Processing equipment for circuit breaker and control method thereof

Abstract

The invention relates to the field of intelligent detection, and particularly discloses processing equipment for a circuit breaker and a control method of the processing equipment. Deep feature mining in a high-dimensional space is carried out on a detection image and a reference image of to-be-cut metal for circuit breaker processing; and comparing the feature consistency of the two features in the high-dimensional feature space to obtain a classification result whether the flatness of the to-be-cut metal surface for manufacturing the circuit breaker meets a preset requirement or not. Therefore, unqualified products can be quickly detected, the processing quality of the products is improved, and the production efficiency is also improved.

Description

Technical Field

[0001] This application relates to the field of intelligent detection, and more specifically, to a processing equipment for circuit breakers and a control method thereof. Background Technology

[0002] A circuit breaker is an electrical device used to switch current in a circuit. A stamping machine is an industrial piece of equipment typically used to manufacture metal parts and components. The general steps in manufacturing circuit breakers using a stamping machine include design, material cutting, material bending, punching, assembly, and testing. The manufacturing process begins with designing the circuit breaker, determining its size and shape. Then, metal sheets are prepared for manufacturing. Next, a cutting machine cuts the metal sheets to the required dimensions. A bending machine then bends the metal sheets to the desired shape. A punch press punches holes in the metal sheets and assembles the components. Finally, the finished circuit breaker undergoes quality inspection. When cutting the material for circuit breakers, insufficient operator experience or skill can lead to unevenness in the cut metal surface. It is crucial to promptly identify and address any non-compliant products to ensure quality. Failure to do so may result in circuit breaker malfunctions during use, affecting its lifespan.

[0003] Therefore, an optimized processing solution for circuit breakers is needed. Summary of the Invention

[0004] To address the aforementioned technical problems, this application is proposed. Embodiments of this application provide a processing equipment and control method for circuit breakers. This method performs deep feature mining in a high-dimensional space on a detection image and a reference image of the metal to be cut used for circuit breaker processing. The consistency of the features between the two images is compared in the high-dimensional feature space to obtain a classification result indicating whether the surface flatness of the metal to be cut meets predetermined requirements for manufacturing circuit breakers. This allows for rapid detection of defective products, improves product processing quality, and increases production efficiency.

[0005] According to one aspect of this application, a processing apparatus for circuit breakers is provided, comprising:

[0006] The image acquisition module is used to acquire a detection image and a reference image of the metal to be cut for manufacturing circuit breakers. The reference image is a metal cross-section image that meets the requirements in the database.

[0007] A Siamese network module is used to pass the detected image and the reference image through a Siamese network model containing a first convolutional neural network and a second convolutional neural network to obtain a detection feature map and a reference feature map, wherein the first convolutional neural network and the second convolutional neural network have the same network structure;

[0008] The feature distribution difference module is used to calculate the difference feature map between the detected feature map and the reference feature map;

[0009] A sparse optimization module is used to perform sparse optimization on the difference feature map based on local feature self-clustering to obtain an optimized difference feature map; and

[0010] The detection result generation module is used to pass the optimized differential feature map through a classifier to obtain a classification result, which is used to indicate whether the flatness of the metal surface to be cut for manufacturing circuit breakers meets the predetermined requirements.

[0011] In the aforementioned circuit breaker processing equipment, the twin network module includes: a detection image feature extraction unit, configured to perform depth convolutional encoding on the detection image using multiple convolutional layers of the first convolutional neural network to output a depth detection feature map from the last layer of the multiple convolutional layers; a first spatial attention unit, configured to input the depth detection feature map into the first spatial attention module of the first convolutional neural network to obtain a first spatial attention map; and a detection feature map generation unit, configured to calculate the positional dot product of the depth detection feature map and the first spatial attention map to obtain the detection feature map.

[0012] In the aforementioned circuit breaker processing equipment, the twin network module further includes: a reference image feature extraction unit, configured to perform depth convolutional encoding on the reference image using multiple convolutional layers of the second convolutional neural network to output a depth reference feature map from the last layer of the multiple convolutional layers; a second spatial attention unit, configured to input the depth reference feature map into the second spatial attention module of the second convolutional neural network to obtain a second spatial attention map; and a reference feature map generation unit, configured to calculate the positional multiplication of the depth reference feature map and the second spatial attention map to obtain the reference feature map.

[0013] In the aforementioned circuit breaker processing equipment, the feature difference module is further configured to: calculate the difference feature map between the detection feature map and the reference feature map using the following difference formula; wherein, the difference formula is:

[0014]

[0015] Wherein, F1 represents the detected feature map, F2 represents the reference feature map, and F represents the differential feature map. This indicates differentiation based on position.

[0016] In the aforementioned circuit breaker processing equipment, the sparse optimization module includes: a matrix feature flattening unit, used to flatten the feature matrices along the channel dimension of the difference feature map to obtain multiple classification feature local feature vectors; a bulldozer distance generation unit, used to calculate the bulldozer distance between each classification feature local feature vector and other classification feature local feature vectors to obtain multiple bulldozer distances of each classification feature local feature vector; a summation calculation unit, used to calculate the summation of the multiple bulldozer distances of each classification feature local feature vector as the local feature self-clustering feature value of each classification feature local feature vector; and a masking processing unit, used to mask the feature matrix of the corresponding channel dimension of the difference feature map based on the local feature self-clustering feature value of each classification feature local feature vector to obtain the optimized difference feature map.

