Intelligent recognition system and method for surface defects of platinum-iridium developer ring based on image analysis

Abstract

The present invention relates to the technical field of image analysis, and discloses an intelligent recognition system and method for surface defects of a platinum-iridium developer ring based on image analysis. The present invention first obtains a surface image of a platinum-iridium developer ring, extracts suspected defects, performs median filtering and clustering processing on the suspected defects, obtains cluster centers, merges suspected defects into defect areas according to the cluster centers, and obtains a surface image of a platinum-iridium developer ring with defects extracted; secondly, morphological processing is performed on the surface image of the platinum-iridium developer ring with defects extracted, and defects are marked using 8 adjacent pixels; then, a kernel principal component analysis algorithm is used to extract defect feature parameters, and the kernel parameters of the kernel principal component analysis algorithm are optimized based on a dung beetle optimization algorithm to obtain principal component parameters, and finally, the principal component parameters are used to train an RBF neural network, and defect recognition results are output to achieve recognition of surface defects of the platinum-iridium developer ring.

Description

Technical Field

[0001] The present invention relates to the technical field of image analysis, and in particular to an intelligent recognition system and method for surface defects of a platinum-iridium developer ring based on image analysis. Background Art

[0002] With the rapid development of interventional therapy, successful positioning of cases has become a topic of widespread concern. Platinum-iridium developer rings, as X-ray developer marking and positioning rings, are difficult to process and prone to surface defects. Traditional methods for detecting surface defects of platinum-iridium developer rings have problems such as slow detection speed and low detection accuracy, and do not use high-tech technologies such as image analysis and neural networks, which is not conducive to screening defective platinum-iridium developer rings. Summary of the invention

[0003] In response to the problems in the related art, the present invention provides a system and method for intelligently identifying surface defects of a platinum-iridium developer ring based on image analysis to overcome the above-mentioned technical problems existing in the existing related art.

[0004] To solve the above technical problems, the present invention is achieved through the following technical solutions:

[0005] The present invention is an intelligent recognition method for surface defects of platinum-iridium developer rings based on image analysis, comprising the following steps:

[0006] S1, obtaining a surface image of a platinum-iridium developer ring, generating a surface matrix of a platinum-iridium developer ring to be processed, extracting suspected defects using a background difference method, obtaining a filtered and denoised surface matrix of a platinum-iridium developer ring by median filtering, determining a clustering center of the filtered and denoised surface matrix of a platinum-iridium developer ring by a clustering algorithm, locating and extracting defect areas, and obtaining a surface image of a platinum-iridium developer ring with defects extracted;

[0007] S2, performing morphological processing on the surface image of the platinum-iridium developing ring extracted from the defect to obtain the surface image of the platinum-iridium developing ring after morphological processing, regenerating the surface matrix of the platinum-iridium developing ring after morphological processing, setting 8 adjacent pixel points, marking the defect, and obtaining the surface matrix of the platinum-iridium developing ring after defect marking;

[0008] S3. Extract the surface defect characteristic parameters of the platinum-iridium developing ring of the surface matrix of the platinum-iridium developing ring after the defect marking, calculate the contribution rate after optimizing the kernel parameters in the characteristic parameters based on the dung beetle optimization algorithm, measure the surface defect characteristic parameters of the platinum-iridium developing ring according to the contribution rate, obtain the principal component parameters, use the principal component parameters to train the RBF neural network to obtain the RBF neural network classifier, output the defect recognition result, and realize the recognition of the surface defects of the platinum-iridium developing ring.

[0009] The invention first obtains the surface image of the platinum-iridium developer ring, uses the background difference method to compare the surface image of the normal platinum-iridium developer ring, extracts the suspected defects, performs median filtering on the suspected defects, replaces the gray value of the pixel with the intermediate value, and then uses a clustering algorithm based on the density peak to cluster the suspected defects, finds the cluster center and uses the judgment index to measure the cluster center, completes the clustering of the suspected defects to obtain the defect area, and the clustering algorithm has a fast search speed and high clustering accuracy; secondly, the surface image of the platinum-iridium developer ring with defects is subjected to morphological processing such as expansion, corrosion and opening operations to remove burrs and false edges in the defect area, and then the image in the surface matrix of the platinum-iridium developer ring after morphological processing is processed. The pixel is processed with 8 adjacent pixels, and the defect is marked by marking the 8 adjacent pixels around the pixel. This method accurately marks the defect area and lays the foundation for the subsequent defect feature extraction. The kernel principal component analysis algorithm is used to extract the surface defect characteristic parameters of the platinum-iridium developer ring. The kernel function is introduced to transform the coordinate system by mapping to achieve the goal of feature parameter extraction, and the kernel parameters are optimized based on the dung beetle optimization algorithm to obtain the principal component parameters. The optimization algorithm has the characteristics of strong search ability and fast convergence speed. By converting more feature parameters into fewer comprehensive indicators, the parameter redundancy is reduced and the problem is simplified. Finally, the RBF neural network is trained to realize the recognition of surface defects of the platinum-iridium developer ring.

[0010] Preferably, the S1 comprises the following steps:

[0011] S11, obtaining a surface image of a platinum-iridium developing ring, selecting i×j data sampling points on the surface image of the platinum-iridium developing ring for sampling, performing quantization processing on the data sampling points, and generating a surface matrix A of the platinum-iridium developing ring to be processed as follows:

[0012] ;

[0013] in, Indicates the i-th pixel in the horizontal direction and the j-th pixel in the vertical direction of the surface image of the platinum-iridium developing ring;

[0014] Acquire a defect-free platinum-iridium developing ring surface image, generate a comparison platinum-iridium developing ring surface matrix, compare the pixel values ​​of the pixel points in the platinum-iridium developing ring surface matrix to be processed with the pixel values ​​of the pixel points in the comparison platinum-iridium developing ring surface matrix, if the pixel values ​​of the pixel points in the platinum-iridium developing ring surface matrix to be processed are the same as the pixel values ​​of the pixel points in the comparison platinum-iridium developing ring surface matrix, then set the pixel values ​​of the pixel points in the platinum-iridium developing ring surface matrix to be processed to 0; otherwise, retain the pixel values ​​of the pixel points in the platinum-iridium developing ring surface matrix to be processed, and obtain a platinum-iridium developing ring surface matrix with suspected defects;

[0015] S12. Select a 3×3 neighborhood matrix in the surface matrix of the platinum-iridium developer ring with suspected defects, record it as the first neighborhood matrix, calculate the average grayscale value of the pixel points in the first neighborhood matrix, record it as the first grayscale value average, and use the first grayscale value average to replace the grayscale value of the central pixel point of the first neighborhood matrix; move the 3×3 neighborhood matrix to obtain a second neighborhood matrix, calculate the average grayscale value of the pixel points in the second neighborhood matrix, record it as the second grayscale value average, and use the second grayscale value average to replace the grayscale value of the central pixel point of the second neighborhood matrix; repeatedly move the 3×3 neighborhood matrix until the grayscale values ​​of the pixel points except the 1st row, the i-th row, the 1st column and the j-th column in the surface matrix of the platinum-iridium developer ring with suspected defects are replaced, and obtain the filtered and denoised platinum-iridium developer ring surface matrix and the filtered and denoised platinum-iridium developer ring surface image;

