CN114022420B - Detection method for automatically identifying defects of photovoltaic cell EL (electro-luminescence) component - Google Patents
Detection method for automatically identifying defects of photovoltaic cell EL (electro-luminescence) component Download PDFInfo
- Publication number
- CN114022420B CN114022420B CN202111228565.1A CN202111228565A CN114022420B CN 114022420 B CN114022420 B CN 114022420B CN 202111228565 A CN202111228565 A CN 202111228565A CN 114022420 B CN114022420 B CN 114022420B
- Authority
- CN
- China
- Prior art keywords
- photovoltaic cell
- component
- network model
- defect
- image
- 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
- 230000007547 defect Effects 0.000 title claims abstract description 105
- 238000001514 detection method Methods 0.000 title claims abstract description 63
- 238000005401 electroluminescence Methods 0.000 title description 67
- 230000004927 fusion Effects 0.000 claims abstract description 26
- 238000000605 extraction Methods 0.000 claims abstract description 13
- 238000012545 processing Methods 0.000 claims abstract description 11
- 238000000034 method Methods 0.000 claims abstract description 10
- 238000012549 training Methods 0.000 claims description 35
- 238000012360 testing method Methods 0.000 claims description 22
- 238000007781 pre-processing Methods 0.000 claims description 15
- 238000010606 normalization Methods 0.000 claims description 13
- 230000006870 function Effects 0.000 claims description 9
- 238000010586 diagram Methods 0.000 claims description 7
- 230000002708 enhancing effect Effects 0.000 claims description 6
- 230000009467 reduction Effects 0.000 claims description 6
- 238000013519 translation Methods 0.000 claims description 5
- 230000004913 activation Effects 0.000 claims description 4
- 238000002372 labelling Methods 0.000 claims description 4
- 238000011176 pooling Methods 0.000 claims description 4
- 238000002360 preparation method Methods 0.000 claims description 4
- 238000012544 monitoring process Methods 0.000 claims description 3
- 238000007670 refining Methods 0.000 claims description 3
- 238000012935 Averaging Methods 0.000 claims 1
- 238000001914 filtration Methods 0.000 claims 1
- 230000007246 mechanism Effects 0.000 abstract description 7
- 238000005516 engineering process Methods 0.000 abstract description 5
- 238000003062 neural network model Methods 0.000 abstract description 5
- 230000008569 process Effects 0.000 abstract description 3
- 238000013135 deep learning Methods 0.000 abstract description 2
- 230000005764 inhibitory process Effects 0.000 abstract 1
- 230000000694 effects Effects 0.000 description 5
- 238000005520 cutting process Methods 0.000 description 3
- 230000002950 deficient Effects 0.000 description 3
- 238000011161 development Methods 0.000 description 3
- 238000004519 manufacturing process Methods 0.000 description 3
- 239000000969 carrier Substances 0.000 description 2
- 238000006243 chemical reaction Methods 0.000 description 2
- 238000005070 sampling Methods 0.000 description 2
- 229910000679 solder Inorganic materials 0.000 description 2
- 206010027146 Melanoderma Diseases 0.000 description 1
- 230000002776 aggregation Effects 0.000 description 1
- 238000004220 aggregation Methods 0.000 description 1
- 208000003464 asthenopia Diseases 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000013527 convolutional neural network Methods 0.000 description 1
- 229910021419 crystalline silicon Inorganic materials 0.000 description 1
- 230000007423 decrease Effects 0.000 description 1
- 230000001419 dependent effect Effects 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 238000003384 imaging method Methods 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 238000013507 mapping Methods 0.000 description 1
- 229910021421 monocrystalline silicon Inorganic materials 0.000 description 1
- 230000000877 morphologic effect Effects 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 230000035945 sensitivity Effects 0.000 description 1
- 239000000126 substance Substances 0.000 description 1
- 230000001629 suppression Effects 0.000 description 1
- 230000009897 systematic effect Effects 0.000 description 1
- 239000013598 vector 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
- G01—MEASURING; TESTING
- G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
- G01N21/00—Investigating or analysing materials by the use of optical means, i.e. using sub-millimetre waves, infrared, visible or ultraviolet light
- G01N21/84—Systems specially adapted for particular applications
- G01N21/88—Investigating the presence of flaws or contamination
- G01N21/8851—Scan or image signal processing specially adapted therefor, e.g. for scan signal adjustment, for detecting different kinds of defects, for compensating for structures, markings, edges
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/25—Fusion techniques
- G06F18/253—Fusion techniques of extracted features
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
- G06N3/084—Backpropagation, e.g. using gradient descent
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
- G01N21/00—Investigating or analysing materials by the use of optical means, i.e. using sub-millimetre waves, infrared, visible or ultraviolet light
- G01N21/84—Systems specially adapted for particular applications
- G01N21/88—Investigating the presence of flaws or contamination
- G01N21/8851—Scan or image signal processing specially adapted therefor, e.g. for scan signal adjustment, for detecting different kinds of defects, for compensating for structures, markings, edges
- G01N2021/8887—Scan or image signal processing specially adapted therefor, e.g. for scan signal adjustment, for detecting different kinds of defects, for compensating for structures, markings, edges based on image processing techniques
-
- 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
- G06T2207/20081—Training; Learning
-
- 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
- G06T2207/20—Special algorithmic details