[0017] In the aforementioned circuit breaker processing equipment, the detection result generation module includes: a feature map unrolling unit, used to unroll the optimized differential feature map into a classification feature vector based on row vectors or column vectors; a fully connected encoding unit, used to perform fully connected encoding on the classification feature vector using the fully connected layer of the classifier to obtain an encoded classification feature vector; and a classification unit, used to pass the encoded classification feature vector through the Softmax classification function of the classifier to obtain the classification result.

[0018] According to another aspect of this application, a processing control method for a circuit breaker is provided, comprising:

[0019] Acquire a detection image and a reference image of the metal to be cut, wherein the reference image is a metal cross-section image that meets the requirements in the database;

[0020] The detected image and the reference image are passed through a Siamese network model containing a first convolutional neural network and a second convolutional neural network to obtain a detection feature map and a reference feature map, wherein the first convolutional neural network and the second convolutional neural network have the same network structure;

[0021] Calculate the difference feature map between the detected feature map and the reference feature map;

[0022] The differential feature map is subjected to sparse optimization based on local feature self-clustering to obtain an optimized differential feature map;

[0023] The optimized differential feature map is passed through a classifier to obtain a classification result, which is used to indicate whether the surface flatness of the metal to be cut for manufacturing circuit breakers meets predetermined requirements.

[0024] According to another aspect of this application, an electronic device is provided, comprising: a processor; and a memory storing computer program instructions, which, when executed by the processor, cause the processor to perform the circuit breaker processing control method as described above.

[0025] According to another aspect of this application, a computer-readable medium is provided having computer program instructions stored thereon, which, when executed by a processor, cause the processor to perform the circuit breaker processing control method as described above.

[0026] Compared with existing technologies, the circuit breaker processing equipment and control method provided in this application perform deep feature mining in a high-dimensional space on the detection image and reference image of the metal to be cut for circuit breaker processing, and compare the feature consistency of the two in the high-dimensional feature space to obtain a classification result on whether the surface flatness of the metal to be cut for manufacturing circuit breakers meets predetermined requirements. This allows for rapid detection of defective products, improves product processing quality, and increases production efficiency. Attached Figure Description

[0027] The above and other objects, features, and advantages of this application will become more apparent from the more detailed description of the embodiments of this application in conjunction with the accompanying drawings. The drawings are provided to further understand the embodiments of this application and constitute a part of the specification. They are used together with the embodiments of this application to explain this application and do not constitute a limitation thereof. In the drawings, the same reference numerals generally represent the same components or steps.

[0028] Figure 1 This is a system block diagram of a circuit breaker processing equipment according to an embodiment of this application.

[0029] Figure 2 This is a schematic diagram of a processing equipment for circuit breakers according to an embodiment of this application.

[0030] Figure 3 This is a block diagram of a sparse optimization module in a circuit breaker processing equipment according to an embodiment of this application.

[0031] Figure 4 A module block diagram is generated for the detection results in the circuit breaker processing equipment according to the embodiments of this application.

[0032] Figure 5 This is a flowchart of a circuit breaker manufacturing control method according to an embodiment of this application.

[0033] Figure 6 This is a block diagram of an electronic device according to an embodiment of this application. Detailed Implementation

[0034] Hereinafter, exemplary embodiments according to this application will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are merely some embodiments of this application, and not all embodiments of this application. It should be understood that this application is not limited to the exemplary embodiments described herein.

[0035] Application Overview

[0036] As mentioned in the background section, in existing circuit breaker manufacturing processes, insufficient operator experience can lead to situations where the flatness of the cut metal surface does not meet predetermined requirements. In such cases, it is necessary to promptly detect non-compliant products to ensure that the quality of the manufactured products meets the requirements; otherwise, it may affect the service life of the circuit breaker. Therefore, an optimized manufacturing solution for circuit breakers is desired.

[0037] Currently, deep learning and neural networks are widely used in fields such as computer vision, natural language processing, and speech signal processing. Furthermore, deep learning and neural networks have demonstrated near-human or even surpassed human-level performance in areas such as image classification, object detection, semantic segmentation, and text translation. In recent years, the development of deep learning and neural networks has provided new solutions and approaches for circuit breaker manufacturing.

[0038] In the technical solution of this application, it is expected that deep feature mining in high-dimensional space will be performed on the detection image and reference image of the metal to be cut for circuit breaker processing, and the feature consistency between the two will be compared in high-dimensional feature space to obtain a classification result on whether the surface flatness of the metal to be cut for manufacturing circuit breakers meets the predetermined requirements. In this way, defective products can be quickly detected, improving the processing quality of products and increasing production efficiency.