[0016] S13, clustering the suspected defects in the filtered and denoised surface image of the platinum-iridium developer ring using a density peak-based clustering algorithm to determine the final clustering center, the specific steps are as follows:

[0017] S131, calculate the distance between the pixels in the surface matrix of the filtered and denoised platinum-iridium developer ring, record it as the pixel distance matrix, obtain the pixel proximity distance and the pixel local density from the pixel distance in the pixel distance matrix, and set the pixel proximity distance to , the local density of pixels is , then the pixel standardized product The calculation formula is as follows:

[0018] ;

[0019] in, Represents the normalized product of the i-th pixel in the horizontal direction and the j-th pixel in the vertical direction of the surface matrix of the platinum-iridium developer ring for filtering and noise reduction, Represents the distance between the i-th pixel in the horizontal direction and the j-th pixel in the vertical direction of the surface matrix of the Pt-Ir development ring for filtering and noise reduction. Represents the local density of the i-th pixel in the horizontal direction and the j-th pixel in the vertical direction of the surface matrix of the Pt-Ir developing ring for filtering and noise reduction;

[0020] A pixel standardized product set is obtained according to the pixel standardized product, and the pixel standardized product in the pixel standardized product set is equally divided into b equally divided intervals. When there is no pixel standardized product between the b-th equally divided interval and the b-1-th equally divided interval, the pixel corresponding to the pixel standardized product in the b-th equally divided interval is used as a possible clustering center, and the b equally divided intervals are scanned in sequence from back to front to obtain a possible clustering center set;

[0021] S132, calculating the distances of possible cluster centers in the possible cluster center set, merging the possible cluster centers with the same distances in the possible cluster center set to obtain a first cluster center set, then calculating the distances of cluster centers in the first cluster center set, selecting the cluster centers with the closest distances in the first cluster center set and merging them into a second cluster center set; obtaining the cluster center set to be processed ,in Represents the cth final cluster center, and sets the final cluster center cohesion in the set of cluster centers to be processed to , the final cluster center separation degree in the set of cluster centers to be processed is , then the calculation formulas for the final cluster center silhouette coefficient and evaluation index in the cluster center set to be processed are as follows:

[0022] , ;

[0023] in, Indicates the first cluster center in the set to be processed The final cluster center silhouette coefficient, Represents the evaluation index, Indicates the first cluster center in the set to be processed The final cluster center cohesion, Indicates the first cluster center in the set to be processed The final cluster center separation degree;

[0024] Set the judgment index threshold to , when the evaluation index is less than When the cluster center set to be processed is the final cluster center set, otherwise S122 is repeated until the evaluation index is less than , get the final cluster center set;

[0025] S14, the number of final cluster centers in the final cluster center set is the number of defect areas. After the filtered and denoised surface image of the platinum-iridium developer ring is clustered, the final cluster center set locates the defect area, and the minimum circumscribed rectangle method is used to select the defect area. The distance threshold is set to , when the distance between the final cluster centers in the final cluster center set is less than , the defect area located by the final cluster center in the final cluster center set is classified into a minimum circumscribed rectangle, and the minimum circumscribed rectangle is masked to obtain the defect area; repeat S13 to complete the extraction of the defect area and obtain the surface image of the defect-extracted platinum-iridium developing ring.

[0026] The invention obtains the surface image of the platinum-iridium developer ring, uses the background difference method to compare the surface image of the normal platinum-iridium developer ring, extracts suspected defects, performs median filtering on the suspected defects, replaces them with the average value of the grayscale value of the pixel points, and then uses a clustering algorithm based on density peaks to cluster the suspected defects. The clustering algorithm has a fast search speed and high clustering accuracy. It finds the cluster center and uses the judgment index to evaluate the cluster center to obtain the defect area.

[0027] Preferably, S2 comprises the following steps:

[0028] S21, performing morphological processing on the surface image of the defect-extracted platinum-iridium developer ring, setting the defect area in the surface image of the defect-extracted platinum-iridium developer ring as , the structural elements and structural factors are mapped as ,and It includes a structural element origin, and the structural element origin continuously moves in a defect area in the defect extracted platinum iridium developing ring surface image, and when it moves to the edge of the defect area in the defect extracted platinum iridium developing ring surface image, it starts to expand the defect area in the defect extracted platinum iridium developing ring surface image to obtain an expanded platinum iridium developing ring surface image;

[0029] The origin of the structural element continuously moves in the expanded platinum-iridium developing ring surface image, and when it moves to the edge of the expanded platinum-iridium developing ring surface image, the structural element and the structural factor image are corroded relative to the redundant part of the expanded platinum-iridium developing ring surface image to obtain the corroded platinum-iridium developing ring surface image; and then the corroded platinum-iridium developing ring surface image is opened to obtain the platinum-iridium developing ring surface image after morphological processing;

[0030] S22, generating a surface matrix of the platinum-iridium developing ring after morphological processing according to the surface image of the platinum-iridium developing ring after morphological processing, selecting the horizontal first image of the platinum-iridium developing ring surface matrix after morphological processing , vertical pixels, with coordinates marked as , then the lateral first , vertical The eight pixels adjacent to the pixel are recorded as eight adjacent pixels, and their coordinates are , , , , , , and ;

[0031] In the platinum-iridium developing ring surface matrix after the morphological processing, starting from the first pixel point in the upper left corner, the 8 adjacent pixels of the first pixel point in the upper left corner are marked. When the 8 adjacent pixels of the first pixel point in the upper left corner are in the defect area, the 8 adjacent pixels of the first pixel point in the upper left corner in the defect area are marked as serial number 1, otherwise the 8 adjacent pixels of the first pixel point in the upper left corner in the defect area are marked as serial number 0; the first pixel point in the upper left corner moves to the right, and when the 8 adjacent pixels in the defect area are repeatedly marked, the 8 adjacent pixels in the defect area are repeatedly marked with the serial number plus 1 until reaching the upper right corner of the platinum-iridium developing ring surface matrix after the morphological processing, and then the platinum-iridium developing ring surface matrix after the morphological processing is traversed in sequence to complete the defect marking and obtain the platinum-iridium developing ring surface matrix after defect marking.

[0032] The invention performs morphological processing such as dilation, corrosion and opening operation on the surface image of the platinum-iridium developer ring after defect extraction to remove burrs and false edges in the defect area for easy subsequent processing, and then marks the pixels in the surface matrix of the platinum-iridium developer ring after morphological processing with 8 adjacent pixels. This method can accurately mark the defect area and complete the defect marking.

