CN116797533B - Appearance defect detection method and system for power adapter - Google Patents
Appearance defect detection method and system for power adapter Download PDFInfo
- Publication number
- CN116797533B CN116797533B CN202310298123.7A CN202310298123A CN116797533B CN 116797533 B CN116797533 B CN 116797533B CN 202310298123 A CN202310298123 A CN 202310298123A CN 116797533 B CN116797533 B CN 116797533B
- Authority
- CN
- China
- Prior art keywords
- classification
- image
- feature
- matrix
- vectors
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
- 238000001514 detection method Methods 0.000 title claims abstract description 129
- 230000007547 defect Effects 0.000 title claims abstract description 92
- 238000012545 processing Methods 0.000 claims abstract description 41
- 239000013598 vector Substances 0.000 claims description 143
- 239000011159 matrix material Substances 0.000 claims description 119
- 238000010586 diagram Methods 0.000 claims description 48
- 238000000034 method Methods 0.000 claims description 33
- 238000013527 convolutional neural network Methods 0.000 claims description 31
- 230000004044 response Effects 0.000 claims description 27
- 230000000903 blocking effect Effects 0.000 claims description 18
- 238000000605 extraction Methods 0.000 claims description 13
- 230000006870 function Effects 0.000 claims description 11
- 230000004913 activation Effects 0.000 claims description 7
- 230000004931 aggregating effect Effects 0.000 claims description 5
- 230000002950 deficient Effects 0.000 claims description 4
- 238000011176 pooling Methods 0.000 claims description 3
- 238000012546 transfer Methods 0.000 claims description 3
- 238000005516 engineering process Methods 0.000 abstract description 10
- 238000004519 manufacturing process Methods 0.000 abstract description 10
- 238000009826 distribution Methods 0.000 abstract description 7
- 238000013473 artificial intelligence Methods 0.000 abstract description 5
- 238000013135 deep learning Methods 0.000 abstract description 5
- 238000003860 storage Methods 0.000 description 11
- 238000004590 computer program Methods 0.000 description 9
- 230000008901 benefit Effects 0.000 description 7
- 230000008569 process Effects 0.000 description 6
- 238000013136 deep learning model Methods 0.000 description 4
- 230000000694 effects Effects 0.000 description 4
- 238000004458 analytical method Methods 0.000 description 2
- 230000004927 fusion Effects 0.000 description 2
- 238000007689 inspection Methods 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 238000003062 neural network model Methods 0.000 description 2
- 230000003287 optical effect Effects 0.000 description 2
- 238000007792 addition Methods 0.000 description 1
- 230000004075 alteration Effects 0.000 description 1
- 238000004891 communication Methods 0.000 description 1
- 238000000354 decomposition reaction Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000002452 interceptive effect Effects 0.000 description 1
- 230000007246 mechanism Effects 0.000 description 1
- 239000013307 optical fiber Substances 0.000 description 1
- 238000005215 recombination Methods 0.000 description 1
- 230000006798 recombination Effects 0.000 description 1
- 230000003252 repetitive effect Effects 0.000 description 1
- 239000004065 semiconductor Substances 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/0002—Inspection of images, e.g. flaw detection
- G06T7/0004—Industrial image inspection
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/40—Extraction of image or video features
- G06V10/46—Descriptors for shape, contour or point-related descriptors, e.g. scale invariant feature transform [SIFT] or bags of words [BoW]; Salient regional features
- G06V10/467—Encoded features or binary features, e.g. local binary patterns [LBP]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/40—Extraction of image or video features
- G06V10/50—Extraction of image or video features by performing operations within image blocks; by using histograms, e.g. histogram of oriented gradients [HoG]; by summing image-intensity values; Projection analysis
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/764—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using classification, e.g. of video objects
- G06V10/765—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using classification, e.g. of video objects using rules for classification or partitioning the feature space
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/82—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using neural networks
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Multimedia (AREA)
- Evolutionary Computation (AREA)
- Artificial Intelligence (AREA)
- Health & Medical Sciences (AREA)
- Computing Systems (AREA)
- Databases & Information Systems (AREA)
- General Health & Medical Sciences (AREA)
- Medical Informatics (AREA)
- Software Systems (AREA)
- Quality & Reliability (AREA)
- Image Analysis (AREA)
Abstract
The application relates to the field of appearance defect detection, and particularly discloses an appearance defect detection method and an appearance defect detection system for a power adapter, which are based on an artificial intelligence technology and an image processing technology of deep learning so as to extract and capture advanced implicit characteristic distribution information about the appearance defects from an appearance detection image of a first side surface of the power adapter to be detected, and realize the purpose of rapidly and accurately detecting the appearance defects of the power adapter through classification processing, thereby improving the production efficiency and the product quality.
Description
Technical Field
The present disclosure relates to the field of appearance defect detection, and more particularly, to an appearance defect detection method of a power adapter and a system thereof.
Background
Along with the rapid updating iteration and development of the intelligent terminal equipment, the production scale of the power adapter is gradually enlarged. However, in the industrial process, defects in the appearance of the power adapter due to improper operation occur, for example, a scratch of dirt on the surface of the power adapter. The appearance defects often have the characteristics of irregularity, tiny targets and the like, and in the existing solution, the detection is usually carried out manually, so that the method is time-consuming and labor-consuming, and has great influence on the production efficiency.
Therefore, an optimized appearance defect detection scheme for a power adapter is desired.
Disclosure of Invention
The present application has been made in order to solve the above technical problems. The embodiment of the application provides an appearance defect detection method and system of a power adapter, which are based on an artificial intelligence technology and an image processing technology of deep learning, so as to extract and capture advanced implicit characteristic distribution information about appearance defects from an appearance detection image of a first side surface of the power adapter to be detected, and realize the purpose of rapidly and accurately detecting the appearance defects of the power adapter through classification processing, thereby improving production efficiency and product quality.
According to an aspect of the present application, there is provided an appearance defect detection method of a power adapter, including: obtaining an appearance detection image of a first side surface of a power adapter to be detected; extracting a local binary pattern diagram and a direction gradient histogram of the appearance detection image of the first side, and aggregating the local binary pattern diagram, the direction gradient histogram and the appearance detection image of the first side along a channel dimension to obtain a multi-channel detection image; performing image blocking processing on the multi-channel detection image, and then obtaining a plurality of image block context semantic feature vectors through a ViT model; two-dimensionally arranging the context semantic feature vectors of the image blocks into a two-dimensional feature matrix, and then obtaining a classification feature map through a convolutional neural network model serving as a feature extractor; performing convolution dictionary contrast response learning on each classification feature matrix of the classification feature map by using the two-dimensional feature matrix to obtain an optimized classification feature map; and passing the optimized classification characteristic diagram through a classifier to obtain a classification result, wherein the classification result is used for indicating whether the appearance of the first side surface of the power adapter to be detected has defects.
In the above method for detecting an appearance defect of a power adapter, performing image blocking processing on the multi-channel detection image, and then obtaining a plurality of context semantic feature vectors of the image block by using a ViT model, including: performing image blocking processing on the multi-channel detection image to obtain a sequence of image blocks; using the embedding layer of the ViT model to respectively carry out embedded coding on each image block in the sequence of the image blocks so as to obtain a sequence of image block embedded vectors; and inputting the sequence of image block embedding vectors into a converter module of the ViT model to obtain the plurality of image block context semantic feature vectors.
