Tire tread production control method based on image recognition

Abstract

The application discloses a tire tread production control method based on image recognition, and particularly relates to the technical field of production control, which comprises the following steps: collecting high-resolution image data of a rubber surface in a tire tread forming process in real time, removing noise to generate a clear tread texture image by using an image enhancement method; obtaining anisotropic spectral feature data and tread texture structure feature data of the tread surface by using a multi-angle spectral response difference analysis method; quantifying the directional features of the molecular chain orientation of the rubber surface by using a spectral directionality gradient analysis method, and outputting a surface molecular orientation feature map; calibrating the position and scale of the texture abnormal area by combining a texture boundary recognition and spatial frequency analysis method, and outputting a tread texture abnormality marking map; establishing a spatial correspondence between the texture abnormal area and the orientation change, outputting an accurate identification result of the tread surface defects, and performing adaptive dynamic adjustment of the process parameters of the production equipment, so as to realize accurate control of the tire tread production process.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of production control, and more particularly, to a tire tread production control method based on image recognition. BACKGROUND

[0002] In the production and manufacturing process of tire tread, rubber material passes through extrusion, calendering and vulcanization forming links. Due to the difference in operating conditions of processing equipment and process parameters in each stage, the directional difference arrangement of the micro-distribution characteristics of the polymer chain structure and filler particles in the tire tread rubber is caused, so that the rubber tire tread surface formed finally presents anisotropic reflection characteristics, which is manifested as the spectral reflection difference of the tire tread surface under a specific observation angle.

[0003] The existing tire production control technology generally focuses on the macroscopic defects or fixed mode texture abnormalities of the rubber tire tread surface appearance, lacks in-depth analysis and effective identification of the anisotropic reflection effect of the rubber tire tread surface caused by the microstructure difference of the material, and is difficult to realize the dynamic adjustment for the microstructure change of the material in the tire tread production process, which affects the accuracy and stability of the tire production quality control. SUMMARY

[0004] In order to overcome the above-mentioned defects of the prior art, the embodiments of the present application provide a tire tread production control method based on image recognition to solve the problems raised in the background art.

[0005] To achieve the above-mentioned purpose, the present application provides the following technical scheme:

[0006] The tire tread production control method based on image recognition comprises the following steps:

[0007] S1: collecting high-resolution image data of the rubber surface in the tire tread forming process, removing image noise by using an image enhancement algorithm, and generating a clear tread texture image;

[0008] S2: based on the multi-angle spectral response difference analysis method, outputting the anisotropic spectral feature data of the tire tread surface and the tread texture structure feature data from the clear tread texture image;

[0009] S3: according to the anisotropic spectral feature data of the tire tread surface, quantifying the directional feature of the molecular chain orientation of the rubber surface by using the spectral directionality gradient analysis method, and outputting a surface molecular orientation feature map;

[0010] S4: according to the tread texture structure feature data, extracting the texture abnormal region position and scale information by using the texture boundary recognition and spatial frequency analysis method, and outputting a tread texture abnormality marking map;

[0011] S5: Based on the surface molecular orientation feature map and the tire tread texture abnormality mark map, the correspondence between the texture abnormal area and the spectral directionality change is established, and the tire tread surface defect accurate recognition result is output;

[0012] S6: According to the tire tread surface defect accurate recognition result, the adaptive dynamic adjustment of the tire tread production control is carried out.

[0013] In a preferred embodiment, S1, specifically:

[0014] In the tire tread forming process, high-resolution image data of the rubber tire tread surface is collected in real time by an industrial high-definition camera;

[0015] Gaussian filter algorithm is used to filter out noise of high-resolution image data of the rubber tire tread surface, and random interference signals generated in the image acquisition process are eliminated;

[0016] The histogram equalization method is used to enhance the brightness of the high-resolution image data after noise filtering, and a clear tire tread texture image is generated.

[0017] In a preferred embodiment, S2, specifically:

[0018] A multi-angle light source irradiation device is used to project visible light from multiple preset directions to the rubber tire tread surface corresponding to the clear tire tread texture image, and the spectral reflectance intensity data of the rubber tire tread surface is obtained;

[0019] The difference between the spectral reflectance intensity data is calculated by the spectral response difference calculation method, and the anisotropic spectral feature data of the rubber tire tread surface is obtained;

[0020] The two-dimensional Fourier transform method is used to analyze the spatial frequency spectrum of the texture structure of the rubber tire tread surface, extract the frequency spectrum distribution characteristics of the rubber tire tread surface texture, and obtain the tire tread texture structure feature data of the rubber tire tread surface.

[0021] In a preferred embodiment, S3, specifically:

[0022] The spectral directionality gradient analysis method is used to calculate the spectral change gradient value of the anisotropic spectral feature data of the rubber tire tread surface in the horizontal and vertical orthogonal directions;

[0023] According to the spectral change gradient value, the mapping relationship between the spectral change gradient value of the rubber tire tread surface and the orientation direction of the polymer chain of the rubber tire tread surface is established;

[0024] According to the mapping relationship, point-by-point quantitative calculation is carried out, the orientation direction and change amplitude information of the rubber molecular chain of the rubber tire tread surface are obtained, and the surface molecular orientation feature map is generated.

[0025] In a preferred embodiment, S4, specifically:

[0026] The spatial frequency characteristic distribution data is obtained by performing frequency domain transformation on the tread texture structure feature data through a spatial frequency analysis method.

[0027] According to the spatial frequency characteristic distribution data, the boundary position of the abnormal texture region on the rubber tread surface is identified.

[0028] The spatial position and spatial scale size information of the abnormal texture region are obtained by performing connectivity analysis on the boundary position of the abnormal texture region using the texture boundary identification method.

[0029] Based on the spatial position and spatial scale size information of the abnormal texture region, the abnormal texture marking map of the tread is generated.

[0030] In a preferred embodiment, S5, specifically:

[0031] According to the spatial position and spatial scale size information of the abnormal texture region in the abnormal texture marking map of the tread, the abnormal texture region is mapped to the corresponding spatial position on the surface molecular orientation feature map.