[0033] Preferably, S3 comprises the following steps:

[0034] S31, using a kernel principal component analysis algorithm to extract the platinum-iridium developer ring surface defect characteristic parameters of the platinum-iridium developer ring surface matrix after the defect marking, and using a dung beetle optimization algorithm to optimize the kernel parameters in the characteristic parameters to obtain a set of platinum-iridium developer ring surface defect characteristic parameters, the specific steps are as follows:

[0035] S311, obtaining a surface image of the platinum-iridium developing ring after defect marking according to the surface matrix of the platinum-iridium developing ring after defect marking, and calculating the defect area, the perimeter of the defect area, the coordinates of the center of gravity of the defect area, the circularity of the defect area, the matrix filling degree of the defect area and the linearity of the defect area according to the surface image of the platinum-iridium developing ring after defect marking, and recording them as the surface defect characteristic parameters of the platinum-iridium developing ring;

[0036] S312, the surface defect characteristic parameters of the platinum-iridium developing ring generate a set of surface defect characteristic parameters of the platinum-iridium developing ring, set the mapping function to be f, and map the surface defect characteristic parameter set of the platinum-iridium developing ring in the feature space through the mapping function to obtain a feature space set, which is recorded as ,in Indicates feature space elements, the covariance matrix calculation formula in the feature space is as follows:

[0037] ;

[0038] in, represents the covariance matrix, represents the transpose of the feature space set;

[0039] Using the characteristic equation corresponding to the covariance matrix to obtain the eigenvalues ​​of the covariance matrix and the eigenvectors of the covariance matrix;

[0040] S313, set the parameter vector to , the kernel function is , the kernel function includes kernel parameters , and , and the kernel function polynomial , substituting the parameter vector and the kernel function into the characteristic equation corresponding to the covariance matrix, optimizing the kernel parameters in the characteristic parameters using the dung beetle optimization algorithm, and obtaining the kernel matrix after transformation;

[0041] Among them, the specific steps of using the dung beetle optimization algorithm to optimize the kernel parameters in the feature parameters are as follows:

[0042] The fitness function is set to a kernel function polynomial. In the dung beetle population, the kth dung beetle rolls the dung ball along a straight line. During the rolling process, the probability coefficient is set to , the probability of a dung beetle encountering an obstacle is , represents a random number and ,when When , the probability coefficient is 1, otherwise the probability coefficient is -1. represents the first deviation coefficient, represents the second deviation coefficient, represents a constant and , the intensity of light is , set the current iteration number to h, and the position of the kth dung beetle in the dung beetle population at the hth iteration is , the position of the kth dung beetle in the dung beetle population at the h-1th iteration is , when the dung beetle encounters an obstacle, the position of the kth dung beetle in the dung beetle population at the h+1th iteration is The calculation formula is as follows:

[0043] ;

[0044] When the dung beetle does not encounter an obstacle, the position of the kth dung beetle in the dung beetle population at the h+1th iteration is The calculation formula is as follows:

[0045] ;

[0046] The dung beetles keep looking for safe areas. At this time, the fitness function value of this iteration is found, and the local optimal position in the dung beetle population is set as , the maximum number of iterations is H, and the upper bound of the dung beetle population activity area is , the lower limit of the dung beetle population activity area is , the iteration coefficient is and , then the lower limit of the safe area of ​​dung beetle population is and the upper limit of the safe area of ​​dung beetle population The calculation formula is as follows:

[0047] , ;

[0048] After the dung beetle finds a safe area, the female dung beetle in the dung beetle population produces a chick ball, and the random independent vector is set as and , for the h+1th iteration position of the kth dung beetle in the dung beetle population Update to get the location , at this time position That is, the position of the chick ball, the calculation formula is as follows:

[0049] ;

[0050] The young ball grows into a small dung beetle, and the small dung beetle starts to look for food at the position of the young ball. At this time, the fitness function value is continuously iterated to find the optimal fitness function value. The food position of the small dung beetle is set as the global optimal position in the dung beetle population. The upper bound of the small dung beetle's optimal foraging is , the optimal foraging lower bound for the dung beetle is , the random vector obeying normal distribution is , the random vector is and , then when the small dung beetle is looking for food, the position of the kth dung beetle in the dung beetle population at the h+1th iteration is Update to get the location , the calculation formula is as follows:

[0051] ;

[0052] The best fitness function value of the hth iteration is found. At this time, there are thieving dung beetles in the dung beetle population. The thieving dung beetles steal the food of other dung beetles. The position of the thieving dung beetles is the position of the best fitness function value, which is recorded as ,set up represents a random vector that follows a normal distribution, represents a constant, which is the position of the kth dung beetle in the h+1th iteration of the dung beetle population. Update to get the location , the calculation formula is as follows:

[0053] ;

[0054] The kth dung beetle in the dung beetle population continuously updates its position. When the current number of iterations is equal to the maximum number of iterations, the iteration is stopped to obtain the final dung beetle position. The coordinate values ​​of the final dung beetle position in the three-dimensional space correspond to the kernel parameters , and ;

[0055] S32, the surface defect characteristic parameter set of the platinum-iridium developer ring is projected on the feature space through the kernel matrix. At this time, the weight factor calculation formula is as follows:

[0056] ;

[0057] in, Represents the first The weight factor of the feature space elements, Represents the first The weight of each feature space element;

[0058] The weight factors of the feature space elements in the feature space set are accumulated and summed to obtain the contribution rate; the surface defects of the platinum-iridium developing ring include spots, scratches and pits, and the surface spot defect images of the platinum-iridium developing ring, the surface scratch images of the platinum-iridium developing ring and the surface pit images of the platinum-iridium developing ring are collected, and the surface spot defect images of the platinum-iridium developing ring, the surface scratch images of the platinum-iridium developing ring and the surface pit images of the platinum-iridium developing ring are processed respectively using the kernel principal component analysis algorithm to obtain the contribution rates of the platinum-iridium developing ring surface defect characteristic parameters corresponding to the spots, scratches and pits respectively, and the platinum-iridium developing ring surface defect characteristic parameters corresponding to the spots are extracted with the first, second and third corresponding contribution rates, which are recorded as the spot principal component parameters; the platinum-iridium developing ring surface defect characteristic parameters corresponding to the scratches are extracted with the first, second and third corresponding contribution rates, which are recorded as the scratch principal component parameters; the platinum-iridium developing ring surface defect characteristic parameters corresponding to the pits are extracted with the first, second and third corresponding contribution rates, which are recorded as the pit principal component parameters;

