Motor stator core weld image enhancement method based on machine vision

Abstract

The invention provides a motor stator core welding seam image enhancement method based on machine vision, which improves the definition of a welding seam image through a global feature extraction module, a local feature fusion module and a detail enhancement module, extracts global semantic information of the welding seam image by combining multi-scale convolution with an attention mechanism, and improves the accuracy of the welding seam image. The overall feature expression ability is enhanced, global features and local details are effectively combined through convolution operation and a specific fusion strategy, the remarkable features of a welding seam area are highlighted, a refining network based on an edge attention mechanism is designed, the edge and texture of a welding seam are emphatically enhanced, and the detail expressive force of an image is improved.

Description

Technical Field

[0001] The invention belongs to the field of image enhancement, and in particular relates to a method for enhancing the weld image of a motor stator core based on machine vision. Background Art

[0002] The quality of the weld of the motor stator core directly affects the motor's operating performance and life. However, due to the complex welding environment, the weld surface is easily affected by noise, blur, reflection, etc. Traditional image processing methods have limited effects on improving the quality of weld images. Contrast enhancement based on simple enhancement algorithms often fails to effectively highlight the features of the weld area. This limitation poses a challenge to the accurate detection of weld defects.

[0003] Existing technologies mostly focus on a single enhancement method, which makes it difficult to balance the processing of global information and local details. For example, only improving the contrast may lead to the amplification of background noise, and only focusing on local details may ignore the overall structure. Therefore, a comprehensive enhancement method that integrates global features and local details is needed, and an attention mechanism is introduced to highlight the key areas of the weld, so as to better adapt to the complex needs of improving the quality of weld images.

[0004] By proposing a weld image enhancement method for the motor stator core based on machine vision, the clarity and detail expression of the weld image can be improved, providing high-quality input data for weld defect detection and quality assessment. The application of this method can not only improve the detection accuracy, but also enhance the automation level of industrial production, which is of great significance to ensuring the quality of motor products and extending their service life. Summary of the invention

[0005] The present invention provides a motor stator core weld image enhancement method based on machine vision, aiming to improve the clarity and detail expression ability of the weld image, so as to more efficiently perform defect detection and quality assessment. The method includes three modules: global feature extraction, local feature fusion and detail enhancement. In the global feature extraction module, multi-scale convolution combined with attention mechanism is adopted to extract the global semantic information of the weld image and enhance the overall feature expression ability. In the local feature fusion module, through convolution operation and specific fusion strategy, the global features are effectively combined with the local details to highlight the significant features of the weld area. In the detail enhancement module, a refinement network based on edge attention mechanism is designed to focus on enhancing the edge and texture of the weld and improve the detail expression of the image. The method can be widely used in the field of industrial weld image detection and enhancement, and provide technical support for weld quality control.

[0006] The present invention aims to propose a weld image enhancement optimization model and provide a motor stator core weld image enhancement method based on machine vision, comprising the following steps: S1. Production of weld image dataset: select high-resolution industrial camera, use ring light source to reduce shadow and spot interference, adapt to the reflective characteristics of welds of different materials, collect images in an environment with uniform lighting and no strong light interference, select factory site and laboratory environment, collect images of cracks of three different materials: steel, aluminum alloy and stainless steel, and obtain weld image dataset of motor stator core; S2. Preprocess the original data set, including rotating and flipping the data set to generate multi-angle and multi-directional images, and obtain the weld image data set. All images have a uniform size of , preliminary image enhancement is performed using grayscale histogram and Gaussian filtering; S3, build attention module, including Function calculation and dot multiplication operation; S4, build a global feature extraction module, including convolution, Convolution operation, Operation and Activation function; S5, build detail enhancement module, including Convolution operation, Activation function, attention module, operate, manipulation and splicing operations; S6: Construct a local feature fusion module, including Convolution operation, Convolution operation, operate, Activation functions and attention modules; S7. Construct a weld image enhancement model of a motor stator core based on machine vision, including a global feature extraction module, a detail enhancement module, a local feature fusion module and a splicing operation; S8. Use the weld image dataset of the motor stator core to train a machine vision-based weld image enhancement model for the motor stator core.