[0032] The orientation direction and variation amplitude information of the rubber molecular chain on the rubber tread surface are extracted at the corresponding spatial position.

[0033] The spatial corresponding relationship between the abnormal texture region and the orientation variation of the rubber molecular chain on the rubber tread surface is determined by performing data correlation analysis on the spatial position and spatial scale size information of the abnormal texture region and the extracted orientation direction and variation amplitude information of the rubber molecular chain, and the accurate identification result of the tread surface defect is output.

[0034] In a preferred embodiment, S6, specifically:

[0035] The defect type, defect position and defect severity information are extracted from the accurate identification result of the tread surface defect.

[0036] According to the defect type and severity information, the pre-set process adjustment rules are called to calculate the target value of the extruder screw speed, the target value of the calender roll pressure, and the target value of the vulcanizing machine heating temperature and vulcanizing time, respectively.

[0037] The target value of the extruder screw speed, the target value of the calender roll pressure, and the target value of the vulcanizing machine heating temperature and vulcanizing time are output as production control instructions to the corresponding equipment.

[0038] The technical effects and advantages of the tire tread production control method based on image recognition are as follows:

[0039] By obtaining the high-resolution tread image filtered by noise, the visibility of texture details is greatly improved; the anisotropic spectrum and texture structure features are extracted under multi-angle illumination, the quantitative of microscopic reflection difference is realized; the distribution of molecular chain orientation is mapped based on spectral gradient analysis, which provides intuitive basis for the internal orientation anomaly of the material; the spatial resolution of defect detection is improved by accurately positioning the position and size information of the texture abnormal area combined with spatial frequency and boundary recognition; the high accuracy of defect identification is realized by establishing the corresponding relationship between the texture abnormal area and the directional change of the spectrum; the accurate identification result of the tread surface defect is converted into the adaptive adjustment instruction of the extrusion, calendering and vulcanization process parameters, forming a closed-loop control, which significantly improves the production stability and product consistency of the tread, and realizes the fine control of the whole process from the microstructure to the process optimization. BRIEF DESCRIPTION OF DRAWINGS

[0040] Figure 1 The figure is a schematic diagram of the tire tread production control method based on image recognition of the present application. DETAILED DESCRIPTION

[0041] The technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor fall within the scope of protection of the present application. EMBODIMENT

[0042] Figure 1 The present application provides a tire tread production control method based on image recognition, which comprises the following steps:

[0043] S1: collecting high-resolution image data of the rubber surface in the tire tread forming process, removing image noise by using image enhancement algorithm, and generating clear tread texture image;

[0044] S2: based on the multi-angle spectral response difference analysis method, outputting the anisotropic spectral feature data and the tread texture structure feature data of the tread surface from the clear tread texture image;

[0045] S3: according to the anisotropic spectral feature data of the tread surface, quantifying the directional features of the molecular chain orientation of the rubber surface by using the spectral directional gradient analysis method, and outputting the surface molecular orientation feature map;

[0046] S4: according to the tread texture structure feature data, extracting the position and scale information of the texture abnormal area by using the texture boundary recognition and spatial frequency analysis method, and outputting the tread texture abnormal marker map;

[0047] S5: Based on the surface molecular orientation feature map and the tire tread texture abnormality mark map, the correspondence between the texture abnormal area and the spectral directionality change is established, and the tire tread surface defect accurate identification result is output;

[0048] S6: According to the tire tread surface defect accurate identification result, the adaptive dynamic adjustment of the tire tread production control is carried out.

[0049] S1: Collecting high-resolution image data of the rubber surface in the tire tread forming process, using image enhancement algorithm to remove image noise, generating clear tire tread texture image, including:

[0050] In the tire tread forming process, high-resolution image data of the rubber tire surface is collected in real time by an industrial high-definition camera;

[0051] Specifically, the tire tread forming process directly determines the shape characteristics of the rubber tire surface, the integrity of the surface texture structure, and the overall performance and quality of the tire. In order to realize effective control and accurate defect identification of the tire tread production process, real-time and high-precision image acquisition of the appearance and structure details of the rubber tire surface in the tire tread forming process is required. On the tire production line, especially near the outlet of the extrusion, calendering and vulcanization forming equipment of the tire tread rubber, an industrial high-definition camera with a resolution of not less than 1920x1080 pixels is installed. By setting the industrial high-definition camera at a fixed position within a distance of 20cm to 50cm from the rubber tire surface, the rubber tire surface in the moving state during the forming process is continuously photographed frame by frame to collect high-resolution image data of the rubber tire surface. By adjusting the lens focal length and exposure time of the industrial high-definition camera, it is ensured that the collected high-resolution image data of the rubber tire surface can contain complete rubber tire texture structure information. For example, on the actual production line, when the tire tread rubber passes through the field of view of the industrial high-definition camera at a speed of 0.2 meters per second to 1 meter per second, the frame rate of the camera is set between 30 frames per second to 120 frames per second, so as to ensure that the spatial displacement between each high-resolution image data does not exceed the set allowable range, and to avoid the blurring or loss of details of the tire tread rubber surface. Through the above operation, a sufficient amount of high-resolution image data of the rubber tire surface can be obtained.

[0052] Gaussian filter algorithm is used to filter out the noise of the high-resolution image data of the rubber tire surface, and the random interference signals generated in the image acquisition process are eliminated;

[0053] Specifically, due to the actual environment of the tire production site is usually more complex, therefore in the process of collecting high-resolution image data of the rubber tread surface by industrial high-definition camera in real time, random interference signals will be introduced to cause noise in the image data. Random interference signals may come from electronic noise of industrial high-definition camera sensor, light changes of workshop lighting equipment, local reflection or scattering generated by small particles or bubbles on the surface of the rubber tread, etc. Therefore, after obtaining the high-resolution image data of the rubber tread surface, it is necessary to use Gaussian filtering algorithm to filter the noise of the high-resolution image data of the rubber tread surface, in order to eliminate the influence of random interference signals on the image clarity. First, define a filter window size for the collected high-resolution image data of the rubber tread surface, for example, a 3x3 pixel or 5x5 pixel Gaussian filter window can be selected, in each filter window, the gray value of the high-resolution image data of the rubber tread surface is weighted and averaged to obtain the filtered smooth gray value, and the weighted average calculation is performed for each pixel position of the high-resolution image data, so as to eliminate the random interference signals caused by the random interference signals on the basis of keeping the tread texture structure contour information. For example, when using a 5x5 pixel Gaussian filter window to filter the noise of the high-resolution image data of the tread surface, by setting appropriate Gaussian function standard deviation parameters, such as between 1.0 and 2.0, the high-frequency noise can be filtered out, so as to obtain the smooth image data of the rubber tread surface, so as to ensure the reliability and stability of the image processing process.