[0059] S33, obtaining a new platinum-iridium developer ring surface image, obtaining new spot principal component parameters, new scratch principal component parameters and new pit principal component parameters, forming a new platinum-iridium developer ring surface detection sample set, dividing the new platinum-iridium developer ring surface detection sample set into a platinum-iridium developer ring surface detection test set and a platinum-iridium developer ring surface detection verification set, setting the RBF neural network input node to 3, the RBF neural network output node to 1, inputting the platinum-iridium developer ring surface detection test set into the RBF neural network, setting the current number of iterations to 1, the maximum number of iterations to L, and when the current number of iterations is equal to the maximum number of iterations, stopping the iteration to obtain a trained RBF neural network; then inputting the platinum-iridium developer ring surface detection verification set into the trained RBF neural network, setting the error threshold to , when the error of the trained RBF neural network is less than When the trained RBF neural network is the RBF neural network classifier, otherwise the weights are adjusted until the error of the trained RBF neural network is less than ;

[0060] The spot principal component parameters, scratch principal component parameters and pit principal component parameters are input into the RBF neural network classifier, the RBF neural network classifier outputs 0, representing that there is no defect on the surface of the platinum-iridium developing ring, the RBF neural network classifier outputs 1, representing that there are spots on the surface of the platinum-iridium developing ring, the RBF neural network classifier outputs 2, representing that there are scratches on the surface of the platinum-iridium developing ring, and the RBF neural network classifier outputs 3, representing that there are pits on the surface of the platinum-iridium developing ring; the recognition of surface defects of the platinum-iridium developing ring is realized.

[0061] The invention extracts the characteristic parameters of the surface defects of the platinum-iridium developer ring by using the kernel principal component analysis algorithm, reduces the parameter redundancy and simplifies the problem by converting more characteristic parameters into fewer comprehensive indicators, thereby achieving the goal of characteristic parameter extraction; and optimizes the kernel parameters based on the dung beetle optimization algorithm to obtain the principal component parameters, which has the characteristics of strong search capability and fast convergence speed; finally, the principal component parameters are used to train the RBF neural network to realize the recognition of the surface defects of the platinum-iridium developer ring.

[0062] The present invention also discloses an intelligent recognition system for surface defects of a platinum-iridium developer ring based on image analysis, which specifically comprises: a platinum-iridium developer ring surface image suspected defect extraction module, a platinum-iridium developer ring surface image defect extraction module, a platinum-iridium developer ring surface image defect marking module and a platinum-iridium developer ring surface image defect recognition module;

[0063] The platinum-iridium developer ring surface image suspected defect extraction module is used to extract suspected defects using a background difference method;

[0064] The platinum-iridium developer ring surface image defect extraction module is used to locate and extract defect areas using median filtering and clustering algorithms;

[0065] The platinum-iridium developer ring surface image defect marking module is used to mark defects using 8 adjacent pixels after morphological processing;

[0066] The platinum-iridium developer ring surface image defect recognition module is used to extract defect characteristic parameters to obtain principal component parameters, and train the RBF neural network to realize the recognition of platinum-iridium developer ring surface defects.

[0067] The present invention has the following beneficial effects:

[0068] 1. The invention obtains the surface image of the platinum-iridium developing ring, uses the background difference method to compare the surface image of the normal platinum-iridium developing ring, extracts suspected defects, performs median filtering on the suspected defects, replaces them with the average value of the grayscale value of the pixel points, and then uses a clustering algorithm based on density peaks to cluster the suspected defects. The clustering algorithm has a fast search speed and high clustering accuracy. It finds the cluster center and uses the judgment index to evaluate the cluster center to obtain the defect area.

[0069] 2. The invention performs morphological processing such as dilation, corrosion and opening operation on the surface image of the platinum-iridium developer ring after defect extraction to remove burrs and false edges in the defect area for easy subsequent processing, and then marks the pixels in the surface matrix of the platinum-iridium developer ring after morphological processing with 8 adjacent pixels. This method can accurately mark the defect area and complete the marking of the defect.

[0070] 3. The invention uses a kernel principal component analysis algorithm to extract the characteristic parameters of the surface defects of the platinum-iridium developer ring, and reduces parameter redundancy and simplifies the problem by converting relatively more characteristic parameters into relatively fewer comprehensive indicators; and optimizes the kernel parameters based on the dung beetle optimization algorithm to obtain the principal component parameters. The optimization algorithm has the characteristics of strong search capability and fast convergence speed; finally, the principal component parameters are used to train the RBF neural network to realize the recognition of the surface defects of the platinum-iridium developer ring.

[0071] Of course, any product implementing the present invention does not necessarily need to achieve all of the advantages described above at the same time. BRIEF DESCRIPTION OF THE DRAWINGS

[0072] In order to more clearly illustrate the technical solutions of the embodiments of the invention, the drawings required for describing the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the invention. For ordinary technicians in this field, they can also obtain drawings based on these drawings without paying creative work.

[0073] Figure 1 A schematic diagram of the process of intelligently identifying surface defects of a platinum-iridium developer ring by the intelligent recognition system for surface defects of a platinum-iridium developer ring based on image analysis provided by the present invention. DETAILED DESCRIPTION

[0074] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0075] In the description of the present invention, it is necessary to understand that the terms "opening", "upper", "lower", "top", "middle", "inside" and the like indicating orientation or positional relationship are only for the convenience of describing the invention and simplifying the description, and do not indicate or imply that the components or elements referred to must have a specific orientation, be constructed and operate in a specific orientation, and therefore cannot be understood as limiting the invention.

[0076] This embodiment discloses an intelligent recognition method for surface defects of platinum-iridium developer rings based on image analysis, which specifically includes the following contents:

[0077] S1, obtaining a surface image of a platinum-iridium developer ring, generating a surface matrix of a platinum-iridium developer ring to be processed, extracting suspected defects using a background difference method, obtaining a filtered and denoised surface matrix of a platinum-iridium developer ring by median filtering, determining a clustering center of the filtered and denoised surface matrix of a platinum-iridium developer ring by a clustering algorithm, locating and extracting defect areas, and obtaining a surface image of a platinum-iridium developer ring with defects extracted;

[0078] The S1 comprises the following steps:

[0079] S11, obtaining a surface image of a platinum-iridium developing ring, selecting i×j data sampling points on the surface image of the platinum-iridium developing ring for sampling, performing quantization processing on the data sampling points, and generating a surface matrix A of the platinum-iridium developing ring to be processed as follows:

[0080] ;

[0081] in, Indicates the i-th pixel in the horizontal direction and the j-th pixel in the vertical direction of the surface image of the platinum-iridium developing ring;

