Intelligent detection method and system for shaft forgings

Abstract

The invention relates to the technical field of image processing, in particular to an intelligent detection method and system for shaft forgings. The method comprises the following steps: acquiring gray values, positions and gradient directions of all pixel points in a surface image of the shaft forging; clustering all pixel points in the surface image by using a condensed hierarchical clustering algorithm to obtain a plurality of regions; calculating the suspicious degree of each region; determining a scratch burr area in the surface image of the shaft forging to realize intelligent detection of the shaft forging; in each round of iteration process of the condensed hierarchical clustering algorithm, for two to-be-combined clusters in each round of iteration process, calculating the separation degree of the two clusters; calculating the feature similarity of the two clustering clusters; determining the number of clustering clusters of the current iteration; a plurality of regions of the surface image of the shaft forging are determined. By dynamically adjusting the number of clusters, the problem of wrong merging or premature iteration stopping of a traditional condensed hierarchical clustering algorithm under the fixed number of clusters is avoided.

Description

Technical Field

[0001] The present invention relates to the technical field of image processing, and in particular to an intelligent detection method and system for shaft forgings. Background Art

[0002] Shaft forgings are key components widely used in the machinery manufacturing industry, especially in the aviation, automobile, energy and other industries. Shaft forgings usually need to withstand large loads and stresses, so the manufacturing quality is crucial to the performance and reliability of the entire equipment. During the forging process, due to the difference in the purity of metal materials, for example, the impurities such as sulfur contained in a batch of billets will hinder the relative sliding between metal grains, thereby reducing the fluidity of the billet, resulting in increased friction between the internal edge structure of the mold and the billet, and different degrees of uneven surface scratches and burrs. Scratches and burrs will appear on the surface of the forging after demolding. If the above problems are not discovered and handled in time, stress concentration will occur in this area of ​​the forging, thereby affecting the performance of the shaft forging. Therefore, it is necessary to divide different scratch and burr areas based on the distribution characteristics of the image, and the idea of ​​the Agglomerative Nesting clustering algorithm (AGNES) is to divide the samples in the image data set into several categories, so that the samples in the same category have a high degree of similarity, to distinguish the scratches and burrs on the surface of the forging, and then evaluate the surface quality of the forging.

[0003] The patent application document with publication number CN116452542A discloses a GIS partial discharge defect diagnosis method based on a neural supervised decision tree, which relates to the technical field of insulation status diagnosis of power equipment, and includes the following steps: performing agglomerative hierarchical clustering based on the AGNES algorithm to construct a decision tree structure, replacing the final linear fully connected layer of the CNN neural network as an induction layer, and using the output vector of the CNN neural network as a supervision layer to perform supervised correction on the output of the induction layer to construct an improved CNN neural network; by collecting the partial discharge phase distribution map of the GIS as a data set, the improved CNN neural network is trained to construct a neural network model for identifying whether the target GIS has partial discharge defects.

[0004] However, the above patent application documents are not aimed at surface inspection of shaft forgings, and the traditional AGNES algorithm obtains the two nearest clusters based on the nearest pixel point during the iteration process, and merges them until the preset number of clusters is reached, thereby distinguishing different scratch and burr areas. The above patent application documents do not solve the problem that when scratches and burrs appear on the surface of the forging due to the influence of the purity of the billet, since the traditional AGNES algorithm uses a preset number of clusters, the AGNES algorithm may have mis-mergers, or stop iterating when it cannot accurately divide different scratch and burr areas, making the final clustering result inaccurate, affecting the quality inspection of shaft forgings. Summary of the invention

[0005] In order to solve the problem that the AGNES algorithm uses a preset number of clusters, which may lead to mis-merging during the iteration process, or the iteration is stopped when different scratch and burr areas cannot be accurately divided, resulting in inaccurate final clustering results and affecting the quality inspection of shaft forgings, the present invention provides an intelligent inspection method and system for shaft forgings.