[0039] Specifically, in the technical solution of this application, firstly, detection images of the metal to be cut used to manufacture circuit breakers and reference images are acquired. The reference images are metal cross-section images that meet the requirements from a database. It should be understood that the detection images of the metal to be cut used to manufacture circuit breakers are used as model input, i.e., as the data source for model training and prediction. By acquiring a large number of different detection images of the metal to be cut and inputting them together with the corresponding reference images (i.e., metal cross-section images that meet the requirements) into the Siamese network module, the model can learn the characteristic variation patterns of the metal surface, thereby improving the accuracy of classification and prediction. The role of the reference images is to provide a reference standard so that the model can determine whether the metal surface to be detected is qualified.

[0040] Next, the detected image and the reference image are processed through a Siamese network model comprising a first convolutional neural network and a second convolutional neural network to obtain a detection feature map and a reference feature map, respectively. The first and second convolutional neural networks have the same network structure. It should be understood that the above feature extraction of the image uses a Siamese network model, where one network is for the detected image and the other is for the reference image. Two convolutional neural networks with the same network structure are used to process the detected image and the reference image of the metal to be cut for circuit breaker processing, respectively, and output their respective feature maps for feature matching and comparison in the subsequent feature distribution difference module. The Siamese network model can model the correlation between two inputs by sharing network parameters, learning the features of the inputs and their differences, and classifying based on these feature differences. Furthermore, compared with other neural networks, the Siamese network model can better handle image variations of the metal to be cut caused by factors such as lighting and tilt angle. It can transform the two input images into a feature space for comparison, rather than directly processing the images, thereby improving the accuracy and reliability of detection.

[0041] Then, a difference feature map is calculated between the detected feature map and the reference feature map. After obtaining the detected feature map and reference feature map from the detected image and the reference image through a Siamese network, they need to be compared using the difference feature map to determine whether the metal surface to be cut meets the requirements. That is, the difference between the detected feature map and the reference feature map in the same feature space is calculated to obtain the difference feature map, and the difference feature map is passed through a classifier to obtain a classification result representing whether the flatness of the metal surface to be cut used to manufacture the circuit breaker meets the predetermined requirements.

[0042] In the technical solution of this application, considering that the differential feature map is a collection of high-dimensional, complex and non-convex local feature matrices, among the local feature matrices along the channel dimension of the differential feature map, some local feature matrices are feature matrices corresponding to noise information or feature matrices corresponding to abnormal features. These local feature matrices will affect the accuracy of the classification judgment of the differential feature map.

[0043] Based on this, in the technical solution of this application, the feature matrices along the channel dimension of the difference feature map are first flattened to obtain multiple classification feature local feature vectors. Then, the bulldozer distance between each classification feature local feature vector and other classification feature local feature vectors is used to represent the similarity of the feature distribution between each classification feature local feature vector and other classification feature local feature vectors; that is, the dimensionality-reduced similarity of the feature distribution between the local feature matrices along the channel dimension of the classification feature matrix. The sum of the multiple bulldozer distances of the calculated local feature vectors is then used as the local feature self-clustering feature value of each local feature vector. Here, the local feature self-clustering feature value represents the global clustering feature value of each feature matrix along the channel dimension of the difference feature map relative to the global feature distribution of the difference feature map.

[0044] Next, based on the self-clustering feature values ​​among the local features of each classification feature local feature vector, the feature matrix of the corresponding channel dimension of the difference feature map is masked to obtain an optimized difference feature map. For example, in a specific example, masking the feature matrix of the corresponding channel dimension of the difference feature map to obtain an optimized difference feature map based on the self-clustering feature values ​​among the local features of each classification feature local feature vector includes: determining whether to zero out the feature matrix of the corresponding channel dimension of the difference feature map based on a comparison between the self-clustering feature values ​​among the local features of each classification feature local feature vector and a predetermined threshold.

[0045] In this way, sparse optimization based on local feature self-clustering is performed on the differential feature map to identify the feature matrix corresponding to noise information or the feature matrix corresponding to abnormal features in the differential feature map based on the correlation and similarity between the data distributions of each local feature distribution. This reduces the effective dimension of the differential feature map while enabling the differential feature map to more effectively reflect the essential characteristics and patterns of the data, thereby improving the accuracy of the classification judgment of the differential feature map.

[0046] Having introduced the basic principles of this application, various non-limiting embodiments of this application will now be described in detail with reference to the accompanying drawings.

[0047] Exemplary System

[0048] Figure 1 This is a system block diagram of a circuit breaker processing apparatus according to an embodiment of this application. Figure 1As shown, the circuit breaker processing equipment 100 includes: an image acquisition module 110, used to acquire a detection image and a reference image of the metal to be cut for manufacturing the circuit breaker, wherein the reference image is a metal cross-section image that meets the requirements in a database; a twin network module 120, used to pass the detection image and the reference image through a twin network model containing a first convolutional neural network and a second convolutional neural network to obtain a detection feature map and a reference feature map, wherein the first convolutional neural network and the second convolutional neural network have the same network structure; a feature distribution difference module 130, used to calculate the difference feature map between the detection feature map and the reference feature map; a sparse optimization module 140, used to perform sparse optimization based on local feature self-clustering on the difference feature map to obtain an optimized difference feature map; and a detection result generation module 150, used to pass the optimized difference feature map through a classifier to obtain a classification result, wherein the classification result is used to indicate whether the surface flatness of the metal to be cut for manufacturing the circuit breaker meets the predetermined requirements.