[0082] Acquire a defect-free platinum-iridium developing ring surface image, generate a comparison platinum-iridium developing ring surface matrix, compare the pixel values ​​of the pixel points in the platinum-iridium developing ring surface matrix to be processed with the pixel values ​​of the pixel points in the comparison platinum-iridium developing ring surface matrix, if the pixel values ​​of the pixel points in the platinum-iridium developing ring surface matrix to be processed are the same as the pixel values ​​of the pixel points in the comparison platinum-iridium developing ring surface matrix, then set the pixel values ​​of the pixel points in the platinum-iridium developing ring surface matrix to be processed to 0; otherwise, retain the pixel values ​​of the pixel points in the platinum-iridium developing ring surface matrix to be processed, and obtain a platinum-iridium developing ring surface matrix with suspected defects;

[0083] S12. Select a 3×3 neighborhood matrix in the surface matrix of the platinum-iridium developer ring with suspected defects, record it as the first neighborhood matrix, calculate the average grayscale value of the pixel points in the first neighborhood matrix, record it as the first grayscale value average, and use the first grayscale value average to replace the grayscale value of the central pixel point of the first neighborhood matrix; move the 3×3 neighborhood matrix to obtain a second neighborhood matrix, calculate the average grayscale value of the pixel points in the second neighborhood matrix, record it as the second grayscale value average, and use the second grayscale value average to replace the grayscale value of the central pixel point of the second neighborhood matrix; repeatedly move the 3×3 neighborhood matrix until the grayscale values ​​of the pixel points except the 1st row, the i-th row, the 1st column and the j-th column in the surface matrix of the platinum-iridium developer ring with suspected defects are replaced, and obtain the filtered and denoised platinum-iridium developer ring surface matrix and the filtered and denoised platinum-iridium developer ring surface image;

[0084] S13, clustering the suspected defects in the filtered and denoised surface image of the platinum-iridium developer ring using a density peak-based clustering algorithm to determine the final clustering center, the specific steps are as follows:

[0085] S131, calculate the distance between the pixels in the surface matrix of the filtered and denoised platinum-iridium developer ring, record it as the pixel distance matrix, obtain the pixel proximity distance and the pixel local density from the pixel distance in the pixel distance matrix, and set the pixel proximity distance to , the local density of pixels is , then the pixel standardized product The calculation formula is as follows:

[0086] ;

[0087] in, Represents the normalized product of the i-th pixel in the horizontal direction and the j-th pixel in the vertical direction of the surface matrix of the platinum-iridium developer ring for filtering and noise reduction, Represents the distance between the i-th pixel in the horizontal direction and the j-th pixel in the vertical direction of the surface matrix of the Pt-Ir development ring for filtering and noise reduction. Represents the local density of the i-th pixel in the horizontal direction and the j-th pixel in the vertical direction of the surface matrix of the Pt-Ir developing ring for filtering and noise reduction;

[0088] A pixel standardized product set is obtained according to the pixel standardized product, and the pixel standardized product in the pixel standardized product set is equally divided into b equally divided intervals. When there is no pixel standardized product between the b-th equally divided interval and the b-1-th equally divided interval, the pixel corresponding to the pixel standardized product in the b-th equally divided interval is used as a possible clustering center, and the b equally divided intervals are scanned in sequence from back to front to obtain a possible clustering center set;

[0089] S132, calculating the distances of possible cluster centers in the possible cluster center set, merging the possible cluster centers with the same distances in the possible cluster center set to obtain a first cluster center set, then calculating the distances of cluster centers in the first cluster center set, selecting the cluster centers with the closest distances in the first cluster center set and merging them into a second cluster center set; obtaining the cluster center set to be processed ,in Represents the cth final cluster center, and sets the final cluster center cohesion in the set of cluster centers to be processed to , the final cluster center separation degree in the set of cluster centers to be processed is , then the calculation formulas for the final cluster center silhouette coefficient and evaluation index in the cluster center set to be processed are as follows:

[0090] , ;

[0091] in, Indicates the first cluster center in the set to be processed The final cluster center silhouette coefficient, Represents the evaluation index, Indicates the first cluster center in the set to be processed The final cluster center cohesion, Indicates the first cluster center in the set to be processed The final cluster center separation degree;

[0092] Set the judgment index threshold to , when the evaluation index is less than When the cluster center set to be processed is the final cluster center set, otherwise S122 is repeated until the evaluation index is less than , get the final cluster center set;

[0093] S14, the number of final cluster centers in the final cluster center set is the number of defect areas. After the filtered and denoised surface image of the platinum-iridium developer ring is clustered, the final cluster center set locates the defect area, and the minimum circumscribed rectangle method is used to select the defect area. The distance threshold is set to , when the distance between the final cluster centers in the final cluster center set is less than , the defect area located by the final cluster center in the final cluster center set is classified into a minimum circumscribed rectangle, and the minimum circumscribed rectangle is masked to obtain the defect area; S13 is repeated to complete the extraction of the defect area, and the surface image of the platinum-iridium developer ring after defect extraction is obtained;

[0094] S2, performing morphological processing on the surface image of the platinum-iridium developing ring extracted from the defect to obtain the surface image of the platinum-iridium developing ring after morphological processing, regenerating the surface matrix of the platinum-iridium developing ring after morphological processing, setting 8 adjacent pixel points, marking the defect, and obtaining the surface matrix of the platinum-iridium developing ring after defect marking;

[0095] The S2 comprises the following steps:

[0096] S21, performing morphological processing on the surface image of the defect-extracted platinum-iridium developer ring, setting the defect area in the surface image of the defect-extracted platinum-iridium developer ring as , the structural elements and structural factors are mapped as ,and It includes a structural element origin, and the structural element origin continuously moves in a defect area in the defect extracted platinum iridium developing ring surface image, and when it moves to the edge of the defect area in the defect extracted platinum iridium developing ring surface image, it starts to expand the defect area in the defect extracted platinum iridium developing ring surface image to obtain an expanded platinum iridium developing ring surface image;

[0097] The origin of the structural element continuously moves in the expanded platinum-iridium developing ring surface image, and when it moves to the edge of the expanded platinum-iridium developing ring surface image, the structural element and the structural factor image are corroded relative to the redundant part of the expanded platinum-iridium developing ring surface image to obtain the corroded platinum-iridium developing ring surface image; and then the corroded platinum-iridium developing ring surface image is opened to obtain the platinum-iridium developing ring surface image after morphological processing;

[0098] S22, generating a surface matrix of the platinum-iridium developing ring after morphological processing according to the surface image of the platinum-iridium developing ring after morphological processing, selecting the horizontal first image of the platinum-iridium developing ring surface matrix after morphological processing , vertical pixels, with coordinates marked as , then the lateral first , vertical The eight pixels adjacent to the pixel are recorded as eight adjacent pixels, and their coordinates are , , , , , , and ;