[0006] In the first aspect, the present invention provides an intelligent detection method for shaft forgings, which adopts the following technical solution: A method for intelligent detection of shaft forgings comprises: obtaining the gray value, position and gradient direction of all pixel points in the surface image of the shaft forging; clustering all pixel points in the surface image using an agglomerative hierarchical clustering algorithm to obtain multiple regions, recording any region as a target region, calculating the suspicious degree of the target region based on the average gray value and number of all pixel points in the target region, determining the scratch and burr region in the surface image of the shaft forging, so as to realize intelligent detection of the shaft forgings; in each round of iterative clustering using the agglomerative hierarchical clustering algorithm, in response to the number of clustering clusters in two consecutive rounds of iteration being the same, the iteration is terminated, and the clusters at the time of the termination of the iteration are clustered. The cluster is recorded as the final cluster cluster, and the surface image of the shaft forging is divided into multiple areas according to the final cluster cluster, including: for two cluster clusters to be merged in each round of iteration, based on the difference in the average grayscale value of all pixels between the two cluster clusters and the spatial distance between the positions of the centroids of the two cluster clusters, the separation of the two cluster clusters is calculated; based on the difference in the standard deviation of the gradient direction of all pixels between the two cluster clusters, the feature similarity of the two cluster clusters is calculated; the ratio of the feature similarity to the separation is used as the merging quality of this round of iteration, and the number of cluster clusters in this round of iteration is determined according to the size of the merging quality of this round of iteration and the previous round of iteration.

[0007] By dynamically adjusting the number of clusters, the problem of mis-merging or premature cessation of iteration that occurs in the traditional AGNES algorithm under a fixed number of clusters is avoided, ensuring that the clustering process is more flexible and accurate; based on the number of pixels and grayscale values ​​of the clusters, the suspiciousness of the defective area can be intelligently evaluated, thereby improving the reliability of detection; combining multiple features such as the grayscale value, gradient direction and spatial position of the pixel, the ability to identify complex surface defects is improved; by calculating the separation degree and feature similarity, the accuracy of the clustering results is ensured, the situation of mis-merging is reduced, and the accuracy of detection is improved.

[0008] Furthermore, the gradient direction is obtained using a Sobel operator.

[0009] Furthermore, the separation degree satisfies the following relationship: ; In the formula, For the The separation degree of the two clusters to be merged during the round iteration process, and Respectively The first and The average gray value of all pixels in a cluster. For the The spatial distance between the centroids of the two clusters to be merged during the round iteration. and Respectively The first The horizontal and vertical coordinates of all pixel locations in the clusters are: and The first The horizontal and vertical coordinates of all pixel locations in the clusters are: is the standard normalization function.

[0010] By comprehensively considering the grayscale differences and spatial position differences between clusters, the clustering process can more comprehensively identify the heterogeneity between different clusters, avoiding inaccurate merging that may be caused by considering only a single feature, helping to improve the accuracy of clustering results, reduce erroneous merging or separation, and thus improve the quality of detection tasks; through normalization processing, different features can be processed uniformly, avoiding the influence of dimensional differences on clustering results, and ensuring the fairness and accuracy of separation calculation.

[0011] Furthermore, the feature similarity satisfies the following relationship: ; In the formula, For the The feature similarity of the two clusters to be merged during the round of iterations, and Respectively The first Clusters and The standard deviation of the gradient direction of all pixels in a cluster, is a natural exponential function.

[0012] The gradient standard deviation as a feature can effectively reduce the impact of noise, because it considers the local changes of pixels in the image rather than the grayscale value of a single pixel, so it can reduce the mismerging or separation caused by noise or outliers; the gradient direction reflects the edge and texture information in the image. By utilizing this feature, the structural consistency in the image can be better maintained during the clustering process, thereby avoiding clustering errors caused by loss of image details; the standard deviation difference is smoothed by an exponential function to ensure that clusters with lower similarity will not be mistakenly merged during the clustering iteration process, making the clustering process more stable.

[0013] Furthermore, the number of clusters satisfies the following relationship: ; In the formula, For the The number of clusters in the round of iteration, is the initial number of clusters, is the number of iterations, For the The feature similarity of the two clusters to be merged during the round of iterations, For the The separation degree of the two clusters to be merged during the round iteration process, For the The feature similarity of the two clusters to be merged during the round of iterations, For the The separation degree of the two clusters to be merged during the iteration process.

[0014] By deciding whether to merge clusters based on the ratio of similarity to separation, unreasonable cluster merging can be avoided, ensuring the accuracy and rationality of the clustering results, thereby improving the clustering quality; dynamically adjusting the number of clusters can adaptively control the clustering process according to the actual structure of the data, making the final clustering results more stable and reducing poor clustering caused by data noise or irregularity; by gradually reducing the number of clusters, the convergence speed of the clustering process can be effectively accelerated, thereby improving computational efficiency while ensuring clustering accuracy.

[0015] Furthermore, the degree of suspicion satisfies the following relationship: ; In the formula, For the The suspiciousness of the target area, For the The number of all pixels in the target area, For the The average gray value of all pixels in the target area.