[0049] Figure 2 This is a schematic diagram of a processing apparatus for circuit breakers according to an embodiment of this application. Figure 2 As shown, in this architecture, firstly, a detection image and a reference image of the metal to be cut are acquired, where the reference image is a metal cross-section image that meets the requirements from a database. Next, the detection image and the reference image are passed through a Siamese network model containing a first convolutional neural network and a second convolutional neural network to obtain a detection feature map and a reference feature map, respectively. The first and second convolutional neural networks have the same network structure. Then, a difference feature map is calculated between the detection feature map and the reference feature map. Furthermore, the difference feature map is subjected to sparse optimization based on local feature self-clustering to obtain an optimized difference feature map. Finally, the optimized difference feature map is passed through a classifier to obtain a classification result, which indicates whether the surface flatness of the metal to be cut, used to manufacture circuit breakers, meets predetermined requirements.

[0050] In the circuit breaker processing equipment 100, the image acquisition module 110 is used to acquire detection images of the metal to be cut for manufacturing the circuit breaker and reference images. The reference images are metal cross-section images that meet the requirements from a database. It should be understood that the detection images of the metal to be cut for manufacturing the circuit breaker are used as model input, i.e., as the data source for model training and prediction. By acquiring a large number of different detection images of the metal to be cut and inputting them together with the corresponding reference images (i.e., metal cross-section images that meet the requirements) into the Siamese network module, the model can learn the characteristic variation patterns of the metal surface, thereby improving the accuracy of classification and prediction. The role of the reference image is to provide a reference standard so that the model can determine whether the metal surface to be inspected is qualified. Here, the detection images of the metal to be cut for manufacturing the circuit breaker are acquired through a camera, while the reference images are obtained from a database.

[0051] In the circuit breaker processing equipment 100, the Siamese network module 120 is used to pass the detected image and the reference image through a Siamese network model containing a first convolutional neural network and a second convolutional neural network to obtain a detection feature map and a reference feature map. The first convolutional neural network and the second convolutional neural network have the same network structure. It should be understood that convolutional neural networks perform excellently in feature extraction, and they consist of an input layer, multiple convolutional layers, pooling layers, and fully connected layers. The convolutional layers perform convolution operations on the input image through multiple filters to extract features of different sizes and orientations; the pooling layers reduce the number of parameters by downsampling while retaining important information; and the fully connected layers map the output of the previous layer to a specified number of categories. In the Siamese network model, two convolutional neural networks with the same structure perform convolution operations on the input detection image and the reference image, respectively obtaining a detection feature map and a reference feature map. These feature maps can reflect features at different levels, such as edges, corners, and textures. Siamese neural network models can model the correlation between two inputs by sharing parameters, learning the features and differences between the inputs, and classifying based on these feature differences. Furthermore, compared to other neural networks, Siamese neural network models can better handle image variations of the metal to be cut due to factors such as lighting and tilt angle. They can transform two input images into a feature space for comparison, rather than directly processing the images, thereby improving the accuracy and reliability of detection.

[0052] Specifically, in the circuit breaker processing equipment 100, the twin network module 120 includes: a detection image feature extraction unit, used to perform depth convolutional encoding on the detection image using multiple convolutional layers of the first convolutional neural network to output a depth detection feature map by the last layer of the multiple convolutional layers; a first spatial attention unit, used to input the depth detection feature map into the first spatial attention module of the first convolutional neural network to obtain a first spatial attention map; and a detection feature map generation unit, used to calculate the positional dot product of the depth detection feature map and the first spatial attention map to obtain the detection feature map.

[0053] Specifically, in the circuit breaker processing equipment 100, the twin network module 120 further includes: a reference image feature extraction unit, used to perform depth convolutional encoding on the reference image using multiple convolutional layers of the second convolutional neural network to output a depth reference feature map by the last layer of the multiple convolutional layers; a second spatial attention unit, used to input the depth reference feature map into the second spatial attention module of the second convolutional neural network to obtain a second spatial attention map; and a reference feature map generation unit, used to calculate the positional dot product of the depth reference feature map and the second spatial attention map to obtain the reference feature map.

[0054] In the circuit breaker processing equipment 100, the feature distribution difference module 130 is used to calculate a difference feature map between the detection feature map and the reference feature map. It should be understood that calculating the difference feature map between the detection feature map and the reference feature map is to compare the differences between the surface features of the metal to be inspected and the surface features of the reference image, thereby determining whether the flatness of the metal to be cut meets predetermined requirements. The difference feature map is obtained by subtracting the detection feature map from the reference feature map. This method can greatly reduce the influence of noise on the surface quality inspection of metal cutting and extract the main features of the metal to be inspected to the greatest extent possible.