- G06T2207/20212—Image combination
- G06T2207/20221—Image fusion; Image merging
-
- 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
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- General Physics & Mathematics (AREA)
- Life Sciences & Earth Sciences (AREA)
- General Engineering & Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Artificial Intelligence (AREA)
- Evolutionary Computation (AREA)
- General Health & Medical Sciences (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Evolutionary Biology (AREA)
- Bioinformatics & Computational Biology (AREA)
- Health & Medical Sciences (AREA)
- Computing Systems (AREA)
- Computational Linguistics (AREA)
- Biophysics (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Biomedical Technology (AREA)
- Molecular Biology (AREA)
- Analytical Chemistry (AREA)
- Biochemistry (AREA)
- Immunology (AREA)
- Pathology (AREA)
- Chemical & Material Sciences (AREA)
- Signal Processing (AREA)
- Quality & Reliability (AREA)
- Image Analysis (AREA)
- Investigating Materials By The Use Of Optical Means Adapted For Particular Applications (AREA)
Abstract
The invention discloses a detection method for automatically identifying defects of a photovoltaic cell EL component, which is based on an improved YOLO v5s neural network model, wherein a self-designed residual channel attention gating mechanism module suitable for detecting defects of the photovoltaic cell EL component is added into a feature fusion part in an original model, so that multi-scale feature fusion is guided, complex background features in a defect position area inhibition image are highlighted, and the identification capability of the defects of the photovoltaic cell EL component is effectively improved. The detection method combines the deep learning technology and the image processing technology, so that the inefficiency and uncertainty of the traditional manual feature extraction are avoided, and the detection process has stronger robustness; when the detection method is adopted to initially set the learning rate to be 0.001, the average accuracy of model classification reaches 94.1%, so that the detection precision is obviously improved, and the detection speed is improved.
Description
Technical Field
The invention belongs to the technical field of defect detection, and particularly relates to a detection method for automatically identifying defects of an EL component of a photovoltaic cell.
Background
The global energy crisis is a very concerned problem of governments of various countries, the traditional fossil energy is not only nonrenewable, but also brings great pollution to the earth, and the development of novel renewable energy is a global subject which has been continued for decades.
Solar energy is a big prop of new energy, and can be converted into heat energy, chemical energy and electric energy, wherein the theoretical basis of the conversion of solar energy into electric energy is the photovoltaic effect, which is discovered by the french scientist becker in 1839. Thereafter, in 1954, scientists in the bell laboratories in the united states made practical single crystal silicon solar cells for the first time. Through the development of nearly 70 years, the solar cell is applied to various fields from the top to aviation aerospace and the bottom to daily life, the production technology of the solar cell is also more and more mature, and the photovoltaic solar industry is also developing rapidly.
The photovoltaic cells can generate various defects in the production process, the defects can affect the photoelectric conversion rate of the cells and also determine the grading of the cells when the cells are sold, so that the defects of the cells are necessary to be detected. Electroluminescence (EL) is an important technique for battery imaging, and the principle is that a forward bias voltage is applied to a crystalline silicon solar battery, a large amount of unbalanced carriers are injected into the battery, the unbalanced carriers continuously and compositely emit light to form infrared light rays, and the infrared light rays are captured by an industrial camera and are displayed as images after being processed by a computer.
Various defects can occur during the production of photovoltaic cell EL modules. Depending on the size of the defect, the battery may be classified into different grades to meet the requirements of different scenes, and this process is still mostly dependent on manual determination, which has the following drawbacks: (1) The standard is not uniformly mastered by different quality inspectors. (2) EL element images have random defect locations and more types, and quality inspectors have not fully identified defects and often consider such failures. (3) Visual fatigue is easy to occur when a quality inspector looks at a large number of pictures for a long time, and erroneous judgment and omission judgment occur. (4) A qualified quality inspector needs to be trained in a very systematic way, and the cultivation period is long.
Therefore, the difficulty brought by the manual assessment method can be effectively improved and solved, and the method is a popular research work of professionals. Under the condition, a detection method capable of intelligently detecting defects of the photovoltaic cell EL component is developed, the overall quality of the photovoltaic cell EL component is improved, the overall cost of the photovoltaic industry is reduced, and therefore the overall development of the photovoltaic industry is promoted.
Disclosure of Invention
Aiming at the defects of the prior art, the invention aims to provide a detection method for automatically identifying defects of an EL component of a photovoltaic cell. The detection method is a photovoltaic cell EL component defect automatic identification detection method based on YOLOv s neural network model improvement, and has the advantages of high defect identification accuracy and high detection speed.
