CN117291900A - Photovoltaic module processing method and system - Google Patents

Photovoltaic module processing method and system Download PDF

Info

Publication number
CN117291900A
CN117291900A CN202311327137.3A CN202311327137A CN117291900A CN 117291900 A CN117291900 A CN 117291900A CN 202311327137 A CN202311327137 A CN 202311327137A CN 117291900 A CN117291900 A CN 117291900A
Authority
CN
China
Prior art keywords
feature map
feature
map
channel
convolution
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.)
Withdrawn
Application number
CN202311327137.3A
Other languages
Chinese (zh)
Inventor
陈文科
李梦宝
王陈
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Fengyang Zhaori Fine Technology Co ltd
Original Assignee
Fengyang Zhaori Fine Technology Co ltd
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Fengyang Zhaori Fine Technology Co ltd filed Critical Fengyang Zhaori Fine Technology Co ltd
Priority to CN202311327137.3A priority Critical patent/CN117291900A/en
Publication of CN117291900A publication Critical patent/CN117291900A/en
Withdrawn legal-status Critical Current

Links

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
    • G06V10/44—Local feature extraction by analysis of parts of the pattern, e.g. by detecting edges, contours, loops, corners, strokes or intersections; Connectivity analysis, e.g. of connected components
    • G06V10/443—Local feature extraction by analysis of parts of the pattern, e.g. by detecting edges, contours, loops, corners, strokes or intersections; Connectivity analysis, e.g. of connected components by matching or filtering
    • G06V10/449—Biologically inspired filters, e.g. difference of Gaussians [DoG] or Gabor filters
    • G06V10/451—Biologically inspired filters, e.g. difference of Gaussians [DoG] or Gabor filters with interaction between the filter responses, e.g. cortical complex cells
    • G06V10/454—Integrating the filters into a hierarchical structure, e.g. convolutional neural networks [CNN]
    • 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
    • G06V10/77—Processing image or video features in feature spaces; using data integration or data reduction, e.g. principal component analysis [PCA] or independent component analysis [ICA] or self-organising maps [SOM]; Blind source separation
    • G06V10/80—Fusion, i.e. combining data from various sources at the sensor level, preprocessing level, feature extraction level or classification level
    • G06V10/806—Fusion, i.e. combining data from various sources at the sensor level, preprocessing level, feature extraction level or classification level of extracted features
    • 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
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/10—Image acquisition modality
    • G06T2207/10048—Infrared image
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/20—Special algorithmic details
    • G06T2207/20084—Artificial neural networks [ANN]
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/30—Subject of image; Context of image processing
    • G06T2207/30108—Industrial image inspection
    • G06T2207/30148—Semiconductor; IC; Wafer
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V2201/00—Indexing scheme relating to image or video recognition or understanding
    • G06V2201/06—Recognition of objects for industrial automation
    • Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
    • Y02E—REDUCTION OF GREENHOUSE GAS [GHG] EMISSIONS, RELATED TO ENERGY GENERATION, TRANSMISSION OR DISTRIBUTION
    • Y02E10/00—Energy generation through renewable energy sources
    • Y02E10/50—Photovoltaic [PV] energy

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Evolutionary Computation (AREA)
  • General Health & Medical Sciences (AREA)
  • Artificial Intelligence (AREA)
  • Health & Medical Sciences (AREA)
  • Multimedia (AREA)
  • Computing Systems (AREA)
  • Databases & Information Systems (AREA)
  • Medical Informatics (AREA)
  • Software Systems (AREA)
  • Biomedical Technology (AREA)
  • Biodiversity & Conservation Biology (AREA)
  • Molecular Biology (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Quality & Reliability (AREA)
  • Image Processing (AREA)

Abstract

The application relates to the field of intelligent processing, and particularly discloses a processing method and a system of a photovoltaic module, wherein an artificial intelligent technology based on a deep neural network model is adopted to obtain near infrared images of the photovoltaic module, key local feature information is extracted through spatial attention convolution after pretreatment, key feature representations among different feature channels are enhanced through channel attention convolution respectively, and dependency relations in a global range are captured through a non-local network after fusion, so that a classification result for representing whether the photovoltaic module has defects or not is obtained. Therefore, the defect detection of the photovoltaic cell panel assembly can be further realized, the cost and the error rate of manual detection are reduced, and the automation level and the product quality of a production line are improved.

