tire footprint image generation apparatus disclosed herein makes it possible to acquire tire footprint images more easily than in the conventional method. [Brief explanation of the drawing]

[0018] [Figure 1] This diagram illustrates the hardware configuration of a footprint image generation device. [Figure 2] This diagram illustrates the functional blocks of a footprint image generation device. [Figure 3] This is a flowchart illustrating the footprint image generation process. [Figure 4] This is an example diagram showing a photograph of the tire's appearance. [Figure 5] This figure shows an example of removing the background from a photograph of the tire's exterior. [Figure 6] It is a diagram showing an example when the image of FIG. 7 is binarized. [Figure 7] It is a diagram explaining the process of finding the center of a binarized appearance image. [Figure 8] It is a diagram showing an example when a binarized image is divided. [Figure 9] It is a diagram explaining the process of performing projective transformation on a divided image. [Figure 10] It is a diagram showing an example when divided images after projective transformation are combined. [Figure 11] It is a diagram illustrating a contact length CNN. [Figure 12] It is a diagram illustrating a contact width CNN. [Figure 13] It is a diagram explaining the extraction range of the tread surface image on the plane of FIG. 10. [Figure 14] It is a diagram illustrating a tread surface image extracted by an extraction frame. [Figure 15] It is a diagram showing an example when R-processing is performed on a rectangular frame. [Figure 16] It is a diagram illustrating a generated footprint image. DETAILED DESCRIPTION OF EMBODIMENTS

[0019] 1. Hardware Configuration A tire footprint image generation apparatus 10 according to the present embodiment is illustrated in FIG. 1. Hereinafter, the tire footprint image generation apparatus 10 is appropriately referred to as "image generation apparatus 10".

[0020] The image generation apparatus 10 is configured of, for example, a computer. The image generation apparatus 10 includes a CPU 11, a RAM 12, a ROM 13, a storage 14, an input / output controller 15, and a display unit 16.

[0021] The CPU 11 is a central processing unit, also called a processor. The RAM 12 is a volatile memory device that temporarily stores data being worked on. The ROM 13 is a memory device that allows data to be read. The storage 14 is a memory device that allows data to be written to and read from. The storage 14 consists of, for example, an HDD (Hard Disk Drive) or an SSD (Solid State Drive).

[0022] 2. Functional Blocks The CPU 11 (processor) executes a program stored in the ROM 13 or storage 14, thereby configuring the image generation device 10 with the functional units shown in Figure 2. Specifically, the image generation device 10 includes an image processing unit 20, a ground length CNN 21, a ground width CNN 22, and an image extraction unit 23 as functional units. In other words, these functional units are implemented in the CPU 11. This means the CPU 11 can execute the processing performed by the above functional units. The details of the processing performed by the above functional units will be described later.

[0023] The image generation device 10 receives data from the input device 17 and the imager 19. The input device 17 is, for example, a keyboard. Also, as will be described later, image data of the tire's appearance is input to the image generation device 10 from the imager 19.

[0024] 3. Footprint image generation process Figure 3 illustrates the footprint image generation flow. This generation flow is executed by the image generation device 10. This generation flow branches into two processes: processing (S10-S17) and setting (S20-S25). The processing and setting processes merge later. Furthermore, the generation flow performs extraction (S30, S31).

[0025] In the processing stage, the curved tread surface image is processed into a flat image. In the setting stage, the extraction range of the flat tread surface image is determined. Furthermore, in the extraction stage, a portion of the flat tread surface image is extracted based on the extraction range. The extracted image is output as a footprint image.

[0026] 3-1. Flattening (processing) of the tread surface image Referring to Figure 2, the image generation device 10 receives an external image of the tire from the imager 19. Figure 4 shows an example of an external image of the tire 100 captured by the imager 19. This external image includes the curved tread surface 110. That is, the tread surface 110 in a non-contact state is included in the captured image.

[0027] The exterior of the tire 100 is imaged while no pressure is applied to the tread surface 110. For example, as shown in Figure 4, the tire 100 is imaged while lying down. As shown in Figure 4, the exterior image of the tire 100 is represented with an arc-shaped contour extending from the longitudinal center to both ends. The image generation device 10 processes this curved image of the tread surface 110 into a planar image (see Figure 10).

[0028] Referring to Figures 2 and 3, the image processing unit 20 requests an image of the tire 100's appearance (S10). For example, the image processing unit 20 displays an image request message on the display unit 16. The image processing unit 20 then checks whether an image of the tire 100's appearance has been input (S11). If no image has been input, the flow returns to step S10.