[0007] Preferably, in S2, for each image in the original data set, it is flipped and rotated at a random angle to obtain a weld image data set of the motor stator core, and the size of each image is unified as , to facilitate the training of the model, histogram equalization is used to process each image to obtain a crack image with contrast enhancement, and Gaussian filtering is used to remove some interference in the image.

[0008] Preferably, in S3, an attention module is constructed, and the input image is recorded as ,in Indicates height, Indicates width, Indicates the number of channels, apply a Function on feature map All channels at each spatial position of the image are obtained to obtain a weight map with the same size as the original image but reflecting the relative importance of each channel , this weight map With the original feature map Multiply element by element to get the weighted feature map ,in Represents an element-wise multiplication operation.

[0009] Preferably, the attention module proposed by S3 is used to enhance the salient features related to the weld, while also suppressing irrelevant or interfering channel information, which helps to improve the performance of subsequent weld detection and quality assessment tasks.

[0010] Preferably, in S4, for the global feature extraction module, the weld image first input is recorded as ,in represents the height of the feature map, represents the width of the feature map, The number of channels representing the feature map, the feature map Processed by convolutional neural unit module, the unit module includes a Convolution kernel, and Activation function, after passing through this module, the feature map is obtained ,in Indicates that the convolution kernel size is The convolution operation, express The convolution kernel, represents the bias term, express operate, express Activation function, next, feature It is further processed by a convolutional neural unit module, which consists of a Convolution kernel, and Activation function, after passing through the unit module, the feature map is obtained ,in Indicates that the convolution kernel size is The convolution operation, express The convolution kernel, represents the bias term, express operate, express Activation function, the feature With features Perform point multiplication to obtain features ,in Represents the dot product operation, feature After being processed by the attention module built by S3, the features are obtained ,in represents the attention module, Indicates that the convolution kernel size is The convolution operation, express The convolution kernel is After the convolutional neural unit module operation, we get ,in Indicates that the convolution kernel size is The convolution operation, express The convolution kernel, represents the bias term, express operate, express The activation function is obtained and Perform a dot multiplication operation to obtain the feature map ,in represents the dot product operation, The features are obtained through the attention module built by S3 ,in represents the attention module, Indicates that the convolution kernel size is The convolution operation, express The convolution kernel is finally The final global feature output is obtained through the convolutional neural unit module ,in Indicates that the convolution kernel size is The convolution operation, express The convolution kernel, represents the bias term, express operate, express Activation function.

[0011] Preferably, the global feature extraction module proposed by S4 is used to combine low-level to high-level feature information through multiple convolution and dot multiplication operations, and through the attention module, it can effectively focus on the key areas in the image, improve the detail expression of the weld in the image, thereby enhancing the quality and usability of the image, helping to extract richer global features and improving the enhancement effect of the weld image.

[0012] Preferably, in S5, for the detail enhancement module, the input feature map ,in represents the height of the feature map, represents the width of the feature map, Indicates the number of channels of the feature map, All channel values ​​at each spatial position are added to obtain a new feature map ,in express No. channels, express No. The value of the channel, next, the feature go through Convolution operation and The activation function performs linear transformation to obtain features ,in Indicates that the convolution kernel size is The convolution operation, Expressed as The convolution kernel, represents the bias term, express Activation function, After being processed by the attention module built by S3, the features are obtained ,in Represents the attention module, repeat the above steps again to get the feature , while the characteristics After two consecutive maximum pooling operations and self-attention mechanism operations, we get ,in Indicates the pooling window size The maximum pooling operation, Represents the self-attention function. Finally, the feature and Stitching to get the final detail enhancement features ,in Represents a concatenation operation.

[0013] Preferably, the detail enhancement module proposed by S5 is used. By splicing features of different scales and levels, the detail enhancement module can provide richer feature representation, which helps to improve the performance of the model. Secondly, through convolution operations and attention mechanisms, the detail information in the weld image can be enhanced. Finally, through reasonable implementation and optimization, high computational efficiency can be maintained while ensuring performance.