Description

Photovoltaic module processing method and system
Technical Field
The application relates to the field of intelligent processing, and more particularly relates to a processing method and a system of a photovoltaic module.
Background
The core part of the photovoltaic module is a solar panel, a certain number of single cells are sealed into the solar module in a serial and parallel mode, and the solar module is responsible for converting sunlight into electric energy. The single solar cell needs to be subjected to defect detection after being assembled.
Defects in the photovoltaic module may result in reduced performance. For example, components that are cracked or damaged may not be able to efficiently convert solar energy to electrical energy, thereby reducing the power generation efficiency of the overall photovoltaic system. May lead to a shortened lifetime of the photovoltaic module. For example, components that are subject to corrosion, moisture intrusion, or other damage may be susceptible to environmental effects, resulting in faster aging and damage, and even electrical leakage. However, the prior art is subjective and some fine places are left missing and careless because the prior art is manually and visually detected.
Thus, an optimized photovoltaic module processing scheme 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 a processing method and a processing system of a photovoltaic module, wherein an artificial intelligence technology based on a deep neural network model is adopted to acquire a near infrared image of a photovoltaic panel module, key local feature information and key feature representation among different feature channels are respectively enhanced through spatial attention convolution after preprocessing, and a dependency relationship in a global range is captured through a non-local network after fusion so as to obtain a classification result for representing whether the photovoltaic panel module has defects. Therefore, the defect detection of the photovoltaic cell panel assembly can be further realized, the cost and the error rate of manual detection are reduced, and the automation level and the product quality of a production line are improved.
According to an aspect of the present application, there is provided a method for processing a photovoltaic module, including:
acquiring a near infrared image of a photovoltaic cell panel assembly;
preprocessing the near infrared image of the photovoltaic cell panel assembly to obtain a near infrared image of the preprocessed photovoltaic cell panel assembly, wherein the preprocessing comprises image denoising, contrast enhancement and brightness adjustment;
the near infrared image of the photovoltaic cell panel assembly after pretreatment is subjected to a first convolution neural network model using spatial attention so as to obtain a spatial enhancement photovoltaic assembly characteristic diagram;
the near infrared image of the photovoltaic cell panel assembly after pretreatment is subjected to a second convolution neural network model using the channel attention so as to obtain a channel enhancement photovoltaic assembly characteristic diagram;
fusing the space enhancement photovoltaic module feature map and the channel enhancement photovoltaic module feature map to obtain a fused feature map;
the fusion feature map is passed through a non-local neural network model to obtain a classification feature map;
performing order priori treatment on the classification characteristic map to obtain an optimized classification characteristic 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 photovoltaic cell panel assembly has defects or not.
In the above processing method of a photovoltaic module, the processing method of obtaining a spatially enhanced photovoltaic module feature map from the preprocessed near infrared image of the photovoltaic panel module by using a first convolutional neural network model of spatial attention includes: performing depth convolution coding on the preprocessed near infrared image of the photovoltaic cell panel assembly by using a convolution coding part of the first convolution neural network model to obtain an initial convolution feature map; inputting the initial convolution feature map into a spatial attention portion of the convolution neural network model to obtain a spatial attention map; -passing said spatial attention map through a Softmax activation function to obtain a spatial attention profile; and calculating the position-wise point multiplication of the spatial attention feature map and the initial convolution feature map to obtain the spatial enhancement photovoltaic module feature map.
In the above processing method of a photovoltaic module, inputting the initial convolution feature map into a spatial attention portion of the convolution neural network model to obtain a spatial attention map includes: respectively carrying out average pooling and maximum pooling along the channel dimension on the initial convolution feature map to obtain an average feature matrix and a maximum feature matrix; cascading and channel adjustment are carried out on the average characteristic matrix and the maximum characteristic matrix to obtain a channel characteristic matrix; and convolutionally encoding the channel feature matrix using a convolution layer of the spatial attention profile to obtain a spatial attention profile.
In the above processing method of a photovoltaic module, the processing method of obtaining a channel enhancement photovoltaic module feature map from the near infrared image of the photovoltaic panel module after pretreatment by using a second convolution neural network model of channel attention includes: input data is processed in forward pass of layers using layers of the second convolutional neural network model: performing convolution processing based on a two-dimensional convolution kernel on the input data to generate a convolution feature map; pooling the convolution feature map to generate a pooled feature map; performing activation processing on the pooled feature map to generate an activated feature map; calculating the global average value of each feature matrix of the activated feature map along the channel dimension to obtain a channel feature vector; calculating the ratio of the eigenvalue of each position in the channel eigenvector relative to the weighted sum of the eigenvalues of all positions of the channel eigenvector to obtain a channel weighted eigenvector; taking the characteristic values of the positions of the channel weighted characteristic vectors as weights to perform point multiplication on the characteristic matrix of the activated characteristic map along the channel dimension to obtain a generated characteristic map; the generated feature map output by the last layer of the second convolutional neural network model is the channel enhancement photovoltaic module feature map.
In the above processing method of a photovoltaic module, the step of obtaining the classification feature map by passing the fusion feature map through a non-local neural network model includes: performing first point convolution processing, second point convolution processing and third point convolution processing on the fusion feature map respectively to obtain a first feature map, a second feature map and a third feature map; calculating a weighted sum of the first feature map and the second feature map by position to obtain a weighted fusion feature map; inputting the weighted fusion feature map into a Softmax function to map feature values of all positions in the weighted fusion feature map into a probability space so as to obtain a normalized fusion feature map; calculating the point-by-point multiplication between the normalized fusion feature map and the third feature map to obtain a re-fusion feature map; embedding the re-fusion feature map into a Gaussian similarity function to obtain a global similarity feature map; performing fourth point convolution processing on the global similar feature map to adjust the number of channels of the global similar feature map so as to obtain a channel-adjusted global similar feature map; and calculating a weighted sum of the channel-adjustment global similarity feature map and the fusion feature map by position to obtain the classification feature map.