[0055] Specifically, in the circuit breaker processing equipment 100, the feature distribution difference module 130 is further configured to: calculate the difference feature map between the detection feature map and the reference feature map using the following difference formula; wherein, the difference formula is:

[0056]

[0057] Wherein, F1 represents the detected feature map, F2 represents the reference feature map, and F represents the differential feature map. This indicates differentiation based on position.

[0058] In the circuit breaker processing equipment 100, the sparse optimization module 140 is used to perform sparse optimization on the differential feature map based on local feature self-clustering to obtain an optimized differential feature map.

[0059] In the technical solution of this application, considering that the differential feature map is a collection of high-dimensional, complex and non-convex local feature matrices, among the local feature matrices along the channel dimension of the differential feature map, some local feature matrices are feature matrices corresponding to noise information or feature matrices corresponding to abnormal features. These local feature matrices will affect the accuracy of the classification judgment of the differential feature map.

[0060] Figure 4 This is a block diagram of a sparse optimization module in a circuit breaker processing equipment according to an embodiment of this application. Figure 4 As shown, the sparse optimization module 140 includes: a matrix feature flattening unit 141, used to flatten the feature matrices along the channel dimension of the difference feature map to obtain multiple classification feature local feature vectors; a bulldozer distance generation unit 142, used to calculate the bulldozer distance between each classification feature local feature vector and other classification feature local feature vectors to obtain multiple bulldozer distances of each classification feature local feature vector; a summation calculation unit 143, used to calculate the summation of the multiple bulldozer distances of each classification feature local feature vector as the local feature self-clustering feature value of each classification feature local feature vector; and a masking processing unit 144, used to mask the feature matrix of the corresponding channel dimension of the difference feature map based on the local feature self-clustering feature value of each classification feature local feature vector to obtain the optimized difference feature map.

[0061] Specifically, the bulldozer distance between local feature vectors of each category and other local feature vectors of multiple classification features represents the similarity of feature distributions between local feature vectors of each category and other local feature vectors of multiple classification features. That is, the dimensionality-reduced similarity of feature distributions between local feature matrices along the channel dimension of the classification feature matrix. The local feature self-clustering feature value is used to represent the global clustering feature value of each feature matrix along the channel dimension of the classification feature map relative to the global feature distribution of the classification feature map.

[0062] For example, in a specific example, based on the self-clustering feature values ​​among the local features of each classification feature local feature vector, the feature matrix of the corresponding channel dimension of the difference feature map is masked to obtain an optimized difference feature map, including: based on the comparison between the self-clustering feature values ​​among the local features of each classification feature local feature vector and a predetermined threshold, determining whether to zero out the feature matrix of the corresponding channel dimension of the difference feature map.

[0063] In this way, sparse optimization based on local feature self-clustering is performed on the differential feature map to identify the feature matrix corresponding to noise information or the feature matrix corresponding to abnormal features in the differential feature map based on the correlation and similarity between the data distributions of each local feature distribution. This reduces the effective dimension of the differential feature map while enabling the differential feature map to more effectively reflect the essential characteristics and patterns of the data, thereby improving the accuracy of the classification judgment of the differential feature map.

[0064] In the circuit breaker processing equipment 100, the detection result generation module 150 is used to pass the optimized differential feature map through a classifier to obtain a classification result. The classification result indicates whether the flatness of the metal surface to be cut, used to manufacture the circuit breaker, meets predetermined requirements. This allows for rapid detection of defective products, improves product processing quality, and increases production efficiency.

[0065] Figure 5 A module block diagram is generated for the test results in the circuit breaker processing equipment according to the embodiments of this application. For example... Figure 5 As shown, the detection result generation module 150 includes: a feature map unrolling unit 151, used to unroll the optimized difference feature map into a classification feature vector based on row vectors or column vectors; a fully connected encoding unit 152, used to perform fully connected encoding on the classification feature vector using the fully connected layer of the classifier to obtain an encoded classification feature vector; and a classification unit 153, used to pass the encoded classification feature vector through the Softmax classification function of the classifier to obtain the classification result.

[0066] In summary, the circuit breaker processing equipment 100 based on the embodiments of this application is explained. It performs deep feature mining in a high-dimensional space on a detection image and a reference image of the metal to be cut used for circuit breaker processing, and compares the feature consistency of the two images in the high-dimensional feature space to obtain a classification result on whether the surface flatness of the metal to be cut used to manufacture the circuit breaker meets predetermined requirements. This allows for rapid detection of defective products, improves product processing quality, and increases production efficiency.

[0067] Exemplary methods

[0068] Figure 5 This is a flowchart of a circuit breaker manufacturing control method according to an embodiment of this application. Figure 5 As shown, the processing control method for circuit breakers includes: S110, acquiring a detection image and a reference image of the metal to be cut for manufacturing circuit breakers, wherein the reference image is a metal cross-section image that meets the requirements in a database; S120, passing the detection image and the reference image through a Siamese network model including a first convolutional neural network and a second convolutional neural network to obtain a detection feature map and a reference feature map, wherein the first convolutional neural network and the second convolutional neural network have the same network structure; S130, calculating a difference feature map between the detection feature map and the reference feature map; S140, performing sparse optimization based on local feature self-clustering on the difference feature map to obtain an optimized difference feature map; and S150, passing the optimized difference feature map through a classifier to obtain a classification result, wherein the classification result is used to indicate whether the surface flatness of the metal to be cut for manufacturing circuit breakers meets predetermined requirements.