[0014] Preferably, in S6, for the local feature fusion module, first, the global feature With detail enhancement features Splice to get ,in Represents the concatenation operation, and then the features It is further processed by a convolutional neural unit module, which consists of a Convolution kernel, and Activation function, after passing through the unit module, the feature map is obtained ,in Indicates that the convolution kernel size is The convolution operation, express The convolution kernel, represents the bias term, express operate, express activation function, and then processed by a convolutional neural unit module, which consists of a Convolution kernel, and Activation function, after passing through this module, the feature map is obtained ,in Indicates that the convolution kernel size is The convolution operation, express The convolution kernel, represents the bias term, express operate, express Activation function, final feature The features are processed by the attention module built by S3 ,in represents the attention module, express The convolution kernel.

[0015] Preferably, the local feature fusion module proposed by S6 is used to achieve an effective combination of global semantic information and local detail features through multi-stage feature processing, which can enhance the key features of texture, edge and defects in weld images.

[0016] Preferably, in S7, for a weld image enhancement model of a motor stator core based on machine vision, the weld image first input is , obtained through the global feature extraction module ,in Represented as the global feature extraction module constructed in S4, Obtained through the detail enhancement module ,in Represented as the detail enhancement module constructed in S5, global features With detail enhancement features Splice to get ,in Represented as the local feature fusion module constructed in S6, and Stitching is performed to obtain the final image enhancement optimization result ,in Represents a concatenation operation.

[0017] In summary, due to the adoption of the technical solution, compared with the prior art, the beneficial effects of the present invention are as follows: the present invention proposes a method for enhancing the weld image of a motor stator core based on machine vision, including image preprocessing, an attention module, a global feature extraction module, a detail enhancement module and a local feature fusion module. The image preprocessing expands the data set, and performs contrast enhancement and Gaussian filtering on the image, thereby improving the quality of the data set and the effect of model training. The attention module is proposed to enhance the significant features related to the weld and suppress irrelevant or interfering channel information, thereby improving the performance of subsequent weld detection and quality assessment tasks. The global feature extraction module is proposed to extract high-order semantic information and low-order detail information, and effectively focus on the key areas in the image, thereby improving the detail performance of the target, thereby enhancing the quality of the image and helping to extract richer global features. The detail enhancement module is proposed to splice image features of different scales to provide richer detail texture features, and combined with the attention mechanism, while enhancing the features, the computational efficiency of the model is improved, and the resource overhead is reduced. The local feature fusion module is proposed to comprehensively enhance the overall details and semantic features of the weld image through the fusion of high-order features and low-order features at multiple stages. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] Figure 1 This is a step-by-step diagram of the motor stator core weld image enhancement method.

[0019] Figure 2 This is the structure diagram of the attention module.

[0020] Figure 3 This is the structure diagram of the global feature extraction module.

[0021] Figure 4 Enhance the module structure diagram for details.

[0022] Figure 5 This is the structure diagram of the local feature fusion module.

[0023] Figure 6 This is the structural diagram of the weld image enhancement model of the motor stator core based on machine vision.

[0024] Figure 7 This is a low-quality image of the weld seam of a motor stator core.

[0025] Figure 8 Image of the weld seam of the motor stator core after optimization and enhancement. DETAILED DESCRIPTION

[0026] The present invention aims to propose a weld image enhancement method for a motor stator core based on machine vision. The core of the method is to expand the diversity of data by performing data enhancement on the data set, improve the image quality, introduce an attention module, strengthen the significant features related to the weld, and suppress irrelevant or interfering channel information, significantly improve the model performance, design a global feature extraction module, which can capture high-order semantic information and low-order detail information, and effectively focus on the key areas in the image, enhance the detail performance of the target, and help to extract richer global features. A detail enhancement module is proposed, which provides richer detail texture information by splicing image features of different scales. Combined with the attention mechanism, while enhancing the feature expression, it also improves the computational efficiency of the model. A local feature fusion module is proposed, which realizes the deep fusion of high-order features and low-order features in multiple stages, and comprehensively enhances the overall details and semantic features of the weld image. The following steps are specifically included, such as Figure 1 shown.

