Brake pad surface defect detection method based on machine vision

Abstract

The invention discloses a brake pad surface defect detection method based on machine vision, and relates to the technical field of image processing. The method comprises the steps that model information and production batch information of a target brake pad are obtained, production scheme configuration is carried out, and a target production scheme is obtained; traversing the equipment processing group set to perform equipment reliability analysis to obtain a processing link reliability factor sequence; determining a surface defect detection bandwidth sequence, and performing surface image acquisition on the brake pad semi-finished product to obtain a processing link surface image set sequence; carrying out image quality enhancement processing to obtain a processing link enhanced surface image set sequence; performing surface defect identification to obtain a processing link surface defect identification result set sequence; and carrying out weighted calculation to obtain a surface defect detection result of the target brake pad. The technical problems that in the prior art, brake pad surface defect detection is low in efficiency and insufficient in accuracy are solved, and the technical effect of improving the detection efficiency and accuracy is achieved.

Description

Technical Field

[0001] The present invention relates to the technical field of image processing, and in particular to a method for detecting surface defects of brake pads based on machine vision. Background Art

[0002] As a key safety component in the automobile braking system, the quality of brake pads directly affects the driving safety of the car. During the production process of brake pads, surface defects (such as cracks, scratches, pores, uneven wear, etc.) may lead to performance degradation and even cause safety accidents. Therefore, surface defect detection of brake pads is a key link to ensure product quality and safety. Existing surface defect detection technology mainly relies on manual detection or traditional machine vision detection methods, which have problems such as low detection accuracy, low efficiency, and susceptibility to subjective factors. Summary of the invention

[0003] The present application provides a brake pad surface defect detection method based on machine vision, which solves the technical problems of low efficiency and insufficient accuracy in brake pad surface defect detection in the prior art.

[0004] In view of the above problems, the present application provides a brake pad surface defect detection method based on machine vision.

[0005] The present application provides a brake pad surface defect detection method based on machine vision, the method comprising: Obtain the model information and production batch information of the target brake pad, configure the production plan based on the model information and the production batch information, and obtain the target production plan, wherein the target production plan includes a processing link sequence and an equipment processing group set, and each processing link corresponds to an equipment processing group; traverse the equipment processing group set to perform equipment reliability analysis to obtain a processing link reliability factor sequence; determine the surface defect detection bandwidth sequence based on the processing link reliability factor sequence, call a high-resolution camera according to the sequence of the processing link sequence, and perform surface image acquisition on the brake pad semi-finished product processed in each processing link according to the surface defect detection bandwidth sequence to obtain a processing link surface image set sequence; traverse the processing link surface image set sequence to perform image quality enhancement processing to obtain a processing link enhanced surface image set sequence; perform surface defect recognition based on the processing link enhanced surface image set sequence to obtain a processing link surface defect recognition result set sequence; perform weighted calculation on the processing link surface defect recognition result set sequence to obtain a surface defect detection result of the target brake pad.

[0006] One or more technical solutions provided in this application have at least the following technical effects or advantages: First, the model information and production batch information of the target brake pad are obtained, and the production plan is configured based on the model information and production batch information to obtain the target production plan, wherein the target production plan includes a processing link sequence and an equipment processing group set, and each processing link corresponds to an equipment processing group. Then, the equipment reliability analysis is performed by traversing the equipment processing group set to obtain the processing link reliability factor sequence. Further, the surface defect detection bandwidth sequence is determined based on the processing link reliability factor sequence, and the high-resolution camera is called to collect the surface image of the brake pad semi-finished product processed by each processing link according to the surface defect detection bandwidth sequence according to the sequence of the processing link sequence, and the processing link surface image set sequence is obtained. Then, the processing link surface image set sequence is traversed to perform image quality enhancement processing to obtain the processing link enhanced surface image set sequence. Next, surface defect recognition is performed based on the processing link enhanced surface image set sequence to obtain the processing link surface defect recognition result set sequence. Finally, the processing link surface defect recognition result set sequence is weightedly calculated to obtain the surface defect detection result of the target brake pad. The technical problems of low efficiency and insufficient accuracy of brake pad surface defect detection in the prior art are solved, and the technical effect of improving detection efficiency and accuracy is achieved. BRIEF DESCRIPTION OF THE DRAWINGS

