Gold finger beveling processing system and method

Abstract

This application relates to the field of intelligent processing, specifically disclosing a beveling processing system and method for a gold finger. The system utilizes artificial intelligence technology based on a deep neural network model to obtain a gold finger bevel reference image. After de-noising the image through a denoiser, the image is passed through a convolutional network comprised of multiple convolutional layers and a feature extractor for multi-level feature extraction to obtain a decoded value representing the desired chamfer angle setting for the mold. This system enables automated and precise angle setting, thereby improving production efficiency and product quality.

Description

Technical Field

[0001] The present application relates to the field of intelligent processing, and more specifically, to a beveling processing system and method for a gold finger. Background Art

[0002] Goldfinger beveling involves beveling the edges or corners of a workpiece to create a chamfered metal interface, commonly used for connectivity and communication in electronic products. Different product requirements, such as appearance, functionality, and assembly, require varying bevel angles. However, existing technologies rely on manual control of the goldfinger beveling angle, which in turn controls the grinding process. This can lead to errors, a low level of intelligence, and slow processing times.

[0003] Therefore, an optimized bevel processing solution for gold fingers is expected. Summary of the Invention

[0004] To address the aforementioned technical issues, the present application is presented. The embodiments of the present application provide a beveling processing system and method for a gold finger. The system utilizes artificial intelligence technology based on a deep neural network model to obtain a gold finger bevel reference image. This image is de-noised by a denoiser and then subjected to multi-level feature extraction using a convolutional network comprised of multiple convolutional layers and a feature extractor. This decodes the angle setting value required for the chamfer of the mold. This allows for automated and precise angle setting, thereby improving production efficiency and product quality.

[0005] According to one aspect of the present application, a gold finger beveling processing system is provided, comprising:

[0006] A reference image acquisition module is used to obtain a reference image of the bevel edge of the gold finger;

[0007] An image denoising module, configured to pass the gold finger bevel edge reference image through a depthwise separable convolutional neural network as a denoiser to obtain a denoised gold finger bevel edge image;

[0008] A gold finger bevel feature extraction module is used to pass the denoised gold finger bevel image through a convolutional neural network model with multiple convolution layers to obtain a denoised gold finger bevel feature map;

[0009] A feature map segmentation module is used to segment the denoised golden finger bevel edge feature map to obtain a plurality of golden finger bevel edge feature sub-feature maps;

[0010] A feature convolutional network module, configured to pass the plurality of gold finger bevel feature sub-feature maps through a convolutional neural network model as a feature extractor to obtain a gold finger bevel feature map;

[0011] A sparse calculation module, used for calculating a sparse golden finger oblique edge feature map of the golden finger oblique edge feature map;

[0012] The chamfer setting module is used to decode and regress the sparse gold finger bevel feature map through a decoder to obtain a decoded value, and the decoded value is used to represent the angle setting value that the grinding tool needs to chamfer.

[0013] In the above-mentioned gold finger beveling processing system, the image denoising module includes: an encoding unit for inputting the gold finger bevel reference image into the encoder of the deep separable convolutional neural network serving as a denoiser, wherein the encoder uses a convolutional layer to perform explicit spatial encoding on the gold finger bevel reference image to obtain image features; and a decoding unit for inputting the image features into the decoder of the deep separable convolutional neural network serving as a denoiser, wherein the decoder uses a deconvolution layer to perform deconvolution processing on the image features to obtain the denoised image of the object to be cleaned; wherein the deep separable convolutional neural network serving as a denoiser is a denoiser based on an automatic codec, the image encoder of the automatic codec includes at least one convolutional layer, and the image decoder of the automatic codec includes at least one deconvolution layer.

[0014] In the above-mentioned gold finger bevel processing system, the gold finger bevel feature extraction module is used to: use each hybrid convolution layer of the convolutional neural network model of the multi-layer convolution layer to respectively perform on the input data in the forward pass of the layer: use a first convolution kernel with a first size to convolutionally encode the denoised gold finger bevel image to obtain a first scale feature map; use a second convolution kernel with a first void rate to convolutionally encode the denoised gold finger bevel image to obtain a second scale feature map; use a third convolution kernel with a second void rate to convolutionally encode the denoised gold finger bevel image to obtain a third scale feature map; aggregate the first scale feature map, the second scale feature map, the third scale feature map and the fourth scale feature map along the channel dimension to obtain an aggregated feature map; perform pooling processing on the aggregated feature map to generate a pooled feature map; and perform activation processing on the pooled feature map to generate an activated feature map; wherein the output of the last layer of the convolutional neural network model including multiple hybrid convolution layers is the denoised gold finger bevel feature map.

[0015] In the above-mentioned gold finger bevel processing system, the gold finger bevel feature extraction module is used to: the first convolution kernel, the second convolution kernel, the third convolution kernel and the fourth convolution kernel have the same size, and the second convolution kernel, the third convolution kernel and the fourth convolution kernel have different void rates.

[0016] In the above-mentioned gold finger bevel processing system, the feature convolution network module is used to: use the various layers of the convolutional neural network model as a feature extractor to perform the following on the input data in the forward pass of the layer: convolution processing on the input data to obtain a convolution feature map; global mean pooling based on the local feature matrix on the convolution feature map to obtain a pooled feature map; nonlinear activation on the pooled feature map to obtain an activation feature map; wherein, the output of the last layer of the convolutional neural network model as a feature extractor is each of the multiple gold finger bevel sub-feature maps of the multiple gold finger bevel sub-feature maps, and the input of the first layer of the convolutional neural network model as a feature extractor is each of the multiple gold finger bevel feature sub-feature maps of the multiple gold finger bevel feature sub-feature maps; and, the multiple gold finger bevel sub-feature maps are cascaded to obtain the gold finger bevel feature map.

[0017] In the above-mentioned bevel processing system for the gold finger, the chamfer setting module is used to: use the decoder to perform decoding regression on the sparse gold finger bevel feature map using the following decoding formula to obtain the decoding value; wherein the decoding formula is: Where X is the sparse cheat sheet feature map, Y is the decoded value, and W is the weight matrix. Represents matrix multiplication.