[0099] In the surface matrix of the platinum-iridium developer ring after the morphological treatment, starting from the first pixel point in the upper left corner, mark the 8 adjacent pixels of the first pixel point in the upper left corner, when the 8 adjacent pixels of the first pixel point in the upper left corner are in the defect area, mark the pixels of the 8 adjacent pixels of the first pixel point in the upper left corner in the defect area as serial number 1, otherwise mark the pixels of the 8 adjacent pixels of the first pixel point in the upper left corner in the defect area as serial number 0; the first pixel point in the upper left corner moves to the right, and when the pixels of the 8 adjacent pixels in the defect area are repeatedly marked, the pixels of the 8 adjacent pixels in the defect area are repeatedly marked with the serial number plus 1 until reaching the upper right corner of the surface matrix of the platinum-iridium developer ring after the morphological treatment, and then traverse the surface matrix of the platinum-iridium developer ring after the morphological treatment in sequence to complete the defect marking and obtain the surface matrix of the platinum-iridium developer ring after the defect marking;

[0100] S3, extracting the platinum-iridium developing ring surface defect characteristic parameters of the platinum-iridium developing ring surface matrix after the defect mark, calculating the contribution rate after optimizing the kernel parameters in the characteristic parameters based on the dung beetle optimization algorithm, measuring the platinum-iridium developing ring surface defect characteristic parameters according to the contribution rate, obtaining the principal component parameters, using the principal component parameters to train the RBF neural network to obtain the RBF neural network classifier, outputting the defect recognition result, and realizing the recognition of the surface defects of the platinum-iridium developing ring;

[0101] The S3 comprises the following steps:

[0102] S31, using a kernel principal component analysis algorithm to extract the platinum-iridium developer ring surface defect characteristic parameters of the platinum-iridium developer ring surface matrix after the defect marking, and using a dung beetle optimization algorithm to optimize the kernel parameters in the characteristic parameters to obtain a set of platinum-iridium developer ring surface defect characteristic parameters;

[0103] The specific steps of S31 are as follows:

[0104] S311, obtaining a surface image of the platinum-iridium developing ring after defect marking according to the surface matrix of the platinum-iridium developing ring after defect marking, and calculating the defect area, the perimeter of the defect area, the coordinates of the center of gravity of the defect area, the circularity of the defect area, the matrix filling degree of the defect area and the linearity of the defect area according to the surface image of the platinum-iridium developing ring after defect marking, and recording them as the surface defect characteristic parameters of the platinum-iridium developing ring;

[0105] S312, the surface defect characteristic parameters of the platinum-iridium developing ring generate a set of surface defect characteristic parameters of the platinum-iridium developing ring, set the mapping function to be f, and map the surface defect characteristic parameter set of the platinum-iridium developing ring in the feature space through the mapping function to obtain a feature space set, which is recorded as ,in Indicates feature space elements, the covariance matrix calculation formula in the feature space is as follows:

[0106] ;

[0107] in, represents the covariance matrix, represents the transpose of the feature space set;

[0108] Using the characteristic equation corresponding to the covariance matrix to obtain the eigenvalues ​​of the covariance matrix and the eigenvectors of the covariance matrix;

[0109] S313, set the parameter vector to , the kernel function is , the kernel function includes kernel parameters , and , and the kernel function polynomial , substituting the parameter vector and the kernel function into the characteristic equation corresponding to the covariance matrix, optimizing the kernel parameters in the characteristic parameters using the dung beetle optimization algorithm, and obtaining the kernel matrix after transformation;

[0110] The specific steps of using the dung beetle optimization algorithm to optimize the kernel parameters in the feature parameters are as follows:

[0111] The fitness function is set to a kernel function polynomial. In the dung beetle population, the kth dung beetle rolls the dung ball along a straight line. During the rolling process, the probability coefficient is set to , the probability of a dung beetle encountering an obstacle is , represents a random number and ,when When , the probability coefficient is 1, otherwise the probability coefficient is -1. represents the first deviation coefficient, represents the second deviation coefficient, represents a constant and , the intensity of light is , set the current iteration number to h, and the position of the kth dung beetle in the dung beetle population at the hth iteration is , the position of the kth dung beetle in the dung beetle population at the h-1th iteration is , when the dung beetle encounters an obstacle, the position of the kth dung beetle in the dung beetle population at the h+1th iteration is The calculation formula is as follows:

[0112] ;

[0113] When the dung beetle does not encounter an obstacle, the position of the kth dung beetle in the dung beetle population at the h+1th iteration is The calculation formula is as follows:

[0114] ;

[0115] The dung beetles keep looking for safe areas. At this time, the fitness function value of this iteration is found, and the local optimal position in the dung beetle population is set as , the maximum number of iterations is H, and the upper bound of the dung beetle population activity area is , the lower limit of the dung beetle population activity area is , the iteration coefficient is and , then the lower limit of the safe area of ​​dung beetle population is and the upper limit of the safe area of ​​dung beetle population The calculation formula is as follows:

[0116] ;

[0117] ;

[0118] After the dung beetle finds a safe area, the female dung beetle in the dung beetle population produces a chick ball, and the random independent vector is set as and , for the h+1th iteration position of the kth dung beetle in the dung beetle population Update to get the location , at this time position That is, the position of the chick ball, the calculation formula is as follows:

[0119] ;

[0120] The young ball grows into a small dung beetle, and the small dung beetle starts to look for food at the position of the young ball. At this time, the fitness function value is continuously iterated to find the optimal fitness function value. The food position of the small dung beetle is set as the global optimal position in the dung beetle population. The upper bound of the small dung beetle's optimal foraging is , the optimal foraging lower bound for the dung beetle is , the random vector obeying normal distribution is , the random vector is and , then when the small dung beetle is looking for food, the position of the kth dung beetle in the dung beetle population at the h+1th iteration is Update to get the location , the calculation formula is as follows:

[0121] ;

[0122] The best fitness function value of the hth iteration is found. At this time, there are thieving dung beetles in the dung beetle population. The thieving dung beetles steal the food of other dung beetles. The position of the thieving dung beetles is the position of the best fitness function value, which is recorded as ,set up represents a random vector that follows a normal distribution, represents a constant, which is the position of the kth dung beetle in the h+1th iteration of the dung beetle population. Update to get the location , the calculation formula is as follows:

[0123] ;

[0124] The kth dung beetle in the dung beetle population continuously updates its position. When the current number of iterations is equal to the maximum number of iterations, the iteration is stopped to obtain the final dung beetle position. The coordinate values ​​of the final dung beetle position in the three-dimensional space correspond to the kernel parameters , and ;

[0125] S32, the surface defect characteristic parameter set of the platinum-iridium developer ring is projected on the feature space through the kernel matrix. At this time, the weight factor calculation formula is as follows:

[0126] ;

[0127] in, Represents the first The weight factor of the feature space elements, Represents the first The weight of each feature space element;

