The Research And Application Of Surface Defect Detection Method Based On Image Processing For Solar Cell

Solar cells is the main carrier of solar power in solar energy-electrical energy conversion system. Its quality is one of the main impact of the solar power generation efficiency. Cells exposed to the air for a long time, will be affected by the sun, rain, temperature, air quality and other environmental factors. There will be a certain degree of damage, which reduces the efficiency of solar power generation. So we need to check the solar panels of solar power station, check the quality of the cell and surface defects.Then replace the damage of solar cell. At present, the solar cell chip surface defect inspection of artificial manner. The traditional manual inspection needs a lot of manpower and material resources. Inspection results are easy to be effected man’s subjective factors. And detection efficiency is low. Surface defect detection methods are mainly based on the electroluminescence imaging technology of defect detection, defect detection based on Fourier reconstruction techniques, defect detection based on gray level co-occurrence matrix texture segmentation, defect detection based on matrix recovery, recovery algorithm of matrix is the key to realize the defect detection. The current matrix recovery algorithm is mainly using matrix decomposition technique to image matrix transformation. Then get no defects of low rank matrix image and defective sparse matrix of image. First, we use the kernel norm and one norm to approximate the rank of a matrix and zero norm. It turns the non-deterministic polynomial problem into a convex optimization problem. Then, getting the decomposition matrix of the minimizing kernel norm and the one norm by using the Accelerated Gradient Approximation algorithm and the Non-exact Lagrange Multiplier method, which correspond to a low rank matrix and a sparse matrix. Finally, getting the defective battery chip image by inversing the decomposed sparse matrix in order to realize the defect detection. For the large computational shortcomings of the singular value decomposition at each iteration of the calculation process, which use the accelerating gradient approximation algorithms and inexact method of Lagrange multipliers to solve convex optimization problems, this paper presents two improved singular value decomposition method. One kind is to redefine the rules of implicit redirect PROPACK package part of the singular value decomposition, the other is to introduce the singular value decomposition (SVD) of a linear time algorithm. In this paper, the improved algorithm and the traditional two algorithms are used in solar cells surface defect detection, and in the process of solving the convex optimization problem with the two algorithms in different singular value decomposition, recording and comparing the running time, the number of iterations each parameter, the number of singular value decomposition. It is concluded that the improved algorithm not only can effectively detect the solar cell surface defects, but also has a higher efficiency than the traditional algorithm.