[0018] According to another aspect of the present application, a method for beveling a gold finger is provided, comprising:

[0019] Get the gold finger bevel reference image;

[0020] Passing the gold finger bevel edge reference image through a depthwise separable convolutional neural network as a denoiser to obtain a denoised gold finger bevel edge image;

[0021] Passing the denoised golden finger bevel edge image through a convolutional neural network model with multiple convolutional layers to obtain a denoised golden finger bevel edge feature map;

[0022] Performing feature map segmentation on the denoised golden finger bevel edge feature map to obtain a plurality of golden finger bevel edge feature sub-feature maps;

[0023] Passing the plurality of gold finger oblique edge feature sub-feature maps through a convolutional neural network model as a feature extractor to obtain a gold finger oblique edge feature map;

[0024] Calculating a sparse golden finger oblique edge feature map of the golden finger oblique edge feature map;

[0025] The sparse gold finger bevel feature map is decoded and regressed through a decoder to obtain a decoded value, and the decoded value is used to represent the angle setting value that the grinding tool needs to chamfer.

[0026] Compared to existing technologies, the present invention provides a beveling system and method for goldfinger edges. This system utilizes artificial intelligence technology based on a deep neural network model to obtain a goldfinger bevel reference image. This image is de-noised by a denoiser, and then subjected to multi-level feature extraction using a convolutional network comprised of multiple convolutional layers and a feature extractor. This decodes the angle setting required for the tool chamfer. This allows for automated and precise angle setting, improving production efficiency and product quality. BRIEF DESCRIPTION OF THE DRAWINGS

[0027] The above and other purposes, features, and advantages of the present application will become more apparent through a more detailed description of the embodiments of the present application in conjunction with the accompanying drawings. The accompanying drawings are intended to provide a further understanding of the embodiments of the present application and constitute a part of the specification. Together with the embodiments of the present application, they are used to explain the present application and do not constitute a limitation of the present application. In the drawings, the same reference numerals generally represent the same components or steps.

[0028] Figure 1 4 is a block diagram of a gold finger beveling processing system according to an embodiment of the present application.

[0029] Figure 2 Schematic diagram of the architecture of a gold finger beveling processing system according to an embodiment of the present application.

[0030] Figure 3 4 is a block diagram of an image noise reduction module in a beveling processing system for a gold finger according to an embodiment of the present application.

[0031] Figure 4 4 is a block diagram of a sparsification calculation module in a beveling processing system for a gold finger according to an embodiment of the present application.

[0032] Figure 5 Flowchart of a method for processing the bevel edge of a gold finger according to an embodiment of the present application.

[0033] Figure 6 is a block diagram of an electronic device according to an embodiment of the present application. DETAILED DESCRIPTION

[0034] Below, the exemplary embodiments according to the present application will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments of the present application, and it should be understood that the present application is not limited to the exemplary embodiments described herein.

[0035] Application Overview

[0036] As mentioned above, the angle of the chamfer of the bevel of the gold finger needs to be chamfered at different angles according to different requirements. However, since the existing technology mostly uses manual setting of the chamfer angle, it is easily affected by personal subjective consciousness and experience, and different operators may have different judgments and settings. This may lead to inconsistency and errors in the angle, affecting the quality and consistency of the product. In addition, during the production process, it may be necessary to adjust the chamfer angle according to different product requirements and process changes. If it relies entirely on manual settings, the adjustment may be cumbersome and time-consuming, limiting the flexibility and responsiveness of production. Therefore, an optimized solution for the bevel processing of gold fingers is expected.

[0037] In response to the above technical problems, the applicant of this application obtains a reference image of the bevel edge of the gold finger, denoises the image through a denoiser, and then performs multi-level feature extraction through a convolutional network of multiple convolutional layers and feature extractors to obtain a decoding value for indicating the angle setting value that the grinding tool needs to chamfer.

[0038] Accordingly, in the technical solution of the present application, it is taken into account that the bevel of the gold finger usually needs to be chamfered at a certain angle. By obtaining a reference image, the angle of the bevel of the gold finger can be measured and set, thereby ensuring the accuracy and consistency of the chamfer. After obtaining the reference image, the image can be analyzed and processed using technologies such as image processing and deep learning to extract the characteristic information of the bevel of the gold finger, thereby realizing automated chamfering. In addition, the reference image can be used for quality control and inspection. By comparing the difference between the bevel of the gold finger after actual processing and the reference image, the processing deviation can be discovered and corrected in time to ensure that the quality of the product meets the requirements. In addition, obtaining a reference image can be used to analyze the shape and characteristics of the bevel of the gold finger, and then optimize the process. By comparing the data of different reference images, the optimal chamfer angle and processing parameters can be found to improve production efficiency and product quality.

[0039] In recent years, deep learning and neural networks have been widely used in fields such as computer vision, natural language processing, and text signal processing. Furthermore, deep learning and neural networks have demonstrated capabilities approaching or even surpassing those of humans in areas such as image decoding, object detection, semantic segmentation, and text translation.

[0040] Specifically, in the technical solution of the present application, first, a gold finger bevel reference image is obtained. Considering that the denoiser convolution can effectively reduce the noise and interference in the image, the gold finger bevel image is clearer and more recognizable. This helps to improve the accuracy of subsequent image processing and analysis. In addition, the denoiser convolution can retain the detail information in the gold finger bevel image as much as possible while removing the noise. This ensures that the calculation and analysis of the chamfer angle are based on accurate image data, thereby improving the reliability of the results. The denoised gold finger bevel image can provide better input data quality in the subsequent feature extraction and decoding process. This helps to improve the accuracy and stability of the decoder, and thus improve the setting effect of the chamfer angle. Therefore, denoising by denoiser convolution can remove noise and interference in the gold finger bevel reference image, improve image quality and accuracy, and improve the subsequent processing effect. This is of great significance for the calculation and setting of the gold finger bevel angle.

[0041] Next, consider that multiple convolutional layers can gradually extract different levels of feature information from the gold finger bevel edge image. Each convolutional layer captures features at different scales and levels of abstraction, from edges and textures to higher-level shapes and structures. This helps to more comprehensively describe the characteristics of the gold finger bevel edge.

[0042] Specifically, the output of multiple convolutional layers can be viewed as a hierarchical representation of the beveled edge image. Earlier convolutional layers primarily capture low-level features, while deeper layers extract higher-level semantic features. This hierarchical representation helps better understand the structure and shape of the beveled edge. Therefore, a convolutional neural network model with multiple convolutional layers can effectively extract feature information from the beveled edge image, enabling automated analysis and processing of the beveled edge.