In the above processing method of a photovoltaic module, the classifying result obtained by passing the optimized classifying feature map through a classifier is used for indicating whether the photovoltaic panel module is defective or not, and includes: expanding the optimized classification feature map into classification feature vectors; performing full-connection coding on the classification feature vectors by using a full-connection layer of the classifier to obtain coded classification feature vectors; and passing the coding classification feature vector through a Softmax classification function of the classifier to obtain the classification result.
According to another aspect of the present application, there is provided a processing system of a photovoltaic module, comprising:
the photovoltaic image acquisition module is used for acquiring a near infrared image of the photovoltaic cell panel assembly;
the image preprocessing module is used for preprocessing the near infrared image of the photovoltaic cell panel assembly to obtain a near infrared image of the photovoltaic cell panel assembly after preprocessing, wherein the preprocessing comprises image denoising, contrast enhancement and brightness adjustment;
the space feature extraction module is used for obtaining a space enhancement photovoltaic module feature map through a first convolution neural network model of the space attention of the near infrared image of the preprocessed photovoltaic cell panel module;
The channel characteristic extraction module is used for obtaining a channel enhancement photovoltaic module characteristic diagram through a second convolution neural network model of channel attention from the near infrared image of the pretreated photovoltaic module;
the fusion module is used for fusing the space enhancement photovoltaic module feature map and the channel enhancement photovoltaic module feature map to obtain a fusion feature map;
the global extraction module is used for enabling the fusion feature images to pass through a non-local neural network model to obtain classification feature images;
the optimizing module is used for carrying out order priori on the classification characteristic map so as to obtain an optimized classification characteristic map;
and the defect judging 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 photovoltaic cell panel assembly has defects or not.
Compared with the prior art, the processing method and the processing system of the photovoltaic module, provided by the application, adopt an artificial intelligence technology based on a deep neural network model to acquire near infrared images of the photovoltaic panel module, extract key local feature information through spatial attention convolution after pretreatment and enhance key feature representation among different feature channels through channel attention convolution respectively, and capture dependency relations in a global range through a non-local network after fusion so as to obtain a classification result for representing whether the photovoltaic panel module is defective. Therefore, the defect detection of the photovoltaic cell panel assembly can be further realized, the cost and the error rate of manual detection are reduced, and the automation level and the product quality of a production line 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 a flowchart of a method of processing a photovoltaic module according to an embodiment of the present application.
Fig. 2 is a schematic diagram of a processing method of a photovoltaic module according to an embodiment of the present application.
Fig. 3 is a flowchart of a method for processing a photovoltaic module according to an embodiment of the present application, where a near infrared image of a photovoltaic panel module after pretreatment is used to obtain a feature map of the photovoltaic module by using a first convolutional neural network model with spatial attention.
Fig. 4 is a flowchart of order priors of the classification feature map to obtain an optimized classification feature map in the processing method of the photovoltaic module according to the embodiment of the present application.
Fig. 5 is a block diagram of a processing system for photovoltaic modules 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.
Exemplary method
Fig. 1 is a flowchart of a method of processing a photovoltaic module according to an embodiment of the present application. As shown in fig. 1, according to an embodiment of the present application, the method includes: s110, acquiring a near infrared image of a photovoltaic cell panel assembly; s120, preprocessing the near infrared image of the photovoltaic cell panel assembly to obtain a near infrared image of the preprocessed photovoltaic cell panel assembly, wherein the preprocessing comprises image denoising, contrast enhancement and brightness adjustment; s130, the preprocessed near infrared image of the photovoltaic cell panel assembly is subjected to a first convolution neural network model using spatial attention so as to obtain a spatial enhancement photovoltaic assembly characteristic diagram; s140, the near infrared image of the photovoltaic cell panel assembly after pretreatment is processed through a second convolution neural network model using the channel attention to obtain a channel enhancement photovoltaic assembly characteristic diagram; s150, fusing the space enhancement photovoltaic module feature map and the channel enhancement photovoltaic module feature map to obtain a fused feature map; s160, the fusion feature map is passed through a non-local neural network model to obtain a classification feature map; s170, carrying out order priori on the classification characteristic map to obtain an optimized classification characteristic map; and S180, 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 photovoltaic cell panel assembly has defects or not.
Fig. 2 is a schematic diagram of a processing method of a photovoltaic module according to an embodiment of the present application. As shown in fig. 2, first, a near infrared image of a photovoltaic panel assembly is acquired. And then, preprocessing the near infrared image of the photovoltaic cell panel assembly to obtain a near infrared image of the preprocessed photovoltaic cell panel assembly, wherein the preprocessing comprises image denoising, contrast enhancement and brightness adjustment. And then, the preprocessed near infrared image of the photovoltaic cell panel assembly is subjected to a first convolution neural network model using spatial attention so as to obtain a spatial enhancement photovoltaic assembly characteristic diagram. And meanwhile, obtaining a channel enhancement photovoltaic module characteristic diagram through a second convolution neural network model of channel attention by using the near infrared image of the photovoltaic cell panel module after pretreatment. And then fusing the space enhancement photovoltaic module feature map and the channel enhancement photovoltaic module feature map to obtain a fused feature map. And then, the fusion characteristic map is passed through a non-local neural network model to obtain a classification characteristic map. Then, order priors are carried out on the classification characteristic diagram to obtain an optimized classification characteristic diagram. 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 photovoltaic cell panel assembly has defects or not.