In the above method for detecting an appearance defect of a power adapter, the embedding layer of the ViT model is used to respectively perform embedded coding on each image block in the sequence of image blocks to obtain a sequence of image block embedded vectors, including: respectively expanding each image block in the sequence of image blocks into one-dimensional pixel input vectors to obtain a plurality of one-dimensional pixel input vectors; and performing full-connection encoding on each one-dimensional pixel input vector in the plurality of one-dimensional pixel input vectors by using an embedding layer of the ViT model to obtain a sequence of the image block embedding vectors.
In the above method for detecting an appearance defect of a power adapter, inputting the sequence of image block embedding vectors into the converter module of the ViT model to obtain the plurality of image block context semantic feature vectors includes: arranging the sequence of the image block embedded vectors into an input vector; respectively converting the input vector into a query vector and a key vector through a learning embedding matrix; calculating the product between the query vector and the transpose vector of the key vector to obtain a self-attention correlation matrix; carrying out standardization processing on the self-attention association matrix to obtain a standardized self-attention association matrix; inputting the standardized self-attention association matrix into a Softmax activation function to activate so as to obtain a self-attention feature matrix; and multiplying the self-attention feature matrix with each image block embedded vector in the sequence of image block embedded vectors as a value vector to obtain the plurality of image block context semantic feature vectors.
In the above method for detecting an appearance defect of a power adapter, the two-dimensionally arranging the context semantic feature vectors of the plurality of image blocks into a two-dimensional feature matrix, and obtaining a classification feature map by using a convolutional neural network model as a feature extractor, includes: each layer using the convolutional neural network model is performed in forward transfer of the layer: carrying out convolution processing on input data to obtain a convolution characteristic diagram; carrying out 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 is the classification characteristic diagram, and the input of the first layer of the convolutional neural network model is the two-dimensional characteristic matrix.
In the above method for detecting an appearance defect of a power adapter, performing convolutional dictionary contrast response learning on each classification feature matrix of the classification feature map by using the two-dimensional feature matrix to obtain an optimized classification feature map, including: performing convolution dictionary contrast response learning on each classification feature matrix of the classification feature map based on the two-dimensional feature matrix to obtain a plurality of optimized classification feature matrices; and cascading the plurality of optimized classification feature matrixes to obtain the optimized classification feature map.
In the above method for detecting an appearance defect of a power adapter, performing convolutional dictionary contrast response learning on each classification feature matrix of the classification feature map based on the two-dimensional feature matrix to obtain a plurality of optimized classification feature matrices, including: performing convolution dictionary contrast response learning on each classification feature matrix of the classification feature map according to the following learning formula based on the two-dimensional feature matrix to obtain a plurality of optimized classification feature matrices; wherein, the learning formula is:wherein (1)>Representing the two-dimensional feature matrix->Representing the +.sup.th in the classification characteristic diagram>A classification feature matrix- >Frobenius norms of the matrix are represented, < >>Representing subtracting by position ++>Representing matrix multiplication +.>Representing the optimized classification feature matrix.
In the above method for detecting appearance defects of a power adapter, the step of passing the optimized classification feature map through a classifier to obtain a classification result, where the classification result is used to indicate whether the appearance of the first side of the power adapter to be detected has defects, includes: performing feature map expansion on the optimized classification feature map to obtain classification feature vectors; inputting the classification feature vector into a Softmax classification function of the classifier to obtain a probability value of the classification feature vector belonging to each classification label; and determining the classification label corresponding to the maximum probability value as the classification result.
According to another aspect of the present application, there is provided an appearance defect detection system of a power adapter, including: the appearance detection image acquisition module is used for acquiring an appearance detection image of the first side surface of the power adapter to be detected; the image information widening module is used for extracting a local binary pattern diagram and a direction gradient histogram of the appearance detection image of the first side surface and converging the local binary pattern diagram, the direction gradient histogram and the appearance detection image of the first side surface along the channel dimension to obtain a multi-channel detection image; the image block context extraction module is used for obtaining a plurality of image block context semantic feature vectors through a ViT model after carrying out image block processing on the multi-channel detection image; the global advanced semantic extraction module is used for performing two-dimensional arrangement on the context semantic feature vectors of the plurality of image blocks into a two-dimensional feature matrix and then obtaining a classification feature map through a convolutional neural network model serving as a feature extractor; the convolution dictionary contrast response learning module is used for carrying out convolution dictionary contrast response learning on each classification characteristic matrix of the classification characteristic map by using the two-dimensional characteristic matrix so as to obtain an optimized classification characteristic map; and the detection result generation module is used for enabling the optimized classification characteristic diagram to pass through a classifier to obtain a classification result, wherein the classification result is used for indicating whether the appearance of the first side face of the power adapter to be detected has defects or not.
According to still another aspect of the present application, there is provided an electronic apparatus including: a processor; and a memory in which computer program instructions are stored which, when executed by the processor, cause the processor to perform the appearance defect detection method of a power adapter as described above.
According to yet another aspect of the present application, there is provided a computer readable medium having stored thereon computer program instructions which, when executed by a processor, cause the processor to perform the method of appearance defect detection of a power adapter as described above.
Compared with the prior art, the appearance defect detection method and the appearance defect detection system for the power adapter are based on the artificial intelligence technology and the image processing technology of deep learning, so that the advanced implicit characteristic distribution information about the appearance defects is extracted and captured from the appearance detection image of the first side face of the power adapter to be detected, and the purpose of rapidly and accurately detecting the appearance defects of the power adapter is achieved through classification processing, so that the production efficiency and the product quality are improved.
Drawings
The foregoing and other objects, features and advantages of the present application will become more apparent from the following more particular description of embodiments of the present application, as illustrated in the accompanying drawings. The accompanying drawings are included to provide a further understanding of embodiments of the application and are incorporated in and constitute a part of this specification, illustrate the application and not constitute a limitation to the application. In the drawings, like reference numerals generally refer to like parts or steps.
Fig. 1 is an application scenario diagram of an appearance defect detection method of a power adapter according to an embodiment of the present application.
Fig. 2 is a flowchart of an appearance defect detection method of a power adapter according to an embodiment of the present application.
Fig. 3 is a schematic diagram of an appearance defect detection method of a power adapter according to an embodiment of the present application.
Fig. 4 is a flowchart of obtaining context semantic feature vectors of a plurality of image blocks by using a ViT model after performing image blocking processing on the multi-channel detection image in the appearance defect detection method of the power adapter according to the embodiment of the application.
Fig. 5 is a block diagram of an appearance defect detection system of a power adapter according to an embodiment of the present application.
Fig. 6 is a block diagram of an electronic device according to an embodiment of the present application.
Detailed Description
Hereinafter, example embodiments according to the present application will be described in detail with reference to the accompanying drawings. It should be apparent that the described embodiments are only some of the embodiments of the present application and not all of the embodiments of the present application, and it should be understood that the present application is not limited by the example embodiments described herein.