[0043] Next, considering that the bevel edge of a gold finger may have different shapes, curves, and details, by splitting the feature map into multiple sub-feature maps, independent feature extraction can be performed on different parts of the bevel edge of the gold finger. This helps to more accurately capture the feature information of each local area. In addition, the features of the bevel edge of a gold finger may exist in smaller local areas. By splitting the feature map, the feature representation of the bevel edge of the gold finger can be subdivided into smaller parts, providing a more refined feature representation. This helps to more accurately describe the shape and structure of the bevel edge of the gold finger. In addition, the bevel edge of a gold finger may have features at different scales and levels. By splitting the feature map, feature extraction and analysis can be performed at different scales. This helps to capture the multi-scale features of the bevel edge of the gold finger and improves the adaptability to bevel edges of different sizes and shapes. Therefore, by splitting the feature map of the bevel edge of the gold finger, the local features and details of the bevel edge of the gold finger can be better captured, providing a more accurate and refined feature representation.

[0044] Next, consider that, through convolution operations, CNNs can extract useful feature information from the input sub-feature maps of the gold finger bevel edge feature. These features can represent the shape, texture, and edges of the gold finger bevel edge, facilitating subsequent analysis and processing. Furthermore, the convolutional neural network, with its multi-layered structure of convolutional and pooling layers, can gradually extract different levels of representation of the gold finger bevel edge feature. Low-level convolutional layers can capture local details of the gold finger bevel edge, while higher-level convolutional layers capture more abstract, overall features. Through multiple layers of convolution, a richer and more diverse representation of the gold finger bevel edge feature can be obtained. Furthermore, the convolution operation performed by the convolutional neural network on the gold finger bevel edge feature sub-feature maps can enhance the robustness and noise resistance of the feature. The convolution operation smoothes the feature map, reduces the impact of noise, and extracts a noise-robust feature representation. Therefore, the convolution operation of the feature extractor can extract useful feature information from multiple sub-feature maps of the gold finger bevel edge feature.

[0045] Furthermore, the gold finger bevel feature map is decoded and regressed through a decoder to obtain a decoded value, and the decoded value is used to represent the angle setting value that the mold needs to chamfer. Considering that the gold finger bevel feature map is an abstract feature representation obtained by convolution of the feature extractor. The role of the decoder is to map these abstract features back to the original angle setting value. Through decoding by the decoder, the feature information of the gold finger bevel can be converted into an interpretable angle setting value. Decoding by the decoder can restore the angle setting value of the gold finger bevel as much as possible. The design goal of the decoder is to minimize the information loss in the feature extraction and decoding process to maintain the precision and accuracy of the decoded value. This helps to ensure the accuracy and consistency of the chamfering operation and facilitates subsequent angle setting and processing operations.

[0046] In particular, considering that the Golden Finger slant edge feature map may contain a large amount of redundant information that is not important for feature representation and expression, sparsification can remove this redundant information, making the feature map more streamlined and compact. Furthermore, the sparsified feature map has fewer non-zero elements, requiring less storage and computing resources for computation. This can speed up model inference and reduce computational costs. At the same time, sparsification can highlight key features in the Golden Finger slant edge feature map, focusing attention on features that are more meaningful to the problem or task. This helps improve model performance and generalization. Sparsification can reduce redundant information and noise in the feature map, thereby reducing the risk of model overfitting. Overfitting refers to the overfitting of the model to the training data, resulting in a decrease in generalization ability on new data. By sparsifying the Golden Finger slant edge feature map, the model's generalization ability can be improved.

[0047] Specifically, calculating the sparse golden finger bevel edge feature map of the golden finger bevel edge feature map includes: determining the size and the number of channels of the golden finger bevel edge feature map, wherein the size of the golden finger bevel edge feature map is H×W, H is the height, W is the width, and the number of channels is C; for each pixel position (i, j) of the golden finger bevel edge feature map, extracting the channel feature vector of the golden finger bevel edge feature map at each pixel position as the feature descriptor of each pixel position; for each pixel position (i, j) of the classification feature map, calculating the feature descriptor of each pixel position (i, j) and the feature descriptor of other pixel positions of the classification feature map. Set the Euclidean distance between the feature descriptors of (k, l), and calculate the sum of multiple Euclidean distances as the score value of each pixel position (i, j) of the classification feature map; perform maximum-based normalization on the score value of each pixel position (i, j) of the classification feature map to obtain the homography probability density value of each pixel position of the classification feature map; use the probability density feature matrix composed of the homography probability density value of each pixel position of the gold finger slant edge feature map as the weight matrix, and perform position point multiplication on the feature matrices of each channel dimension of the gold finger slant edge feature map to obtain the sparse gold finger slant edge feature map.

[0048] In the technical solution of the present application, a continuous and smooth probability distribution is obtained by calculating the homography probability density space representation of the golden finger oblique edge feature map, thereby avoiding problems such as threshold selection and post-processing common in traditional methods. At the same time, according to the distribution of pixels of different categories in the probability space, the weight of the golden finger oblique edge feature map is adaptively adjusted, thereby enhancing the flexibility and adaptability of the feature expression of the golden finger oblique edge feature map, and being able to adapt to the needs of different scenarios and tasks.

[0049] Based on this, the present application provides a gold finger bevel processing system, which includes: a reference image acquisition module for acquiring a gold finger bevel reference image; an image denoising module for passing the gold finger bevel reference image through a depth-separable convolutional neural network as a denoiser to obtain a denoised gold finger bevel image; a gold finger bevel feature extraction module for passing the denoised gold finger bevel image through a convolutional neural network model with multiple convolution layers to obtain a denoised gold finger bevel feature map; a feature map segmentation module for segmenting the denoised gold finger bevel feature map. The feature map is segmented to obtain a plurality of gold finger bevel feature sub-feature maps; a feature convolution network module is used to pass the plurality of gold finger bevel feature sub-feature maps through a convolutional neural network model as a feature extractor to obtain a gold finger bevel feature map; a sparse calculation module is used to calculate a sparse gold finger bevel feature map of the gold finger bevel feature map; and a chamfer setting module is used to decode and regress the sparse gold finger bevel feature map through a decoder to obtain a decoding value, and the decoding value is used to indicate the angle setting value that the grinding tool needs to chamfer.

[0050] After introducing the basic principles of the present application, various non-limiting embodiments of the present application will be described in detail with reference to the accompanying drawings.