[0054] The high-resolution image data after noise filtering is enhanced in brightness by using histogram equalization method to generate a clear tread texture image.

[0055] Specifically, after the high-resolution image data of the rubber tread surface is filtered by the Gaussian filtering algorithm, relatively smooth image data can be obtained, but the surface texture structure information of the rubber tread may still be difficult to effectively reflect due to insufficient image gray contrast, especially at the details of the tread texture structure. Therefore, it is necessary to enhance the brightness of the high-resolution image data after noise filtering, so as to highlight the surface texture structure information of the rubber tread. The brightness enhancement method is to use histogram equalization method, that is, to perform pixel-by-pixel gray mapping on the high-resolution image data after Gaussian filtering, so that the gray value distribution of the image data is more uniform. For example, the gray distribution histogram of all pixel points in the filtered high-resolution image data is counted, and the cumulative probability density function of each gray value is calculated, and the gray value of the image data is remapped according to the cumulative probability density function, so that the original gray value is more uniformly distributed in the gray space, thereby obtaining the image data of the rubber tread surface after brightness enhancement, forming a clear tread texture image.

[0056] S2: based on the multi-angle spectral response difference analysis method, outputting anisotropy spectral feature data and tire tread texture structure feature data of the tire tread surface, including:

[0057] A multi-angle light source irradiation device is used to project visible light from multiple preset directions to the rubber tire tread surface corresponding to the tire tread texture clear image, and obtain spectral reflectance intensity data of the rubber tire tread surface;

[0058] Specifically, in the tire tread production control process, in order to accurately reflect the anisotropy characteristics of the rubber tire tread surface, it is necessary to collect spectral reflectance intensity data of the rubber tire tread surface under different directions of irradiation. The multi-angle light source irradiation device includes at least three independently arranged visible light sources, each visible light source has a wavelength range of 380 nanometers to 780 nanometers, and is arranged on a circle with equal intervals with the center of the rubber tire tread surface as the center. The included angle between the light sources is set to be between 30° and 90°, for example, it can be set to 60°. When projecting visible light on the rubber tire tread surface, each visible light source is turned on in turn to make the light rays project onto the rubber tire tread surface from different angles. Each pixel position of the tire tread texture clear image corresponds to a spatial position of the rubber tire tread surface, that is, the spectral reflectance intensity value of each spatial position on the rubber tire tread surface under different irradiation angles can be recorded one by one. For example, when the first visible light source projects visible light from 0° azimuth, the second visible light source projects visible light from 60° azimuth, and the third visible light source projects visible light from 120° azimuth, and so on, at least three spectral reflectance intensity data of the rubber tire tread surface under different directions can be obtained. Due to the difference in the interaction between visible light and the microstructure of the rubber tire tread surface under different irradiation directions, the spectral reflectance intensity data of each spatial position will show obvious differences.

[0059] The difference between the spectral reflectance intensity data is calculated by a spectral response difference calculation method to obtain anisotropy spectral feature data existing on the rubber tire tread surface;

[0060] Specifically, in order to quantify the anisotropic spectral characteristics of the rubber tread surface, spectral reflectance intensity data from multiple directions need to be calculated and analyzed. The spectral response difference calculation method is to calculate the difference between the spectral reflectance intensity data of the same spatial position on the rubber tread surface in different directions. For example, take the spectral reflectance intensity data of visible light projected from 0° azimuth angle as the reference data, and calculate the difference between it and the spectral reflectance intensity data of visible light projected from other directions in turn. Through difference calculation, the spectral reflectance intensity difference of the rubber tread surface under irradiation from different directions can be accurately quantified, forming the anisotropic spectral characteristic data of the rubber tread surface. For example, there is a certain spatial position on the rubber tread surface, when irradiated from 0° direction, the reflectance intensity value of this position is 120 gray units, and when irradiated from 60° direction, the reflectance intensity value of this position is 135 gray units, then the spectral response difference value of this spatial position is 15 gray units; and the difference between other direction data is calculated in turn, finally obtaining a comprehensive data set describing the spectral response difference of each spatial position on the rubber tread surface, i.e. the anisotropic spectral characteristic data.

[0061] The two-dimensional Fourier transform method is used to analyze the spatial frequency spectrum of the texture structure of the rubber tread surface, extract the frequency spectrum distribution characteristics of the rubber tread surface texture, and obtain the tread texture structure characteristic data of the rubber tread surface.

[0062] Specifically, in order to obtain the structural characteristic information of the rubber tread surface, the texture structure characteristics presented by the rubber tread surface in the tread texture clear image need to be analyzed by frequency spectrum. The two-dimensional Fourier transform method is to convert the gray data of each spatial position in the tread texture clear image from spatial domain to frequency domain by two-dimensional Fourier transform formula. For example, for a tread texture clear image with a pixel size of 1024x1024, the frequency spectrum components corresponding to each spatial position are calculated one by one by using the two-dimensional Fourier transform formula, so as to obtain the frequency spectrum characteristic distribution data of the rubber tread surface. The frequency spectrum characteristic distribution data can directly reflect the spatial frequency characteristics of the texture structure of the rubber tread surface. For example, when the rubber tread surface texture presents regular and clear strip shape, the frequency spectrum distribution characteristics will present relatively concentrated peak value in the frequency domain; when the texture structure of the rubber tread surface is chaotic or there is an abnormal area, the frequency spectrum characteristic distribution will present dispersed frequency spectrum peak value. Therefore, according to the specific peak position, peak number and distribution of the frequency spectrum distribution characteristics, the spatial frequency spectrum characteristics of the texture structure of the rubber tread surface can be accurately extracted, and the tread texture structure characteristic data can be obtained.