Summary of the application: for the technical problems, the technical conception of the application is as follows: the artificial intelligence technology and the image processing technology based on deep learning are used for extracting and capturing advanced implicit characteristic distribution information about the appearance defects from the appearance detection image of the first side face of the power adapter to be detected, and the aim of rapidly and accurately detecting the appearance defects of the power adapter is fulfilled through classification processing, so that the production efficiency and the product quality are improved.
Specifically, in the technical scheme of the application, first, an appearance detection image of a first side surface of a power adapter to be detected is obtained. In a specific implementation, acquiring the appearance inspection image of the first side of the power adapter to be inspected may be accomplished by using a camera or other image acquisition device. When the image is collected, the acquired image needs to be ensured to be clear and have sufficient light, and the problems of background interference, image distortion and the like are avoided as much as possible.
And then, extracting a local binary pattern diagram and a directional gradient histogram of the appearance detection image of the first side, and aggregating the local binary pattern diagram, the directional gradient histogram and the appearance detection image of the first side along a channel dimension to obtain a multi-channel detection image. Here, a Local Binary Pattern (LBP) is an image feature for local texture analysis, which can capture important information of textures; a direction gradient Histogram (HOG) is a feature describing image contour and edge information, and can model the shape of an object to be detected, thereby effectively distinguishing different objects. In the appearance defect detection of the power adapter, the combination of the two characteristics can effectively describe the texture and appearance shape information of the power adapter to be detected, the detection precision and robustness can be improved, and the possibility of misjudgment is reduced. Therefore, in the technical scheme of the application, the local binary pattern diagram, the direction gradient histogram and the appearance detection image of the first side face are aggregated along the channel dimension so as to better describe different characteristics and information of the power adapter to be detected, and therefore a multi-channel detection image is obtained. The fusion mode can improve the expression capability and the distinguishing degree of the features, so that the deep learning model can learn the effective features more easily, and the detection accuracy and the detection robustness are improved.
And then, performing image blocking processing on the multi-channel detection image, and then obtaining a plurality of image block context semantic feature vectors through a ViT model. Here, after the image blocking processing is performed on the multi-channel detection image, the original large-size image can be divided into a plurality of small-size image blocks through the ViT model, so that local features and detailed information of different areas of the object to be detected are better captured, and interference and noise of global features are reduced. The ViT model is a transform-based deep learning model, and can extract context semantic feature vectors from an input image sequence to represent the relation and dependence among different image blocks. The feature vectors can effectively express the appearance form and structure information of the power adapter to be detected, and meanwhile, the feature vectors have strong interpretability and robustness. In the detection of the appearance defects of the power adapter, the ViT model can improve the accuracy and efficiency of feature extraction, and further enhance the performance and stability of a detection algorithm.
And then, the context semantic feature vectors of the image blocks are two-dimensionally arranged into a two-dimensional feature matrix, and then the two-dimensional feature matrix is used as a convolutional neural network model of a feature extractor to obtain a classification feature map. Here, the context semantic feature vectors of the image blocks are two-dimensionally arranged into a two-dimensional feature matrix, and then the high-level semantic features of the image can be further extracted through a convolutional neural network model. It should be appreciated that since the ViT model results in a serialized feature vector, it is necessary to transform them into a two-dimensional feature matrix that can be input to the convolutional neural network model.
The convolution neural network model can fully utilize local information and spatial structures in the image to abstract and integrate the characteristics in multiple layers and multiple scales, so that more effective characteristic extraction and classification are realized. The method combining ViT and convolutional neural network can exert their advantages to the greatest extent, so that the performance and the accuracy of detecting the appearance defects of the power adapter are improved.
After the classification characteristic diagram is obtained, the classification characteristic diagram is passed through a classifier to obtain a classification result, wherein the classification result is used for indicating whether the appearance of the first side surface of the power adapter to be detected is defective or not. In the technical scheme of the application, the classification result obtained by passing the classification feature map through the classifier is to perform two classifications on whether defects exist in the appearance of the first side surface of the power adapter to be detected, namely, defects or non-defects. Here, the classifier may further process and analyze the high-level semantic features contained in the classification feature map, and output a corresponding classification result. In this way, the appearance defect of the power adapter is automatically, accurately and efficiently detected, so that the production quality is ensured. Meanwhile, in practical application, classification results can be further refined and classified, for example, specific defect types, defect positions and the like can be detected.
In the technical scheme of the application, when the classification feature map is obtained through a convolutional neural network model serving as a feature extractor after the context semantic feature vectors of the plurality of image blocks are arranged in two dimensions into a two-dimensional feature matrix, because each of the context semantic feature vectors of the plurality of image blocks expresses context image semantic information of a single image block, the context image semantic information is expected to be still utilized to the greatest extent while feature extraction is carried out.
Based on this, the applicant of the present application uses the image block context semantic feature matrix in which the plurality of image block context semantic feature vectors are two-dimensionally arranged, for example, denoted asFor each classification feature matrix of the classification feature map (e.g., # along the channel dimension>The individual classification feature matrix is marked->) Performing convolutional dictionary contrast response learning to optimize the classification feature matrix, e.g., optimized +.>The individual classification feature matrix is marked->，/>The concrete steps are as follows:wherein->Representing the Frobenius norm of the matrix.
That is, based on the image block context semantic feature matrixNeighborhood operator attribute characterized by convolution kernel of a convolutional neural network, for each classification feature matrix +_ of the classification feature map by convolutional dictionary contrast learning based on differential feature flow between corresponding features >The n-level (n-hop) neighbors of the eigenvalue of the feature level (n-hop) are subjected to eigenvalue expression of an eigenvalue prior structure, and prior knowledge under low-rank expression is used as a characteristic response reference of high-dimensional characteristic distribution, so that an interpretable response to the characteristic level information expression is learned, and the classification characteristic matrix of the optimized classification characteristic diagram can be improved>And expressing the expression effect of the context image semantic information contained in the context semantic feature matrix of the image block, thereby improving the accuracy of the classification result obtained by the classifier of the classification feature map.
Fig. 1 is an application scenario diagram of an appearance defect detection method of a power adapter according to an embodiment of the present application. As shown in fig. 1, in this application scenario, first, an appearance detection image (e.g., I as illustrated in fig. 1) of a first side (e.g., P as illustrated in fig. 1) of a power adapter to be detected (e.g., a as illustrated in fig. 1) acquired by a camera (e.g., C as illustrated in fig. 1) is acquired. Further, the appearance detection image is input to a server (e.g., S as illustrated in fig. 1) in which an appearance defect detection algorithm of a power adapter is disposed, wherein the server is capable of processing the appearance detection image based on the appearance defect detection algorithm of the power adapter to obtain a classification result indicating whether or not an appearance of a first side of the power adapter to be detected is defective.
Having described the basic principles of the present application, various non-limiting embodiments of the present application will now be described in detail with reference to the accompanying drawings.