[0128] The weight factors of the feature space elements in the feature space set are accumulated and summed to obtain the contribution rate; the surface defects of the platinum-iridium developing ring include spots, scratches and pits, and the surface spot defect images of the platinum-iridium developing ring, the surface scratch images of the platinum-iridium developing ring and the surface pit images of the platinum-iridium developing ring are collected, and the surface spot defect images of the platinum-iridium developing ring, the surface scratch images of the platinum-iridium developing ring and the surface pit images of the platinum-iridium developing ring are processed respectively using the kernel principal component analysis algorithm to obtain the contribution rates of the platinum-iridium developing ring surface defect characteristic parameters corresponding to the spots, scratches and pits respectively, and the platinum-iridium developing ring surface defect characteristic parameters corresponding to the spots are extracted with the first, second and third corresponding contribution rates, which are recorded as the spot principal component parameters; the platinum-iridium developing ring surface defect characteristic parameters corresponding to the scratches are extracted with the first, second and third corresponding contribution rates, which are recorded as the scratch principal component parameters; the platinum-iridium developing ring surface defect characteristic parameters corresponding to the pits are extracted with the first, second and third corresponding contribution rates, which are recorded as the pit principal component parameters;

[0129] S33, obtaining a new platinum-iridium developer ring surface image, obtaining new spot principal component parameters, new scratch principal component parameters and new pit principal component parameters, forming a new platinum-iridium developer ring surface detection sample set, dividing the new platinum-iridium developer ring surface detection sample set into a platinum-iridium developer ring surface detection test set and a platinum-iridium developer ring surface detection verification set, setting the RBF neural network input node to 3, the RBF neural network output node to 1, inputting the platinum-iridium developer ring surface detection test set into the RBF neural network, setting the current number of iterations to 1, the maximum number of iterations to L, and when the current number of iterations is equal to the maximum number of iterations, stopping the iteration to obtain a trained RBF neural network; then inputting the platinum-iridium developer ring surface detection verification set into the trained RBF neural network, setting the error threshold to , when the error of the trained RBF neural network is less than When the trained RBF neural network is the RBF neural network classifier, otherwise the weights are adjusted until the error of the trained RBF neural network is less than ;

[0130] The spot principal component parameters, scratch principal component parameters and pit principal component parameters are input into the RBF neural network classifier, the RBF neural network classifier outputs 0, representing that there is no defect on the surface of the platinum-iridium developing ring, the RBF neural network classifier outputs 1, representing that there are spots on the surface of the platinum-iridium developing ring, the RBF neural network classifier outputs 2, representing that there are scratches on the surface of the platinum-iridium developing ring, and the RBF neural network classifier outputs 3, representing that there are pits on the surface of the platinum-iridium developing ring; the recognition of surface defects of the platinum-iridium developing ring is realized.

[0131] The present invention also discloses an intelligent recognition system for surface defects of a platinum-iridium developer ring based on image analysis, which specifically comprises: a platinum-iridium developer ring surface image suspected defect extraction module, a platinum-iridium developer ring surface image defect extraction module, a platinum-iridium developer ring surface image defect marking module and a platinum-iridium developer ring surface image defect recognition module;

[0132] The platinum-iridium developer ring surface image suspected defect extraction module is used to extract suspected defects using a background difference method;

[0133] The platinum-iridium developer ring surface image defect extraction module is used to locate and extract defect areas using median filtering and clustering algorithms;

[0134] The platinum-iridium developer ring surface image defect marking module is used to mark defects using 8 adjacent pixels after morphological processing;

[0135] The platinum-iridium developer ring surface image defect recognition module is used to extract defect characteristic parameters to obtain principal component parameters, and train the RBF neural network to realize the recognition of platinum-iridium developer ring surface defects.

[0136] In the description of this specification, the description with reference to the terms "one embodiment", "example", "specific example", etc. means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the invention. In this specification, the schematic representation of the above terms does not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described can be combined in any one or more embodiments or examples in a suitable manner.

[0137] The preferred embodiments of the invention disclosed above are only used to help explain the invention. The preferred embodiments do not describe all the details in detail, nor do they limit the invention to the specific implementation methods described. Obviously, many modifications and changes can be made according to the content of this specification. This specification selects and specifically describes these embodiments in order to better explain the principles and practical applications of the invention, so that those skilled in the art can understand and use the invention well.