[0069] In one example, in the above-described circuit breaker processing control method, the step of passing the detected image and the reference image through a Siamese network model comprising a first convolutional neural network and a second convolutional neural network to obtain a detection feature map and a reference feature map, wherein the first convolutional neural network and the second convolutional neural network have the same network structure, includes: performing depth convolutional encoding on the detected image using multiple convolutional layers of the first convolutional neural network to output a depth detection feature map by the last layer of the multiple convolutional layers; inputting the depth detection feature map into a first spatial attention module of the first convolutional neural network to obtain a first spatial attention map; and calculating a positional dot product of the depth detection feature map and the first spatial attention map to obtain the detection feature map.

[0070] In one example, in the above-described circuit breaker processing control method, the step of passing the detected image and the reference image through a Siamese network model comprising a first convolutional neural network and a second convolutional neural network to obtain a detection feature map and a reference feature map, wherein the first convolutional neural network and the second convolutional neural network have the same network structure, further includes: performing deep convolutional encoding on the reference image using multiple convolutional layers of the second convolutional neural network to output a deep reference feature map by the last layer of the multiple convolutional layers; inputting the deep reference feature map into a second spatial attention module of the second convolutional neural network to obtain a second spatial attention map; and calculating a positional dot product of the deep reference feature map and the second spatial attention map to obtain the reference feature map.

[0071] In one example, in the above-described processing control method for circuit breakers, calculating the differential feature map between the detection feature map and the reference feature map includes: calculating the differential feature map between the detection feature map and the reference feature map using the following differential formula; wherein, the differential formula is:

[0072]

[0073] Wherein, F1 represents the detected feature map, F2 represents the reference feature map, and F represents the differential feature map. This indicates differentiation based on position.

[0074] In one example, in the above-mentioned circuit breaker processing control method, the step of performing sparse optimization based on local feature self-clustering on the differential feature map to obtain an optimized differential feature map includes: flattening the feature matrices along the channel dimension of the differential feature map to obtain multiple classification feature local feature vectors; calculating the bulldozer distance between each classification feature local feature vector and other classification feature local feature vectors to obtain multiple bulldozer distances of each classification feature local feature vector; calculating the sum of the multiple bulldozer distances of each classification feature local feature vector as the local feature self-clustering feature value of each classification feature local feature vector; and masking the feature matrix of the corresponding channel dimension of the differential feature map based on the local feature self-clustering feature value of each classification feature local feature vector to obtain the optimized differential feature map.

[0075] In one example, in the above-mentioned circuit breaker processing control method, the step of passing the optimized differential feature map through a classifier to obtain a classification result, the classification result being used to indicate whether the flatness of the metal surface to be cut for manufacturing the circuit breaker meets predetermined requirements, includes: expanding the optimized differential feature map into a classification feature vector based on row vectors or column vectors; performing fully connected encoding on the classification feature vector using the fully connected layer of the classifier to obtain an encoded classification feature vector; and passing the encoded classification feature vector through the Softmax classification function of the classifier to obtain the classification result.

[0076] In summary, the circuit breaker processing control method based on the embodiments of this application is explained. It involves performing deep feature mining in a high-dimensional space on a detection image and a reference image of the metal to be cut used in circuit breaker processing, and comparing the feature consistency between the two images in the high-dimensional feature space to obtain a classification result indicating whether the surface flatness of the metal to be cut used to manufacture the circuit breaker meets predetermined requirements. This allows for rapid detection of defective products, improves product processing quality, and increases production efficiency.

[0077] Exemplary electronic devices

[0078] Below, for reference Figure 6 This describes an electronic device according to embodiments of the present application.

[0079] Figure 6 This is a block diagram of an electronic device according to an embodiment of this application.

[0080] like Figure 6 As shown, the electronic device 10 includes one or more processors 11 and memory 12.

[0081] The processor 11 may be a central processing unit (CPU) or other form of processing unit with data processing capabilities and / or instruction execution capabilities, and may control other components in the electronic device 10 to perform desired functions.

[0082] The memory 12 may include one or more computer program products, which may include various forms of computer-readable storage media, such as volatile memory and / or non-volatile memory. The volatile memory may include, for example, random access memory (RAM) and / or cache memory. The non-volatile memory may include, for example, read-only memory (ROM), hard disk, flash memory, etc. One or more computer program instructions may be stored on the computer-readable storage medium, and the processor 11 may execute the program instructions to implement the circuit breaker processing control methods and / or other desired functions described in the various embodiments of this application above. The computer-readable storage medium may also store various contents such as detection images of the metal to be cut for manufacturing circuit breakers, reference images, etc.