An exemplary method is: fig. 2 is a flowchart of an appearance defect detection method of a power adapter according to an embodiment of the present application. As shown in fig. 2, the method for detecting the appearance defect of the power adapter according to the embodiment of the application includes: s110, obtaining an appearance detection image of a first side surface of a power adapter to be detected; s120, extracting a local binary pattern diagram and a direction gradient histogram of the appearance detection image of the first side, and aggregating the local binary pattern diagram, the direction gradient histogram and the appearance detection image of the first side along a channel dimension to obtain a multi-channel detection image; s130, performing image blocking processing on the multi-channel detection image, and then obtaining a plurality of image block context semantic feature vectors through a ViT model; s140, performing two-dimensional arrangement on the context semantic feature vectors of the image blocks to form a two-dimensional feature matrix, and then obtaining a classification feature map through a convolutional neural network model serving as a feature extractor; s150, performing convolution dictionary contrast response learning on each classification feature matrix of the classification feature map by using the two-dimensional feature matrix to obtain an optimized classification feature map; and S160, enabling the optimized classification characteristic diagram to pass through a classifier to obtain a classification result, wherein the classification result is used for indicating whether the appearance of the first side face of the power adapter to be detected is defective or not.
Fig. 3 is a schematic diagram of an appearance defect detection method of a power adapter according to an embodiment of the present application. As shown in fig. 3, in this architecture, first, an appearance detection image of a first side of a power adapter to be detected is acquired; then, extracting a local binary pattern diagram and a direction gradient histogram of the appearance detection image of the first side, and aggregating the local binary pattern diagram, the direction gradient histogram and the appearance detection image of the first side along a channel dimension to obtain a multi-channel detection image; then, performing image blocking processing on the multi-channel detection image, and obtaining a plurality of image block context semantic feature vectors through a ViT model; then, the context semantic feature vectors of the image blocks are two-dimensionally arranged into a two-dimensional feature matrix, and then a classified feature map is obtained through a convolutional neural network model serving as a feature extractor; performing convolution dictionary contrast response learning on each classification feature matrix of the classification feature map by using the two-dimensional feature matrix to obtain an optimized classification feature map; and finally, the optimized classification characteristic diagram is passed through a classifier to obtain a classification result, wherein the classification result is used for indicating whether the appearance of the first side surface of the power adapter to be detected has defects.
In step S110, an appearance detection image of a first side of the power adapter to be detected is acquired. In a specific implementation, acquiring the appearance inspection image of the first side of the power adapter to be inspected may be accomplished by using a camera or other image acquisition device. When the image is collected, the acquired image needs to be ensured to be clear and have sufficient light, and the problems of background interference, image distortion and the like are avoided as much as possible.
In step S120, a local binary pattern diagram and a directional gradient histogram of the appearance detection image of the first side are extracted, and the local binary pattern diagram, the directional gradient histogram, and the appearance detection image of the first side are aggregated along a channel dimension to obtain a multi-channel detection image. Here, a Local Binary Pattern (LBP) is an image feature for local texture analysis, which can capture important information of textures; a direction gradient Histogram (HOG) is a feature describing image contour and edge information, and can model the shape of an object to be detected, thereby effectively distinguishing different objects. In the appearance defect detection of the power adapter, the combination of the two characteristics can effectively describe the texture and appearance shape information of the power adapter to be detected, the detection precision and robustness can be improved, and the possibility of misjudgment is reduced. Therefore, in the technical scheme of the application, the local binary pattern diagram, the direction gradient histogram and the appearance detection image of the first side face are aggregated along the channel dimension so as to better describe different characteristics and information of the power adapter to be detected, and therefore a multi-channel detection image is obtained. The fusion mode can improve the expression capability and the distinguishing degree of the features, so that the deep learning model can learn the effective features more easily, and the detection accuracy and the detection robustness are improved.
In step S130, the multi-channel detection image is subjected to image blocking processing, and then a ViT model is used to obtain a plurality of image block context semantic feature vectors. Here, after the image blocking processing is performed on the multi-channel detection image, the original large-size image can be divided into a plurality of small-size image blocks through the ViT model, so that local features and detailed information of different areas of the object to be detected are better captured, and interference and noise of global features are reduced. The ViT model is a transform-based deep learning model, and can extract context semantic feature vectors from an input image sequence to represent the relation and dependence among different image blocks. The feature vectors can effectively express the appearance form and structure information of the power adapter to be detected, and meanwhile, the feature vectors have strong interpretability and robustness. In the detection of the appearance defects of the power adapter, the ViT model can improve the accuracy and efficiency of feature extraction, and further enhance the performance and stability of a detection algorithm.
Fig. 4 is a flowchart of obtaining context semantic feature vectors of a plurality of image blocks by using a ViT model after performing image blocking processing on the multi-channel detection image in the appearance defect detection method of the power adapter according to the embodiment of the application. As shown in fig. 4, after performing image blocking processing on the multi-channel detection image, a ViT model is used to obtain a plurality of image block context semantic feature vectors, which includes the steps of: s210, performing image blocking processing on the multi-channel detection image to obtain a sequence of image blocks; s220, respectively performing embedded coding on each image block in the sequence of image blocks by using an embedded layer of the ViT model to obtain a sequence of image block embedded vectors; and S230, inputting the sequence of image block embedding vectors into a converter module of the ViT model to obtain the plurality of image block context semantic feature vectors.
Specifically, in the embodiment of the present application, the encoding process for performing embedded encoding on each image block in the sequence of image blocks by using the embedded layer of the ViT model to obtain the sequence of image block embedded vectors includes: firstly, respectively expanding each image block in the sequence of the image blocks into one-dimensional pixel input vectors to obtain a plurality of one-dimensional pixel input vectors; then, the embedding layer of the ViT model is used for carrying out full-connection coding on each one-dimensional pixel input vector in the plurality of one-dimensional pixel input vectors so as to obtain a sequence of the image block embedding vectors.
Specifically, in the embodiment of the present application, the encoding process of inputting the sequence of the image block embedded vectors into the converter module of the ViT model to obtain the plurality of image block context semantic feature vectors includes: firstly, arranging a sequence of the image block embedded vectors into input vectors; then, the input vector is respectively converted into a query vector and a key vector through a learning embedding matrix; then, calculating the product between the query vector and the transpose vector of the key vector to obtain a self-attention correlation matrix; then, carrying out standardization processing on the self-attention association matrix to obtain a standardized self-attention association matrix; subsequently, the standardized self-attention association matrix is input into a Softmax activation function to be activated so as to obtain a self-attention feature matrix; finally, the self-attention feature matrix is multiplied by each image block embedding vector in the sequence of image block embedding vectors as a value vector to obtain the plurality of image block context semantic feature vectors.
In step S140, the plurality of image block context semantic feature vectors are two-dimensionally arranged into a two-dimensional feature matrix, and then a classification feature map is obtained by using a convolutional neural network model as a feature extractor. Here, the context semantic feature vectors of the image blocks are two-dimensionally arranged into a two-dimensional feature matrix, and then the high-level semantic features of the image can be further extracted through a convolutional neural network model. It should be appreciated that since the ViT model results in a serialized feature vector, it is necessary to transform them into a two-dimensional feature matrix that can be input to the convolutional neural network model.
The convolution neural network model can fully utilize local information and spatial structures in the image to abstract and integrate the characteristics in multiple layers and multiple scales, so that more effective characteristic extraction and classification are realized. The method combining ViT and convolutional neural network can exert their advantages to the greatest extent, so that the performance and the accuracy of detecting the appearance defects of the power adapter are improved.