In step S110, a near infrared image of the photovoltaic panel assembly is acquired. It is contemplated that near infrared images may provide information about the thermal profile and reflective properties of the material. By analyzing the near infrared image, defects such as cracks, bubbles, contamination, etc., on the surface of the panel assembly can be detected. In addition, near infrared light can be transmitted through some materials and reflected back, with high sensitivity to the thermal profile and composition of the materials. By analyzing the near infrared image, minute defects or anomalies can be detected. Therefore, acquiring the near infrared image of the photovoltaic cell panel assembly can provide information about the surface defects of the assembly, and high-sensitivity defect detection is realized. This helps to improve quality control and efficiency of the photovoltaic panel assembly line.
In step S120, the near infrared image of the photovoltaic cell panel assembly is preprocessed to obtain a near infrared image of the preprocessed photovoltaic cell panel assembly, where the preprocessing includes image denoising, contrast enhancement, and brightness adjustment. Considering that near infrared images may be affected by ambient noise, sensor noise or signal interference, these noise can degrade the quality of the image and interfere with subsequent processing steps. By removing noise, the definition and detail of the image can be improved, so that the subsequent processing and analysis are more accurate. In addition, defects in near infrared images tend to appear as differences from surrounding areas. By enhancing the contrast of the image, the difference between the defect and normal area can be made more pronounced, helping to better detect and segment the defect area. In addition, the near infrared image may have non-uniformity in brightness, and some areas may be too bright or too dark. By adjusting the brightness, details in the image can be made more clearly visible, which is helpful for accurately capturing and analyzing defects. Through the preprocessing steps, near infrared images of the photovoltaic cell panel assembly can be optimized, the quality and usability of the images are improved, and more reliable input is provided for subsequent feature extraction and classification steps. Therefore, the accuracy and the reliability of defect detection can be improved, and the production line is helped to realize high-quality automatic control.
In step S130, the near infrared image of the photovoltaic panel assembly after pretreatment is passed through a first convolutional neural network model using spatial attention to obtain a spatially enhanced photovoltaic assembly feature map. Considering that defects often appear as anomalies or irregularities in localized areas in the near infrared image of the photovoltaic panel assembly. By using spatial attention convolution, the network can be made to pay more attention to these critical local features, improving defect detection capability. In particular, spatial attention convolution may help the network to better perceive context information in the image. By weighting the features in different locations, the network can better understand the relationship between different areas in the image, thereby improving the perceptibility of the defect. In addition, the spatial attention convolution may adaptively adjust the weights of features according to the image content. For near infrared images of photovoltaic panel assemblies, the importance of different regions may be different, and some regions may be more differentiated for defect detection. By using spatial attention convolution, the network can automatically learn and weight the characteristics of different areas, and the defect detection accuracy is improved.
Fig. 3 is a flowchart of a method for processing a photovoltaic module according to an embodiment of the present application, where a near infrared image of a photovoltaic panel module after pretreatment is used to obtain a feature map of the photovoltaic module by using a first convolutional neural network model with spatial attention. Specifically, in the embodiment of the present application, as shown in fig. 3, the method for obtaining a feature map of a spatially enhanced photovoltaic module by using a first convolutional neural network model of spatial attention on a near infrared image of the preprocessed photovoltaic cell panel assembly includes: s210, performing depth convolution coding on the near infrared image of the photovoltaic panel assembly after pretreatment by using a convolution coding part of the first convolution neural network model to obtain an initial convolution feature map; s220, inputting the initial convolution feature map into a spatial attention part of the convolution neural network model to obtain a spatial attention map; s230, the spatial attention is subjected to a Softmax activation function to obtain a spatial attention profile; and S240, calculating the position-wise point multiplication of the spatial attention characteristic diagram and the initial convolution characteristic diagram to obtain the spatial enhancement photovoltaic module characteristic diagram.
More specifically, in an embodiment of the present application, inputting the initial convolution feature map into a spatial attention portion of the convolution neural network model to obtain a spatial attention map includes: respectively carrying out average pooling and maximum pooling along the channel dimension on the initial convolution feature map to obtain an average feature matrix and a maximum feature matrix; cascading and channel adjustment are carried out on the average characteristic matrix and the maximum characteristic matrix to obtain a channel characteristic matrix; and convolutionally encoding the channel feature matrix using a convolution layer of the spatial attention profile to obtain a spatial attention profile.
In step S140, the near infrared image of the photovoltaic panel assembly after pretreatment is processed by using a second convolutional neural network model of channel attention to obtain a channel enhancement photovoltaic assembly feature map. It is contemplated that in the near infrared image of the photovoltaic panel assembly, different channels may have different contributions to the representation and differentiation of defects. By using channel attention convolution, the network can pay more attention to the characteristic channels which are more important for defect detection, and the expression capability of the defects is improved. In particular, channel attention convolution can help the network better distinguish importance between different characteristic channels. By weighting the characteristics of the different channels, the network can better mine and utilize those characteristic channels that distinguish defective areas, improving the accuracy of defect detection. In addition, the channel attention convolution may adaptively adjust the weights of the feature channels according to the image content. For near infrared images of photovoltaic panel assemblies, the importance of different characteristic channels may be different, and some channels may be more discriminating against defect detection. By using the channel attention convolution, the network can automatically learn and weight the characteristics of different channels, and the defect detection accuracy is improved.