[0063] S3: According to the anisotropic spectral characteristic data of the tread surface, the directional characteristics of the molecular chain orientation of the rubber surface are quantified by using the spectral direction gradient analysis method, and the surface molecular orientation characteristic map is output, including:

[0064] The spectral direction gradient analysis method is used to calculate the spectral variation gradient values of the anisotropic spectral feature data of the rubber tread surface in the horizontal and vertical orthogonal directions;

[0065] Specifically, in order to accurately obtain the polymer chain orientation characteristics of the rubber tread surface, it is necessary to determine the variation trend of the spectral features in different spatial directions in each local region on the rubber tread surface. The spectral direction gradient analysis method refers to the point-by-point calculation of the local spatial gradient of the anisotropic spectral feature data of the rubber tread surface to determine the variation rate of the spectral features of the rubber tread surface. The spectral direction gradient analysis includes: selecting two orthogonal directions of the horizontal direction and the vertical direction as the reference direction in the anisotropic spectral feature data of the rubber tread surface; at each spatial position of the anisotropic spectral feature data of the rubber tread surface, the spectral variation gradient value is calculated along the horizontal direction and the vertical direction by using the finite difference calculation method. For example, in the horizontal direction, for the spectral feature data of the spatial position coordinates (x, y), the difference between the spectral feature data of the coordinate position (x+1, y) and the spectral feature data of the coordinate position (x-1, y) is divided by the horizontal spacing of the coordinate position; in the vertical direction, the difference between the spectral feature data of the coordinate position (x, y+1) and the spectral feature data of the coordinate position (x, y-1) is divided by the vertical spacing of the coordinate position. The above calculation process is executed on all spatial positions on the rubber tread surface one by one, thereby forming the horizontal spectral variation gradient value data and the vertical spectral variation gradient value data of the complete rubber tread surface. The horizontal spectral variation gradient value data and the vertical spectral variation gradient value data can accurately describe the directional variation information of the spectral features of the rubber tread surface.

[0066] According to the spectral variation gradient value, a mapping relationship between the spectral variation gradient value of the rubber tread surface and the orientation direction of the polymer chain of the rubber tread surface is established;

[0067] Specifically, the orientation direction of the polymer chains on the rubber tread surface can be inferred and determined by the spectral variation gradient value, and the inference is based on the direct physical correspondence between the arrangement direction of the polymer chains on the rubber tread surface and the directional variation of the surface spectral response. During the extrusion, calendering and vulcanization processes of the rubber material on the rubber tread surface, the polymer chains will form directional arrangement characteristics under the joint action of pressure, flow shear force and temperature gradient, thereby exhibiting anisotropic spectral reflection characteristics on the macro level. In order to effectively quantify the corresponding relationship, it is necessary to establish a mapping relationship between the spectral variation gradient value data of the rubber tread surface and the actual polymer chain orientation direction: a plurality of sample regions at different positions on the rubber tread surface are preselected, for example, no less than 50 sample regions are selected, and a conventional microscopic method such as a scanning electron microscope is used to observe and record the actual polymer chain orientation direction in each region to obtain the real polymer chain orientation direction information of the sample region; at the same time, the horizontal and vertical spectral variation gradient value data of the corresponding position of the sample region are recorded, and a quantitative mapping relationship between the horizontal and vertical spectral variation gradient value and the real observed polymer chain orientation direction is established by a data fitting method such as linear regression or polynomial fitting. After the mapping relationship is established, it can be applied to all other spatial positions on the rubber tread surface to infer and determine the orientation direction of the polymer chains at any position.

[0068] According to the mapping relationship, point-by-point quantitative calculation is performed to obtain the orientation direction and variation amplitude information of the rubber molecular chains on the rubber tread surface, and a surface molecular orientation feature map is generated;

[0069] Specifically, in order to realize the quantitative evaluation of the orientation direction and the orientation amplitude of the polymer chains on the entire rubber tread surface, it is necessary to use the mapping relationship between the spectral variation gradient value of the rubber tread surface and the orientation direction of the polymer chains on the rubber tread surface to quantitatively calculate each spatial position on the rubber tread surface point by point: read the corresponding horizontal and vertical spectral variation gradient value data of each spatial position on the rubber tread surface one by one, substitute it into the mapping relationship for calculation, and obtain the specific value of the orientation direction of the polymer chains at the corresponding position, for example, the chain orientation direction is represented in the form of angle. If the horizontal spectral variation gradient value at a certain spatial position is 10 and the vertical spectral variation gradient value is 15, then through the calculation of the mapping relationship, the corresponding spatial position of the polymer chain orientation direction can be obtained. 56 degrees. In order to obtain the orientation change amplitude information, the spatial variation rate of the orientation direction of the polymer chains in the local area of the rubber tread surface is calculated, that is, the change degree of the orientation direction between adjacent spatial positions. For example, in each local 3x3 pixel area, the chain orientation direction angle of the center position and the surrounding 8 positions is compared and calculated to obtain the local change amplitude of the chain orientation direction. After the above calculation is performed one by one, a comprehensive data set including the orientation direction and the change amplitude information is formed, which is output in the form of a two-dimensional spatial distribution graph to generate a surface molecular orientation feature map, which intuitively reflects the spatial distribution characteristics of the polymer chain orientation on the rubber tread surface in a graphical manner.