[0007] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0008] Figure 1 A schematic flow chart of a brake pad surface defect detection method based on machine vision provided in an embodiment of the present application; Figure 2 A schematic diagram of a process flow for obtaining a set of enhanced surface images of a processing step in a machine vision-based brake pad surface defect detection method provided in an embodiment of the present application. DETAILED DESCRIPTION

[0009] The present application solves the technical problems of low efficiency and insufficient accuracy in brake pad surface defect detection in the prior art by providing a brake pad surface defect detection method based on machine vision.

[0010] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of this application.

[0011] It should be noted that the terms "including" and "having" are intended to cover non-exclusive inclusions. For example, a process, method, system, product or server that includes a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or modules that are not explicitly listed or are inherent to these processes, methods, products or devices.

[0012] Examples, such as Figure 1 As shown, the embodiment of the present application provides a brake pad surface defect detection method based on machine vision, wherein the method includes: Obtain the model information and production batch information of the target brake pad, configure the production plan based on the model information and the production batch information, and obtain the target production plan, wherein the target production plan includes a processing link sequence and an equipment processing group set, and each processing link corresponds to an equipment processing group.

[0013] The model information and production batch information of the target brake pad are obtained through the production management system. The model information includes the specific specifications of the brake pad, such as the size of the brake pad (such as diameter, thickness, friction surface width, etc.), material (such as organic, semi-metallic, ceramic, etc.), and possible other special requirements (such as heat resistance, wear resistance, etc.). The production batch information includes the production quantity and batch number of the current batch. According to the model information and production batch information, the production task is configured to generate the target production plan. The target production plan includes the processing link sequence and the equipment processing group set. Among them, the processing link sequence refers to the sequence of each processing step in the production process of the brake pad, including cutting, forming, grinding, spraying and other links. The equipment processing group set refers to the equipment group that performs the tasks of each processing link. Each processing link corresponds to an equipment processing group; each processing link is associated with its corresponding equipment processing group to ensure that each processing link can be matched with the equipment resources in the established order, thereby forming a complete target production plan and providing a basis for the execution of the subsequent production process.

[0014] Furthermore, configuring a production plan based on the model information and the production batch information to obtain a target production plan includes: Acquire a historical model information set, a historical production batch information set and a corresponding historical production plan set as training data; construct a network framework based on a convolutional neural network, use the training data to perform framework supervised training, learn the two-to-one mapping relationship between model information and production batch information and production plans, update the network parameters of the framework according to the output results until the training converges, and obtain a trained plan configurator; use the plan configurator to configure the production plan for the model information and the production batch information, and obtain the target production plan.

[0015] Specifically, a historical model information set, a historical production batch information set and a corresponding historical production plan set are obtained from a database, and these data sets are used as training data, wherein the historical model information and production batch information contain the specification parameters and production quantity of brake pads of different models, and the historical production plan set corresponds to the processing link sequence and equipment processing group configuration plan adopted under each model and batch information; a network framework is constructed based on a convolutional neural network (CNN) to learn the two-to-one mapping relationship between model information, production batch information and production plan, the training data is input into the network framework, and supervised learning is used for training, the error is calculated by the back propagation algorithm, and the parameters of the network framework are iteratively updated based on the output results, and the network weights and parameters are gradually optimized until the network training reaches a convergence state, thereby obtaining an accurate and stable production plan configurator; using the trained plan configurator, the model information and production batch information of the target brake pad are used as input to automatically generate a matching target production plan, which includes an adapted processing link sequence and equipment processing group set to ensure that the production process can be executed efficiently and meet quality requirements.

[0016] The equipment processing group set is traversed to perform equipment reliability analysis to obtain a processing link reliability factor sequence.