[0051] Exemplary Systems

[0052] Figure 1 FIG. 1 is a block diagram of a bevel processing system for a gold finger according to an embodiment of the present application. Figure 1 As shown, according to the embodiment of the present application, the bevel processing system 100 of the gold finger includes: a reference image acquisition module 110, which is used to obtain a gold finger bevel reference image; an image denoising module 120, which is used to pass the gold finger bevel reference image through a depth-separable convolutional neural network as a denoiser to obtain a denoised gold finger bevel image; a gold finger bevel feature extraction module 130, which is used to pass the denoised gold finger bevel image through a convolutional neural network model with multiple convolution layers to obtain a denoised gold finger bevel feature map; a feature map segmentation module 140, which is used to extract the denoised gold finger bevel feature map. The bevel feature map is segmented to obtain a plurality of gold finger bevel feature sub-feature maps; a feature convolution network module 150 is used to pass the plurality of gold finger bevel feature sub-feature maps through a convolutional neural network model as a feature extractor to obtain a gold finger bevel feature map; a sparse calculation module 160 is used to calculate a sparse gold finger bevel feature map of the gold finger bevel feature map; and a chamfer setting module 170 is used to decode and regress the sparse gold finger bevel feature map through a decoder to obtain a decoded value, and the decoded value is used to indicate the angle setting value that the grinding tool needs to chamfer.

[0053] Figure 2 FIG. 1 is a schematic diagram of the structure of the beveling processing system of the gold finger according to the embodiment of the present application. Figure 2 As shown, first, a gold finger bevel reference image is obtained. Then, the gold finger bevel reference image is passed through a depth-separable convolutional neural network as a denoiser to obtain a denoised gold finger bevel image. Then, the denoised gold finger bevel image is passed through a convolutional neural network model with multiple convolutional layers to obtain a denoised gold finger bevel feature map. Then, the denoised gold finger bevel feature map is subjected to feature map segmentation to obtain a plurality of gold finger bevel feature sub-feature maps. Then, the plurality of gold finger bevel feature sub-feature maps are passed through a convolutional neural network model as a feature extractor to obtain a gold finger bevel feature map. Then, a sparse gold finger bevel feature map of the gold finger bevel feature map is calculated. Finally, the sparse gold finger bevel feature map is decoded and regressed through a decoder to obtain a decoded value, which is used to represent the angle setting value that the grinding tool needs to chamfer.

[0054] In an embodiment of the present application, the reference image acquisition module 110 is used to obtain a reference image of the bevel of the gold finger. Considering that the bevel of the gold finger usually needs to be chamfered at a certain angle. By acquiring the reference image, the angle of the bevel of the gold finger can be measured and set, thereby ensuring the accuracy and consistency of the chamfer. After acquiring the reference image, the image can be analyzed and processed using technologies such as image processing and deep learning to extract the characteristic information of the bevel of the gold finger, thereby realizing automated chamfering. In addition, the reference image can be used for quality control and inspection. By comparing the difference between the bevel of the gold finger after actual processing and the reference image, the processing deviation can be discovered and corrected in time to ensure that the quality of the product meets the requirements. In addition, the reference image can be used to analyze the shape and characteristics of the bevel of the gold finger, and then perform process optimization. By comparing the data of different reference images, the optimal chamfer angle and processing parameters can be found to improve production efficiency and product quality.

[0055] In this embodiment of the present application, the image denoising module 120 is configured to pass the gold finger bevel edge reference image through a depthwise separable convolutional neural network (DSN) as a denoiser to obtain a denoised gold finger bevel edge image. Considering that the denoiser convolution can effectively reduce noise and interference in the image, the gold finger bevel edge image is clearer and more recognizable. This helps improve the accuracy of subsequent image processing and analysis. Furthermore, the denoiser convolution can maximize the preservation of detail information in the gold finger bevel edge image while removing noise. This ensures that the calculation and analysis of the chamfer angle are based on accurate image data, improving the reliability of the results. The denoised gold finger bevel edge image provides better input data quality for subsequent feature extraction and decoding processes. This helps improve the accuracy and stability of the decoder, thereby enhancing the chamfer angle setting effect. Therefore, denoising through the denoiser convolution can remove noise and interference from the gold finger bevel edge reference image, improve image quality and accuracy, and enhance subsequent processing results. This is of great significance for the calculation and setting of the gold finger bevel edge angle.

[0056] Figure 3 FIG. 1 is a block diagram of an image noise reduction module in a bevel processing system for a gold finger according to an embodiment of the present application. Specifically, in the embodiment of the present application, Figure 3 As shown, the image denoising module 120 includes: an encoding unit 121, which is used to input the golden finger bevel reference image into the encoder of the deep separable convolutional neural network serving as a denoiser, wherein the encoder uses a convolution layer to perform explicit spatial encoding on the golden finger bevel reference image to obtain image features; and a decoding unit 122, which is used to input the image features into the decoder of the deep separable convolutional neural network serving as a denoiser, wherein the decoder uses a deconvolution layer to perform deconvolution processing on the image features to obtain the denoised image of the object to be cleaned; wherein the deep separable convolutional neural network serving as a denoiser is a denoiser based on an automatic codec, the image encoder of the automatic codec includes at least one convolution layer, and the image decoder of the automatic codec includes at least one deconvolution layer.

[0057] In this embodiment of the present application, the gold finger bevel feature extraction module 130 is configured to pass the denoised gold finger bevel image through a convolutional neural network model with multiple convolutional layers to obtain a denoised gold finger bevel feature map. Considering that multiple convolutional layers can extract different levels of feature information from the gold finger bevel image layer by layer, each convolutional layer can capture features at different scales and levels of abstraction, from edges and textures to higher-level shapes and structures. This helps to more comprehensively describe the characteristics of the gold finger bevel. Specifically, the output of the multiple convolutional layers can be viewed as a hierarchical representation of the gold finger bevel image. Earlier convolutional layers primarily capture low-level features, while deeper convolutional layers can extract higher-level semantic features. This hierarchical representation helps to better understand the structure and shape of the gold finger bevel. Therefore, the convolutional neural network model with multiple convolutional layers can effectively extract feature information from the gold finger bevel image, enabling automated analysis and processing of the gold finger bevel.