[0070] S4: According to the tread texture structure feature data, the texture boundary recognition and spatial frequency analysis method is used to extract the texture abnormal area position and scale information, and output the tread texture abnormal marker map, including:

[0071] The spatial frequency analysis method is used to perform frequency domain transformation on the tread texture structure feature data to obtain spatial frequency feature distribution data;

[0072] Specifically, the spatial frequency analysis method comprises: first obtaining the tire texture structure feature data of the rubber tire surface, such as the gray spectrum feature data obtained by the two-dimensional Fourier transform method; applying the spatial frequency analysis method to the tire texture structure feature data, such as processing again by using the two-dimensional Fourier transform formula, that is, through the mathematical transform formula, the texture feature data in the spatial domain is converted into the spatial frequency feature data in the frequency domain. For example, for the rubber tire texture structure feature data image with a pixel size of 1024*1024, the gray value of each spatial position of the rubber tire surface is represented by two-dimensional coordinates (p, q), and the two-dimensional Fourier transform formula is used to calculate the spectrum of each spatial position, thereby obtaining the spectrum component in the corresponding frequency domain. The spectrum component can be represented by frequency domain coordinates (u, v), wherein the amplitude of the spectrum component represents the intensity feature information of the corresponding spatial position texture structure feature at a specific spatial frequency, and the distribution of the spectrum component represents the overall feature of the texture structure feature of the rubber tire surface in the entire frequency domain. For example, when the rubber tire texture is a regular periodic stripe, the spectrum component obtained by the spatial frequency analysis will be obviously concentrated near a specific frequency, showing a high amplitude peak value; if the rubber tire texture appears an abnormal or irregular area, the spectrum distribution will show multiple scattered peak values. Therefore, the spatial frequency analysis method can obtain complete spatial frequency feature distribution data, reflecting the overall spatial feature of the rubber tire surface texture structure.

[0073] According to the spatial frequency feature distribution data, the boundary position of the texture abnormal area on the rubber tire surface is identified;

[0074] Specifically, specific spectrum components representing the texture structure abnormal feature are selected from the spatial frequency feature distribution data, including spectrum components with an amplitude lower than a predetermined normal value or deviating from the normal spectrum range; the frequency domain data is restored to the spatial domain by inverse two-dimensional Fourier transform calculation of the selected spectrum components, generating a spatial domain image highlighting only the texture abnormal area; in the spatial domain image, the change gradient value of the gray value between the rubber tire surface texture abnormal area and the normal area is analyzed point by point by a boundary detection algorithm, such as an edge gradient operator algorithm, to determine the boundary position of the tire texture abnormal area. For example, when the texture structure of a certain area of the rubber tire surface is abnormal, the corresponding spectrum component data after inverse transform shows a clear gray change gradient on the spatial domain image, and the boundary profile of the abnormal area can be detected by the boundary detection algorithm at this time; while the gray change of the normal tire texture structure area in the spatial domain image is relatively flat and has a small gradient, so it will not be detected as a boundary position.

[0075] The texture boundary recognition method is used to analyze the connectivity of the boundary positions of the texture abnormal area, to obtain the spatial position and spatial scale size information of the texture abnormal area.

[0076] Specifically, after obtaining the boundary positions of the tire tread texture abnormal area, quantitative analysis of the spatial position and scale size of the texture abnormal area boundary positions is needed. The texture boundary recognition method includes: analyzing the connectivity of each detected texture abnormal area boundary position, that is, by pixel-by-pixel checking, the spatial connectivity between each abnormal area boundary position is determined; for example, an 8-connected region analysis method can be used, for any boundary position pixel in the tire tread texture abnormal area, it is checked whether the 8 spatially adjacent positions also belong to the abnormal area boundary position, so as to determine the complete boundary contour of each texture abnormal area; the geometric center coordinates of each texture abnormal area are determined, and the spatial position of the abnormal area is defined by the geometric center coordinates. According to the spatial size of each tire tread texture abnormal area boundary contour, for example, based on the maximum width and maximum height of the area boundary, the spatial scale size information of each tire tread texture abnormal area is calculated. For example, when an oval-shaped texture abnormal area with a length of 50 pixels and a width of 20 pixels appears on the surface of the rubber tire, the connectivity analysis method can determine all the boundary positions of the abnormal area boundary contour, and through the spatial size measurement, the texture abnormal area spatial scale size information is calculated and recorded as 50 pixels by 20 pixels.

[0077] Based on the spatial position and spatial scale size information of the texture abnormal area, a tire tread texture abnormality marking map is generated;

[0078] Specifically, the method for generating the tire texture abnormality marked map includes: first, marking the spatial position and scale size information of the tire texture abnormality region in a blank image with a uniform coordinate system, the blank image being consistent in size with the tire texture clear image; for each texture abnormality region, marking its position in the blank image with a specific pattern, such as a rectangle or an ellipse, according to the geometric barycenter coordinate position determined by the connectivity analysis; then, determining the size of the marked pattern according to the spatial scale size information of the abnormal region, such as the length and width dimensions corresponding to the length and width information of the abnormal region, respectively, with a specific pixel size; finally, repeating the above marking operation for all texture abnormality regions, thereby generating a tire texture abnormality marked map containing the spatial position and scale size of all abnormal regions. For example, if the tire texture clear image is 1024x1024 pixels, the spatial position coordinates of the tire texture abnormality region are (300, 400), and the scale size of the abnormal region is 60x30 pixels, then when marking the tire texture abnormality region in the tire texture abnormality marked map, a rectangular marked region with a length of 60 pixels and a width of 30 pixels is drawn with the coordinate position (300, 400) as the center, thereby presenting the accurate position and size information of the abnormal region. The tire texture abnormality marked map can directly show the spatial position, scale size, and distribution of all abnormal regions on the rubber tire surface.