The technical scheme adopted by the invention for solving the technical problems is as follows: a detection method for automatically identifying defects of an EL component of a photovoltaic battery is designed, and is characterized by comprising the following steps:
The first step: image preprocessing
1) Acquiring a defect image library of the photovoltaic cell EL component: firstly, acquiring an original image of a photovoltaic cell EL component through an industrial camera, and then performing operations such as rotation, translation, contrast deepening and the like on the original image to obtain a pre-processed image; finally, performing size normalization operation on the original image and the preprocessed image, removing images without defects, and forming a defect image library of the photovoltaic cell EL component by the rest images;
2) Sample set data preparation: selecting 50-80% of the defects in the EL component defect image library of the photovoltaic cell as a training sample set at random, marking images in each sample set respectively for the rest of the test sample set, and adding defect type labels;
and a second step of: training of photovoltaic cell EL component defect detection network model based on improved Yolov s network model
1) Sample set preprocessing
Preprocessing a training sample set in a mode of enhancing Mosaic data;
2) Parameter setting
Initializing all weight values, bias values and batch normalization scale factor values, setting initial learning rate and batch_size of a network, inputting initialized parameter data into the network, and setting the initial learning rate of the network to be 0.001;
3) Network model training
Inputting the preprocessed training sample set into an improved Yolov s network model with well-set initialization parameters, wherein the network model is obtained by replacing a Concat module in the Yolov5s network model with a RCAG module; under the processing of the improved Yolov s network model, firstly extracting features, combining RCAG modules to perform multi-scale fusion, then refining and fusing feature graphs obtained by the last three layers of extraction through a convolution layer to reduce dimensionality, and predicting tensors by utilizing a classification regression network, wherein the tensors comprise position, confidence and classification score prediction values; comparing the generated predicted value with the label information to generate a loss value, then carrying out back propagation, updating parameters of a backbone network and a classified regression network until the loss value accords with the preset value, and completing training of network model parameters;
4) Network model testing
Inputting the test sample set into the network model with parameter training completed in the step 3), and obtaining a tensor predicted value of the test sample set; comparing the tensor predicted value with the labeling information, testing the reliability of the network model, and monitoring whether the model is fitted or not so as to determine whether the training is stopped and the parameters are readjusted;
And a third step of: defect detection for photovoltaic cell EL component
And (3) carrying out size normalization operation on the image of the photovoltaic cell EL component to be detected in the step (1) as the first step, and then inputting the image into a defect detection network model of the photovoltaic cell EL component which is tested to be reliable in the step (II) to obtain defect tensor information of the image of the photovoltaic cell EL component to be detected, wherein the defect tensor information comprises defect positions, defect types and confidence.
Compared with the prior art, the invention has the beneficial effects that:
The detection method is based on an improved YOLO v5s neural network model, a self-designed residual channel attention gating mechanism module suitable for defect detection of the photovoltaic cell EL component is added to a feature fusion part in an original model, multi-scale feature fusion is guided, complex background features in a defect position area suppression image are highlighted, and the identification capability of the photovoltaic cell EL component defect is effectively improved. The detection method combines the deep learning technology and the image processing technology, so that the inefficiency and uncertainty of the traditional manual feature extraction are avoided, and the detection process has stronger robustness; the original YOLOv s network model is adopted to detect defects of the photovoltaic cell EL component, the average recognition rate is 90.2%, and when the initial learning rate is set to be 0.001, the model classification average accuracy rate reaches 94.1%, so that the detection precision is obviously improved, and the detection speed is improved.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required in the description of the embodiments or the prior art will be briefly described below, it being obvious that the drawings in the following description are only some embodiments of the present invention, and that other drawings may be obtained according to these drawings without invasive efforts to those skilled in the art.
FIG. 1 is a flow chart of an embodiment of the detection method of the present invention.
FIG. 2 is a diagram showing a network model structure for detecting defects of a photovoltaic cell EL element based on a modified Yolov s network model according to an embodiment of the detection method of the present invention.
FIG. 3 is a schematic diagram of the operation of the Focus module of the photovoltaic cell EL element defect detection network model based on the improved Yolov s network model according to one embodiment of the detection method of the present invention.
FIG. 4 is a schematic diagram of the Mosaic enhancement of a photovoltaic cell EL component defect detection network model based on a modified Yolov s network model in accordance with one embodiment of the detection method of the present invention.
Fig. 5 is a schematic structural diagram of a csp1_x module of a photovoltaic cell EL member defect detection network model based on a modified Yolov s network model according to one embodiment of the detection method of the present invention.
Fig. 6 is a schematic structural diagram of a csp2_x module of a photovoltaic cell EL member defect detection network model based on a modified Yolov s network model according to one embodiment of the detection method of the present invention.
FIG. 7 is a schematic diagram of a residual channel attention gating mechanism module of a photovoltaic cell EL element defect detection network model based on a modified Yolov s network model according to one embodiment of the detection method of the present invention.
Fig. 8 is an image of a photovoltaic cell EL assembly to be inspected.
Fig. 9 is a graph of the result of detecting the image of fig. 8 using the detection method of the present invention.