[0016] By combining the number of pixels and the average grayscale value in the target area, it is possible to accurately determine whether each area is suspicious. When there are fewer pixels in the area or the grayscale value distribution is abnormal, the area will be more suspicious, helping to discover potential errors or inappropriate clusters; when the number of pixels in a certain area is too small or the grayscale value is abnormal, the suspicion level will increase, thereby automatically identifying abnormal areas and improving detection accuracy.

[0017] Furthermore, the determination of scratch and burr areas in the surface image of shaft forgings to achieve intelligent detection of shaft forgings includes: in response to the suspicion level being greater than a preset abnormality threshold, identifying the target area as a scratch and burr area; in response to the number of scratch and burr areas being greater than a preset quality threshold, identifying that there are significant scratch and burr areas on the surface of the shaft forging, issuing an early warning prompt, and completing the intelligent detection of shaft forgings.

[0018] In a second aspect, the present invention provides an intelligent detection system for shaft forgings, which adopts the following technical solution: A shaft forging intelligent detection system comprises: a processor and a memory, wherein the memory stores computer program instructions, and when the computer program instructions are executed by the processor, the above-mentioned shaft forging intelligent detection method is implemented.

[0019] By adopting the above technical solution, the above-mentioned intelligent detection method for shaft forgings is generated into a computer program and stored in a memory to be loaded and executed by a processor, so that a terminal device is made according to the memory and the processor for easy use.

[0020] The present invention has the following technical effects: In each iteration of the agglomerative hierarchical clustering algorithm, multiple factors are comprehensively considered, such as the difference in average grayscale value of pixels between clusters, the spatial distance of centroids, the difference in standard deviation of pixel gradient direction, etc., the merging quality is calculated, and the number of clusters is determined based on the comparison with the previous iteration. This reduces the occurrence of mismerging and can more accurately divide different scratch and burr areas, thereby improving the accuracy of the final clustering results. Since the clustering results are more accurate, the identification and distinction of various features in the surface images of shaft forgings are more accurate, which helps to more reliably realize the intelligent detection of shaft forgings, provide more powerful data support for the quality detection of shaft forgings, improve the accuracy and effectiveness of quality detection, and better control the quality of shaft forgings. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] By reading the detailed description below with reference to the accompanying drawings, the above and other purposes, features and advantages of the exemplary embodiments of the present invention will become readily understood. In the accompanying drawings, several embodiments of the present invention are shown in an exemplary and non-restrictive manner, and the same or corresponding numbers are the same or corresponding parts.

[0022] Figure 1 It is a flow chart of a method for intelligent detection of shaft forgings in an embodiment of the present invention.

[0023] Figure 2 It is a method flow chart of step S2 in an intelligent detection method for shaft forgings according to an embodiment of the present invention. DETAILED DESCRIPTION

[0024] 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 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 those skilled in the art without creative work are within the scope of protection of the present invention.

[0025] It should be understood that when the terms "first", "second", etc. are used in the claims, descriptions, and drawings of the present invention, they are only used to distinguish different objects, rather than to describe a specific order. The terms "include" and "comprise" used in the description and claims of the present invention indicate the presence of the described features, wholes, steps, operations, elements, and / or components, but do not exclude the presence or addition of one or more other features, wholes, steps, operations, elements, components, and / or their collections.

[0026] The embodiment of the present invention discloses an intelligent detection method for shaft forgings, referring to Figure 1 , comprising steps S1 to S4: S1: Obtain the grayscale value, position and gradient direction of all pixels in the surface image of shaft forgings.

[0027] By using an industrial camera to collect images in a uniformly illuminated environment, and selecting the appropriate shooting angle and distance according to the size and shape of the shaft forging surface, we can ensure that the surface details can be fully captured. The surface image is grayed to obtain the gray value of each pixel and its position in the image.

[0028] Specifically, the gradient direction is obtained by using a Sobel Operator (Sobel).

[0029] S2: Using the agglomerative hierarchical clustering algorithm, all pixels in the surface image are clustered to obtain multiple regions.

[0030] In each round of clustering using the agglomerative hierarchical clustering algorithm, refer to Figure 2 , step S2 includes steps S201 to S204, which are specifically as follows: S201: For two clusters to be merged in each round of iteration, the separation degree of the two clusters is calculated.