Specifically, in an embodiment of the present application, the obtaining a channel enhancement photovoltaic module feature map by using the second convolutional neural network model of channel attention on the near infrared image of the photovoltaic panel module after preprocessing includes: input data is processed in forward pass of layers using layers of the second convolutional neural network model: performing convolution processing based on a two-dimensional convolution kernel on the input data to generate a convolution feature map; pooling the convolution feature map to generate a pooled feature map; performing activation processing on the pooled feature map to generate an activated feature map; calculating the global average value of each feature matrix of the activated feature map along the channel dimension to obtain a channel feature vector; calculating the ratio of the eigenvalue of each position in the channel eigenvector relative to the weighted sum of the eigenvalues of all positions of the channel eigenvector to obtain a channel weighted eigenvector; taking the characteristic values of the positions of the channel weighted characteristic vectors as weights to perform point multiplication on the characteristic matrix of the activated characteristic map along the channel dimension to obtain a generated characteristic map; the generated feature map output by the last layer of the second convolutional neural network model is the channel enhancement photovoltaic module feature map.
In step S150, the space enhancement photovoltaic module feature map and the channel enhancement photovoltaic module feature map are fused to obtain a fused feature map. Considering that the spatial feature map mainly focuses on spatial locations and local structures in the image, the channel feature map focuses on key features between different feature channels. By combining the two feature diagrams, the space information and the channel information can be integrated, and the features of the photovoltaic module can be more comprehensively described. Specifically, fusing the spatial feature map and the channel feature map may enhance the expressive power of the features. The spatial signature may provide more detailed local structural information, while the channel signature may provide a richer representation of the features. By fusing them, a feature map having more distinguishing and expressing ability can be obtained, which contributes to more accurately detecting and classifying defects of the photovoltaic module.
In step S160, the fused feature map is passed through a non-local neural network model to obtain a classification feature map. Considering that the non-local neural network model can capture global context information between different locations in the image, it helps to better understand the structure and relationships in the image. In particular, in the present technical solution, global context information is very important for understanding the structure and relationship in the image in defect detection of the photovoltaic module. By using the non-local neural network model, the dependency relationship in the global scope can be captured, so that the network can better understand the relationship between the features at different positions in the image, and the expression capability of the classification feature map is improved. In addition, the non-local neural network model can expand the receptive field, enabling the network to receive wider information. This is important for defect detection of photovoltaic modules, as defects may be distributed at different locations throughout the image. By expanding the receptive field, the non-local neural network model can better capture the spatial distribution characteristics of the defects, and the accuracy of the classification characteristic map is improved.
Specifically, in the embodiment of the present application, the step of passing the fused feature map through a non-local neural network model to obtain a classification feature map includes: performing first point convolution processing, second point convolution processing and third point convolution processing on the fusion feature map respectively to obtain a first feature map, a second feature map and a third feature map; calculating a weighted sum of the first feature map and the second feature map by position to obtain a weighted fusion feature map; inputting the weighted fusion feature map into a Softmax function to map feature values of all positions in the weighted fusion feature map into a probability space so as to obtain a normalized fusion feature map; calculating the point-by-point multiplication between the normalized fusion feature map and the third feature map to obtain a re-fusion feature map; embedding the re-fusion feature map into a Gaussian similarity function to obtain a global similarity feature map; performing fourth point convolution processing on the global similar feature map to adjust the number of channels of the global similar feature map so as to obtain a channel-adjusted global similar feature map; and calculating a weighted sum of the channel-adjustment global similarity feature map and the fusion feature map by position to obtain the classification feature map.
In step S170, the classification characteristic map is rank-priors to obtain an optimized classification characteristic map.
In the technical scheme of the application, the classification characteristic diagram is taken into consideration as a characteristic set to meet the preset prior distribution, but the type of the prior distribution has uncertainty; on the other hand, the classification characteristic map is obtained from the original data, noise is inevitably introduced into the original data when the data is acquired, and noise information is inevitably learned in the process of utilizing the deep neural network model to perform characteristic learning, so that the characteristic distribution and expression of the noise information also exist in the classification characteristic map.
In view of the above technical problems, in the technical solution of the present application, feature decoupling is performed on the classification feature map along a channel dimension to obtain a sequence of feature matrices, and a per-position average value matrix of all feature matrices in the sequence of feature matrices is calculated to obtain a pseudo prior center feature matrix. And performing feature deconstructment on the classification feature map, and calculating a center matrix of all feature matrix subsets in the classification feature map as implicit feature expression of self prior distribution of the classification feature map. And then, calculating information entropy between the pseudo prior central feature matrix and each feature matrix in the sequence of feature matrices to obtain a plurality of information entropy. Here, the information entropy is used to measure the similarity between each feature matrix of the classification feature map along the channel dimension and its own prior distribution hidden matrix, it should be understood that the feature encoding is essentially to map data into a specific feature domain, and the feature expression of the data before and after the domain mapping still maintains the predetermined pattern features, so in the technical solution of the present application, the degree of the predetermined pattern features is maintained by the information entropy quantization measure.