[0058] Specifically, in an embodiment of the present application, the golden finger bevel feature extraction module is used to: use each hybrid convolution layer of the convolutional neural network model of the multi-layer convolutional layer to perform the following on the input data in the forward pass of the layer: use a first convolution kernel with a first size to perform convolution encoding on the denoised golden finger bevel image to obtain a first scale feature map; use a second convolution kernel with a first void rate to perform convolution encoding on the denoised golden finger bevel image to obtain a second scale feature map; use a third convolution kernel with a second void rate to perform convolution encoding on the denoised golden finger bevel image to obtain a third scale feature map; aggregate the first scale feature map, the second scale feature map, the third scale feature map and the fourth scale feature map along the channel dimension to obtain an aggregated feature map; perform pooling processing on the aggregated feature map to generate a pooled feature map; and, perform activation processing on the pooled feature map to generate an activation feature map; wherein the output of the last layer of the convolutional neural network model comprising multiple hybrid convolution layers is the denoised golden finger bevel feature map.

[0059] More specifically, in an embodiment of the present application, the golden finger bevel edge feature extraction module is used to: the first convolution kernel, the second convolution kernel, the third convolution kernel and the fourth convolution kernel have the same size, and the second convolution kernel, the third convolution kernel and the fourth convolution kernel have different void rates.

[0060] In this embodiment of the present application, the feature map segmentation module 140 is configured to segment the denoised gold finger bevel edge feature map to obtain multiple gold finger bevel edge feature sub-feature maps. Considering that gold finger bevel edges may have different shapes, curves, and details, segmenting the feature map into multiple sub-feature maps allows for independent feature extraction of different parts of the gold finger bevel edge. This helps more accurately capture the feature information of each local area. Furthermore, the features of the gold finger bevel edge may exist in smaller local areas. Feature map segmentation allows the feature representation of the gold finger bevel edge to be subdivided into smaller parts, providing a more refined feature representation. This helps more accurately describe the shape and structure of the gold finger bevel edge. Furthermore, the gold finger bevel edge may have features at different scales and levels. Feature map segmentation allows for feature extraction and analysis at different scales. This helps capture the multi-scale features of the gold finger bevel edge and improves adaptability to gold finger bevel edges of varying sizes and shapes. Therefore, segmenting the gold finger bevel edge feature map can better capture the local features and details of the gold finger bevel edge, providing a more accurate and refined feature representation.

[0061] In this embodiment of the present application, the feature convolutional network module 150 is configured to pass the multiple gold finger bevel edge feature sub-feature maps through a convolutional neural network model acting as a feature extractor to obtain a gold finger bevel edge feature map. Considering that, through convolution operations, the CNN can extract useful feature information from the input gold finger bevel edge feature sub-feature maps. These features can represent the shape, texture, edges, and other characteristics of the gold finger bevel edge, facilitating subsequent analysis and processing. Furthermore, the convolutional neural network has a multi-layered structure of convolutional and pooling layers, which can gradually extract different levels of representation of the gold finger bevel edge features. Low-level convolutional layers can capture local details of the gold finger bevel edge, while higher-level convolutional layers can capture more abstract, overall features. Through multiple layers of convolution, a richer and more diverse gold finger bevel edge feature representation can be obtained. Furthermore, the convolutional neural network's convolution operations on the gold finger bevel edge feature sub-feature maps can enhance the robustness and noise resistance of the features. Through the convolution operation, the feature map can be smoothed, the impact of noise can be reduced, and a feature representation that is robust to noise can be extracted. Therefore, useful feature information can be extracted from multiple golden finger bevel feature sub-feature maps through the convolution operation of the feature extractor.

[0062] Specifically, in an embodiment of the present application, the feature convolution network module is used to: use the layers of the convolutional neural network model as a feature extractor to perform the following on the input data in the forward pass of the layer: convolution processing on the input data to obtain a convolution feature map; global mean pooling based on the local feature matrix on the convolution feature map to obtain a pooled feature map; nonlinear activation on the pooled feature map to obtain an activation feature map; wherein the output of the last layer of the convolutional neural network model as a feature extractor is each of the multiple gold finger bevel sub-feature maps of the multiple gold finger bevel sub-feature maps, and the input of the first layer of the convolutional neural network model as a feature extractor is each of the multiple gold finger bevel feature sub-feature maps of the multiple gold finger bevel feature sub-feature maps; and, the multiple gold finger bevel sub-feature maps are cascaded to obtain the gold finger bevel feature map.

[0063] In the embodiment of the present application, the sparsification calculation module 160 is used to calculate the sparse gold finger bevel edge feature map of the gold finger bevel edge feature map.

[0064] In particular, considering that the Golden Finger slant edge feature map may contain a large amount of redundant information that is not important for feature representation and expression, sparsification can remove this redundant information, making the feature map more streamlined and compact. Furthermore, the sparsified feature map has fewer non-zero elements, requiring less storage and computing resources for computation. This can speed up model inference and reduce computational costs. At the same time, sparsification can highlight key features in the Golden Finger slant edge feature map, focusing attention on features that are more meaningful to the problem or task. This helps improve model performance and generalization. Sparsification can reduce redundant information and noise in the feature map, thereby reducing the risk of model overfitting. Overfitting refers to the overfitting of the model to the training data, resulting in a decrease in generalization ability on new data. By sparsifying the Golden Finger slant edge feature map, the model's generalization ability can be improved.

[0065] Figure 4 FIG. 1 is a block diagram of a sparse calculation module in a bevel processing system for a gold finger according to an embodiment of the present application. Specifically, in the embodiment of the present application, Figure 4As shown, the sparsification calculation module 160 includes: a size and channel number determination unit 161, which is used to determine the size and channel number of the golden finger bevel feature map, wherein the size of the golden finger bevel feature map is H×W, H is the height, W is the width, and the number of channels is C; a channel feature vector extraction unit 162, which is used to extract the channel feature vector of the golden finger bevel feature map at each pixel position (i, j) as the feature descriptor of each pixel position; a Euclidean distance calculation unit 163, which is used to calculate the distance between the feature descriptor of each pixel position (i, j) of the classification feature map and other images of the classification feature map. The Euclidean distance between the feature descriptors of the pixel position (k, l) is calculated, and the sum of multiple Euclidean distances is calculated as the score value of each pixel position (i, j) of the classification feature map; the homography probability density value calculation unit 164 is used to perform maximum value-based normalization processing on the score value of each pixel position (i, j) of the classification feature map to obtain the homography probability density value of each pixel position of the classification feature map; and the weight applying unit 165 is used to use the probability density feature matrix composed of the homography probability density value of each pixel position of the gold finger bevel feature map as the weight matrix, and perform position point multiplication on the feature matrix of each channel dimension of the gold finger bevel feature map to obtain the sparse gold finger bevel feature map.