[0031] It should be noted that the traditional AGNES clustering algorithm continuously merges the two closest clusters until the preset number of clusters is reached; however, due to the influence of the purity of the billet, scratches and burrs of varying degrees appear on the surface of the forging and are unevenly distributed. With the continuous iteration of clustering, two merged clusters are obtained. If the categories generated by a round of iterative merging can already distinguish different scratch and burr areas on the surface, but the preset number of stop clusters has not been reached, it is necessary to continue iterative merging, which will cause some subtle burrs or scratch areas to be merged with normal areas. Therefore, for each of the two clusters that meet the merging conditions, the overall pixel distribution difference between the two clusters is analyzed. Since the grayscale difference at the centroid of the burr or scratch area is significantly different from the centroid of the normal area and the spatial distribution distance is also different, if an incorrect merger occurs, the grayscale value and spatial position corresponding to the centroid of the two clusters will be significantly different. Therefore, the separation between the two clusters can be evaluated by analyzing their differences.

[0032] For the two clusters to be merged in each round of iteration, the separation degree of the two clusters is calculated based on the difference in the average grayscale values ​​of all pixels between the two clusters and the spatial distance between the centroids of the two clusters.

[0033] Specifically, the separation degree satisfies the following relationship: ; In the formula, For the The separation degree of the two clusters to be merged during the round iteration process, and Respectively The first and The average gray value of all pixels in a cluster. For the The spatial distance between the centroids of the two clusters to be merged during the round iteration. and Respectively The first The horizontal and vertical coordinates of all pixel locations in the clusters are: and The first The horizontal and vertical coordinates of all pixel locations in the clusters are: is the standard normalization function.

[0034] in, Indicates The absolute difference of the average grayscale values ​​of all pixels in the two clusters to be merged during the iteration. The larger the value, the more significant the difference in grayscale between the two clusters. Indicates The spatial distance between the centroids of the two clusters to be merged during the iteration. The larger the value, the farther the centroids of the two clusters are distributed in space. In this case, if there is a significant difference in grayscale between the two clusters, it increases the possibility that the two clusters belong to different types of features, which means that the first During the iteration process, the separation degree of the two clusters to be merged is large.

[0035] S202: Calculate feature similarity between two clusters.

[0036] It should be noted that when scratches and burrs appear on the surface of actual shaft forgings, they are mostly narrow and uneven, and are distributed in the image in various forms. If the two clusters to be merged in a certain round happen to be adjacent to the normal area and one end of the scratch and burr area in the image, since the abnormal area appears to be narrow and thin, the spatial distance between the centroids of the two clusters is relatively large, and the obtained separation is also large, which is reasonable. If the normal area and the scratch and burr area are parallel and adjacent, the spatial distance between the centroids of the two clusters may be relatively small, resulting in a smaller separation, which is obviously not in line with expectations. Scratches and burrs are generated by the friction of the internal edge structure of the mold, that is, the scratches and burrs in the same area have a clear and consistent gradient direction, while the surface morphology of the normal area is usually smoother and has no obvious directional characteristics. Therefore, the difference in the stability of the gradient features of the two clusters can be analyzed to further evaluate the feature similarity between the two clusters.

[0037] The feature similarity between the two clusters is calculated based on the difference in the standard deviation of the gradient direction of all pixels between the two clusters.

[0038] Specifically, the feature similarity satisfies the following relationship: ; In the formula, For the The feature similarity of the two clusters to be merged during the round of iterations, and Respectively The first Clusters and The standard deviation of the gradient direction of all pixels in a cluster, is a natural exponential function.

[0039] in, Indicates The absolute difference of the standard deviation of the gradient direction of all pixels in the two clusters to be merged during the iteration. The larger the value, the greater the discreteness of the gradient direction of one cluster, that is, there is a large fluctuation in the gradient direction, and the gradient direction of the other cluster is relatively consistent, which means that the first cluster is more discrete. In the iterative process, the two clusters to be merged do not belong to the same type of surface features, so the feature similarity of the two clusters is lower.

[0040] S203: Determine the number of clusters in this round of iteration.