[0029] When an image of the tire 100 is input, the image processing unit 20 removes the background from the image (S12). For example, the background image is removed using the Grub Cut Algorithm.

[0030] The grab-cut algorithm is a well-known image processing technique. Therefore, the grab-cut algorithm will be briefly explained below. First, a foreground region is specified from the external image of the tire 100 (see Figure 4). For example, the operator of the input device 17 specifies the foreground region. In this embodiment, the foreground region is the image region of the tire 100. Based on the specification of the foreground region, the image processing unit 20 divides the image into a foreground region and a background region.

[0031] The foreground image after the splitting process is displayed on the display unit 16. The operator of the input device 17 confirms the displayed image. If the foreground is included in the background, or if the background is included in the foreground, the input device 17 specifies the background area in the forward-leaning position and the foreground area in the background. This interactive extraction process is repeated until the background is removed from the image in Figure 4. The tire image after the deletion process in step S12 is illustrated in Figure 5.

[0032] Next, the image processing unit 20 binarizes the tire image in Figure 5 (S13). In the binarization process, the pixel values ​​of pixels on the tire image are divided into either 0 (black) or 255 (white). Figure 6 shows an example of a binarized tire image. For example, the tire image is binarized by adaptive binarization. Adaptive binarization is a well-known binarization method. Therefore, the process using adaptive binarization will be briefly explained below.

[0033] In adaptive binarization, the average value of the pixel to be transformed and its surrounding pixels becomes the binarization threshold. For example, suppose a pixel (pixel A1) and another pixel (pixel A2) have the same pixel value (e.g., 120). If the average value of pixel A1 and its surrounding pixels is lower than that of pixel A1 (e.g., 100), pixel A1 is transformed to white (120 → 255). Also, if the average value of pixel A2 and its surrounding pixels is higher than that of pixel A2 (e.g., 140), pixel A2 is transformed to black (120 → 0). In this way, the binarization threshold is adaptively set according to the ambient brightness, etc. By binarizing tire images, geometric calculations like the following can be performed with high accuracy.

[0034] The image processing unit 20 determines the center line of the binarized tire image (S14). Note that in Figures 7, 9, 10, and 13-16, the XY coordinate system is used as the image coordinate system. The X and Y axes are orthogonal.

[0035] Referring to Figure 7, for example, the image processing unit 20 determines the endpoint P1 on the longitudinal axis (X-axis) of the binarized image of the tire 100. For example, the image processing unit 20 extracts the pixel with the maximum X-axis coordinate value from among the pixels with a pixel value of 255 (white), and sets that pixel as endpoint P1. Similarly, the image processing unit 20 extracts the pixel with the minimum X-axis coordinate value from among the pixels with a pixel value of 255 (white), and sets that pixel as endpoint P2.

[0036] Furthermore, the image processing unit 20 determines the center point P3 of the line L1 connecting the endpoints P1 and P2. Next, the image processing unit 20 determines the perpendicular line from the center point P3 to the line L1. This perpendicular line becomes the center line L2.

[0037] Referring to Figure 8, the image processing unit 20 then divides the tire's appearance image into images F1 and F2, using the center line L2 as a dividing line (boundary line) (S15). Next, the image processing unit 20 performs a projection transformation on each of the divided images F1 and F2 (S16).

[0038] Figure 9 shows an example of a projective transformation applied to the segmented image F1. A projective transformation is also called a homographic transformation. In a projective transformation, one quadrilateral is transformed into another quadrilateral. For example, a trapezoid is transformed into a rectangle.

[0039] In homography transformation, the coordinates of the four vertices of a quadrilateral (x11, y11), (x12, y12), (x13, y13), and (x14, y14) are moved to (x21, y21), (x22, y22), (x23, y23), and (x24, y24). This movement of vertices is expressed by the following equation (eq. 1).

[0040]

number

[0041] The vertices before transformation (x11, y11), (x12, y12), (x13, y13), (x14, y14) and the vertices after transformation (x21, y21), (x22, y22), (x23, y23), (x24, y24) are specified in advance by the operator or other relevant parties. Therefore, equation (eq. 1) becomes an eight-variable equation containing homography parameters h11, h12, h13, h21, h22, h23, h31, h32. In equation (eq. 1), the eight equations are represented as a matrix. In other words, by substituting the values ​​of (x11, y11), (x12, y12), (x13, y13), (x14, y14), (x21, y21), (x22, y22), (x23, y23), and (x24, y24) into equation (eq. 1), the homography parameters h11, h12, h13, h21, h22, h23, h31, and h32 can be calculated.