[0066] In the technical solution of the present application, a continuous and smooth probability distribution is obtained by calculating the homography probability density space representation of the golden finger oblique edge feature map, thereby avoiding problems such as threshold selection and post-processing common in traditional methods. At the same time, according to the distribution of pixels of different categories in the probability space, the weight of the golden finger oblique edge feature map is adaptively adjusted, thereby enhancing the flexibility and adaptability of the feature expression of the golden finger oblique edge feature map, and being able to adapt to the needs of different scenarios and tasks.

[0067] In an embodiment of the present application, the chamfer setting module 170 is used to decode and regress the sparse gold finger bevel feature map through a decoder to obtain a decoded value, and the decoded value is used to represent the angle setting value that the mold needs to chamfer. Considering that the role of the decoder is to map these abstract features back to the original angle setting value. Through decoding by the decoder, the feature information of the gold finger bevel can be converted into an interpretable angle setting value. Decoding by the decoder can restore the angle setting value of the gold finger bevel as much as possible. The design goal of the decoder is to minimize the information loss in the feature extraction and decoding process to maintain the precision and accuracy of the decoded value. This helps to ensure the accuracy and consistency of the chamfering operation and facilitates subsequent angle setting and processing operations.

[0068] Specifically, in an embodiment of the present application, the chamfer setting module is used to: use the decoder to perform decoding regression on the sparse golden finger bevel feature map using the following decoding formula to obtain the decoding value; wherein the decoding formula is: Where X is the sparse cheat sheet feature map, Y is the decoded value, and W is the weight matrix. Represents matrix multiplication.

[0069] In summary, the gold finger beveling processing system 100 according to the embodiment of the present application is described. It utilizes artificial intelligence technology based on a deep neural network model to obtain a gold finger bevel reference image, de-noise the image, and then perform multi-level feature extraction through a convolutional network of multiple convolutional layers and feature extractors to obtain a decoded value representing the desired chamfer angle setting for the mold. This enables automated and precise angle setting, thereby improving production efficiency and product quality.

[0070] Exemplary Methods

[0071] Figure 5 FIG. 1 is a flow chart of a method for processing the bevel edge of a gold finger according to an embodiment of the present application. Figure 5 As shown, the bevel processing method of the gold finger according to the embodiment of the present application includes: S110, obtaining a gold finger bevel reference image; S120, passing the gold finger bevel reference image through a depth-separable convolutional neural network as a denoiser to obtain a denoised gold finger bevel image; S130, passing the denoised gold finger bevel image through a convolutional neural network model with multiple convolutional layers to obtain a denoised gold finger bevel feature map; S140, performing feature map segmentation on the denoised gold finger bevel feature map to obtain a plurality of gold finger bevel feature sub-feature maps; S150, passing the plurality of gold finger bevel feature sub-feature maps through a convolutional neural network model as a feature extractor to obtain a gold finger bevel feature map; S260, calculating a sparse gold finger bevel feature map of the gold finger bevel feature map; S170, decoding and regressing the sparse gold finger bevel feature map through a decoder to obtain a decoding value, which is used to represent the angle setting value that the grinding tool needs to chamfer.

[0072] Here, those skilled in the art will appreciate that the specific operations of each step in the above-mentioned beveling method of the gold finger have been described in detail in the above reference. Figures 1 to 4 The description of the beveling processing system of the gold finger has been introduced in detail, and therefore, its repeated description will be omitted.

[0073] Exemplary electronic devices

[0074] Below, reference Figure 6 To describe the electronic device according to the embodiment of the present application. Figure 6FIG is a block diagram of an electronic device according to an embodiment of the present application. Figure 6 As shown, the electronic device 10 includes one or more processors 11 and a memory 12 .

[0075] The processor 11 may be a central processing unit (CPU) or other forms of processing modules having data processing capabilities and / or instruction execution capabilities, and may control other components in the electronic device 10 to perform desired functions.

[0076] 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), a hard disk, a 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 functions of the gold finger beveling processing system and method thereof of the various embodiments of the present application described above and / or other desired functions. Various contents such as a gold finger bevel reference image may also be stored in the computer-readable storage medium.

[0077] In one example, the electronic device 10 may further include an input device 13 and an output device 14 , and these components are interconnected via a bus system and / or other forms of connection mechanisms (not shown).

[0078] The input device 13 may include, for example, a keyboard, a mouse, and the like.

[0079] The output device 14 can output various information to the outside, including decoded values, etc. The output device 14 can include, for example, a display, a speaker, a printer, a communication network and a remote output device connected thereto, and the like.

[0080] Of course, to simplify, Figure 6 Only some of the components related to the present application in the electronic device 10 are shown, and components such as a bus, an input / output interface, etc. are omitted. In addition, the electronic device 10 may further include any other appropriate components according to specific application scenarios.

[0081] Exemplary computer program products and computer-readable storage media

[0082] In addition to the above-mentioned methods and devices, an embodiment of the present application may also be a computer program product, which includes computer program instructions, which, when executed by a processor, enable the processor to execute the steps in the functions of the beveling method of the gold finger according to various embodiments of the present application described in the above-mentioned "Exemplary Method" section of this specification.

[0083] The computer program product may be written in any combination of one or more programming languages ​​to implement the program code for performing the operations of the embodiments of the present application, including object-oriented programming languages ​​such as Java, C++, and conventional procedural programming languages ​​such as "C" or similar programming languages. The program code may 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.

[0084] In addition, an embodiment of the present application may also be a computer-readable storage medium having computer program instructions stored thereon, which, when executed by a processor, enables the processor to execute the steps in the functions of the beveling method of the gold finger according to various embodiments of the present application described in the above "Exemplary Method" section of this specification.

[0085] The computer-readable storage medium can adopt any combination of one or more readable media. The readable medium can be a readable signal medium or a readable storage medium. The readable storage medium can, for example, include but is not limited to a system, device or component of electricity, magnetism, light, electromagnetic, infrared, or semiconductor, or any combination thereof. More specific examples (non-exhaustive list) of readable storage media include: an electrical connection with one or more wires, a portable disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination thereof.