[0079] S5: Based on the surface molecular orientation feature map and the tire texture abnormality marked map, a corresponding relationship between the texture abnormality region and the spectral directionality change is established, and an accurate tire surface defect recognition result is output, including:

[0080] According to the spatial position and spatial scale size information of the texture abnormality region in the tire texture abnormality marked map, the texture abnormality region is mapped to the corresponding spatial position on the surface molecular orientation feature map;

[0081] Specifically, in order to analyze the spatial correspondence between the texture abnormal area and the change of the rubber molecular chain orientation, the spatial position of each texture abnormal area in the tire tread texture abnormal marking map needs to be mapped to the corresponding spatial position in the surface molecular orientation feature map. The mapping method includes: the tire tread texture abnormal marking map and the surface molecular orientation feature map are generated based on the same tire tread texture clear image, so they have the same image size and spatial coordinate system; taking the spatial position of the texture abnormal area in the tire tread texture abnormal marking map as the coordinate reference, the spatial coordinate position of each abnormal area is determined; using the spatial coordinate matching method, the coordinate position in the surface molecular orientation feature map that corresponds to the spatial position in the tire tread texture abnormal marking map is found, so that the position of each texture abnormal area in the surface molecular orientation feature map is accurately determined. For example, if there is a texture abnormal area in the tire tread texture abnormal marking map, the geometric center coordinates of the texture abnormal area are (400, 500), and the area size is 30 pixels wide and 60 pixels high, first, the spatial range of the texture abnormal area in the tire tread texture abnormal marking map is determined, that is, the horizontal coordinates are between 385 pixels and 415 pixels, and the vertical coordinates are between 470 pixels and 530 pixels; in the surface molecular orientation feature map, the corresponding spatial range is found, that is, the coordinates between 385 pixels and 415 pixels horizontally and 470 pixels and 530 pixels vertically, so that the spatial position of the tire tread texture abnormal area is accurately mapped to the surface molecular orientation feature map. In this way, each texture abnormal area is mapped in turn to ensure the consistency of the spatial position of the texture abnormal area between the two images.

[0082] extracting the rubber molecular chain orientation direction and change amplitude information of the rubber tire tread surface at the corresponding spatial position;

[0083] Specifically, after the mapping of the spatial position is completed, it is necessary to extract the orientation direction and variation amplitude information of the rubber molecular chain at the corresponding spatial position of each mapped texture abnormal area: for each mapped texture abnormal area, read the orientation direction value of each spatial position in the texture abnormal area from the surface molecular orientation feature map point by point; statistical analysis is performed on the orientation direction values of all spatial positions in each texture abnormal area, and the statistical method is, for example, to calculate the average value, maximum value and minimum value of the orientation direction values in each area to determine the overall orientation direction feature of the rubber molecular chain. In order to obtain the orientation variation amplitude information of the rubber molecular chain, the difference between the orientation directions of adjacent spatial positions is analyzed in each texture abnormal area, for example, the absolute value of the orientation angle difference between each spatial position and its adjacent position is calculated, and then the local variation amplitude of the orientation direction of the rubber molecular chain in the entire texture abnormal area is determined by taking the average. Assuming that in the surface molecular orientation feature map, there are 100 spatial pixels in a certain mapped texture abnormal area, 100 corresponding rubber molecular chain orientation direction values are obtained after reading, and the orientation direction value range is distributed between 30 degrees and 80 degrees, then the average angle of the orientation direction of the rubber molecular chain in the texture abnormal area can be obtained by taking the average of the orientation angle values of all spatial positions in the texture abnormal area, for example, the average value is 55 degrees; at the same time, the orientation angle difference between each spatial position and its adjacent position is calculated point by point, for example, the angle of a certain spatial position is 60 degrees, and the angle of the adjacent position is 65 degrees, then the variation amplitude is 5 degrees, and then the average variation amplitude of the orientation direction of the rubber molecular chain is obtained by statistical analysis of all adjacent position angle differences, for example, the average value of the variation amplitude is 8 degrees. Through the above method, the orientation direction and variation amplitude information of the rubber molecular chain at the corresponding position of each texture abnormal area can be extracted.

[0084] The spatial position and spatial scale size information of the texture abnormal area and the extracted orientation direction and variation amplitude information of the rubber molecular chain are analyzed for data correlation to determine the spatial correspondence between the texture abnormal area of the rubber tire surface and the orientation variation of the rubber molecular chain, and output the accurate identification result of the tire surface defect;

[0085] Specifically, the data correlation analysis includes: taking the spatial position and the spatial scale size information of each texture abnormal area as the basic information of the spatial dimension; taking the rubber molecular chain orientation direction and the change amplitude information extracted from the corresponding area as the information of the orientation dimension; using a data analysis method, such as correlation coefficient analysis method or spatial data clustering analysis method, to calculate the numerical correlation degree between the spatial dimension and the orientation dimension; for example, using the correlation coefficient analysis method, the scale size of each abnormal area can be taken as the spatial information, and the orientation change amplitude can be taken as the orientation information, to calculate the linear correlation coefficient between the scale size and the orientation change amplitude; if the spatial data clustering analysis method is used, the spatial position and the scale information, and the orientation change amplitude information can be used to judge the spatial distribution clustering rule of each texture abnormal area in the orientation direction change characteristics by using the clustering algorithm. For example, when it is analyzed that the orientation change amplitude of the rubber molecular chain corresponding to the texture abnormal area with large scale on the surface of the rubber tire tread is significantly higher than that of the area with small scale, it can be concluded that there is a high correlation between the spatial scale size of the texture abnormal area and the orientation change amplitude of the rubber molecular chain. Through the above analysis, the accurate spatial corresponding relationship between each texture abnormal area on the surface of the rubber tire tread and the orientation change of the rubber molecular chain is finally determined, and the accurate identification result of the tire tread surface defect is output. The accurate identification result of the tire tread surface defect includes the texture abnormal area position, the spatial scale size information and the rubber molecular chain orientation change amplitude.