[0042] Figure 9 illustrates the process of determining the four vertices in a projection transformation. Vertices C and D are set on the center line L2, which is the center of the tire. The image processing unit 20 searches for pixels on the center line L2 whose pixel value is 255 (white). Furthermore, from among those pixels, the image processing unit 20 sets the uppermost and lowermost points on the center line L2 as vertices C and D. Vertices C and D become fixed points that do not move before and after the projection transformation. That is, for vertex C, the relationship (x11, y11) = (x21, y21) holds. Also, for vertex D, the relationship (x12, y12) = (x22, y22) holds.

[0043] Next, the image processing unit 20 approximates the boundary edge of the curved tread surface 110 in the divided image F1 with a curve. In other words, the image processing unit 20 approximates the boundary line between the tread surface 110 and the background with a curve. Since the tread surface 110 has boundaries with the background on the top and bottom, two approximation curves are obtained.

[0044] The two approximation curves intersect at vertices C and D. The image processing unit 20 finds the tangent L10 to the approximation curve at vertex C. Similarly, the image processing unit 20 finds the tangent L12 to the approximation curve at vertex D. The tangents L10 and L12 can be said to be tangents to the boundary edge of the tread surface 110. Vertices A and B are set on the tangents L10 and L12. For example, the intersection points of the tangents L10 and L12 with the side edges of the divided image F1 become vertices A and B.

[0045] Next, the image processing unit 20 determines perpendiculars L13 and L14 to the center line L2. Here, perpendiculars L13 and L14 intersect with vertices C and D. The intersections of perpendiculars L13 and L14 with the side edges of the divided image F1 become vertices A' and B'. Furthermore, the image processing unit 20 determines homography parameters that move vertices A and B to vertices A' and B', with vertices C and D as fixed points.

[0046] In this way, by performing a projection transformation according to the curved shape of the tread surface 110, the tread surface 110 can be flattened with high accuracy.

[0047] The projection transformation illustrated in Figure 9 is also performed on the segmented image F2 (see Figure 8). The image processing unit 20 combines the segmented images F1' and F2' after the projection transformation. Figure 10 shows the combined image, which is a planar tread surface image F10.

[0048] 3-2. Determining the extraction frame (setting process) Referring to Figure 2, the CPU 11 of the image generation device 10 is equipped with a ground length CNN 21 (ground length neural network) and a ground width CNN 22 (ground width neural network). CNN is an abbreviation for Convolutional Neural Network.

[0049] Figure 11 illustrates a ground-length CNN21. The ground-length CNN21 comprises an input layer, a hidden layer, and an output layer. Each layer contains multiple nodes.

[0050] The input layer receives specification data for tire 100 and vehicle weight data. The specification data includes data such as the dimensions, weight, material, and function of tire 100. For example, the specification data for tire 100 includes tire weight, tread rubber hardness, tire size, tire weight, rolling resistance coefficient, tread rubber material, load index, static load radius, aspect ratio, rim diameter, and rim width. The specification data for tire 100 can be obtained from the tire 100 supplier.

[0051] Vehicle weight data can be obtained from the vehicle supplier. Note that instead of entering the numerical value listed in the specifications sheet, a value obtained by reducing that value by a predetermined percentage may be entered.

[0052] The output layer outputs the contact length of tire 100. The output layer has multiple nodes. Each node outputs the contact length and accuracy.

[0053] Figure 12 illustrates a ground-width CNN 22. Similar to the ground-length CNN 21, the ground-width CNN 22 comprises an input layer, a hidden layer, and an output layer. Each layer has multiple nodes.

[0054] Similar to the contact length CNN21, the input layer of the contact width CNN22 receives the specifications data of tire 100 and the vehicle weight data. The output layer outputs the contact width of tire 100. The output layer has multiple nodes. Each node outputs the contact width and accuracy.

[0055] Conventional tire characteristic analysis records tire specifications, vehicle weight, contact length, and contact width. Therefore, training data can be created from past tire characteristic analysis results. In this training data, the input data is tire specifications and vehicle weight. The correct data is contact length and contact width. A contact length CNN 21 and a contact width CNN 22, trained using this training data, are implemented in the CPU 11 of the image generation device 10 (see Figure 1).

[0056] Referring to Figure 3, the image generation device 10 requests specification data for the tire 100 (S20). For example, the image generation device 10 displays a specification data request message on the display unit 16 (see Figure 2). The specification data for the tire 100 is input from the input device 17. The image generation device 10 then checks whether the specification information for the tire 100 has been input (S21). If the specification information has not been input, the flow returns to step S20.