Specifically, in the embodiment of the present application, after two-dimensionally arranging the context semantic feature vectors of the plurality of image blocks into a two-dimensional feature matrix, obtaining a classification feature map through a convolutional neural network model serving as a feature extractor, including: each layer using the convolutional neural network model is performed in forward transfer of the layer: carrying out convolution processing on input data to obtain a convolution characteristic diagram; carrying out 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 is the classification characteristic diagram, and the input of the first layer of the convolutional neural network model is the two-dimensional characteristic matrix.
In step S150, a convolutional dictionary contrast response learning is performed on each classification feature matrix of the classification feature map using the two-dimensional feature matrix to obtain an optimized classification feature map. In the technical scheme of the application, when the classification feature map is obtained through a convolutional neural network model serving as a feature extractor after the context semantic feature vectors of the plurality of image blocks are arranged in two dimensions into a two-dimensional feature matrix, because each of the context semantic feature vectors of the plurality of image blocks expresses context image semantic information of a single image block, the context image semantic information is expected to be still utilized to the greatest extent while feature extraction is carried out.
Based on this, the applicant of the present application uses the image block context semantic feature matrix in which the plurality of image block context semantic feature vectors are two-dimensionally arranged, for example, denoted asFor each classification feature matrix of the classification feature map (e.g., # along the channel dimension>The individual classification feature matrix is marked->) Performing convolutional dictionary contrast response learning to optimize the classification feature matrix, e.g., optimized +.>The individual classification feature matrix is marked- >，/>The concrete steps are as follows:wherein (1)>Representing the two-dimensional feature matrix->Representing the +.sup.th in the classification characteristic diagram>A classification feature matrix->Frobenius norms of the matrix are represented, < >>Representing subtracting by position ++>Representing matrix multiplication +.>Representing the optimized classification feature matrix.
That is, based on the image block context semantic feature matrixNeighborhood operator attribute characterized by a convolution kernel based on a convolutional neural network of (2)Convolved dictionary contrast learning of differential feature flows between corresponding features for each classification feature matrix of the classification feature map>The n-level (n-hop) neighbors of the eigenvalue of the feature level (n-hop) are subjected to eigenvalue expression of an eigenvalue prior structure, and prior knowledge under low-rank expression is used as a characteristic response reference of high-dimensional characteristic distribution, so that an interpretable response to the characteristic level information expression is learned, and the classification characteristic matrix of the optimized classification characteristic diagram can be improved>And expressing the expression effect of the context image semantic information contained in the context semantic feature matrix of the image block, thereby improving the accuracy of the classification result obtained by the classifier of the classification feature map.
Specifically, in the embodiment of the present application, the coding process for performing convolutional dictionary contrast response learning on each classification feature matrix of the classification feature map by using the two-dimensional feature matrix to obtain an optimized classification feature map includes: firstly, performing convolution dictionary contrast response learning on each classification feature matrix of the classification feature map based on the two-dimensional feature matrix to obtain a plurality of optimized classification feature matrices; and then, cascading the plurality of optimized classification feature matrixes to obtain the optimized classification feature map.
In step S160, the optimized classification feature map is passed through a classifier to obtain a classification result, where the classification result is used to indicate whether the appearance of the first side of the power adapter to be detected has a defect. In the technical scheme of the application, the classification result obtained by passing the classification feature map through the classifier is to perform two classifications on whether defects exist in the appearance of the first side surface of the power adapter to be detected, namely, defects or non-defects. Here, the classifier may further process and analyze the high-level semantic features contained in the classification feature map, and output a corresponding classification result. In this way, the appearance defect of the power adapter is automatically, accurately and efficiently detected, so that the production quality is ensured. Meanwhile, in practical application, classification results can be further refined and classified, for example, specific defect types, defect positions and the like can be detected.
Specifically, in the embodiment of the present application, the optimizing classification feature map is passed through a classifier to obtain a classification result, where the classification result is used to indicate whether the appearance of the first side of the power adapter to be detected has a defect, and the encoding process includes: firstly, carrying out feature map expansion on the optimized classification feature map to obtain classification feature vectors; then, inputting the classification feature vector into a Softmax classification function of the classifier to obtain a probability value of the classification feature vector belonging to each classification label; and then, determining the classification label corresponding to the maximum probability value as the classification result.
In summary, the method for detecting the appearance defects of the power adapter according to the embodiment of the application is explained, which is based on the artificial intelligence technology and the image processing technology of deep learning, so as to extract and capture the advanced implicit characteristic distribution information about the appearance defects from the appearance detection image of the first side of the power adapter to be detected, and realize the purpose of quickly and accurately detecting the appearance defects of the power adapter through classification processing, thereby improving the production efficiency and the product quality.
Exemplary System: fig. 5 is a block diagram of an appearance defect detection system of a power adapter according to an embodiment of the present application. As shown in fig. 5, an appearance defect detection system 100 of a power adapter according to an embodiment of the present application includes: an appearance detection image obtaining module 110, configured to obtain an appearance detection image of a first side of the power adapter to be detected; the image information widening module 120 is configured to extract a local binary pattern diagram and a directional gradient histogram of the appearance detection image of the first side, and aggregate the local binary pattern diagram, the directional gradient histogram, and the appearance detection image of the first side along a channel dimension to obtain a multi-channel detection image; the image block context extraction module 130 is configured to perform image blocking processing on the multi-channel detection image, and then obtain a plurality of image block context semantic feature vectors through a ViT model; the global advanced semantic extraction module 140 is configured to two-dimensionally arrange the context semantic feature vectors of the plurality of image blocks into a two-dimensional feature matrix, and then obtain a classification feature map through a convolutional neural network model serving as a feature extractor; a convolution dictionary contrast response learning module 150, configured to perform convolution dictionary contrast response learning on each classification feature matrix of the classification feature map by using the two-dimensional feature matrix to obtain an optimized classification feature map; and a detection result generating module 160, configured to pass the optimized classification feature map through a classifier to obtain a classification result, where the classification result is used to indicate whether the appearance of the first side of the power adapter to be detected has a defect.
Here, it will be understood by those skilled in the art that the specific functions and operations of the respective units and modules in the above-described external defect detection system 100 for a power adapter have been described in detail in the above description of the external defect detection method for a power adapter with reference to fig. 1 to 4, and thus, repetitive descriptions thereof will be omitted.
As described above, the appearance defect detection system 100 of the power adapter according to the embodiment of the present application can be implemented in various terminal devices, such as a server or the like for appearance defect detection of the power adapter. In one example, the appearance defect detection system 100 of the power adapter according to the embodiments of the present application may be integrated into the terminal device as one software module and/or hardware module. For example, the appearance defect detection system 100 of the power adapter may be a software module in the operating system of the terminal device, or may be an application developed for the terminal device; of course, the appearance defect detection system 100 of the power adapter can also be one of a plurality of hardware modules of the terminal device.
Alternatively, in another example, the appearance defect detection system 100 of the power adapter and the terminal device may be separate devices, and the appearance defect detection system 100 of the power adapter may be connected to the terminal device through a wired and/or wireless network and transmit the interactive information in a agreed data format.