[0017] Traverse the set of equipment processing groups, and perform equipment reliability analysis on the equipment processing groups corresponding to each processing link, including evaluating key indicators such as the equipment's operating status, historical fault records, processing accuracy, load stability, and service life. Obtain the reliability coefficient of each device through weighted calculation. On this basis, combine the weight and comprehensive performance of each device in the equipment group in the processing link to generate the reliability factor of the corresponding processing link, arrange the reliability factors of all processing links in sequence, and finally obtain the processing link reliability factor sequence, which provides a basis for the dynamic adjustment of the subsequent surface defect detection bandwidth.

[0018] A surface defect detection bandwidth sequence is determined based on a reliability factor sequence of processing links, and a high-resolution camera is called to collect surface images of the brake pad semi-finished products processed in each processing link according to the surface defect detection bandwidth sequence according to the sequence of processing links, so as to obtain a collection sequence of surface images of the processing links.

[0019] The surface defect detection bandwidth sequence is determined based on the reliability factor sequence of the processing link. First, the reliability factor of the processing link is evaluated. The larger the reliability factor, the higher the stability and reliability of the processing link, and the lower the probability of surface defects. Therefore, a larger detection bandwidth is allocated to the link, and the detection frequency is reduced to improve the detection efficiency. On the contrary, the smaller the reliability factor, the poorer the stability of the processing link, and the higher the probability of surface defects. Therefore, a smaller detection bandwidth is allocated, the detection frequency is increased, and the accuracy and coverage of image acquisition are improved. Subsequently, according to the sequence of the processing links, a high-resolution camera is called to collect surface images of the semi-finished brake pads processed in each processing link. The camera acquisition parameters are dynamically adjusted according to the detection bandwidth of each link to ensure that high-resolution and more detailed images are captured in the key links, and the acquisition efficiency is improved in the stable link. Finally, the collected images are stored in the order of the processing links to form a collection sequence of surface images of the processing links, providing accurate visual data support for subsequent image processing and defect recognition.

[0020] The processing link surface image set sequence is traversed to perform image quality enhancement processing to obtain the processing link enhanced surface image set sequence.

[0021] Traverse the surface image collection sequence of the processing link, and perform image quality enhancement processing on each of them in turn. First, through the image preprocessing steps, including denoising, contrast enhancement and edge detail optimization, the overall quality of the image is improved; then the image quality enhancement algorithm is used to further improve the clarity and texture details of the image, making the surface defect area more prominent, which is convenient for the accuracy of subsequent defect identification. Finally, the images that have been processed with image quality enhancement are stored in the order of the original processing links to generate a processing link enhanced surface image collection sequence, ensuring that the image quality of each processing link reaches the best state, providing high-quality input data for subsequent surface defect identification.

[0022] Furthermore, if Figure 2 As shown, the processing link surface image set sequence is traversed to perform image quality enhancement processing to obtain the processing link enhanced surface image set sequence, including: Acquire an image quality enhancement model, wherein the image quality enhancement model includes an image quality diffuser and an image quality enhancer; Using the image quality diffuser to perform noise interference on the images in the processing link surface image set sequence to obtain the processing link diffuse surface image set sequence; The image quality enhancer is used to enhance and denoise the diffusion surface image set sequence of the processing link to obtain the enhanced surface image set sequence of the processing link.

[0023] Preferably, an image quality enhancement model is obtained, which consists of an image quality diffuser and an image quality enhancer, which are respectively used for image noise simulation and denoising and enhancement processing; the image is interfered with by noise by the image quality diffuser, and a processing link diffuse surface image set sequence is generated by introducing simulated noise into each image in the processing link surface image set sequence; the diffuse image is enhanced and denoised by the image quality enhancer, and the processing link diffuse surface image set sequence is input into the image quality enhancer for processing.

[0024] Further, including: Obtain a diffusion function, wherein the diffusion function is: Among them, L ij is the jth processing link diffusion surface image of the i-th processing link diffusion surface image set in the processing link diffusion surface image set sequence, y ij is the jth processing link surface image of the i-th processing link surface image set in the processing link surface image set sequence, is the preset diffusion intensity, The Gaussian noise is preset; and the image quality diffuser is constructed based on the diffusion function.