[0027] S1. The images of the motor stator core welds taken by a high-resolution industrial camera were sorted to obtain the initial motor stator core weld image dataset. The images were collected under uniform lighting and without strong light interference, and were divided into two backgrounds: factory site and laboratory environment. Under each background, crack images of three different materials, steel, aluminum alloy and stainless steel, were collected. There were 200 images of each material, and the initial dataset had a total of 1,200 images. The training set and test set were divided in a ratio of 4:1.

[0028] S2. Preprocess the obtained original data set, including rotating and flipping the data set, generating multi-angle and multi-directional images, obtaining the weld image data set, and performing preliminary image enhancement using grayscale histogram and Gaussian filtering.

[0029] Furthermore, in S2, the obtained initial data set is preprocessed. First, each image is rotated at a random angle in the range of -170° to 170° to obtain 4 additional rotated images. Operation , unify the image size to Size, and then histogram equalization is performed to obtain a contrast-enhanced image , and perform Gaussian filtering to remove interference in some images to obtain the preprocessed image The preprocessed images constitute the data set used for training, and the data set contains 6000 images in total.

[0030] S3, build an attention module, including a Function calculation and a dot product operation.

[0031] Furthermore, in S3, for the attention module, its structure is as follows Figure 2 As shown, for the input image , where 3 is the number of channels, 224 and 224 are the width and height of the input image. First do it once Function calculation to obtain the channel weight feature vector , express The weight of each pixel on different channels. The higher the weight, the more important the corresponding pixel is for enhancement. and Perform pixel-by-pixel multiplication on each channel to obtain weighted features ,in, , , .

[0032] S4, build a global feature extraction module, including two Convolution, four Convolution operation, four Operation and four Activation function.

[0033] Furthermore, in S4, for the global feature extraction module, its structure is as follows Figure 3 As shown, for the features input into this module , , , , the eigenvector First, through a convolution, and The neural unit module of the activation function processes the initial extracted features to obtain the extracted feature vector , ,in , , , thereafter Enter the included convolution, and The convolutional neural unit of the activation function calculates the feature vector , , ,in , , , and then and Dot multiplication to get , ,in, , , , thereafter After one attention module processing, , ,in , , , thereby further extracting effective features, and then repeating The processing of the neural unit module is obtained , The size and Same, again and Dot multiplication to get , ,in , , , and then Perform the channel attention operation again to further enhance the weight of important channels and reduce the weight of non-critical channels, and obtain , After a After convolution adjusts the number of channels, the final feature extraction is performed through the neural unit module. , ,in , , which is the final global feature output.

[0034] S5, build detail enhancement module, including two Convolution operation, two activation function, two attention modules, two Operation, two operation and a concatenation operation.

[0035] Further, in S5, for the detail enhancement module, its structure is as follows Figure 4 As shown, for the input of this module First, add the elements at the same position in the channel dimension to obtain the feature map after fusion feature , , First go through Convolution operation and The activation function performs linear transformation to obtain features , whose channels have been expanded, , and then use the attention module to Process and extract effective information to obtain , , and then repeat the above operation to further extract features ,at the same time, After two consecutive maximum pooling and self-attention mechanism operations, we can get , , the last feature and After concatenation in the channel dimension, we get , .

[0036] S6: Construct a local feature fusion module, which includes two Convolution operation, a Convolution operation, two Operation, two Activation functions and an attention module.

[0037] Furthermore, in S6, for the local feature fusion module, its structure is as follows Figure 5 As shown, the input features of this module are the global features extracted by the global feature extraction module. And the detail enhancement features output by the detail enhancement module , and First, we concatenate in the channel dimension to get , , thereafter First, use a Convolution kernel, and The convolutional neural unit module of the activation function performs preliminary feature fusion to obtain the features after preliminary fusion of features , , and then the characteristics Use includes a Convolution kernel, and The convolutional neural unit module of the activation function is further integrated to obtain , , thereafter After a After convolution fusion features and an attention module operation, we get , , which is the enhanced image.

[0038] S7. Construct a weld image enhancement model of the motor stator core based on machine vision, including a global feature extraction module, a detail enhancement module, a local feature fusion module and a splicing operation.