Exemplary electronic device: next, an electronic device according to an embodiment of the present application is described with reference to fig. 6. Fig. 6 is a block diagram of an electronic device according to an embodiment of the present application. As shown in fig. 6, the electronic device 10 includes one or more processors 11 and a memory 12.
The processor 11 may be a Central Processing Unit (CPU) or other form of processing unit having data processing and/or instruction execution capabilities, and may control other components in the electronic device 10 to perform desired functions.
Memory 12 may include one or more computer program products that 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 (cache), and the like. The non-volatile memory may include, for example, read Only Memory (ROM), hard disk, flash memory, and the like. One or more computer program instructions may be stored on the computer readable storage medium that may be executed by the processor 11 to perform the functions in the appearance defect detection method of the power adapter of the various embodiments of the present application described above and/or other desired functions. Various contents such as appearance detection images may also be stored in the computer-readable storage medium.
In one example, the electronic device 10 may further include: an input device 13 and an output device 14, which are interconnected by a bus system and/or other forms of connection mechanisms (not shown).
The input means 13 may comprise, for example, a keyboard, a mouse, etc.
The output device 14 may output various information including the classification result and the like to the outside. The output means 14 may include, for example, a display, speakers, a printer, and a communication network and remote output devices connected thereto, etc.
Of course, only some of the components of the electronic device 10 that are relevant to the present application are shown in fig. 6 for simplicity, components such as buses, input/output interfaces, etc. are omitted. In addition, the electronic device 10 may include any other suitable components depending on the particular application.
Exemplary computer program product and computer readable storage Medium: in addition to the methods and apparatus described above, embodiments of the present application may also be a computer program product comprising computer program instructions which, when executed by a processor, cause the processor to perform steps in the functionality of the appearance defect detection method of a power adapter according to various embodiments of the present application described in the "exemplary methods" section of the present specification.
The computer program product may write program code for performing the operations of embodiments of the present application in any combination of one or more programming languages, including an object oriented programming language such as Java, C++ or the like and conventional procedural programming languages, such as the "C" programming language or similar programming languages. The program code may execute entirely on the user's computing device, partly on the user's device, as a stand-alone software package, partly on the user's computing device, partly on a remote computing device, or entirely on the remote computing device or server.
Furthermore, embodiments of the present application may also be a computer-readable storage medium, having stored thereon computer program instructions that, when executed by a processor, cause the processor to perform steps in the functionality in the appearance defect detection method of a power adapter according to various embodiments of the present application described in the above "exemplary method" section of the present specification.
The computer readable storage medium may employ any combination of one or more readable media. The readable medium may be a readable signal medium or a readable storage medium. The readable storage medium may include, for example, but is not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or a combination of any of the foregoing. More specific examples (a non-exhaustive list) of the readable storage medium would include the following: an electrical connection having one or more wires, a portable disk, a hard disk, random Access Memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.
The basic principles of the present application have been described above in connection with specific embodiments, however, it should be noted that the advantages, benefits, effects, etc. mentioned in the present application are merely examples and not limiting, and these advantages, benefits, effects, etc. are not to be considered as necessarily possessed by the various embodiments of the present application. Furthermore, the specific details disclosed herein are for purposes of illustration and understanding only, and are not intended to be limiting, as the application is not intended to be limited to the details disclosed herein as such.
The block diagrams of the devices, apparatuses, devices, systems referred to in this application are only illustrative examples and are not intended to require or imply that the connections, arrangements, configurations must be made in the manner shown in the block diagrams. As will be appreciated by one of skill in the art, the devices, apparatuses, devices, systems may be connected, arranged, configured in any manner. Words such as "including," "comprising," "having," and the like are words of openness and mean "including but not limited to," and are used interchangeably therewith. The terms "or" and "as used herein refer to and are used interchangeably with the term" and/or "unless the context clearly indicates otherwise. The term "such as" as used herein refers to, and is used interchangeably with, the phrase "such as, but not limited to.
It is also noted that in the apparatus, devices and methods of the present application, the components or steps may be disassembled and/or assembled. Such decomposition and/or recombination should be considered as equivalent to the present application.
The previous 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 generic principles defined herein may be applied to other aspects without departing from the scope of the application. Thus, the present application is not intended to be limited to the aspects shown herein but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.
The foregoing description has been presented for purposes of illustration and description. Furthermore, this description is not intended to limit the embodiments of the application to the form disclosed herein. Although a number of example aspects and embodiments have been discussed above, a person of ordinary skill in the art will recognize certain variations, modifications, alterations, additions, and subcombinations thereof.
Claims (6)
1. A method for detecting an appearance defect of a power adapter, comprising:
obtaining an appearance detection image of a first side surface of a power adapter to be detected;
Extracting a local binary pattern diagram and a direction gradient histogram of the appearance detection image of the first side, and aggregating the local binary pattern diagram, the direction gradient histogram and the appearance detection image of the first side along a channel dimension to obtain a multi-channel detection image;
performing image blocking processing on the multi-channel detection image, and then obtaining a plurality of image block context semantic feature vectors through a ViT model;
two-dimensionally arranging the context semantic feature vectors of the image blocks into a two-dimensional feature matrix, and then obtaining a classification feature map through a convolutional neural network model serving as a feature extractor;
performing convolution dictionary contrast response learning on each classification feature matrix of the classification feature map by using the two-dimensional feature matrix to obtain an optimized classification feature map; and
the optimized classification characteristic diagram is passed through a classifier to obtain a classification result, wherein the classification result is used for indicating whether the appearance of the first side surface of the power adapter to be detected has defects;
the method for obtaining the context semantic feature vectors of the image blocks through the ViT model after performing image blocking processing on the multi-channel detection image comprises the following steps:
Performing image blocking processing on the multi-channel detection image to obtain a sequence of image blocks;
using the embedding layer of the ViT model to respectively carry out embedded coding on each image block in the sequence of the image blocks so as to obtain a sequence of image block embedded vectors; and
inputting the sequence of image block embedding vectors into a converter module of the ViT model to obtain the plurality of image block context semantic feature vectors;
the embedding layer of the ViT model is used for respectively carrying out embedded coding on each image block in the sequence of the image blocks to obtain a sequence of image block embedded vectors, and the method comprises the following steps:
respectively expanding each image block in the sequence of image blocks into one-dimensional pixel input vectors to obtain a plurality of one-dimensional pixel input vectors; and
performing full-connection encoding on each one-dimensional pixel input vector in the plurality of one-dimensional pixel input vectors by using an embedding layer of the ViT model to obtain a sequence of the image block embedding vectors;
wherein inputting the sequence of image block embedding vectors into a converter module of the ViT model to obtain the plurality of image block context semantic feature vectors comprises:
arranging the sequence of the image block embedded vectors into an input vector;
Respectively converting the input vector into a query vector and a key vector through a learning embedding matrix;
calculating the product between the query vector and the transpose vector of the key vector to obtain a self-attention correlation matrix;
carrying out standardization processing on the self-attention association matrix to obtain a standardized self-attention association matrix;
inputting the standardized self-attention association matrix into a Softmax activation function to activate so as to obtain a self-attention feature matrix; and
and multiplying the self-attention feature matrix with each image block embedded vector in the sequence of image block embedded vectors as a value vector to obtain the plurality of image block context semantic feature vectors.