[0086] The basic principles of the present application have been described above in conjunction with specific embodiments. However, it should be noted that the advantages, strengths, and effects mentioned in this application are merely illustrative and not restrictive, and it should not be assumed that these advantages, strengths, and effects are required of each embodiment of this application. In addition, the specific details disclosed above are merely illustrative and facilitating understanding, and are not restrictive. The above details do not limit this application to necessarily being implemented using the above specific details.

[0087] The block diagrams of the devices, devices, equipment, 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 will be appreciated by those skilled in the art, these devices, devices, equipment, and systems can be connected, arranged, or configured in any manner. Words such as "include," "comprise," "have," and the like are open-ended words, meaning "including but not limited to," and can be used interchangeably therewith. The words "or" and "and" used herein refer to the words "and / or" and can be used interchangeably therewith, unless the context clearly indicates otherwise. The word "such as" used herein refers to the phrase "such as but not limited to," and can be used interchangeably therewith.

[0088] It should also be noted that in the apparatus, device, and system of the present application, each component or each step can be decomposed and / or recombined. Such decomposition and / or recombination should be regarded as equivalent solutions of the present application.

[0089] The above description of the disclosed aspects is provided to enable any person skilled in the art to make or use the present application. Various modifications to these aspects will be readily apparent to those skilled in the art, and the general principles defined herein may be applied to other aspects without departing from the scope of the present application. Therefore, the present 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.

[0090] The above description has been provided for the purpose of illustration and description. In addition, this description is not intended to limit the embodiments of the present application to the forms disclosed herein. Although a number of example aspects and embodiments have been discussed above, those skilled in the art will recognize certain variations, modifications, alterations, additions, and sub-combinations thereof.

[0091] The basic principles of the present application have been described above in conjunction with specific embodiments. However, it should be noted that the advantages, strengths, and effects mentioned in this application are merely illustrative and not restrictive, and it should not be assumed that these advantages, strengths, and effects are required of each embodiment of this application. In addition, the specific details disclosed above are merely illustrative and facilitating understanding, and are not restrictive. The above details do not limit this application to necessarily being implemented using the above specific details.

[0092] The block diagrams of the devices, devices, equipment, 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 will be appreciated by those skilled in the art, these devices, devices, equipment, and systems can be connected, arranged, or configured in any manner. Words such as "include," "comprise," "have," and the like are open-ended words, meaning "including but not limited to," and can be used interchangeably therewith. The words "or" and "and" used herein refer to the words "and / or" and can be used interchangeably therewith, unless the context clearly indicates otherwise. The word "such as" used herein refers to the phrase "such as but not limited to," and can be used interchangeably therewith.

[0093] It should also be noted that in the apparatus, device, and system of the present application, each component or each step can be decomposed and / or recombined. Such decomposition and / or recombination should be regarded as equivalent solutions of the present application.

[0094] The above description of the disclosed aspects is provided to enable any person skilled in the art to make or use the present application. Various modifications to these aspects will be readily apparent to those skilled in the art, and the general principles defined herein may be applied to other aspects without departing from the scope of the present application. Therefore, the present 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.

[0095] The above description has been provided for the purpose of illustration and description. In addition, this description is not intended to limit the embodiments of the present application to the forms disclosed herein. Although a number of example 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 beveling processing system for a gold finger, characterized in that: include: A reference image acquisition module is used to obtain a reference image of the bevel edge of the gold finger; An image denoising module, configured to pass the gold finger bevel edge reference image through a depthwise separable convolutional neural network as a denoiser to obtain a denoised gold finger bevel edge image; A gold finger bevel feature extraction module is used to pass the denoised gold finger bevel image through a convolutional neural network model with multiple convolution layers to obtain a denoised gold finger bevel feature map; A feature map segmentation module is used to segment the denoised golden finger bevel edge feature map to obtain a plurality of golden finger bevel edge feature sub-feature maps; A feature convolutional network module, configured to pass the plurality of gold finger bevel feature sub-feature maps through a convolutional neural network model as a feature extractor to obtain a gold finger bevel feature map; A sparse calculation module, used for calculating a sparse golden finger oblique edge feature map of the golden finger oblique edge feature map; A chamfer setting module is used to decode and regress the sparse gold finger bevel feature map through a decoder to obtain a decoded value, and the decoded value is used to indicate the angle setting value of the grinding tool that needs to be chamfered; The sparse calculation module includes: A size and channel number determination unit, configured to determine the size and channel number of the gold finger bevel edge feature map, wherein the size of the gold finger bevel edge feature map is H×W, where H is the height, W is the width, and the number of channels is C; A channel feature vector extraction unit is used to extract, for each pixel position (i, j) of the golden finger bevel edge feature map, the channel feature vector of the golden finger bevel edge feature map at each pixel position as a feature descriptor of each pixel position; a Euclidean distance calculation unit, configured to calculate, for each pixel position (i, j) of the classification feature map, a Euclidean distance between a feature descriptor at the pixel position (i, j) and feature descriptors at other pixel positions (k, l) of the classification feature map, and calculate a sum of multiple Euclidean distances as a score value for each pixel position (i, j) of the classification feature map; a homography probability density value calculation unit, configured to perform a maximum-based normalization process on the score value of each pixel position (i, j) of the classification feature map to obtain a homography probability density value of each pixel position of the classification feature map; The weight applying unit is used to use the probability density feature matrix composed of the homography probability density value of each pixel position of the golden finger bevel edge feature map as the weight matrix, and multiply the feature matrix of each channel dimension of the golden finger bevel edge feature map by position point to obtain the sparse golden finger bevel edge feature map.

2. The beveling processing system of the gold finger according to claim 1, characterized in that: The image noise reduction module includes: an encoding unit, configured to input the golden finger bevel edge reference image into an encoder of the depthwise separable convolutional neural network serving as a denoiser, wherein the encoder uses a convolutional layer to perform explicit spatial encoding on the golden finger bevel edge reference image to obtain image features; A decoding unit, configured to input the image features into a decoder of the depthwise separable convolutional neural network serving as a denoiser, wherein the decoder uses a deconvolution layer to perform deconvolution processing on the image features to obtain the denoised golden finger bevel edge image; The deep separable convolutional neural network used as a denoiser is a denoiser based on an autocoder, wherein the image encoder of the autocoder includes at least one convolutional layer, and the image decoder of the autocoder includes at least one deconvolutional layer.