[0039] Further, in S7, for the weld image enhancement model of the motor stator core based on machine vision, its structure is as follows: Figure 6 As shown, the weld image is first input as ,in , , , obtained through the global feature extraction module ,in Represented as the global feature extraction module constructed in S4, ,in , , , Obtained through the detail enhancement module ,in Represented as a detail enhancement module built in S5, , global features With detail enhancement features Splice to get ,in Represented as the local feature fusion module constructed in S6, , and Stitching is performed to obtain the final image enhancement optimization result ,in Represents a splicing operation, .

[0040] S8. Train a motor stator core weld image enhancement optimization model. Use a motor stator core weld image dataset to train a motor stator core weld image enhancement optimization model.

[0041] Furthermore, in S8, the operating system platform used for training the model is Centos system, the language is Python3.9.11, and the PyTorch deep learning framework is used for training. The hardware platform is NVIDIARTX 3090 with 24G video memory. The data set contains 6,000 crack images of three different materials: steel, aluminum alloy, and stainless steel under two backgrounds: factory site and laboratory environment. The training set and test set are divided into training set and test set in a ratio of 4:1. During the training process, SGD is used as the optimizer, with an initial learning rate of 0.001. After every 50 rounds of training, the learning rate decays to the original 0.1, and a total of 500 rounds of training are conducted.

[0042] Motor stator core weld image processing, to obtain the motor stator core weld image that needs to be optimized and enhanced, such as Figure 7 As shown, Figure 7 The low-quality motor stator core weld image is shown and input into the motor stator core weld image enhancement optimization model to obtain the optimized and enhanced motor stator core weld image, as shown in Figure 8 As shown, Figure 8 The image of the motor stator core weld is shown after optimization and enhancement by the motor stator core weld image enhancement optimization model.

[0043] The above are only preferred embodiments of the present invention. It should be pointed out that a person skilled in the art can make several modifications and improvements without departing from the inventive concept of the present invention, and these all fall within the protection scope of the present invention.

Claims

1. A method for enhancing the weld image of a motor stator core based on machine vision, characterized in that: The following steps are involved: S1. Preparation of weld image dataset: Select high-resolution industrial camera to collect images in an environment with uniform lighting and no strong light interference. Select factory site and laboratory environment to collect images of cracks in three different materials: steel, aluminum alloy and stainless steel, and obtain the weld image dataset of the motor stator core. S2. Preprocess the original data set, including rotating and flipping the original images in the data set to obtain the weld image data set. All images are of the same size. , preliminary image enhancement is performed using grayscale histogram and Gaussian filtering; S3, build attention module, including Function calculation and dot multiplication operation; S4, build a global feature extraction module, including convolution, Convolution operation, Operation and Activation function; S5, build detail enhancement module, including Convolution operation, Activation function, attention module, operate, manipulation and splicing operations; S6: Construct a local feature fusion module, including Convolution operation, Convolution operation, operate, Activation functions and attention modules; S7. Construct a weld image enhancement model of a motor stator core based on machine vision, including a global feature extraction module, a detail enhancement module, a local feature fusion module and a splicing operation; S8. Use the weld image dataset of the motor stator core to train a machine vision-based weld image enhancement model for the motor stator core.

2. The method for enhancing the weld seam image of a motor stator core based on machine vision according to claim 1, characterized in that: In S2, the collected weld image dataset is preprocessed. First, each image in the dataset is rotated and flipped to obtain images with different orientations, thereby expanding the size of the dataset. Second, the size of each image is processed into size, and finally use histogram equalization and Gaussian filtering to process the image to enhance the contrast of the image and remove interference in the image, thereby improving the quality of the data set.

3. The method for enhancing the weld seam image of a motor stator core based on machine vision according to claim 2, characterized in that: In S3, for the attention module, the input image is first recorded as ,in Indicates height, Indicates width, Indicates the number of channels, apply a Function on feature map All channels at each spatial position of the image are obtained to obtain a weight map with the same size as the original image but reflecting the relative importance of each channel , this weight map With the original feature map Multiply element by element to get the weighted feature map ,in Represents an element-wise multiplication operation.