Then, an order priors mask feature vector is generated based on a comparison between the plurality of information entropies and a predetermined threshold, and the classification feature map is processed to obtain the optimized classification feature map based on the order priors mask feature vector. Specifically, if the ith information entropy is greater than or equal to a predetermined threshold, the eigenvalue of the ith position in the order priors mask eigenvector is 1; if the ith information entropy is smaller than a preset threshold value, the feature value of the ith position in the order priori mask feature vector is 0, and in the process of processing the classification feature map based on the order priori mask feature vector to obtain the optimized classification feature map, if the feature value of the corresponding position is 1, the feature matrix of the classification feature map is reserved, and if the feature value of the corresponding position is 0, the corresponding feature matrix of the classification feature map is set to be a 0-value matrix. It should be noted that this is essentially based on the information entropy between the implicit matrix of the self prior distribution of the classification feature map and the respective original feature matrices of the classification feature map, so as to perform feature masking screening to reject noise information, thereby improving the signal-to-noise ratio of the information expression of the feature distribution, and improving the accuracy of classification judgment of the optimized classification feature map by the classifier.
Specifically, in the embodiment of the present application, based on the comparison between the plurality of information entropies and the predetermined threshold, an order-prioritised mask feature vector is generated, and if the i-th information entropy is greater than or equal to the predetermined threshold, the feature value of the i-th position in the order-prioritised mask feature vector is 1; if the ith information entropy is less than the predetermined threshold, the rank priors the eigenvalue of the ith position in the mask eigenvector to 0.
More specifically, in the embodiment of the present application, based on the order priors mask feature vector, the classification feature map is processed to obtain the optimized classification feature map, if the feature value of the corresponding position is 1, the feature matrix of the classification feature map is reserved, and if the feature value of the corresponding position is 0, the corresponding feature matrix of the classification feature map is set to be a 0-value matrix.
Fig. 4 is a flowchart of order priors of the classification feature map to obtain an optimized classification feature map in the processing method of the photovoltaic module according to the embodiment of the present application. More specifically, in the embodiment of the present application, as shown in fig. 4, the order priors of the classification feature map is performed to obtain an optimized classification feature map, including: s310, performing feature decoupling on the classification feature map along the channel dimension to obtain a sequence of feature matrixes; s320, calculating the position-based mean value matrix of all feature matrices in the sequence of the feature matrices to obtain a pseudo prior center feature matrix; s330, calculating information entropy between the pseudo prior central feature matrix and each feature matrix in the sequence of the feature matrix to obtain a plurality of information entropy; s340, generating order priori mask feature vectors based on the comparison between the information entropies and the preset threshold values; and S350, processing the classification characteristic map based on the order priori mask characteristic vector to obtain the optimized classification characteristic map.
In step S180, 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 photovoltaic cell panel assembly is defective.
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 photovoltaic cell panel assembly is defective, and the method includes: expanding the optimized classification feature map into classification feature vectors; performing full-connection coding on the classification feature vectors by using a full-connection layer of the classifier to obtain coded classification feature vectors; and passing the coding classification feature vector through a Softmax classification function of the classifier to obtain the classification result.
In summary, a processing method of a photovoltaic module based on an embodiment of the application is explained, an artificial intelligence technology based on a deep neural network model is adopted to obtain a near infrared image of a photovoltaic panel module, key local feature information and key feature representation among different feature channels are respectively extracted through spatial attention convolution after preprocessing, and a dependency relationship in a global range is captured through a non-local network after fusion, so that a classification result for representing whether the photovoltaic panel module has defects or not is obtained. Therefore, the defect detection of the photovoltaic cell panel assembly can be further realized, the cost and the error rate of manual detection are reduced, and the automation level and the product quality of a production line are improved.
Exemplary System
Fig. 5 is a block diagram of a processing system for photovoltaic modules according to an embodiment of the present application. As shown in fig. 5, a processing system 100 of a photovoltaic module according to an embodiment of the present application includes: the photovoltaic image acquisition module 110 is used for acquiring a near infrared image of the photovoltaic cell panel assembly; the image preprocessing module 120 is configured to preprocess the near infrared image of the photovoltaic cell panel assembly to obtain a preprocessed near infrared image of the photovoltaic cell panel assembly, where the preprocessing includes image denoising, contrast enhancement, and brightness adjustment; the spatial feature extraction module 130 is configured to obtain a spatial enhancement photovoltaic module feature map by using a first convolutional neural network model of spatial attention on the near infrared image of the preprocessed photovoltaic cell panel assembly; the channel feature extraction module 140 is configured to obtain a channel enhancement photovoltaic module feature map by using a second convolutional neural network model of channel attention from the near infrared image of the pretreated photovoltaic panel module; the fusion module 150 is configured to fuse the space enhancement photovoltaic module feature map and the channel enhancement photovoltaic module feature map to obtain a fused feature map; the global extraction module 160 is configured to pass the fused feature map through a non-local neural network model to obtain a classification feature map; an optimization module 170, configured to order the classification feature map prior to obtain an optimized classification feature map; and a defect judging module 180, 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 photovoltaic panel assembly has defects.
Here, it will be understood by those skilled in the art that the specific operations of the respective steps in the above-described processing system of the photovoltaic module have been described in detail in the above description of the processing method of the photovoltaic module with reference to fig. 1 to 4, and thus, repetitive descriptions thereof will be omitted.
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 can be executed by the processor 11 to perform the functions described above in the methods of processing photovoltaic modules of the various embodiments of the present application and/or other desired functions. Various contents such as near infrared images of the photovoltaic panel assembly 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 functions in the processing methods of photovoltaic modules according to the various embodiments of the present application described in the "exemplary methods" section of this 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, on which computer program instructions are stored, which, when being executed by a processor, cause the processor to perform steps in the functions in the processing method of a photovoltaic module 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.
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 (10)