[0086] S6: According to the accurate identification result of the tire tread surface defect, the adaptive dynamic adjustment of the tire tread production control is performed, including:

[0087] extracting the defect type, defect position and defect severity information from the accurate identification result of the tire tread surface defect;

[0088] Specifically, the spatial position information of each texture abnormal area is extracted from the tire tread surface defect accurate identification result, and the spatial position information is represented as the geometric center coordinate position of the texture abnormal area. For example, in a tire tread image with a size of 1024x1024 pixels, the geometric center position coordinate of the defect area may be (350, 420). According to the spectral feature, the spectral response feature and the rubber molecular chain orientation change feature of the texture abnormal area, the defect type of the texture abnormal area is identified and classified. The classification standard is, for example: if the spectral change amplitude of the area is greater than a preset threshold and the orientation change is obvious, the defect type is determined as a molecular orientation abnormal defect; if the spectral feature of the area appears significant dispersion or irregular feature, the defect type is determined as a structure abnormal defect. For example, the spectral response change amplitude of a certain texture abnormal area is 20 gray units, which exceeds the preset threshold of 15 gray units, and it is determined that the defect type of this area is a molecular orientation abnormal defect. According to the spatial scale size and the orientation change amplitude information of the texture abnormal area, the defect severity is evaluated. The evaluation standard is as follows: if the scale of the area is above 50 pixels and the orientation change amplitude exceeds 10 degrees, it is defined as a serious defect; if the scale of the area is between 20 pixels and 50 pixels and the orientation change amplitude is between 5 degrees and 10 degrees, it is defined as a general defect; if the scale of the area is less than 20 pixels and the orientation change amplitude is less than 5 degrees, it is defined as a slight defect. For example, the spatial scale of a certain defect area is 55 pixels and the orientation change amplitude is 12 degrees, and the defect severity is determined as a serious defect.

[0089] According to the defect type and severity information, the preset process adjustment rules are called to calculate the target value of the extruder screw speed, the target value of the calender roll pressure, and the target value of the curing machine heating temperature and curing time, respectively.

[0090] Specifically, in the extrusion, calendering and vulcanization forming process of the tire tread rubber, the process parameters of various equipment directly affect the structure quality and molecular chain orientation state of the tread surface, therefore, after identifying the defect type and severity of the tread, corresponding process adjustment measures are taken for defects of different types and severities to improve the quality of the production process. The setting of process adjustment rules includes: first, establish a correlation model between defect type and equipment process parameters, for example, for molecular orientation abnormal defects, the molecular chain arrangement state of the rubber compound needs to be improved by adjusting the screw speed of the extruder and the roll pressure of the calender, and for structural abnormal defects, the structural integrity needs to be improved mainly by adjusting the temperature and time of the curing machine; a quantitative relationship model between various defect types and equipment process parameters is established using experimental data, for example, through a large number of test data, when a serious molecular orientation abnormal defect occurs, the target value of the extruder screw speed needs to be reduced by 5 revolutions per minute to 15 revolutions per minute based on the basic speed, and the target value of the calender roll pressure needs to be increased by 0.1 megapascal to 0.3 megapascal; when a serious structural abnormal defect occurs, the heating temperature of the curing machine needs to be increased by 3 degrees Celsius to 5 degrees Celsius based on the basic temperature, and the curing time needs to be extended by 10 seconds to 20 seconds; based on this, a complete process adjustment rule database is established. For example, when the defect type of the tread is a serious molecular orientation abnormal defect and the defect severity is serious, then according to the pre-set rules, the target value of the extruder screw speed is reduced by 10 revolutions per minute, and the target value of the calender roll pressure is increased by 0.2 megapascal; if the defect type is a structural abnormal defect and the defect severity is general, then the target value of the heating temperature of the curing machine is increased by 3 degrees Celsius, and the target value of the curing time is extended by 15 seconds. Through the above method, the determination of the dynamic adjustment target value of the equipment process parameters based on the defect information is realized.

[0091] The target values of the extruder screw speed, the calender roll pressure, and the heating temperature and curing time of the curing machine are output as production control instructions to the corresponding equipment;

[0092] Specifically, in order to realize effective closed-loop control of the production process of the tread rubber, ensure that the tread rubber forming quality is significantly improved in the production batch, and send the process parameter target value in the form of digital production control instructions to the corresponding extruder, calender and vulcanizing machine equipment on the production line. First, the calculated target value of the extruder screw speed is output to the extruder control terminal through the production line automatic control system, for example, if the target value of the extruder screw speed is 85 revolutions per minute, the target value is transmitted to the extruder equipment through the digital interface of the automatic control system, and the extruder equipment automatically adjusts the screw speed to 85 revolutions per minute according to the received target value; at the same time, the target value of the calender roll pressure is output to the control terminal of the calender, for example, the target value is 3.2 megapascals, and the automatic control system of the calender receives the target value and automatically adjusts the roll pressure to 3.2 megapascals; and the target values of the heating temperature and curing time of the vulcanizing machine are output to the control terminal of the vulcanizing machine, for example, the target values are temperature 165 degrees Celsius and curing time 420 seconds, and the temperature control system and timing system of the vulcanizing machine automatically complete the adjustment control of the temperature and curing time according to the target values respectively. After each device completes the adjustment of the target value, the adjusted process parameter data is recorded in real time through the production line control system to ensure real-time monitoring and data tracing in the production process. Therefore, the closed-loop control of the device process parameter adjustment instruction can effectively realize the continuous optimization of the tread rubber forming process quality and the real-time improvement of the defect problem, and finally output the tread rubber product with stable quality.

[0093] The above formulas are dimensionless numerical calculations, and the formulas are obtained by collecting a large amount of data to simulate the most recent real situation, and the preset parameters and threshold values in the formula are set by a person skilled in the art according to the actual situation.

[0094] The above-described embodiments can be implemented in whole or in part by software, hardware, firmware, or any combination thereof. When implemented by software, the above-described embodiments can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, the processes or functions described in the embodiments of the present application are wholly or partially generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable devices. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another computer-readable storage medium, for example, the computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center through a wired (for example, infrared, wireless, microwave, etc.) manner. The computer-readable storage medium can be any available medium that can be accessed by a computer or a data storage device such as a server, data center, etc. containing one or more available medium collections. The available medium can be a magnetic medium (for example, a floppy disk, a hard disk, a magnetic tape), an optical medium (for example, a DVD), or a semiconductor medium. The semiconductor medium can be a solid-state disk.

[0095] Those of ordinary skill in the art can realize that the modules and algorithm steps of the examples described in conjunction with the embodiments disclosed herein can be implemented by electronic hardware or a combination of computer software and electronic hardware. Whether the functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of the present application.

[0096] Those skilled in the art can clearly understand that, for the convenience and brevity of the description, the specific working processes of the above-described system, device, and module can refer to the corresponding processes in the foregoing method embodiments, which will not be described here.