2. The method for detecting an appearance defect of a power adapter according to claim 1, wherein the two-dimensionally arranging the plurality of image block context semantic feature vectors into a two-dimensional feature matrix and obtaining a classification feature map by a convolutional neural network model as a feature extractor comprises: each layer using the convolutional neural network model is performed in forward transfer of the layer:
carrying out convolution processing on input data to obtain a convolution characteristic diagram;
Carrying out mean pooling based on a local feature matrix on the convolution feature map to obtain a pooled feature map; and
non-linear activation is carried out on the pooled feature map so as to obtain an activated feature map;
the output of the last layer of the convolutional neural network model is the classification characteristic diagram, and the input of the first layer of the convolutional neural network model is the two-dimensional characteristic matrix.
3. The method of claim 2, wherein performing convolutional dictionary contrast response learning on each classification feature matrix of the classification feature map using the two-dimensional feature matrix to obtain an optimized classification feature map, comprises:
performing convolution dictionary contrast response learning on each classification feature matrix of the classification feature map based on the two-dimensional feature matrix to obtain a plurality of optimized classification feature matrices; and
and cascading the plurality of optimized classification feature matrixes to obtain the optimized classification feature map.
4. The method of claim 3, wherein performing convolutional dictionary contrast response learning on each classification feature matrix of the classification feature map based on the two-dimensional feature matrix to obtain a plurality of optimized classification feature matrices, comprises:
Performing convolution dictionary contrast response learning on each classification feature matrix of the classification feature map according to the following learning formula based on the two-dimensional feature matrix to obtain a plurality of optimized classification feature matrices;
wherein, the learning formula is:
wherein M is 1 Representing the two-dimensional feature matrix, M 2i Representing an ith classification feature matrix in the classification feature graph, |·|| F The Frobenius norm of the matrix is represented,representing subtracting by position ++>Representing matrix multiplication, M 2 ' represents the optimized classification feature matrix.
5. The method for detecting the appearance defect of the power adapter according to claim 4, wherein the optimizing the classification characteristic map through a classifier to obtain a classification result, wherein the classification result is used for indicating whether the appearance of the first side surface of the power adapter to be detected is defective or not, and the method comprises the following steps:
performing feature map expansion on the optimized classification feature map to obtain classification feature vectors;
inputting the classification feature vector into a Softmax classification function of the classifier to obtain a probability value of the classification feature vector belonging to each classification label; and
and determining the classification label corresponding to the maximum probability value as the classification result.
6. An appearance defect detection system of a power adapter, comprising:
the appearance detection image acquisition module is used for acquiring an appearance detection image of the first side surface of the power adapter to be detected;
the image information widening module is used for extracting a local binary pattern diagram and a direction gradient histogram of the appearance detection image of the first side surface and converging the local binary pattern diagram, the direction gradient histogram and the appearance detection image of the first side surface along the channel dimension to obtain a multi-channel detection image;
the image block context extraction module is used for obtaining a plurality of image block context semantic feature vectors through a ViT model after carrying out image block processing on the multi-channel detection image;
the global advanced semantic extraction module is used for performing two-dimensional arrangement on the context semantic feature vectors of the plurality of image blocks into a two-dimensional feature matrix and then obtaining a classification feature map through a convolutional neural network model serving as a feature extractor;
the convolution dictionary contrast response learning module is used for carrying out convolution dictionary contrast response learning on each classification characteristic matrix of the classification characteristic map by using the two-dimensional characteristic matrix so as to obtain an optimized classification characteristic map; and
The detection result generation module is used for enabling the optimized classification characteristic diagram to pass through a classifier to obtain a classification result, wherein the classification result is used for indicating whether the appearance of the first side face of the power adapter to be detected has a defect or not;
the image block context extraction module is used for:
performing image blocking processing on the multi-channel detection image to obtain a sequence of image blocks;
using the embedding layer of the ViT model to respectively carry out embedded coding on each image block in the sequence of the image blocks so as to obtain a sequence of image block embedded vectors; and
inputting the sequence of image block embedding vectors into a converter module of the ViT model to obtain the plurality of image block context semantic feature vectors;
the embedding layer of the ViT model is used for respectively carrying out embedded coding on each image block in the sequence of the image blocks to obtain a sequence of image block embedded vectors, and the method comprises the following steps:
respectively expanding each image block in the sequence of image blocks into one-dimensional pixel input vectors to obtain a plurality of one-dimensional pixel input vectors; and
performing full-connection encoding on each one-dimensional pixel input vector in the plurality of one-dimensional pixel input vectors by using an embedding layer of the ViT model to obtain a sequence of the image block embedding vectors;
Wherein inputting the sequence of image block embedding vectors into a converter module of the ViT model to obtain the plurality of image block context semantic feature vectors comprises:
arranging the sequence of the image block embedded vectors into an input vector;
respectively converting the input vector into a query vector and a key vector through a learning embedding matrix;
calculating the product between the query vector and the transpose vector of the key vector to obtain a self-attention correlation matrix;
carrying out standardization processing on the self-attention association matrix to obtain a standardized self-attention association matrix;
inputting the standardized self-attention association matrix into a Softmax activation function to activate so as to obtain a self-attention feature matrix; and
and multiplying the self-attention feature matrix with each image block embedded vector in the sequence of image block embedded vectors as a value vector to obtain the plurality of image block context semantic feature vectors.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310298123.7A CN116797533B (en) | 2023-03-24 | 2023-03-24 | Appearance defect detection method and system for power adapter |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310298123.7A CN116797533B (en) | 2023-03-24 | 2023-03-24 | Appearance defect detection method and system for power adapter |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN116797533A CN116797533A (en) | 2023-09-22 |
| CN116797533B true CN116797533B (en) | 2024-01-23 |
Family
ID=88046937