1. The processing method of the photovoltaic module is characterized by comprising the following steps of:
acquiring a near infrared image of a photovoltaic cell panel assembly;
preprocessing the near infrared image of the photovoltaic cell panel assembly to obtain a near infrared image of the preprocessed photovoltaic cell panel assembly, wherein the preprocessing comprises image denoising, contrast enhancement and brightness adjustment;
the near infrared image of the photovoltaic cell panel assembly after pretreatment is subjected to a first convolution neural network model using spatial attention so as to obtain a spatial enhancement photovoltaic assembly characteristic diagram;
the near infrared image of the photovoltaic cell panel assembly after pretreatment is subjected to a second convolution neural network model using the channel attention so as to obtain a channel enhancement photovoltaic assembly characteristic diagram;
fusing the space enhancement photovoltaic module feature map and the channel enhancement photovoltaic module feature map to obtain a fused feature map;
The fusion feature map is passed through a non-local neural network model to obtain a classification feature map;
performing order priori treatment on the classification characteristic map to obtain an optimized classification characteristic 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 photovoltaic cell panel assembly has defects or not.
2. The method for processing a photovoltaic module according to claim 1, wherein the step of obtaining the spatially enhanced photovoltaic module feature map from the preprocessed near infrared image of the photovoltaic panel module by using the first convolutional neural network model of spatial attention comprises:
performing depth convolution coding on the preprocessed near infrared image of the photovoltaic cell panel assembly by using a convolution coding part of the first convolution neural network model to obtain an initial convolution feature map;
inputting the initial convolution feature map into a spatial attention portion of the convolution neural network model to obtain a spatial attention map;
-passing said spatial attention map through a Softmax activation function to obtain a spatial attention profile;
and calculating the position-based point multiplication of the spatial attention characteristic diagram and the initial convolution characteristic diagram to obtain the spatial enhancement photovoltaic module characteristic diagram.
3. The method of processing a photovoltaic module of claim 2, wherein inputting the initial convolution signature into the spatial attention portion of the convolution neural network model to obtain a spatial attention map comprises:
respectively carrying out average pooling and maximum pooling along the channel dimension on the initial convolution feature map to obtain an average feature matrix and a maximum feature matrix;
cascading and channel adjustment are carried out on the average characteristic matrix and the maximum characteristic matrix to obtain a channel characteristic matrix;
the channel feature matrix is convolutionally encoded using a convolution layer of the spatial attention feature map to obtain a spatial attention map.
4. The method for processing a photovoltaic module according to claim 3, wherein the step of obtaining the channel enhancement photovoltaic module feature map from the near infrared image of the pretreated photovoltaic panel module by using the second convolutional neural network model of channel attention comprises:
input data is processed in forward pass of layers using layers of the second convolutional neural network model:
performing convolution processing based on a two-dimensional convolution kernel on the input data to generate a convolution feature map;
Pooling the convolution feature map to generate a pooled feature map;
performing activation processing on the pooled feature map to generate an activated feature map;
calculating the global average value of each feature matrix of the activated feature map along the channel dimension to obtain a channel feature vector;
calculating the ratio of the eigenvalue of each position in the channel eigenvector relative to the weighted sum of the eigenvalues of all positions of the channel eigenvector to obtain a channel weighted eigenvector;
taking the characteristic values of each position of the channel weighted characteristic vector as weights to carry out point multiplication on the characteristic matrix of the activated characteristic map along the channel dimension to obtain a generated characteristic map;
the generated feature map output by the last layer of the second convolutional neural network model is the channel enhancement photovoltaic module feature map.
5. The method of claim 4, wherein the step of passing the fused feature map through a non-local neural network model to obtain a classification feature map comprises:
performing first point convolution processing, second point convolution processing and third point convolution processing on the fusion feature map respectively to obtain a first feature map, a second feature map and a third feature map;
Calculating a weighted sum of the first feature map and the second feature map by position to obtain a weighted fusion feature map;
inputting the weighted fusion feature map into a Softmax function to map feature values of all positions in the weighted fusion feature map into a probability space so as to obtain a normalized fusion feature map;
calculating the point-by-point multiplication between the normalized fusion feature map and the third feature map to obtain a re-fusion feature map;
embedding the re-fusion feature map into a Gaussian similarity function to obtain a global similarity feature map;
performing fourth point convolution processing on the global similar feature map to adjust the number of channels of the global similar feature map so as to obtain a channel-adjusted global similar feature map;
and calculating a weighted sum of the channel adjustment global similarity feature map and the fusion feature map according to positions to obtain the classification feature map.
6. The method of claim 5, wherein order priors the classification signature to obtain an optimized classification signature, comprising:
performing feature decoupling on the classification feature map along a channel dimension to obtain a sequence of feature matrixes;
Calculating the mean value matrix according to the position of all feature matrixes in the sequence of the feature matrixes to obtain a pseudo prior center feature matrix;
calculating information entropy between each feature matrix in the pseudo prior center feature matrix and the sequence of the feature matrix to obtain a plurality of information entropy;
generating an order-priors mask feature vector based on a comparison between the plurality of information entropies and a predetermined threshold;
and processing the classification characteristic map based on the order prior mask characteristic vector to obtain the optimized classification characteristic map.
7. The method according to claim 6, wherein the step of passing the optimized classification feature map through a classifier to obtain a classification result, wherein the classification result is used for indicating whether the photovoltaic panel assembly is defective or not, and the method comprises the steps of:
expanding the optimized classification feature map into classification feature vectors;
performing full-connection coding on the classification feature vectors by using a full-connection layer of the classifier to obtain coded classification feature vectors;
and the coding classification feature vector is passed through a Softmax classification function of the classifier to obtain the classification result.
8. A photovoltaic module processing system, comprising:
The photovoltaic image acquisition module is used for acquiring a near infrared image of the photovoltaic cell panel assembly;
the image preprocessing module is used for preprocessing the near infrared image of the photovoltaic cell panel assembly to obtain a near infrared image of the photovoltaic cell panel assembly after preprocessing, wherein the preprocessing comprises image denoising, contrast enhancement and brightness adjustment;
the space feature extraction module is used for obtaining a space enhancement photovoltaic module feature map through a first convolution neural network model of the space attention of the near infrared image of the preprocessed photovoltaic cell panel module;
the channel characteristic extraction module is used for obtaining a channel enhancement photovoltaic module characteristic diagram through a second convolution neural network model of channel attention from the near infrared image of the pretreated photovoltaic module;
the fusion module is used for fusing the space enhancement photovoltaic module feature map and the channel enhancement photovoltaic module feature map to obtain a fusion feature map;
the global extraction module is used for enabling the fusion feature images to pass through a non-local neural network model to obtain classification feature images;
the optimizing module is used for carrying out order priori on the classification characteristic map so as to obtain an optimized classification characteristic map;
And the defect judging 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 photovoltaic cell panel assembly has defects or not.
9. The photovoltaic module processing system of claim 8, wherein the spatial feature extraction module is configured to:
performing depth convolution coding on the preprocessed near infrared image of the photovoltaic cell panel assembly by using a convolution coding part of the first convolution neural network model to obtain an initial convolution feature map;
inputting the initial convolution feature map into a spatial attention portion of the convolution neural network model to obtain a spatial attention map;
-passing said spatial attention map through a Softmax activation function to obtain a spatial attention profile;
and calculating the position-based point multiplication of the spatial attention characteristic diagram and the initial convolution characteristic diagram to obtain the spatial enhancement photovoltaic module characteristic diagram.
10. The photovoltaic module processing system of claim 9, wherein the channel feature extraction module is configured to:
input data is processed in forward pass of layers using layers of the second convolutional neural network model:
Performing convolution processing based on a two-dimensional convolution kernel on the input data to generate a convolution feature map;
pooling the convolution feature map to generate a pooled feature map;
performing activation processing on the pooled feature map to generate an activated feature map;
calculating the global average value of each feature matrix of the activated feature map along the channel dimension to obtain a channel feature vector;
calculating the ratio of the eigenvalue of each position in the channel eigenvector relative to the weighted sum of the eigenvalues of all positions of the channel eigenvector to obtain a channel weighted eigenvector;
taking the characteristic values of each position of the channel weighted characteristic vector as weights to carry out point multiplication on the characteristic matrix of the activated characteristic map along the channel dimension to obtain a generated characteristic map;
the generated feature map output by the last layer of the second convolutional neural network model is the channel enhancement photovoltaic module feature map.
CN202311327137.3A 2023-10-13 2023-10-13 Photovoltaic module processing method and system Withdrawn CN117291900A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202311327137.3A CN117291900A (en) 2023-10-13 2023-10-13 Photovoltaic module processing method and system