Detailed Description
The following description of the embodiments of the present invention will be made with reference to the accompanying drawings, in which it is evident that the embodiments described are only some embodiments of the present invention, but not all embodiments. All other embodiments, which can be made by those skilled in the art based on the embodiments of the invention without making any inventive effort, are intended to be within the scope of the invention.
The invention relates to a detection method for automatically identifying defects of a photovoltaic cell EL component, which is used for detecting the photovoltaic cell EL component and comprises the following steps:
The first step: image preprocessing
1) Acquiring a defect image library of the photovoltaic cell EL component: firstly, acquiring an original image (including defective and nondefective images) of a photovoltaic cell EL component through an industrial camera, and then performing operations such as rotation, translation, contrast deepening and the like on the original image to obtain a preprocessed image; finally, performing size normalization operation on the original image and the preprocessed image, removing images without defects, and forming a defect image library of the photovoltaic cell EL component by the rest images;
2) Sample set data preparation: selecting 50-80% of the defects in the EL component defect image library of the photovoltaic cell as a training sample set at random, marking images in each sample set respectively for the rest of the test sample set, and adding defect type labels;
second step, training of photovoltaic cell EL component defect detection network model based on improved Yolov s network model
3) Sample set preprocessing
And preprocessing the training sample set by adopting a mode of enhancing Mosaic data.
The mosaics data enhancement adopts a mode of random scaling, random cutting and random arrangement to splice images, and has good detection effect on small targets. The original input image resolution is then adjusted, adaptively scaling the picture according to the aspect ratio of the picture.
4) Parameter setting
Initializing all weight values, bias values and batch normalization scale factor values, setting initial learning rate and batch_size (batch processing parameters) of a network, inputting initialized parameter data into the network, and setting the initial learning rate of the network to be 0.001;
3) Network model training
Inputting the preprocessed training sample set into an improved Yolov s network model with well-set initialization parameters, wherein the network model is obtained by replacing a Concat module (a splicing module) in the Yolov s network model with a RCAG module (a residual channel attention gating mechanism module); under the processing of the improved Yolov s network model, the preprocessed training sample set is firstly subjected to feature extraction and multi-scale fusion by combining with a RCAG module, then feature images obtained by the last three layers of extraction are refined, fused and dimension-reduced through a convolution layer, and the tensor is predicted by utilizing a classification regression network, wherein the tensor comprises a position, a confidence coefficient and a classification score predicted value. And comparing the generated predicted value with the label information to generate a loss value, and then carrying out back propagation to update parameters of the backbone network and the classified regression network until the loss value accords with the preset, and completing the training of the network model parameters.
4) Network model testing
Inputting the test sample set into the network model with parameter training completed in the step 3), and obtaining a tensor predicted value of the test sample set; and comparing the tensor predicted value with the labeling information, and testing the reliability of the network model, wherein the reliability is used for monitoring whether the model is subjected to fitting so as to determine whether training is required to be stopped and parameters are readjusted. And when the accuracy of the tensor predicted value is more than 90%, the network model is considered to be reliable.
And a third step of: defect detection for photovoltaic cell EL component
And (3) carrying out size normalization operation on the image of the photovoltaic cell EL component to be detected in the step (1) as the first step, and then inputting the image into a defect detection network model of the photovoltaic cell EL component which is tested to be reliable in the step (II) to obtain defect tensor information of the image of the photovoltaic cell EL component to be detected, wherein the defect tensor information comprises defect positions, defect types and confidence.
Example 1
The detection method for automatically identifying and detecting defects of the photovoltaic cell EL component is used for detecting the photovoltaic cell EL component and comprises the following steps:
The first step: image preprocessing
1) Acquiring a defect image library of the photovoltaic cell EL component: firstly, acquiring an original image (including defective and nondefective images) of a photovoltaic cell EL component through an industrial camera, and then performing operations such as rotation, translation, contrast deepening and the like on the original image to obtain a preprocessed image; finally, performing size normalization operation on the original image and the preprocessed image, removing images without defects, and forming a defect image library of the photovoltaic cell EL component by the rest images;
Because the number of the defect images acquired by normal shooting is obviously less than that of the normal images, the acquired photovoltaic cell EL component images are subjected to pretreatment operations such as rotation, translation, contrast deepening and the like so as to expand a photovoltaic cell EL component database; performing size normalization operation on all collected photovoltaic cell EL component images and the photovoltaic cell EL component images obtained by preprocessing to obtain uniform-scale images, and finally removing unnecessary parts (without defects) in the images by adopting regional morphological treatment to form a photovoltaic cell EL component defect image library;
The original picture size used in this embodiment is 3600×618, and the picture size in the training is 641×589; since the original picture has the picture size of 3600×618 and the pixel value is too large, each picture is cut into 6 pictures, the size of the cut picture is 641×589, the images without defects are removed, and finally a defect image library of the photovoltaic cell EL component with the number of 7440 pictures is obtained.