[0057] When the specifications of tire 100 are input, the image generation device 10 requests vehicle weight data (S22). For example, the image generation device 10 displays a message requesting vehicle weight data on the display unit 16. The vehicle weight data is input from the input device 17. The image generation device 10 then checks whether vehicle weight information has been input (S23). If weight information has not been input, the flow returns to step S22.

[0058] When the specifications data of tire 100 and the vehicle weight data are input, these data are input to the input layers of the contact length CNN 21 and the contact width CNN 22 (S24). The image extraction unit 23 obtains the contact length and contact width values ​​with the highest accuracy from the output layers of the contact length CNN 21 and the contact width CNN 22 (S25). Based on the obtained contact length and contact width, the image extraction range is determined.

[0059] 3-3. Extraction of footprint images (extraction process) Referring to Figures 2 and 13, the image processing unit 20 determines the distance between vertex C and vertex D from the tire size. For example, the tire size is expressed as "250 / 45 / R18 100W". In this example, the tire width is 250 mm. The image processing unit 20 sets the distance between vertex C and vertex D to the tire width (for example, CD = 250 mm).

[0060] The image processing unit 20 determines the size of the extraction frame FL1 based on the tire width value. The horizontal dimension (X-axis length) and vertical dimension (Y-axis length) of the extraction frame FL1 are the contact length and contact width obtained in step S25. For example, the image processing unit 20 calculates the image length (scale) of the contact length L31 and contact width L30 when the distance between vertex C and vertex D is the tire width.

[0061] Referring further to Figure 13, the image processing unit 20 determines the center point P10. The center point P10 is the midpoint of vertices C and D. The image processing unit 20 aligns the center of the extraction frame FL1 with the center point P10.

[0062] Based on the extraction frame FL1, the image processing unit 20 extracts a portion of the planar tread surface image F10 (S30). In other words, as illustrated in Figure 14, the image region within the extraction frame FL1 is extracted from the planar tread surface image F10.

[0063] In Figure 14, the extracted image is rectangular. However, the actual footprint image has rounded corners. Therefore, as illustrated in Figure 15, the extracted frame FL1 may be given a radius of curvature. For example, the radius of curvature is set to 1 / 4 of the shorter of the contact width L30 and contact length L31.

[0064] An example of an image extracted using the R-processed extraction frame FL1 is shown in Figure 16. The image processing unit 20 outputs the extracted image as a footprint image 60 (S31). For example, the footprint image 60 is displayed on the display unit 16 (see Figure 2).

[0065] In the above-described embodiment, an image captured by the camera 19 was used as the external image of the tire 100. Instead of an captured image, 3D data from 3D-CAD or the like may be input as the external image of the tire 100. [Explanation of Symbols]

[0066] 10 Image generation device, 11 CPU (processor), 20 Image processing unit, 21 Ground contact length CNN, 22 Ground contact width CNN, 23 Image extraction unit, 60 Footprint image, 100 Tire, 110 Tread surface.

Claims

1. A device for generating tire footprint images, equipped with a processor, The aforementioned processor, A processing method that transforms a flat tread surface image from an external image of a tire that includes a curved tread surface, The setting process for determining the extraction range of the planar tread surface image, An extraction process that extracts a portion of the planar tread surface image based on the extraction range and outputs it as a footprint image, Execute, The aforementioned processor, In the setting process described above, the tire's contact width and contact length are determined based on the tire's specifications and vehicle weight data. Furthermore, in the setting process, the extraction range is determined based on the grounding width and the grounding length. A device for generating tire footprint images.

2. A tire footprint image generating apparatus according to claim 1, The processor, in the processing process, Determine the center line of the aforementioned tire's external image, Using the aforementioned center line as a dividing line, the external image of the tire is divided, For each of the segmented images, perform a projection transformation. By combining the segmented images after the projection transformation, the planar tread surface image is generated. A device for generating tire footprint images.

3. A tire footprint image generating apparatus according to claim 2, In the setting process, the processor Based on the grounding width and grounding length, an extraction frame is generated. R-processing is performed on the extraction frame. A device for generating tire footprint images.

4. A tire footprint image generating apparatus according to claim 2 or 3, In the projection transformation, the processor sets the vertices of the projection transformation on the tangents to the boundary edges of the curved tread surface included in the segmented image before the transformation. A device for generating tire footprint images.

5. A tire footprint image generation apparatus according to claim 1, wherein the processor includes: A contact length neural network is provided, with the tire specifications data and vehicle weight data as input layers and the contact length as the output layer. A contact width neural network is provided, with the tire specifications data and vehicle weight data as input layers and the contact width as output layers. It has been implemented. A device for generating tire footprint images.

Images