[0041] It should be noted that the traditional algorithm uses the number of clusters preset as the iteration stop condition based on experience, so as to divide different regions according to the clustering results. In the actual processing process, the fluidity of the billet is easily affected by its purity, which leads to different degrees of friction in the internal structure of the mold, resulting in uneven scratches and burrs on the surface of the forging. According to experience, it is impossible to obtain the number of different categories on the current forging surface. If the number of preset clusters is too small, the algorithm may have multiple mergers, which will cause some scratch and burr areas to be mistakenly merged into the normal area later; if the number of preset clusters is too large, the algorithm will merge less times, and may stop the iteration without fully dividing different scratch and burr areas, resulting in inaccurate final clustering results. Therefore, it is necessary to adaptively obtain the number of clusters at the end of the iteration according to the surface distribution characteristics of the shaft forgings. Considering that the algorithm is in the process of iterative merging, each time the merging condition is met, the separation degree of the two clusters to be merged and the feature similarity of the current two clusters are combined to comprehensively evaluate the quality of each merger, and the number of subsequent final clusters is adaptively obtained according to the change of the merger quality of adjacent iterations.

[0042] The ratio of the feature similarity to the separation degree is used as the merging quality of this round of iteration, and the number of clusters of this round of iteration is determined according to the merging quality of this round of iteration and the previous round of iteration.

[0043] Specifically, the number of clusters satisfies the following relationship: ; In the formula, For the The number of clusters in the round of iteration, is the initial number of clusters, is the number of iterations, For the The feature similarity of the two clusters to be merged during the round of iterations, For the The separation degree of the two clusters to be merged during the round iteration process, For the The feature similarity of the two clusters to be merged during the round of iterations, For the The separation degree of the two clusters to be merged during the iteration process.

[0044] in, Indicates The quality of the round of iterative merging is The smaller the separation between the two clusters in the round of iterative merging, and the higher the feature similarity between the two clusters, the greater the possibility that the two clusters belong to the same type of features, which means that the quality of this round of merging is higher; , that is, If the quality of the merging in the first round of iterations is better or unchanged than that in the previous round of merging, it means that the If the two clusters to be merged in the iteration belong to the same region, the merging operation can be continued iteratively. , that is, The worse the quality of the merging in the first round of iterations is than the quality of the merging in the previous round, the The two clusters to be merged in the iteration belong to different class areas. If there is a mis-merge, the merging should be stopped and the iteration should be terminated.

[0045] S204: Determine multiple areas of the surface image of the shaft forging.

[0046] In response to the fact that the number of clusters in two consecutive iterations is the same, the iteration is terminated. When the iteration is terminated, the two clusters to be merged in the iteration process stop being merged. The cluster at the time of the iteration termination is recorded as the final cluster. According to the final cluster, the surface image of the shaft forging is divided into multiple regions, and all pixel points in each cluster form a region.

[0047] S3: Calculate the suspiciousness level of each area.

[0048] Any area is recorded as the target area, and the suspiciousness of the target area is calculated based on the average gray value and number of all pixels in the target area.

[0049] Specifically, the degree of suspicion satisfies the following relationship: ; In the formula, For the The suspiciousness of the target area, For the The number of all pixels in the target area, For the The average gray value of all pixels in the target area.

[0050] Among them, since the scratch burr area is often narrowly distributed and darker than the normal area, that is, the clustering in the scratch burr area is characterized by a small number of pixels and a low grayscale mean. The smaller the value, the more likely the target cluster is the scratch burr area, that is, the more suspicious the target cluster is.

[0051] S4: Determine the scratch and burr areas in the surface image of shaft forgings to achieve intelligent detection of shaft forgings.

[0052] Specifically, the method of determining the scratch burr area in the surface image of the shaft forging to realize intelligent detection of the shaft forging includes: In response to the suspicious degree being greater than a preset abnormal threshold, determining the target area to be a scratch burr area; In response to the fact that the number of scratch and burr areas is greater than a preset quality threshold, it is determined that there are significant scratch and burr areas on the surface of the shaft forging, indicating that there are quality problems with the shaft forging and that it needs to be dealt with in a timely manner. An early warning prompt is issued to complete the intelligent detection of shaft forgings.

[0053] Implementers can set the anomaly threshold and quality threshold according to specific implementation circumstances. For example, the anomaly threshold is 0.8 and the quality threshold is 3.

[0054] An embodiment of the present invention further discloses an intelligent detection system for shaft forgings, including a processor and a memory, wherein the memory stores computer program instructions, and when the computer program instructions are executed by the processor, an intelligent detection method for shaft forgings according to the present invention is implemented.

[0055] The above system also includes other components well known to those skilled in the art, such as a communication bus and a communication interface, and their configuration and functions are known in the art, so they will not be described in detail here.

[0056] In the present invention, the aforementioned memory may be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, apparatus, or device. For example, a computer-readable storage medium may be any suitable magnetic storage medium or magneto-optical storage medium, or any other medium that can be used to store the required information and can be accessed by an application, module, or both. Any such computer storage medium may be part of the device or accessible or connectable to the device.