4. The method for enhancing the weld seam image of a motor stator core based on machine vision according to claim 3, characterized in that: In S4, for the global feature extraction module, the weld image is first input as ,in represents the height of the feature map, represents the width of the feature map, The number of channels representing the feature map, the feature map Processed by convolutional neural unit module, the unit module includes a Convolution kernel, and Activation function, after passing through this module, the feature map is obtained ,in Indicates that the convolution kernel size is The convolution operation, express The convolution kernel, represents the bias term, express operate, express Activation function, next, feature It is further processed by a convolutional neural unit module, which consists of a Convolution kernel, and Activation function, after passing through the unit module, the feature map is obtained ,in Indicates that the convolution kernel size is The convolution operation, express The convolution kernel, represents the bias term, express operate, express Activation function, the feature With features Perform point multiplication to obtain features ,in Represents the dot product operation, feature After being processed by the attention module built by S3, the features are obtained ,in represents the attention module, Indicates that the convolution kernel size is The convolution operation, express The convolution kernel is After the convolutional neural unit module operation, we get ,in Indicates that the convolution kernel size is The convolution operation, express The convolution kernel, represents the bias term, express operate, express The activation function is obtained and Perform a dot multiplication operation to obtain the feature map ,in represents the dot product operation, The features are obtained through the attention module built by S3 ,in represents the attention module, Indicates that the convolution kernel size is The convolution operation, express The convolution kernel is finally The final global feature output is obtained through the convolutional neural unit module ,in Indicates that the convolution kernel size is The convolution operation, express The convolution kernel, represents the bias term, express operate, express Activation function.

5. The method for enhancing the weld seam image of a motor stator core based on machine vision according to claim 4, characterized in that: In S5, for the detail enhancement module, the input feature map ,in represents the height of the feature map, represents the width of the feature map, Indicates the number of channels of the feature map, All channel values ​​at each spatial position are added to obtain a new feature map ,in express No. channels, express No. The value of the channel, next, the feature go through Convolution operation and The activation function performs linear transformation to obtain features ,in Indicates that the convolution kernel size is The convolution operation, Expressed as The convolution kernel, represents the bias term, express Activation function, After being processed by the attention module built by S3, the features are obtained ,in Represents the attention module, repeat the above steps again to get the feature , while the characteristics After two consecutive maximum pooling operations and self-attention mechanism operations, we get ,in Indicates the pooling window size The maximum pooling operation, Represents the self-attention function. Finally, the feature and Stitching to get the final detail enhancement features ,in Represents a concatenation operation.

6. The method for enhancing the weld seam image of a motor stator core based on machine vision according to claim 5, characterized in that: In S6, for the local feature fusion module, the global feature With detail enhancement features Splice to get ,in Represents the concatenation operation, and then the features It is further processed by a convolutional neural unit module, which consists of a Convolution kernel, and Activation function, after passing through the unit module, the feature map is obtained ,in Indicates that the convolution kernel size is The convolution operation, express The convolution kernel, represents the bias term, express operate, express activation function, and then processed by a convolutional neural unit module, which consists of a Convolution kernel, and Activation function, after passing through this module, the feature map is obtained ,in Indicates that the convolution kernel size is The convolution operation, express The convolution kernel, represents the bias term, express operate, express Activation function, final feature The features are processed by the attention module built by S3 ,in represents the attention module, express The convolution kernel.

7. The method for enhancing the weld seam image of a motor stator core based on machine vision according to claim 6, characterized in that: In S7, for the weld image enhancement model of the motor stator core based on machine vision, the weld image is first input as , obtained through the global feature extraction module ,in Represented as the global feature extraction module constructed in S4, Obtained through the detail enhancement module ,in Represented as the detail enhancement module constructed in S5, global features With detail enhancement features Splice to get ,in Represented as the local feature fusion module constructed in S6, and Stitching to get the final image enhancement optimization result ,in Represents a concatenation operation.

8. The method for enhancing the weld seam image of a motor stator core based on machine vision according to claim 7, characterized in that: In S1, a ring light source is used to reduce shadow and spot interference, while adapting to the reflective characteristics of welds of different materials. Crack images of three different materials, steel, aluminum alloy and stainless steel, are collected. The background includes the factory site and laboratory environment, and the weld image dataset of the motor stator core is obtained.

Images