[0097] In several embodiments provided in the present application, it should be understood that the disclosed system, device, and method can be implemented in other ways. For example, the above-described device embodiments are only schematic, for example, the division of the modules is only a logical function division, and actual implementation can have another division manner, for example, multiple modules or components can be combined or integrated into another system, or some features can be ignored or not executed. In addition, the coupling or direct coupling or communication connection between the displayed or discussed ones can be indirect coupling or communication connection through some interfaces, devices, or modules, which can be electrical, mechanical, or other forms.

[0098] The modules described as separate components may or may not be physically separate, and the components shown as modules may or may not be physical modules, and may be located in one place or distributed on multiple network modules. Part or all of the modules can be selected according to actual needs to achieve the purpose of the embodiment.

[0099] In addition, the functional modules in each embodiment of the present application can be integrated into one processing module, or each module can exist physically alone, or two or more modules can be integrated into one module.

[0100] If the functions are realized in the form of software function modules and sold or used as independent products, they can be stored in a computer readable storage medium. Based on this understanding, the technical solutions of the present application essentially or the part that contributes to the prior art or part of the technical solutions can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes a number of instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present application. The aforementioned storage medium includes: a U disk, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk, and various program code storage media.

[0101] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art can easily think of changes or replacements within the technical scope disclosed in the present application, which should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.

[0102] Finally, the above is only the preferred embodiment of the present application and is not used to limit the present application. Any modification, equivalent replacement, improvement, etc. within the spirit and principles of the present application should be included in the protection scope of the present application.

Claims

1. A method for controlling the production of a tire tread based on image recognition, characterized in that, The method comprises the following steps: S1: collecting high-resolution image data of the rubber surface in the tire tread forming process, removing image noise by using an image enhancement algorithm, and generating a clear tread texture image; S2: based on the multi-angle spectral response difference analysis method, outputting the anisotropic spectral feature data and the tread texture structure feature data of the tread surface; S3: according to the anisotropic spectral feature data of the tread surface, using the spectral direction gradient analysis method to quantize the directionality feature of the molecular chain orientation of the rubber surface, and outputting the surface molecular orientation feature map, specifically: using the spectral direction gradient analysis method, calculating the spectral change gradient value of the anisotropic spectral feature data of the rubber tread surface in the horizontal and vertical orthogonal directions; according to the spectral change gradient value, establishing the mapping relationship between the spectral change gradient value of the rubber tread surface and the orientation direction of the polymer chain of the rubber tread surface; according to the mapping relationship, point-by-point quantization calculation is performed to obtain the orientation direction and change amplitude information of the rubber molecular chain of the rubber tread surface, and a surface molecular orientation feature map is generated; S4: according to the tread texture structure feature data, using the texture boundary recognition and spatial frequency analysis method, extracting the texture abnormal area position and scale information, and outputting the tread texture abnormal marker map; S5: based on the surface molecular orientation feature map and the tread texture abnormal marker map, establishing the corresponding relationship between the texture abnormal area and the spectral directionality change, and outputting the accurate identification result of the tread surface defects; S6: according to the accurate identification result of the tread surface defects, performing adaptive dynamic adjustment of the tire tread production control.

2. The image recognition-based tire tread production control method according to claim 1, characterized by, S1, specifically: in the tire tread forming process, real-time collection of high-resolution image data of the rubber tread surface by an industrial high-definition camera; using a Gaussian filter algorithm to filter noise of the high-resolution image data of the rubber tread surface, and eliminating random interference signals generated in the image acquisition process; using a histogram equalization method to enhance the brightness of the high-resolution image data after noise filtering, and generating a clear tread texture image.

3. The image recognition-based tire tread production control method according to claim 2, characterized by, S2, specifically: using a multi-angle light source illumination device to project visible light from multiple preset directions to the rubber tread surface corresponding to the clear tread texture image, and obtaining spectral reflectance intensity data of the rubber tread surface; calculating the difference between the spectral reflectance intensity data by a spectral response difference calculation method, obtaining anisotropic spectral feature data of the rubber tread surface; using a two-dimensional Fourier transform method to analyze the spatial frequency spectrum of the texture structure of the rubber tread surface, extracting the frequency spectrum distribution feature of the rubber tread surface texture, and obtaining the tread texture structure feature data of the rubber tread surface.

4. The image recognition-based tire tread production control method according to claim 3, characterized by, S4, specifically: performing frequency domain transformation on the tread texture structure feature data by a spatial frequency analysis method, obtaining spatial frequency feature distribution data; according to the spatial frequency feature distribution data, identifying the boundary position of the texture abnormal area in the rubber tread surface; using a texture boundary recognition method to analyze the connectivity of the boundary position of the texture abnormal area, obtaining the spatial position and spatial scale size information of the texture abnormal area; Based on the spatial position and spatial scale size information of the texture abnormal area, a tire tread texture abnormality marking map is generated.

5. The image recognition-based tire tread production control method according to claim 4, characterized by, S5, specifically: According to the spatial position and spatial scale size information of the texture abnormal area in the tire tread texture abnormality marking map, the texture abnormal area is mapped to the corresponding spatial position on the surface molecular orientation feature map; At the corresponding spatial position, the orientation direction and change amplitude information of the rubber molecular chain of the rubber tire tread surface are extracted; The spatial position and spatial scale size information of the texture abnormal area and the extracted orientation direction and change amplitude information of the rubber molecular chain are analyzed for data correlation to determine the spatial correspondence between the texture abnormal area and the orientation change of the rubber molecular chain on the rubber tire tread surface, and an accurate tire tread surface defect recognition result is output.

6. The image recognition-based tire tread production control method according to claim 5, characterized by, S6, specifically: From the accurate tire tread surface defect recognition result, the defect type, defect position and defect severity information are extracted; According to the defect type and severity information, the pre-set process adjustment rules are called to calculate the target value of the extruder screw speed, the target value of the calender roll pressure, and the target value of the curing machine heating temperature and curing time, respectively; The target value of the extruder screw speed, the target value of the calender roll pressure, and the target value of the curing machine heating temperature and curing time are output as production control instructions to the corresponding equipment.

Images