[0025] The image quality diffuser uses a diffusion function to process each image y in the surface image set sequence of the processing link ij Perform pixel-by-pixel weighted processing, introduce noise disturbance, and generate a diffusion surface image set L with noise interference characteristics. ij .when When it is larger, the noise has less impact and the features of the original image remain more obvious; When it is smaller, noise dominates, the noise interference characteristics of the image are more obvious, and the robustness of the model is enhanced.

[0026] Through the image quality diffuser constructed based on the diffusion function, the surface image set sequence of the processing link is processed with noise interference to generate a diffuse surface image set sequence. This not only simulates the noise interference in the actual production environment, but also provides high-quality input for the denoising optimization of the subsequent image quality enhancer, effectively improving the processing effect and defect recognition accuracy of the final enhanced image.

[0027] Further, including: Obtain an image quality enhancement function, wherein the image quality enhancement function is: ;in, is the jth processing link enhanced surface image of the i-th processing link enhanced surface image set in the processing link enhanced surface image set sequence, To strengthen the empirical coefficient, , It is the square of the Euclidean distance between the diffusion surface image of the j-th processing link and the surface image of the j-th processing link.

[0028] Using image quality enhancement function , calculate the diffusion image L pixel by pixel ij With the original image y ij The Euclidean distance is mapped to a weight value through an exponential function. The larger the weight value, the more significant the difference of the pixel and the higher the degree of enhancement. After all the diffusion images and original images of the processing links are processed by the image quality enhancement function, a set of processing link enhanced surface images is generated in sequence. The noise of the images in the sequence is effectively removed, and the edge details and texture information are significantly enhanced, providing high-quality input data for subsequent surface defect recognition.

[0029] The difference between the diffused image and the original image is calculated through the image quality enhancement function, and the normalized weight and the enhancement empirical coefficient are combined to achieve effective enhancement of the local details and significant areas of the image, and generate a clear and detail-rich processing link enhanced surface image set sequence, which helps to improve the accuracy and stability of subsequent surface defect detection.

[0030] Surface defect recognition is performed based on the processing step enhanced surface image set sequence to obtain the processing step surface defect recognition result set sequence.

[0031] Surface defect recognition is performed based on a set of enhanced surface images of processing links, specifically including: first, obtaining a pre-trained surface defect recognition model, the surface defect recognition model is trained based on a historical defect image data set through a deep learning algorithm (such as a convolutional neural network CNN), and the characteristics of different types of defects, including edge, texture, color distribution and other information; the set of enhanced surface images of processing links is input into the surface defect recognition model in sequence, and features are extracted and analyzed for each image to identify surface defect areas; based on the recognition results, defects are classified to determine the types of defects, including cracks, scratches, pores, pits, etc., and the specific location, size and distribution information of the defects are marked; the recognition results are sorted to form a set sequence of surface defect recognition results of processing links, so as to provide accurate data support for subsequent weighted calculation and overall defect assessment.

[0032] A weighted calculation is performed on the surface defect recognition result set sequence of the processing link to obtain the surface defect detection result of the target brake pad.

[0033] Based on the reliability factor sequence of the processing link, different weights are assigned to the surface defect recognition results of each processing link. The smaller the reliability factor, the worse the processing stability, and the higher the weight of the defect recognition result of this link, and vice versa. The defect results identified in each processing link, including the number, type, location and size of defects, are weighted and summed according to the corresponding weights to generate the surface defect detection results of the target brake pad.

[0034] Furthermore, weighted calculation is performed on the surface defect recognition result set sequence of the processing link to obtain the surface defect detection result of the target brake pad, including: A first processing link surface defect identification result set is randomly extracted from the processing link surface defect identification result set sequence; the same type of defect indicator feature value extraction is performed on the first processing link surface defect identification result set sequence according to the preset surface defect indicator set to obtain K first same type indicator feature value sets, wherein K is an integer greater than or equal to 1; the center points of the K first same type indicator feature value sets are respectively identified to obtain K first same type indicator feature center values; the K first same type indicator feature center values ​​are weightedly calculated to obtain the first processing link surface defect detection result; the processing link surface defect identification result set sequence is traversed to perform link internal defect identification to obtain a processing link surface defect detection result set; the processing link surface defect detection result set is weightedly calculated to obtain the surface defect detection result.