2) Sample set data preparation: manually and randomly selecting 80% of the images in the defect image library of the photovoltaic cell EL component as a training sample set, wherein the rest 20% of the images are test sample sets, respectively marking the images in each sample set, and adding defect type labels;
The marked object is an image defect area (including black spots, virtual welding, hidden cracks, broken grids and linear defects), and LabelImg is manually used for marking the defect area; in the embodiment, the training sample set and the test sample set are selected from the 7440 defective pictures, wherein, the black spots 1488 are marked as heiban; the dummy solder 1488 pieces, labeled xuhan; the hidden crack 1488 sheets, marked yinlie; broken gate 1488 sheets, labeled duanshan; linear defect 1488 is labeled xianzhuangquexian.
Second step, training of photovoltaic cell EL component defect detection network model based on improved Yolov s network model
1) Sample set preprocessing
And preprocessing the training sample set by adopting a mode of enhancing Mosaic data.
The mosaics data enhancement adopts a mode of random scaling, random cutting and random arrangement to splice images, and has good detection effect on small targets. The original input image resolution is then adjusted, adaptively scaling the picture according to the aspect ratio of the picture.
2) Parameter setting
Initializing all weight values, bias values and batch normalization scale factor values, setting the initial learning rate and batch_size of the network to 64, inputting initialized parameter data into the network, and setting the initial learning rate of the network to 0.001;
The model parameters are initialized as follows: the maximum iteration number (epoch) is set to 200, the first 50 epoch learning rates are set to 0.001, the last 150 epoch learning rates are set to 0.0001, the decline factor of the learning rate is 0.1, and the weight decay of the regularization term is 0.0005.
3) Network model training
Inputting the preprocessed training sample set into an improved Yolov s network model with well-set initialization parameters, wherein the network model is obtained by replacing a Concat module (a splicing module) in the Yolov s network model with a RCAG module (a residual channel attention gating mechanism module); under the processing of the improved Yolov s network model, the preprocessed training sample set is firstly subjected to feature extraction and multi-scale fusion by combining with a RCAG module, then feature images obtained by the last three layers of extraction are refined, fused and dimension-reduced through a convolution layer, and the tensor is predicted by utilizing a classification regression network, wherein the tensor comprises position, defect type and confidence degree predicted values. And comparing the generated predicted value with the label information to generate a loss value, and then carrying out back propagation to update parameters of the backbone network and the classified regression network until the loss value accords with the preset, and completing the training of the network model parameters.
4) Network model testing
Preprocessing a test sample set in the manner of enhancing the Mosaic data in the step 1), and inputting the preprocessed test sample set into a network model which is subjected to parameter training in the step 3) to obtain a tensor predicted value of the test sample set; and comparing the tensor predicted value with the labeling information to obtain single-type defects with accuracy more than 90%, and confirming that the network model is an effective model by detailed table 1.
TABLE 1 detection Effect
| Defect type | Black spot | Dummy solder joint | Hidden crack | Broken grid | Linear defect |
| Accuracy rate of | 96.2％ | 94.3％ | 91.6％ | 93.4％ | 91.0％ |
And a third step of: defect detection for photovoltaic cell EL component
And (3) carrying out size normalization operation on the image of the photovoltaic cell EL component to be detected in the step 1) in the first step, preprocessing the image in the manner of enhancing the Mosaic data in the step 1) in the second step, and inputting the preprocessed image into a reliable defect detection network model of the photovoltaic cell EL component tested in the step two to obtain defect tensor information of the image of the photovoltaic cell EL component to be detected, wherein the defect tensor information comprises defect positions, defect types and confidence.
The defect detection method adopts Yolov s neural network model of improved version, has high detection precision and high identification speed: specifically, the neural network model structure is shown in fig. 3, and mainly comprises a data enhancement part, a feature extraction part, a feature fusion part and a classification regression part. The data enhancement part adopts Mosaic data enhancement, and the photovoltaic cell EL component images are spliced in a random scaling, random cutting and random arrangement mode, so that the detection capability of small targets is improved. Feature extraction aspects input the photovoltaic cell EL assembly image into a Focus structure and CSPDARKNET network, through convolution, pooling, etc., and extract a feature map therefrom that can be shared for subsequent feature fusion portions. And the feature fusion part adopts FPN (feature pyramid) and PAN (pyramid attention network structure) to generate pooled feature vectors with different fixed sizes, strengthen the expression capability of features and have good effect on the detection of the same object with different sizes. The classification regression part uses GIOU _los as a Loss function of Bounding box, so that the problem of non-coincidence of boundary boxes is effectively solved, and the speed and the accuracy of regression of a prediction box are improved.