3. The beveling processing system of the gold finger according to claim 2, characterized in that: The gold finger bevel feature extraction module is used to: Each hybrid convolutional layer of the convolutional neural network model using the multi-layer convolutional layer performs the following operations on the input data in the forward pass of the layer: Using a first convolution kernel with a first size to perform convolution encoding on the denoised golden finger oblique edge image to obtain a first scale feature map; Using a second convolution kernel with a first dilation rate to perform convolution encoding on the denoised golden finger oblique edge image to obtain a second scale feature map; Using a third convolution kernel with a second dilation rate to perform convolution encoding on the denoised golden finger oblique edge image to obtain a third-scale feature map; Aggregating the first scale feature map, the second scale feature map, and the third scale feature map along the channel dimension to obtain an aggregated feature map; Performing pooling processing on the aggregated feature map to generate a pooled feature map; Performing activation processing on the pooled feature map to generate an activation feature map; Among them, the output of the last layer of the convolutional neural network model including multiple mixed convolutional layers is the denoised golden finger bevel edge feature map.

4. The beveling processing system of the gold finger according to claim 3, characterized in that: The gold finger bevel feature extraction module is used to: The first convolution kernel, the second convolution kernel, and the third convolution kernel have the same size, and the second convolution kernel and the third convolution kernel have different void rates.

5. The beveling processing system of the gold finger according to claim 4, characterized in that: The feature convolutional network module is used to: Each layer of the convolutional neural network model used as a feature extractor performs the following on the input data in the forward pass of the layer: Perform convolution on the input data to obtain a convolution feature map; Performing global mean pooling based on a local feature matrix on the convolution feature map to obtain a pooled feature map; Performing nonlinear activation on the pooled feature map to obtain an activated feature map; The output of the last layer of the convolutional neural network model as a feature extractor is each of the multiple gold finger bevel sub-feature maps, and the input of the first layer of the convolutional neural network model as a feature extractor is each of the multiple gold finger bevel feature sub-feature maps; The plurality of golden finger bevel edge sub-feature maps are cascaded to obtain the golden finger bevel edge feature map.

6. The beveling processing system for the gold finger according to claim 5, characterized in that: The chamfer setting module is used to: The decoder is used to perform decoding regression on the sparse golden finger oblique edge feature map using the following decoding formula to obtain the decoding value; wherein the decoding formula is: Where X is the sparse cheat sheet feature map, Y is the decoded value, and W is the weight matrix. Represents matrix multiplication.

7. A method for processing the bevel of a gold finger, characterized in that: include: Get the gold finger bevel reference image; Passing the gold finger bevel edge reference image through a depthwise separable convolutional neural network as a denoiser to obtain a denoised gold finger bevel edge image; Passing the denoised golden finger bevel edge image through a convolutional neural network model with multiple convolutional layers to obtain a denoised golden finger bevel edge feature map; Performing feature map segmentation on the denoised golden finger bevel edge feature map to obtain a plurality of golden finger bevel edge feature sub-feature maps; Passing the plurality of gold finger oblique edge feature sub-feature maps through a convolutional neural network model as a feature extractor to obtain a gold finger oblique edge feature map; Calculating the sparse gold finger bevel edge feature map of the gold finger bevel edge feature map, including: determining the size and the number of channels of the gold finger bevel edge feature map, wherein the size of the gold finger bevel edge feature map is H×W, H is the height, W is the width, and the number of channels is C; for each pixel position (i, j) of the gold finger bevel edge feature map, extracting the channel feature vector of the gold finger bevel edge feature map at each pixel position as the feature descriptor of each pixel position; for each pixel position (i, j) of the classification feature map, calculating the feature descriptor of each pixel position (i, j) and other pixel positions of the classification feature map (k, l) feature descriptors, and calculate the sum of multiple Euclidean distances as the score value of each pixel position (i, j) of the classification feature map; the score value of each pixel position (i, j) of the classification feature map is normalized based on the maximum value to obtain the homography probability density value of each pixel position of the classification feature map; the probability density feature matrix composed of the homography probability density value of each pixel position of the gold finger oblique edge feature map is used as the weight matrix, and the feature matrices of each channel dimension of the gold finger oblique edge feature map are multiplied by position point to obtain the sparse gold finger oblique edge feature map; The sparse gold finger bevel feature map is decoded and regressed through a decoder to obtain a decoded value, and the decoded value is used to represent the angle setting value that the grinding tool needs to chamfer.

8. The method for processing the bevel edge of a gold finger according to claim 7, characterized in that: The method of passing the gold finger bevel edge reference image through a depthwise separable convolutional neural network as a denoiser to obtain a denoised gold finger bevel edge image comprises: Inputting the golden finger bevel reference image into the encoder of the depthwise separable convolutional neural network as a denoiser, wherein the encoder uses a convolutional layer to perform explicit spatial encoding on the golden finger bevel reference image to obtain image features; Inputting the image features into a decoder of the depthwise separable convolutional neural network serving as a denoiser, wherein the decoder uses a deconvolution layer to perform deconvolution processing on the image features to obtain the denoised golden finger bevel edge image; The deep separable convolutional neural network used as a denoiser is a denoiser based on an autocoder, wherein the image encoder of the autocoder includes at least one convolutional layer, and the image decoder of the autocoder includes at least one deconvolutional layer.

9. The method for processing the bevel edge of a gold finger according to claim 8, characterized in that: The denoised golden finger bevel edge image is passed through a convolutional neural network model with multiple convolutional layers to obtain a denoised golden finger bevel edge feature map, including: Each hybrid convolutional layer of the convolutional neural network model using the multi-layer convolutional layer performs the following operations on the input data in the forward pass of the layer: Using a first convolution kernel with a first size to perform convolution encoding on the denoised golden finger oblique edge image to obtain a first scale feature map; Using a second convolution kernel with a first dilation rate to perform convolution encoding on the denoised golden finger oblique edge image to obtain a second scale feature map; Using a third convolution kernel with a second dilation rate to perform convolution encoding on the denoised golden finger oblique edge image to obtain a third-scale feature map; Aggregating the first scale feature map, the second scale feature map, and the third scale feature map along the channel dimension to obtain an aggregated feature map; Performing pooling processing on the aggregated feature map to generate a pooled feature map; Performing activation processing on the pooled feature map to generate an activation feature map; Among them, the output of the last layer of the convolutional neural network model including multiple mixed convolutional layers is the denoised golden finger bevel edge feature map.

Images