[0035] Preferably, a first processing link surface defect identification result set is randomly extracted from the processing link surface defect identification result set sequence, and is used as the defect analysis data of the current processing link; according to a preset surface defect indicator set (such as defect type, location, size and severity, etc.), the indicator feature values ​​of the same type of defects are extracted from the first processing link surface defect identification result set to obtain K first indicator feature value sets of the same type, where K is an integer greater than or equal to 1, indicating the number of different defect indicators of the same type identified; the center points of the K first indicator feature value sets of the same type are respectively identified, that is, the defect data is clustered or statistically analyzed, and the indicator feature center value of each defect of the same type is extracted to obtain K The first characteristic center value of the same type of indicator; perform weighted calculation on the K first characteristic center values ​​of the same type of indicator according to the weights, and obtain the surface defect detection result of the current first processing link, reflecting the overall defect situation of the processing link; traverse the sequence of surface defect identification result sets of the processing links, perform similar feature analysis and weighted calculation on the defect identification results within each processing link, and obtain the surface defect detection result sets of the processing links in turn; perform weighted calculation on the surface defect detection result sets of the processing links according to the weights of each processing link, and generate the final target brake pad surface defect detection result, including the overall defect severity, defect distribution and key quality assessment indicators, as the final basis for brake pad quality assessment.

[0036] Further, including: Initialize the aggregation window of the K first characteristic value sets of the same type of indicators to obtain K initial aggregation window centers and K initial aggregation windows, wherein the K window radii of the K initial aggregation windows are determined by the size of the volatility variance of the K first characteristic value sets of the same type of indicators, and the K initial aggregation window centers are the mean values ​​of the K first characteristic values ​​of the same type of indicators of the K first characteristic value sets of the same type of indicators; randomly iterate the K initial aggregation window centers according to the K window radii to obtain K iterative aggregation window centers, and construct K iterative aggregation windows based on the K window radii; respectively determine whether the K iterative aggregation amounts of the K iterative aggregation windows are greater than or equal to the K initial aggregation amounts of the K initial aggregation windows, and if so, continue to iterate based on the K iterative aggregation window centers until a preset number of iterations is met, then stop iterating, and obtain K first characteristic center values ​​of the same type of indicators.

[0037] Specifically, the K first sets of characteristic values ​​of indicators of the same type are analyzed, and the fluctuation variance of each set is calculated to determine the corresponding window radius; the mean of each set of characteristic values ​​of the first indicators of the same type is used as the center of the initial aggregation window to obtain the K initial aggregation window centers; according to the size of the fluctuation variance, the window radius of the K initial aggregation windows is determined to complete the initialization of the K initial aggregation windows. On the basis of the initial aggregation window, the centers of the K initial aggregation windows are randomly iterated and calculated according to the K window radii, and the centers of the aggregation windows are gradually updated to obtain the centers of the K iterative aggregation windows; based on the K window radii, the K iterative aggregation windows are reconstructed based on the centers of the iterative aggregation windows. The aggregation amount of each iterative aggregation window is calculated, that is, the number or weighted density of the characteristic values ​​of the indicators in the window; the aggregation amounts of the K iterative aggregation windows are compared with the aggregation amounts of the corresponding initial aggregation windows respectively to determine whether the aggregation amount of the iterative aggregation window is greater than or equal to the aggregation amount of the initial aggregation window. If the condition is met, the iterative update is continued based on the iterative aggregation window center until the preset maximum number of iterations is reached and the iteration is stopped, and finally the K first characteristic center values ​​of the same type of indicators are obtained, providing accurate feature aggregation results for subsequent defect analysis.

[0038] Further, if not, the K iteration directions for obtaining the K iterative aggregation window centers are set as disabled iteration directions, and iteration continues based on the K initial aggregation window centers until the preset number of iterations is met, then the iteration is stopped to obtain the K first center values ​​of the same type of indicator features.