[0057] Although this specification has shown and described a number of embodiments of the present invention, it will be apparent to those skilled in the art that such embodiments are provided by way of example only. Those skilled in the art will conceive of many modifications, changes and alternatives without departing from the ideas and spirit of the present invention. It should be understood that in the practice of the present invention, various alternatives to the embodiments of the present invention described herein may be employed.

[0058] The above are all preferred embodiments of the present invention, and are not intended to limit the protection scope of the present invention. Therefore, any equivalent changes made based on the structure, shape, and principle of the present invention should be included in the protection scope of the present invention.

Claims

1. An intelligent detection method for shaft forgings, characterized in that: include: Obtain the gray value, position and gradient direction of all pixels in the surface image of shaft forgings; Using the agglomerative hierarchical clustering algorithm, all pixels in the surface image are clustered to obtain multiple regions. Any region is recorded as the target region. Based on the average gray value and number of all pixels in the target region, the suspicious degree of the target region is calculated to determine the scratch and burr region in the surface image of the shaft forging, so as to realize the intelligent detection of shaft forgings. In each round of iteration of clustering using the agglomerative hierarchical clustering algorithm, in response to the number of clusters in two consecutive iterations being the same, the iteration is terminated, and the cluster at the time of the termination of the iteration is recorded as the final cluster. According to the final cluster, the surface image of the shaft forging is divided into multiple regions, including: For the two clusters to be merged in each round of iteration, the separation of the two clusters is calculated based on the difference in the average grayscale values ​​of all pixels between the two clusters and the spatial distance between the centroids of the two clusters; the feature similarity of the two clusters is calculated based on the difference in the standard deviation of the gradient direction of all pixels between the two clusters; The ratio of the feature similarity to the separation degree is used as the merging quality of this round of iteration, and the number of clusters of this round of iteration is determined according to the merging quality of this round of iteration and the previous round of iteration.

2. The intelligent detection method for shaft forgings according to claim 1 is characterized in that: The gradient direction is obtained using the Sobel operator.

3. The intelligent detection method for shaft forgings according to claim 1 is characterized in that: The separation degree satisfies the following relationship: ； In the formula, For the The separation degree of the two clusters to be merged during the round iteration process, and Respectively The first and The average gray value of all pixels in a cluster. For the The spatial distance between the centroids of the two clusters to be merged during the round iteration. and Respectively The first The horizontal and vertical coordinates of all pixel locations in the clusters are: and Respectively The first The horizontal and vertical coordinates of all pixel locations in the clusters are: is the standard normalization function.

4. The intelligent detection method for shaft forgings according to claim 1 is characterized in that: The feature similarity satisfies the following relationship: ； In the formula, For the The feature similarity of the two clusters to be merged during the round of iterations, and Respectively The first Clusters and The standard deviation of the gradient direction of all pixels in a cluster, is a natural exponential function.

5. The intelligent detection method for shaft forgings according to claim 1 is characterized in that: The number of clusters satisfies the following relationship: ； In the formula, For the The number of clusters in the round of iteration, is the initial number of clusters, is the number of iterations, For the The feature similarity of the two clusters to be merged during the round of iterations, For the The separation degree of the two clusters to be merged during the round iteration process, For the The feature similarity of the two clusters to be merged during the round of iterations, For the The separation degree of the two clusters to be merged during the iteration process.

6. The intelligent detection method for shaft forgings according to claim 1 is characterized in that: The degree of suspicion satisfies the following relationship: ； In the formula, For the The suspiciousness of the target area, For the The number of all pixels in the target area, For the The average gray value of all pixels in the target area.

7. The intelligent detection method for shaft forgings according to claim 1 is characterized in that: The method of determining the scratch burr area in the surface image of the shaft forging to realize intelligent detection of the shaft forging includes: In response to the suspicious degree being greater than a preset abnormal threshold, determining the target area to be a scratch burr area; In response to the fact that the number of scratch and burr areas is greater than a preset quality threshold, it is determined that there are significant scratch and burr areas on the surface of the shaft forging, and an early warning prompt is issued to complete the intelligent detection of the shaft forging.

8. An intelligent detection system for shaft forgings, characterized in that: include: A processor and a memory, wherein the memory stores computer program instructions, and when the computer program instructions are executed by the processor, an intelligent detection method for shaft forgings according to any one of claims 1-7 is implemented.

Images