[0083] In one example, the electronic device 10 may also include an input device 13 and an output device 14, which are interconnected via a bus system and / or other forms of connection mechanism (not shown).

[0084] The input device 13 may include, for example, a keyboard, a mouse, etc.

[0085] The output device 14 can output various information to the outside, including the result of determining whether the flatness of the metal surface to be cut for manufacturing circuit breakers meets predetermined requirements. The output device 14 may include, for example, a display, a speaker, a printer, and a communication network and its connected remote output devices, etc.

[0086] Of course, for the sake of simplicity, Figure 6 Only some of the components of the electronic device 10 relevant to this application are shown in this illustration; components such as buses, input / output interfaces, etc., are omitted. In addition, the electronic device 10 may include any other suitable components depending on the specific application.

[0087] Exemplary computer program products and computer-readable storage media

[0088] In addition to the methods and apparatus described above, embodiments of this application may also be computer program products, which include computer program instructions that, when executed by a processor, cause the processor to perform the steps in the circuit breaker processing control methods according to various embodiments of this application as described in the "Exemplary Methods" section above.

[0089] The computer program product can be written in any combination of one or more programming languages ​​to perform the operations of the embodiments of this application. The programming languages ​​include object-oriented programming languages ​​such as Java and C++, as well as conventional procedural programming languages ​​such as C or similar languages. The program code can be executed entirely on the user's computing device, partially on the user's computing device, as a standalone software package, partially on the user's computing device and partially on a remote computing device, or entirely on a remote computing device or server.

[0090] Furthermore, embodiments of this application may also be computer-readable storage media storing computer program instructions thereon, which, when executed by a processor, cause the processor to perform the steps in the circuit breaker processing control method according to various embodiments of this application described in the "Exemplary Methods" section above.

[0091] The computer-readable storage medium may be any combination of one or more readable media. A readable medium may be a readable signal medium or a readable storage medium. A readable storage medium may, for example, include, but is not limited to, electrical, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatuses, or devices, or any combination thereof. More specific examples of readable storage media (a non-exhaustive list) include: electrical connections having one or more wires, portable disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fibers, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof.

[0092] The basic principles of this application have been described above with reference to specific embodiments. However, it should be noted that the advantages, benefits, and effects mentioned in this application are merely examples and not limitations, and should not be considered as essential features of each embodiment of this application. Furthermore, the specific details disclosed above are for illustrative and facilitative purposes only, and are not limitations. These details do not limit the application to the necessity of employing the aforementioned specific details for implementation.

[0093] The block diagrams of devices, apparatuses, devices, and systems involved in this application are merely illustrative examples and are not intended to require or imply that they must be connected, arranged, or configured in the manner shown in the block diagrams. As those skilled in the art will recognize, these devices, apparatuses, devices, and systems can be connected, arranged, and configured in any manner. Words such as “comprising,” “including,” “having,” etc., are open-ended terms meaning “including but not limited to,” and are used interchangeably with them. The terms “or” and “and” as used herein refer to the terms “and / or,” and are used interchangeably with them unless the context clearly indicates otherwise. The term “such as” as used herein refers to the phrase “such as but not limited to,” and is used interchangeably with it.

[0094] It should also be noted that in the apparatus, equipment, and methods of this application, the components or steps can be disassembled and / or recombined. These disassemblies and / or recombinations should be considered as equivalent solutions of this application.

[0095] The above description of the disclosed aspects is provided to enable any person skilled in the art to make or use this application. Various modifications to these aspects will be readily apparent to those skilled in the art, and the general principles defined herein can be applied to other aspects without departing from the scope of this application. Therefore, this application is not intended to be limited to the aspects shown herein, but rather to be accorded the widest scope consistent with the principles and novel features disclosed herein.

[0096] The above description has been given for purposes of illustration and description. Furthermore, this description is not intended to limit the embodiments of this application to the forms disclosed herein. Although numerous exemplary aspects and embodiments have been discussed above, those skilled in the art will recognize certain variations, modifications, alterations, additions, and sub-combinations thereof.

Claims

1. A processing equipment for circuit breakers, characterized in that, include: The image acquisition module is used to acquire a detection image and a reference image of the metal to be cut for manufacturing circuit breakers. The reference image is a metal cross-section image that meets the requirements in the database. A Siamese network module is used to pass the detected image and the reference image through a Siamese network model containing a first convolutional neural network and a second convolutional neural network to obtain a detection feature map and a reference feature map, wherein the first convolutional neural network and the second convolutional neural network have the same network structure; The feature distribution difference module is used to calculate the difference feature map between the detected feature map and the reference feature map; The sparse optimization module is used to perform sparse optimization on the difference feature map based on local feature self-clustering to obtain an optimized difference feature map. as well as The detection result generation module is used to pass the optimized differential feature map through a classifier to obtain a classification result, which is used to indicate whether the flatness of the metal surface to be cut for manufacturing circuit breakers meets the predetermined requirements.