[0039] If it is determined that the aggregation amount of the K iterative aggregation windows is not greater than or equal to the aggregation amount of the K initial aggregation windows, the iteration direction of the currently obtained K iterative aggregation window centers is set to a disabled iteration direction to limit further iterations in this direction; subsequently, based on the K initial aggregation window centers, the disabled direction is excluded, and random iterative optimization is re-executed to gradually update the aggregation window centers and construct new iterative aggregation windows; when the number of iterations reaches the preset maximum number of iterations, the iteration is stopped, and finally the K first characteristic center values ​​of the same type of indicators are obtained, thereby ensuring the stability and accuracy of the aggregation results, while improving the computational efficiency and convergence speed of characteristic value aggregation.

[0040] In summary, the embodiments of the present application have at least the following technical effects: First, the model information and production batch information of the target brake pad are obtained, and the production plan is configured based on the model information and production batch information to obtain the target production plan, wherein the target production plan includes a processing link sequence and an equipment processing group set, and each processing link corresponds to an equipment processing group. Then, the equipment reliability analysis is performed by traversing the equipment processing group set to obtain the processing link reliability factor sequence. Further, the surface defect detection bandwidth sequence is determined based on the processing link reliability factor sequence, and the high-resolution camera is called to collect the surface image of the brake pad semi-finished product processed by each processing link according to the surface defect detection bandwidth sequence according to the sequence of the processing link sequence, and the processing link surface image set sequence is obtained. Then, the processing link surface image set sequence is traversed to perform image quality enhancement processing to obtain the processing link enhanced surface image set sequence. Next, surface defect recognition is performed based on the processing link enhanced surface image set sequence to obtain the processing link surface defect recognition result set sequence. Finally, the processing link surface defect recognition result set sequence is weightedly calculated to obtain the surface defect detection result of the target brake pad. The technical problems of low efficiency and insufficient accuracy of brake pad surface defect detection in the prior art are solved, and the technical effect of improving detection efficiency and accuracy is achieved.

[0041] It should be noted that the above-mentioned sequence of the embodiments of the present application is only for description and does not represent the advantages and disadvantages of the embodiments. And the above-mentioned specific embodiments of this specification are described. The processes depicted in the accompanying drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0042] The above description is only a preferred embodiment of the present application and is not intended to limit the present application. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present application shall be included in the protection scope of the present application.

[0043] This specification and drawings are merely exemplary illustrations of the present application and are deemed to cover any and all modifications, variations, combinations or equivalents within the scope of the present application. Obviously, a person skilled in the art may make various modifications and variations to the present application without departing from the scope of the present application. Thus, if these modifications and variations of the present application fall within the scope of the present application and its equivalents, the present application intends to include these modifications and variations.

Claims

1. A brake pad surface defect detection method based on machine vision, characterized in that: The method comprises: Obtaining model information and production batch information of a target brake pad, configuring a production plan based on the model information and the production batch information, and obtaining a target production plan, wherein the target production plan includes a processing link sequence and an equipment processing group set, and each processing link corresponds to an equipment processing group; Traversing the equipment processing group set to perform equipment reliability analysis to obtain a processing link reliability factor sequence; Determine a surface defect detection bandwidth sequence based on a processing link reliability factor sequence, call a high-resolution camera according to the sequence of the processing links, and collect surface images of the brake pad semi-finished product processed in each processing link according to the surface defect detection bandwidth sequence to obtain a processing link surface image set sequence; Traversing the processing link surface image set sequence to perform image quality enhancement processing to obtain the processing link enhanced surface image set sequence; Based on the processing link enhanced surface image set sequence, surface defect recognition is performed to obtain a processing link surface defect recognition result set sequence; A weighted calculation is performed on the surface defect recognition result set sequence of the processing link to obtain the surface defect detection result of the target brake pad.

2. The method for detecting brake pad surface defects based on machine vision according to claim 1, characterized in that: Traversing the processing link surface image set sequence to perform image quality enhancement processing to obtain the processing link enhanced surface image set sequence, including: Acquire an image quality enhancement model, wherein the image quality enhancement model includes an image quality diffuser and an image quality enhancer; Using the image quality diffuser to perform noise interference on the images in the processing link surface image set sequence to obtain the processing link diffuse surface image set sequence; The image quality enhancer is used to enhance and denoise the diffusion surface image set sequence of the processing link to obtain the enhanced surface image set sequence of the processing link.