Claims

1. An intelligent method for identifying surface defects of platinum-iridium developer rings based on image analysis, characterized in that: The steps include: S1, obtaining a surface image of a platinum-iridium developer ring, generating a surface matrix of a platinum-iridium developer ring to be processed, extracting suspected defects using a background difference method, obtaining a filtered and denoised surface matrix of a platinum-iridium developer ring by median filtering, determining a clustering center of the filtered and denoised surface matrix of a platinum-iridium developer ring by a clustering algorithm, locating and extracting defect areas, and obtaining a surface image of a platinum-iridium developer ring with defects extracted; S2, performing morphological processing on the surface image of the platinum-iridium developing ring extracted from the defect to obtain the surface image of the platinum-iridium developing ring after morphological processing, regenerating the surface matrix of the platinum-iridium developing ring after morphological processing, setting 8 adjacent pixel points, marking the defect, and obtaining the surface matrix of the platinum-iridium developing ring after defect marking; S3, extracting the platinum-iridium developing ring surface defect characteristic parameters of the platinum-iridium developing ring surface matrix after the defect mark, calculating the contribution rate after optimizing the kernel parameters in the characteristic parameters, measuring the platinum-iridium developing ring surface defect characteristic parameters according to the contribution rate, obtaining the principal component parameters, using the principal component parameters to train the RBF neural network to obtain the RBF neural network classifier, outputting the defect recognition result, and realizing the recognition of the surface defects of the platinum-iridium developing ring; S31, using a kernel principal component analysis algorithm to extract the platinum-iridium developer ring surface defect characteristic parameters of the platinum-iridium developer ring surface matrix after the defect marking, and using a dung beetle optimization algorithm to optimize the kernel parameters in the characteristic parameters to obtain a set of platinum-iridium developer ring surface defect characteristic parameters; S32, calculating the contribution rate of the surface defect characteristic parameter set of the platinum-iridium developer ring to obtain the principal component parameters, wherein the principal component parameters include the spot principal component parameters, the scratch principal component parameters and the pit principal component parameters; S33, training the RBF neural network to obtain an RBF neural network classifier, inputting the spot principal component parameters, the scratch principal component parameters, and the pit principal component parameters into the RBF neural network classifier to realize the recognition of surface defects of the platinum-iridium developer ring; The method of optimizing the kernel parameters in the feature parameters using the dung beetle optimization algorithm comprises the following steps: Set the fitness function to a kernel function polynomial. A dung beetle rolls a dung ball along a straight line. When the dung beetle rolls the dung ball, the probability coefficient is set to , the probability of a dung beetle encountering an obstacle is , represents a random number and ,when When , the probability coefficient is 1, otherwise the probability coefficient is -1. represents the first deviation coefficient, represents the second deviation coefficient, represents a constant and , the intensity of light is , set the current number of iterations to ， Dung beetle population k Dung beetle h The iteration position is , the first in the dung beetle population k Dung beetle h -1 iteration position is When dung beetles encounter an obstacle, the first k Dung beetle h +1 iteration position The calculation formula is as follows: ； When the dung beetles do not encounter obstacles, the first k Dung beetle h +1 iteration position The calculation formula is as follows: ； The dung beetles keep looking for safe areas. At this time, the fitness function value of this iteration is found, and the local optimal position in the dung beetle population is set as , the maximum number of iterations is H , the upper limit of the dung beetle population activity area is , the lower limit of the dung beetle population activity area is , the iteration coefficient is , then the lower limit of the safe area of ​​dung beetle population is and the upper limit of the safe area of ​​dung beetle population The calculation formula is as follows: ； After the dung beetles find a safe area, the female dung beetles in the dung beetle population produce a chick ball, and the chick ball is located at The chick ball grows into a small dung beetle, and the small dung beetle begins to look for food at the chick ball position. At this time, the fitness function value is continuously iterated to find the optimal fitness function value. When the small dung beetle looks for food, the first k Dung beetle h +1 iteration position Update to get the location ; There are thieving dung beetles in the dung beetle population. The position of the thieving dung beetle is the position of the best fitness function value. k Dung beetle h +1 iteration position Update to get the location ; Dung beetle population k The dung beetle continuously updates its position. When the current number of iterations is equal to the maximum number of iterations, the iteration is stopped to obtain the final dung beetle position. The coordinate values ​​of the final dung beetle position in three-dimensional space correspond to the kernel parameters ; The S33 comprises the following steps: S331, acquiring a new platinum-iridium developer ring surface image, obtaining new spot principal component parameters, new scratch principal component parameters and new pit principal component parameters, forming a new platinum-iridium developer ring surface detection sample set, dividing the new platinum-iridium developer ring surface detection sample set into a platinum-iridium developer ring surface detection test set and a platinum-iridium developer ring surface detection verification set, setting the RBF neural network input node to 3, the RBF neural network output node to 1, inputting the platinum-iridium developer ring surface detection test set into the RBF neural network, setting the current number of iterations to 1, the maximum number of iterations to L, and when the current number of iterations is equal to the maximum number of iterations, stopping the iteration to obtain a trained RBF neural network; Then, the surface detection verification set of the platinum-iridium developer ring is input into the trained RBF neural network, and the error threshold is set to , when the error of the trained RBF neural network is less than When the trained RBF neural network is the RBF neural network classifier, otherwise the weights are adjusted until the error of the trained RBF neural network is less than ; S332. Input the spot principal component parameters, scratch principal component parameters and pit principal component parameters into the RBF neural network classifier. The RBF neural network classifier outputs 0, indicating that there is no defect on the surface of the platinum-iridium developing ring. The RBF neural network classifier outputs 1, indicating that there are spots on the surface of the platinum-iridium developing ring. The RBF neural network classifier outputs 2, indicating that there are scratches on the surface of the platinum-iridium developing ring. The RBF neural network classifier outputs 3, indicating that there are pits on the surface of the platinum-iridium developing ring. The surface defects of the platinum-iridium developing ring are identified.

2. The method for intelligently identifying surface defects of platinum-iridium developer rings based on image analysis according to claim 1, characterized in that: The S1 comprises the following steps: S11, obtaining a surface image of a platinum-iridium developing ring, generating a surface matrix of a platinum-iridium developing ring to be processed, extracting suspected defects using a background difference method, and obtaining a surface matrix of a platinum-iridium developing ring with suspected defects; S12, using median filtering to process the surface matrix of the platinum-iridium developing ring with suspected defects to obtain a surface matrix of the platinum-iridium developing ring with filtered noise reduction; S13, using a clustering algorithm to determine the final cluster center of the filtered and denoised platinum-iridium development ring surface matrix; S14, extracting the defect area through the final cluster center to obtain a surface image of the platinum-iridium developing ring after defect extraction.

3. The method for intelligently identifying surface defects of platinum-iridium developer rings based on image analysis according to claim 2 is characterized in that: The S13 comprises the following steps: S131, calculating the distance and local density of pixels in the surface matrix of the filtered and denoised platinum-iridium developing ring to obtain a standardized product of pixels; dividing the interval according to the standardized product of pixels to obtain a possible cluster center set; S132, calculating the distances of the possible cluster centers in the set of possible cluster centers, merging the possible cluster centers in the set of possible cluster centers with the same distances, and determining the final cluster centers of the filtered and denoised platinum-iridium developing ring surface matrix.

4. The method for intelligently identifying surface defects of platinum-iridium developer rings based on image analysis according to claim 3 is characterized in that: The S2 comprises the following steps: S21, performing expansion, corrosion and opening operations on the surface image of the platinum-iridium developer ring extracted from the defect, completing morphological processing, and obtaining the surface image of the platinum-iridium developer ring after morphological processing; S22. Generate a surface matrix of the platinum-iridium developing ring after morphological processing according to the surface image of the platinum-iridium developing ring after morphological processing, and use 8 adjacent pixels to mark defects on the surface matrix of the platinum-iridium developing ring after morphological processing to obtain a surface matrix of the platinum-iridium developing ring after defect marking.

5. The method for intelligently identifying surface defects of platinum-iridium developer rings based on image analysis according to claim 4 is characterized in that: The method of using 8 adjacent pixels to mark defects on the surface matrix of the platinum-iridium developer ring after morphological processing comprises the following steps: Mark the 8 adjacent pixels of the first pixel in the upper left corner of the platinum-iridium developer ring surface matrix after morphological processing, obtain the marking number, move 8 adjacent pixels to the right, update the marking number until reaching the upper right corner of the platinum-iridium developer ring surface matrix after morphological processing, and then traverse the platinum-iridium developer ring surface matrix after morphological processing in sequence to complete the defect marking.

6. A system for realizing the method for intelligently identifying surface defects of platinum-iridium developer rings based on image analysis as described in any one of claims 1 to 5, characterized in that: Specifically include: A platinum-iridium developing ring surface image suspected defect extraction module, a platinum-iridium developing ring surface image defect extraction module, a platinum-iridium developing ring surface image defect marking module and a platinum-iridium developing ring surface image defect recognition module; The platinum-iridium developer ring surface image suspected defect extraction module is used to extract suspected defects using a background difference method; The platinum-iridium developer ring surface image defect extraction module is used to locate and extract defect areas using median filtering and clustering algorithms; The platinum-iridium developer ring surface image defect marking module is used to mark defects using 8 adjacent pixels after morphological processing; The platinum-iridium developer ring surface image defect recognition module is used to extract defect characteristic parameters to obtain principal component parameters, and train the RBF neural network to realize the recognition of platinum-iridium developer ring surface defects.

Images