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202311327137.3A CN117291900A (en) 2023-10-13 2023-10-13 Photovoltaic module processing method and system

Publications (1)

Publication Number Publication Date
CN117291900A true CN117291900A (en) 2023-12-26

Family

ID=89257024

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202311327137.3A Withdrawn CN117291900A (en) 2023-10-13 2023-10-13 Photovoltaic module processing method and system

Country Status (1)

Country Link
CN (1) CN117291900A (en)

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN117790353A (en) * 2024-02-27 2024-03-29 徐州太一世纪能源科技有限公司 EL detection system and EL detection method
CN118037279A (en) * 2024-04-12 2024-05-14 海宁昱天新能源科技有限公司 Automatic operation and maintenance management system and method for photovoltaic equipment based on computer vision

Cited By (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN117790353A (en) * 2024-02-27 2024-03-29 徐州太一世纪能源科技有限公司 EL detection system and EL detection method
CN117790353B (en) * 2024-02-27 2024-05-28 徐州太一世纪能源科技有限公司 EL detection system and EL detection method
CN118037279A (en) * 2024-04-12 2024-05-14 海宁昱天新能源科技有限公司 Automatic operation and maintenance management system and method for photovoltaic equipment based on computer vision

Similar Documents

Publication Publication Date Title
CN117291900A (en) Photovoltaic module processing method and system
CN114862814B (en) Solar panel defect detection method and system, storage medium and terminal
CN113516656B (en) Defect image data processing simulation method based on ACGAN and Cameralink cameras
CN117036271A (en) Production line quality monitoring method and system thereof
CN116858789A (en) Food safety detection system and method thereof
CN116030018A (en) Incoming material qualification inspection system and method for door processing
CN117115715A (en) Video anomaly detection method based on combination of stream reconstruction and frame prediction
CN116824256A (en) Visual inspection system and method for grinding machine
CN116843677A (en) Appearance quality detection system and method for sheet metal part
CN117218533A (en) Monitoring system and method for carbon sink of oil tea forest
CN117173154A (en) Online image detection system and method for glass bottle
CN114800229A (en) Double-surface double-glass surface polishing device and polishing method thereof
CN113554630A (en) Chip surface defect detection method, system, computer equipment and storage medium
CN117030129A (en) Paper cup on-line leakage detection method and system thereof
CN112733785A (en) Stability detection method of information equipment based on layer depth and receptive field
CN112164025B (en) Threaded connection defect detection method and device, electronic equipment and storage medium
CN116994049A (en) Full-automatic flat knitting machine and method thereof
CN116363136B (en) On-line screening method and system for automatic production of motor vehicle parts
CN116728162A (en) Automatic metal part processing method and system
CN113591771B (en) Training method and equipment for object detection model of multi-scene distribution room
CN116958697A (en) Image-based power safety inspection system and method thereof
CN116977129A (en) Intelligent property management system and method thereof
CN116468894A (en) Distance self-adaptive mask generation method for supervised learning of lithium battery pole piece
CN112380369B (en) Training method, device, equipment and storage medium of image retrieval model
CN116141178B (en) Processing system and method for semiconductor valve

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
WW01 Invention patent application withdrawn after publication
WW01 Invention patent application withdrawn after publication

Application publication date: 20231226