The feature extraction stage of YOLOv s adopts a Focus and CSPDARKNET53 structure, an original 640 x3 image is input into the Focus structure, a slicing operation is adopted, a feature map of 320 x 12 is firstly changed, and then a convolution operation of 32 convolution kernels is carried out, so that a feature map of 320 x 32 is finally changed. CSPDARKNET53 is that a CSP module (cross-phase local module) is added on the basis of Darket53, and the CSP module is composed of a convolutional layer and a residual structure Concate in a Resnet network. Darket53 total 53 convolutions of 53 layers, the last FC (fully connected layer, actually realized by a 1x1 convolution) is removed for a total of 52 convolutions to be used as the principal network. The 52 convolutional layers are composed of: first, a convolution kernel of 1 filter and then 5 sets of repeated residual units resblock _body (the 5 sets of residual units, each consisting of 1 single convolution layer and a set of repeatedly performed convolution layers, the repeatedly performed convolution layers being repeated 1,2, 8, 4 times, respectively; in each repeatedly performed convolution layer, a convolution operation of 1x1 is performed first, a convolution operation of 3x3 is performed, the number of filters is halved first, and then restored), and the total is 52 layers.
The CSPDARKNET is based on Darknet53 using a CSP module, and two structures of csp1_x and csp2_x are designed in Yolov s, wherein the feature extraction part comprises the csp1_x module, and the feature fusion part comprises the csp2_x module. The convolution kernel in front of each CSP module is 3*3 in size, the step length is 2, and the function of downsampling can be achieved. The backbone network adopts a CSP1_1 structure and a CSP1_3 structure, the neck adopts a CSP2_1 structure, the CSP_x divides the input into two parts, one part firstly carries out x times of residual operations and then carries out convolution operation, the other part directly carries out convolution operation, the purpose of the convolution operation is to reduce the number of channels by half, and then the two parts are spliced again and then output to strengthen the feature fusion between the networks, so that the learning capability of the CNN is enhanced, and the accuracy is maintained while the weight is reduced.
RCAG module (residual channel attention gating mechanism module) design: aiming at the feature fusion part of low-level features and high-level features in the original network structure, the low-level features contain rich position information, and the high-level features contain rich semantic information, but are equally treated among channels in the fusion, so that the representation capability of the network is blocked. The RCAG module designed by the invention carries out up-sampling on low-level features and then adds the low-level features with the high-level features, strengthens the fusion of the features through a convolution operation, predicts the importance of each channel through global average pooling operation and a full connection layer, obtains the importance of different channels, and can improve the sensitivity of the model to the channel features by multiplying the weight values to the reinforced features. Then, short connection in the residual structure is added on the basis, so that richer characteristic information can be allowed to be directly transmitted backwards through identity mapping, and the information flow can be ensured. The residual channel attention gating mechanism is used to replace the stitching operation in the FPN (feature pyramid) module and the PAN (path aggregation) module, and can guide multi-scale feature fusion and highlight complex background features in the defect location area suppressed image.
The working principle of RCAG modules is as follows: the low-level feature p after up-sampling and the high-level feature q are respectively converted into two feature spaces f and g through a convolution kernel of 1 multiplied by 1, the fusion feature S is obtained after element summation and ReLu function activation, and then the fusion feature S is filtered through convolution operation. Extracting global features by using a Global Average Pooling (GAP) layer, realizing dimension reduction, inputting the global features into a multi-layer perceptron (MLP) layer for processing, and then generating a channel level attention graph A by using a sigmoid function for processing; the channel level attention map A is multiplied by the fusion feature S to carry out feature re-weighting to obtain a weighted feature, and then the weighted feature and the fusion feature S are subjected to pixel addition through a residual error connection operation to obtain a final output feature. The MLP mainly comprises two full-connection layers (the first layer is provided with C/r channels, r is a reduction ratio, the dimension reduction operation is realized, the second layer is provided with C channels) and a ReLu activation function, the two full-connection layers aim at modeling the relation between the channels, the object can be highlighted and the complex background can be restrained, the ReLu function is used for refining global features, the MLP is used for encoding fusion features S, and the dependence between the channels is learned.
In the third step of this embodiment, a group of images of the photovoltaic cell EL assembly to be detected is shown in fig. 9, and the obtained detection structures are shown in fig. 9, heiban, xuhan, yinlie, duanshan, xianzhuangquexian are defect types, the frame represents the defect position, the defect position is marked by the frame accurately, the number on the frame represents the confidence of the detection result, the maximum value is 1, and the higher the number value is, the more accurate the detection result is.
The test result shows that the single type accuracy rate of the defect detection of the photovoltaic cell EL component exceeds 90% when the method is applied to the defect detection of the photovoltaic cell EL component by adopting an improved Yolov s model, the average accuracy rate can reach 94.1%, and the rapid and intelligent detection of the photovoltaic cell EL component can be realized.
The embodiments of the present invention have been described above with reference to the accompanying drawings, but the present invention is not limited to the above-described embodiments, which are merely illustrative and not restrictive, and many forms may be made by those having ordinary skill in the art without departing from the spirit of the present invention and the scope of the protection of the claims, which fall within the protection of the present invention.
The invention is applicable to the prior art where it is not described.