2. The processing equipment for circuit breakers according to claim 1, characterized in that, The twin network module includes: The detection image feature extraction unit is used to perform depth convolution encoding on the detection image using multiple convolutional layers of the first convolutional neural network so that the last layer of the multiple convolutional layers outputs a depth detection feature map. A first spatial attention unit is used to input the depth detection feature map into the first spatial attention module of the first convolutional neural network to obtain a first spatial attention map. The detection feature map generation unit is used to calculate the position-based dot product of the depth detection feature map and the first spatial attention map to obtain the detection feature map.

3. The processing equipment for circuit breakers according to claim 2, characterized in that, The twin network module also includes: A reference image feature extraction unit is used to perform depth convolutional encoding on the reference image using multiple convolutional layers of the second convolutional neural network to output a depth reference feature map by the last layer of the multiple convolutional layers; A second spatial attention unit is used to input the depth reference feature map into the second spatial attention module of the second convolutional neural network to obtain a second spatial attention map; and A reference feature map generation unit is used to calculate the reference feature map by multiplying the depth reference feature map and the second spatial attention map by position.

4. The processing equipment for circuit breakers according to claim 3, characterized in that, The feature difference module is further configured to: calculate the difference feature map between the detected feature map and the reference feature map using the following difference formula; The difference formula is as follows: Wherein, F1 represents the detected feature map, F2 represents the reference feature map, and F represents the differential feature map. This indicates differentiation based on position.

5. The processing equipment for circuit breakers according to claim 4, characterized in that, The sparse optimization module includes: The matrix feature flattening unit is used to flatten the feature matrices along the channel dimension of the difference feature map to obtain multiple classification feature local feature vectors. A bulldozer distance generation unit is used to calculate the bulldozer distance between each local feature vector of a classification feature and other local feature vectors of classification features to obtain multiple bulldozer distances of each local feature vector of classification features. The summation calculation unit is used to calculate the summation of multiple bulldozer distances of the local feature vectors of each classification feature as the local feature self-clustering feature value of the local feature vectors of each classification feature; and The masking processing unit is used to mask the feature matrix of the corresponding channel dimension of the difference feature map based on the self-clustering feature values ​​between local features of each classification feature local feature vector to obtain the optimized difference feature map.

6. The processing equipment for circuit breakers according to claim 5, characterized in that, The detection result generation module includes: The feature map unpacking unit is used to unpack the optimized difference feature map into a classification feature vector based on row vectors or column vectors; A fully connected encoding unit is configured to perform fully connected encoding on the classification feature vector using the fully connected layer of the classifier to obtain an encoded classification feature vector; and A classification unit is used to pass the encoded classification feature vector through the Softmax classification function of the classifier to obtain the classification result.

7. A processing control method for circuit breakers, characterized in that, include: Acquire detection images and reference images of the metal to be cut for manufacturing circuit breakers, wherein the reference images are metal cross-section images that meet the requirements in the database; The detected image and the reference image are passed through a Siamese network model containing a first convolutional neural network and a second convolutional neural network to obtain a detection feature map and a reference feature map, wherein the first convolutional neural network and the second convolutional neural network have the same network structure; Calculate the difference feature map between the detected feature map and the reference feature map; The differential feature map is subjected to sparse optimization based on local feature self-clustering to obtain an optimized differential feature map; as well as The optimized differential feature map is passed through a classifier to obtain a classification result, which is used to indicate whether the flatness of the metal surface to be cut for manufacturing circuit breakers meets the predetermined requirements.

8. The processing control method for circuit breakers according to claim 7, characterized in that, The detected image and the reference image are processed through a Siamese network model comprising a first convolutional neural network and a second convolutional neural network to obtain a detection feature map and a reference feature map. The first convolutional neural network and the second convolutional neural network have the same network structure, including: The detection image is depthwise convolutionally encoded using multiple convolutional layers of the first convolutional neural network so that the last layer of the multiple convolutional layers outputs a depth detection feature map. The depth detection feature map is input into the first spatial attention module of the first convolutional neural network to obtain the first spatial attention map. The detection feature map is obtained by multiplying the depth detection feature map and the first spatial attention map by their positions.

9. The processing control method for circuit breakers according to claim 8, characterized in that, The detected image and the reference image are processed through a Siamese network model comprising a first convolutional neural network and a second convolutional neural network to obtain a detection feature map and a reference feature map, wherein the first convolutional neural network and the second convolutional neural network have the same network structure, and further comprising: The reference image is depthwise convolutionally encoded using multiple convolutional layers of the second convolutional neural network to output a depth reference feature map from the last layer of the multiple convolutional layers; The depth reference feature map is input into the second spatial attention module of the second convolutional neural network to obtain the second spatial attention map; and The reference feature map is obtained by multiplying the depth reference feature map and the second spatial attention map by their positions.

10. The processing control method for circuit breakers according to claim 9, characterized in that, Calculating the difference feature map between the detected feature map and the reference feature map includes: calculating the difference feature map between the detected feature map and the reference feature map using the following difference formula; The difference formula is as follows: Wherein, F1 represents the detected feature map, F2 represents the reference feature map, and F represents the differential feature map. This indicates differentiation based on position.

Images