3. The method for detecting brake pad surface defects based on machine vision according to claim 2, characterized in that: include: Obtain a diffusion function, wherein the diffusion function is: ； Among them, L ij is the jth processing link diffusion surface image of the i-th processing link diffusion surface image set in the processing link diffusion surface image set sequence, y ij is the jth processing link surface image of the i-th processing link surface image set in the processing link surface image set sequence, is the preset diffusion intensity, To preset Gaussian noise; The image quality diffuser is constructed based on the diffusion function.

4. The method for detecting brake pad surface defects based on machine vision according to claim 3, characterized in that: include: Obtain an image quality enhancement function, wherein the image quality enhancement function is: ； in, is the jth processing link enhanced surface image of the i-th processing link enhanced surface image set in the processing link enhanced surface image set sequence, To strengthen the empirical coefficient, , It is the square of the Euclidean distance between the diffusion surface image of the j-th processing link and the surface image of the j-th processing link.

5. The method for detecting brake pad surface defects based on machine vision according to claim 1, characterized in that: The surface defect detection result of the target brake pad is obtained by performing weighted calculation on the surface defect recognition result set sequence of the processing link, including: Randomly extracting a first processing link surface defect recognition result set from the processing link surface defect recognition result set sequence; Extracting the characteristic values ​​of the same type of defect indicators from the surface defect identification result set sequence of the first processing link according to the preset surface defect indicator set, and obtaining K first characteristic value sets of the same type of indicators, where K is an integer greater than or equal to 1; Identify the center points of the K first indicator feature value sets of the same type respectively to obtain K first indicator feature center values ​​of the same type; Performing weighted calculation on the K first characteristic center values ​​of the same type of indicators to obtain a surface defect detection result of the first processing link; Traversing the processing link surface defect recognition result set sequence to identify internal defects of the link, and obtaining the processing link surface defect detection result set; A weighted calculation is performed on the surface defect detection result set of the processing link to obtain the surface defect detection result.

6. The method for detecting brake pad surface defects based on machine vision according to claim 5, characterized in that: include: Initialize the aggregation window for the K first characteristic value sets of the same type of indicators to obtain K initial aggregation window centers and K initial aggregation windows, wherein the K window radii of the K initial aggregation windows are determined by the size of the fluctuation variance of the K first characteristic value sets of the same type of indicators, and the K initial aggregation window centers are the mean values ​​of the K first characteristic values ​​of the same type of indicators of the K first characteristic value sets of the same type of indicators; Randomly iterating the K initial aggregation window centers according to the K window radii to obtain K iterative aggregation window centers, and constructing K iterative aggregation windows based on the K window radii; Determine whether the K iterative aggregation amounts of the K iterative aggregation windows are greater than or equal to the K initial aggregation amounts of the K initial aggregation windows. If so, continue to iterate based on the centers of the K iterative aggregation windows until a preset number of iterations is met, then stop iterating and obtain K first center values ​​of the same type of indicator features.

7. The method for detecting brake pad surface defects based on machine vision according to claim 6, characterized in that: If not, the K iteration directions for obtaining the K iteration aggregation window centers are set as disabled iteration directions, and iteration is continued based on the K initial aggregation window centers until the preset number of iterations is met, then the iteration is stopped to obtain the K first center values ​​of the same type of indicator features.

8. The method for detecting brake pad surface defects based on machine vision according to claim 1, characterized in that: The production plan is configured based on the model information and the production batch information to obtain a target production plan, including: Obtaining a historical model information set, a historical production batch information set, and a corresponding historical production plan set as training data; Construct a network framework based on a convolutional neural network, use training data to perform supervised training on the framework, learn the two-to-one mapping relationship between model information and production batch information and production plans, and update the network parameters of the framework according to the output results until the training converges to obtain a trained plan configurator; The production plan is configured for the model information and the production batch information using a plan configurator to obtain the target production plan.

Images