Detection Of Hidden Defects Of Crystalline Silicon Photovoltaic Cells Based On Image Reconstruction

As the global energy crisis and environmental pollution continue to intensify,green and sustainable development of energy has attracted more and more attention.Solar energy is an attractive alternative energy source,and solar photovoltaic cells are the foundation of solar power.generation systems.Therefore,ensuring high-quality production of solar photovoltaic cells is of great significance for environmental protection and alleviating energy shortages.The solar photovoltaic cells in the production process will produce many defects.Due to the complex types of defects,it has brought many difficulties to the quality inspection of products in the production process of solar photovoltaic cells.In order to realize the automatic detection of battery quality,this paper takes photovoltaic cell images after optical imaging as the research object,and uses image processing methods to detect battery defects.To this end,we studied the hidden defect detection scheme of crystalline silicon photovoltaic cells for image reconstruction.Firstly,th e independent component analysis algorithm was used to classify the images with and without defects in the solar photovoltaic cell images,and the separated defect images were further used independently The component analysis method detects the specific location of the defect,and then for images with more complex defect types,the background of the image is reconstructed by the case deletion model,and the defect is detected by comparing the difference between the image to be inspected and the background.The main contents of th is article are:(1)A method of solar photovoltaic cell image classification and defect detection based on independent component analysis algorithm is proposed.Independent component analysis algorithms are widely used in many fields such as signal processing,image denoising,and fault diagnosis.In order to realize the classification of solar photovoltaic cell images and the detection of defect positions,a set of non-defective images is trained to obtain a set of image bases according to the independent component analysis algorithm,so that any image to be inspected can get a set of coefficients through this set of base representations,Use the obtained coefficients to estimate the background of the image to be inspected.In the image classification stage,first use this set of image bases to represent the training image and take the average of the obtained coefficients.According to the average of the coefficients,a most representative defect-free background image can be estimated,and then any image to be inspected can be calculated.With the residual value of the reconstructed background,the residual value of each image to be inspected is classified by a predetermined threshold,so that the solar photovoltaic cell image is divided into two types with defects and without defects.In the defect detection stage,the defect image obtained during the classification process is used to construct the background of the hidden defect image of the photovoltaic cell using the trained image base,and the defect area is subtracted by the subtraction of the image to be tested and the background image to obtain the defect.Binary image of position and shape.In this way,the independent component analysis algorithm is used to realize the classification of hidden defect images and the detection of defect positions in photovoltaic cells.Among them,the classification accuracy rate of 98 images(38 defect-free and 60 defects)in the classification process reached 93.2%,and the defect detection rate of 223 images with multiple defect types in the detection stage reached 95.79%.(2)In order to detect hidden defects in photovoltaic cells without training samples,a defect detection method based on block case deletion model is proposed.First,a nonlinear regression model between the gray value and coordinates of the solar photovoltaic cell image is established to obtain the model coefficients.Then,the image is divided into blocks,and each time a block is deleted,a nonlinear regression model between the deleted gray value and its coordinates is established,and the deleted coefficient is calculated.By calculating the Cook distance between coefficients before and after block deletion and using the above cutoff point as the threshold,the image blocks corresponding to all abnormal coefficients are screened out,and the background of the reconstructed image is removed after removing all abnormal image blocks found.Finally,the difference between the original image and the obtained background image is used to highlight the defect area,so as to achieve the purpose of defect detection.Experimental results show that the proposed method can effectively detect many types of defects such as cracks,broken grids and debris in solar photovoltaic cells.This method has a stable detection effect on solar photovoltaic cell images without defects.This method was used to experiment on 313 solar photovoltaic cell images.Among them,158 non-defective images did not detect defects,while the other 155 images contained cracks,only 5 images of defects such as broken grids were misdetected,and the defect detection rate reached 96.77%.