Claims (3)
1. A detection method for automatically identifying defects of an EL assembly of a photovoltaic cell, comprising the steps of:
The first step: image preprocessing
1) Acquiring a defect image library of the photovoltaic cell EL component: firstly, acquiring an original image of a photovoltaic cell EL component through an industrial camera, and then performing rotation, translation and contrast deepening operation on the original image to obtain a preprocessed image; finally, performing size normalization operation on the original image and the preprocessed image, removing images without defects, and forming a defect image library of the photovoltaic cell EL component by the rest images;
2) Sample set data preparation: randomly selecting 50-80% of the defects in the photovoltaic cell EL component image library as training sample sets, wherein the rest are test sample sets, respectively marking the images in each sample set, and adding defect type labels;
And a second step of: training of photovoltaic cell EL component defect detection network model based on improved Yolov s network model
1) Sample set preprocessing
Preprocessing a training sample set in a mode of enhancing Mosaic data;
2) Parameter setting
Initializing all weight values, bias values and batch normalization scale factor values, setting the initial learning rate and batch_size of the network, inputting initialized parameter data into the network, and setting the initial learning rate of the network to be 0.001;
3) Network model training
Inputting the preprocessed training sample set into an improved Yolov s network model with well-set initialization parameters, wherein the network model is obtained by replacing a Concat module in the Yolov5s network model with a RCAG module; under the processing of the improved Yolov s network model, firstly, carrying out feature extraction and multi-scale fusion by combining with a RCAG module, then, refining and fusing the feature images obtained by the last three layers of extraction through a convolution layer to reduce dimensionality, and predicting tensors by utilizing a classification regression network, wherein the tensors comprise position, confidence and classification score prediction values; comparing the generated predicted value with the label information to generate a loss value, then carrying out back propagation, updating parameters of a backbone network and a classified regression network until the loss value accords with the preset, and completing training of network model parameters;
the working principle of RCAG modules is as follows: the method comprises the steps of converting up-sampled low-level features and high-level features into two feature spaces through a convolution kernel of 1 multiplied by 1 respectively, obtaining fusion features after element summation and ReLu function activation, and then filtering the fusion features through convolution operation; extracting global features by using a global averaging pooling layer, realizing dimension reduction work, inputting the global features into a multi-layer perceptron layer for processing, and then generating a channel level attention diagram through sigmoid function processing; the channel level attention is multiplied by the fusion feature to re-weight the feature to obtain a weighted feature, and then the weighted feature and the fusion feature are subjected to pixel addition through a residual error connection operation to obtain a final output feature;
4) Network model testing
Inputting the test sample set into the network model with parameter training completed in the step 3) to obtain a tensor predicted value of the test sample set; comparing the tensor predicted value with the labeling information, testing the reliability of the network model, and monitoring whether the model is fitted or not so as to determine whether the training is stopped and readjusting parameters are needed;
And a third step of: defect detection for photovoltaic cell EL component
And (3) carrying out the same size normalization operation as in the step (1) in the first step on the photovoltaic cell EL component image to be detected, and then inputting the photovoltaic cell EL component image to be detected into a photovoltaic cell EL component defect detection network model which is tested to be reliable in the step (II) to obtain defect tensor information of the photovoltaic cell EL component image to be detected, wherein the defect tensor information comprises defect positions, defect types and confidence degrees.
2. The method according to claim 1, wherein in the second step, step 4), the network model is considered to be reliable when the accuracy of the tensor predicted value is greater than 90%.
3. The method for automatically identifying defects in an EL device of a photovoltaic cell according to claim 1, wherein the multi-layer sensor layer comprises two fully-connected layers and a ReLu activation function, the first fully-connected layer has C/r channels, r is a reduction ratio, a dimension reduction operation is implemented, and the second fully-connected layer has C channels.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111228565.1A CN114022420B (en) | 2021-10-21 | 2021-10-21 | Detection method for automatically identifying defects of photovoltaic cell EL (electro-luminescence) component |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111228565.1A CN114022420B (en) | 2021-10-21 | 2021-10-21 | Detection method for automatically identifying defects of photovoltaic cell EL (electro-luminescence) component |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN114022420A CN114022420A (en) | 2022-02-08 |
| CN114022420B true CN114022420B (en) | 2024-05-31 |
Family
ID=80057050
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202111228565.1A Active CN114022420B (en) | 2021-10-21 | 2021-10-21 | Detection method for automatically identifying defects of photovoltaic cell EL (electro-luminescence) component |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114022420B (en) |
Families Citing this family (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114580568A (en) * | 2022-03-24 | 2022-06-03 | 华南理工大学 | Fish species identification method based on deep learning |
| CN114594103B (en) * | 2022-04-12 | 2023-05-16 | 四川大学 | Automatic detection and report generation method and system for surface defects of nuclear industrial equipment |
| CN114529546A (en) * | 2022-04-24 | 2022-05-24 | 科大天工智能装备技术（天津）有限公司 | Roof panel defect detection method and system |
| CN114820609A (en) * | 2022-06-29 | 2022-07-29 | 南昌大学 | Photovoltaic module EL image defect detection method |
| CN115144406B (en) * | 2022-09-07 | 2022-11-11 | 广州珐玛珈智能设备股份有限公司 | Omnibearing online industrial visual detection and screening device and method |
| CN115754108B (en) * | 2022-11-23 | 2023-06-09 | 福建省杭氟电子材料有限公司 | Acidity determination system and method for electronic grade hexafluorobutadiene |
| CN116542980B (en) * | 2023-07-06 | 2023-11-03 | 宁德时代新能源科技股份有限公司 | Defect detection method, defect detection apparatus, defect detection program, storage medium, and defect detection program |
| CN116912605A (en) * | 2023-09-12 | 2023-10-20 | 法施达(天津)智能科技有限公司 | Fault detection method and system based on anchor bolt image data feature recognition |
| CN118037740A (en) * | 2024-04-15 | 2024-05-14 | 宁德时代新能源科技股份有限公司 | Pole piece detection method, device, equipment, storage medium and computer program product |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110570410A (en) * | 2019-09-05 | 2019-12-13 | 河北工业大学 | Detection method for automatically identifying and detecting weld defects |
| CN112906794A (en) * | 2021-02-22 | 2021-06-04 | 珠海格力电器股份有限公司 | Target detection method, device, storage medium and terminal |
| WO2021139069A1 (en) * | 2020-01-09 | 2021-07-15 | 南京信息工程大学 | General target detection method for adaptive attention guidance mechanism |
-
2021
- 2021-10-21 CN CN202111228565.1A patent/CN114022420B/en active Active
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110570410A (en) * | 2019-09-05 | 2019-12-13 | 河北工业大学 | Detection method for automatically identifying and detecting weld defects |
| WO2021139069A1 (en) * | 2020-01-09 | 2021-07-15 | 南京信息工程大学 | General target detection method for adaptive attention guidance mechanism |
| CN112906794A (en) * | 2021-02-22 | 2021-06-04 | 珠海格力电器股份有限公司 | Target detection method, device, storage medium and terminal |
Non-Patent Citations (1)
| Title |
|---|
| 改进的CNN在太阳能电池板缺陷检测中的应用;周颖;毛立;张燕;陈海永;;计算机仿真;20200315(第03期);全文 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN114022420A (en) | 2022-02-08 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN114022420B (en) | Detection method for automatically identifying defects of photovoltaic cell EL (electro-luminescence) component | |
| Akram et al. | CNN based automatic detection of photovoltaic cell defects in electroluminescence images | |
| CN114973032B (en) | Deep convolutional neural network-based photovoltaic panel hot spot detection method and device | |
| Yahya et al. | Applied imagery pattern recognition for photovoltaic modules’ inspection: A review on methods, challenges and future development | |
| CN113837994B (en) | Photovoltaic panel defect diagnosis method based on edge detection convolutional neural network | |
| Chindarkkar et al. | Deep learning based detection of cracks in electroluminescence images of fielded PV modules | |
| CN115984543A (en) | Target detection algorithm based on infrared and visible light images | |
| Montanez et al. | Photovoltaic module segmentation and thermal analysis tool from thermal images | |
| Demirci et al. | Defective PV cell detection using deep transfer learning and EL imaging | |
| CN111612803B (en) | Vehicle image semantic segmentation method based on image definition | |
| CN116485802A (en) | Insulator flashover defect detection method, device, equipment and storage medium | |
| CN116051530A (en) | Semi-supervised photovoltaic cell surface anomaly detection method based on image restoration | |
| Chen et al. | Automatic crack segmentation and feature extraction in electroluminescence images of solar modules | |
| CN114463686B (en) | Moving target detection method and system based on complex background | |
| Ruan et al. | Deep learning-based method for PV panels segmentation and defects detection with infrared images | |
| Huang et al. | Fault diagnosis for solar panels using convolutional neural network | |
| CN114612468A (en) | Equipment external defect detection method based on positive sample | |
| Kaligambe et al. | A Deep Learning-Based Framework for Automatic Detection of Defective Solar Photovoltaic Cells in Electroluminescence Images Using Transfer Learning | |
| Vo | Realization and Verification of Deep Learning Models for Fault Detection and Diagnosis of Photovoltaic Modules | |
| CN113780434B (en) | Deep learning-based solar cell module defect EL detection method | |
| Binomairah et al. | Detection of microcracks and dark spots in monocrystalline PERC cells using photoluminescene imaging and YOLO-based CNN with spatial pyramid pooling | |
| CN115375677B (en) | Wine bottle defect detection method and system based on multi-path and multi-scale feature fusion | |
| Tang et al. | Deep learning based model for Defect Detection of Mono-Crystalline-Si Solar PV Module Cells in Electroluminescence Images Using Data Augmentation | |
| Wang et al. | Accurate detection and intelligent classification of solar cells defects based on photoluminescence images: A novel study on the optimized YOLOv5 model | |
| Xi et al. | Defect Detection Algorithm of Photovoltaic Module EL Image Based on Two-stage YOLOts |
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 |