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202310298123.7A Active CN116797533B (en) | 2023-03-24 | 2023-03-24 | Appearance defect detection method and system for power adapter |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN116797533B (en) |
Families Citing this family (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117326557A (en) * | 2023-09-28 | 2024-01-02 | 连云港市沃鑫高新材料有限公司 | Preparation method of silicon carbide high-purity micro powder for reaction sintering ceramic structural part |
| CN117636045A (en) * | 2023-12-07 | 2024-03-01 | 湖州练市漆宝木业有限公司 | Wood defect detection system based on image processing |
Citations (15)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105117739A (en) * | 2015-07-29 | 2015-12-02 | 南京信息工程大学 | Clothes classifying method based on convolutional neural network |
| CN107945179A (en) * | 2017-12-21 | 2018-04-20 | 王华锋 | A kind of good pernicious detection method of Lung neoplasm of the convolutional neural networks of feature based fusion |
| CN108052972A (en) * | 2017-12-11 | 2018-05-18 | 奕响（大连）科技有限公司 | A kind of single pass picture similar determination method of Two bors d's oeuveres |
| CN108427920A (en) * | 2018-02-26 | 2018-08-21 | 杭州电子科技大学 | A kind of land and sea border defense object detection method based on deep learning |
| WO2019104767A1 (en) * | 2017-11-28 | 2019-06-06 | 河海大学常州校区 | Fabric defect detection method based on deep convolutional neural network and visual saliency |
| WO2020173022A1 (en) * | 2019-02-25 | 2020-09-03 | 平安科技（深圳）有限公司 | Vehicle violation identifying method, server and storage medium |
| CN111930985A (en) * | 2020-07-08 | 2020-11-13 | 泰康保险集团股份有限公司 | Image retrieval method and device, electronic equipment and readable storage medium |
| CN112528905A (en) * | 2020-12-18 | 2021-03-19 | 上海海事大学 | Image processing method and device and computer storage medium |
| WO2021114809A1 (en) * | 2020-05-27 | 2021-06-17 | 平安科技（深圳）有限公司 | Vehicle damage feature detection method and apparatus, computer device, and storage medium |
| CN113420767A (en) * | 2021-07-22 | 2021-09-21 | 凌云光技术股份有限公司 | Method, system and device for extracting features for font classification |
| WO2022052181A1 (en) * | 2020-09-09 | 2022-03-17 | 华北电力大学（保定） | Insulator defect detection method and system based on zero-shot learning |
| CN115082745A (en) * | 2022-08-22 | 2022-09-20 | 深圳市成天泰电缆实业发展有限公司 | Image-based cable strand quality detection method and system |
| CN115294060A (en) * | 2022-08-05 | 2022-11-04 | 海尔数字科技(青岛)有限公司 | Method and device for detecting appearance defects of electrical equipment, electronic equipment and medium |
| CN115311248A (en) * | 2022-08-26 | 2022-11-08 | 上海致景信息科技有限公司 | Defect model matching method based on gray fabric texture, medium, and electronic device |
| CN115496740A (en) * | 2022-10-10 | 2022-12-20 | 湖北华鑫光电有限公司 | Lens defect detection method and system based on convolutional neural network |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US9275306B2 (en) * | 2013-11-13 | 2016-03-01 | Canon Kabushiki Kaisha | Devices, systems, and methods for learning a discriminant image representation |
-
2023
- 2023-03-24 CN CN202310298123.7A patent/CN116797533B/en active Active
Patent Citations (15)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105117739A (en) * | 2015-07-29 | 2015-12-02 | 南京信息工程大学 | Clothes classifying method based on convolutional neural network |
| WO2019104767A1 (en) * | 2017-11-28 | 2019-06-06 | 河海大学常州校区 | Fabric defect detection method based on deep convolutional neural network and visual saliency |
| CN108052972A (en) * | 2017-12-11 | 2018-05-18 | 奕响（大连）科技有限公司 | A kind of single pass picture similar determination method of Two bors d's oeuveres |
| CN107945179A (en) * | 2017-12-21 | 2018-04-20 | 王华锋 | A kind of good pernicious detection method of Lung neoplasm of the convolutional neural networks of feature based fusion |
| CN108427920A (en) * | 2018-02-26 | 2018-08-21 | 杭州电子科技大学 | A kind of land and sea border defense object detection method based on deep learning |
| WO2020173022A1 (en) * | 2019-02-25 | 2020-09-03 | 平安科技（深圳）有限公司 | Vehicle violation identifying method, server and storage medium |
| WO2021114809A1 (en) * | 2020-05-27 | 2021-06-17 | 平安科技（深圳）有限公司 | Vehicle damage feature detection method and apparatus, computer device, and storage medium |
| CN111930985A (en) * | 2020-07-08 | 2020-11-13 | 泰康保险集团股份有限公司 | Image retrieval method and device, electronic equipment and readable storage medium |
| WO2022052181A1 (en) * | 2020-09-09 | 2022-03-17 | 华北电力大学（保定） | Insulator defect detection method and system based on zero-shot learning |
| CN112528905A (en) * | 2020-12-18 | 2021-03-19 | 上海海事大学 | Image processing method and device and computer storage medium |
| CN113420767A (en) * | 2021-07-22 | 2021-09-21 | 凌云光技术股份有限公司 | Method, system and device for extracting features for font classification |
| CN115294060A (en) * | 2022-08-05 | 2022-11-04 | 海尔数字科技(青岛)有限公司 | Method and device for detecting appearance defects of electrical equipment, electronic equipment and medium |
| CN115082745A (en) * | 2022-08-22 | 2022-09-20 | 深圳市成天泰电缆实业发展有限公司 | Image-based cable strand quality detection method and system |
| CN115311248A (en) * | 2022-08-26 | 2022-11-08 | 上海致景信息科技有限公司 | Defect model matching method based on gray fabric texture, medium, and electronic device |
| CN115496740A (en) * | 2022-10-10 | 2022-12-20 | 湖北华鑫光电有限公司 | Lens defect detection method and system based on convolutional neural network |
Non-Patent Citations (1)
| Title |
|---|
| 基于机器视觉的枪弹外观缺陷检测系统研究;史进伟;郭朝勇;刘红宁;;组合机床与自动化加工技术(04);全文 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN116797533A (en) | 2023-09-22 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN116797533B (en) | Appearance defect detection method and system for power adapter | |
| CN110929622A (en) | Video classification method, model training method, device, equipment and storage medium | |
| CN110704842A (en) | Malicious code family classification detection method | |
| CN115860271B (en) | Scheme management system for artistic design and method thereof | |
| CN115620303B (en) | Personnel file intelligent management system | |
| CN116015837A (en) | Intrusion detection method and system for computer network information security | |
| CN117058622A (en) | Intelligent monitoring system and method for sewage treatment equipment | |
| CN116343301B (en) | Personnel information intelligent verification system based on face recognition | |
| Qian et al. | FESSD: SSD target detection based on feature fusion and feature enhancement | |
| CN117030129A (en) | Paper cup on-line leakage detection method and system thereof | |
| CN116758360B (en) | Land space use management method and system thereof | |
| CN111291712B (en) | Forest fire recognition method and device based on interpolation CN and capsule network | |
| CN116467485B (en) | Video image retrieval construction system and method thereof | |
| CN116402777B (en) | Power equipment detection method and system based on machine vision | |
| CN116858789A (en) | Food safety detection system and method thereof | |
| CN113269171B (en) | Lane line detection method, electronic device and vehicle | |
| CN109902720A (en) | The image classification recognition methods of depth characteristic estimation is carried out based on Subspace Decomposition | |
| CN111127502B (en) | Method and device for generating instance mask and electronic equipment | |
| Tomaszewski et al. | Object description based on local features repeatability | |
| Zhang et al. | A spatial–spectral adaptive learning model for textile defect images recognition with few labeled data | |
| CN112734016A (en) | Training method of neural network for detecting smoothness of surface of steel bar | |
| CN114998990B (en) | Method and device for identifying safety behaviors of personnel on construction site | |
| CN117456284B (en) | Image classification method, device, equipment and storage medium | |
| CN113627556B (en) | Method and device for realizing image classification, electronic equipment and storage medium | |
| CN116341396B (en) | Complex equipment digital twin modeling method based on multi-source data fusion |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |