Multispectral Fusion Of Complex Surface Images Of Solar Cells

Being one of the most important forms of the energy,solar energy has become attractive to human beings due to its clean,safe,rich and renewable features.And the usage of solar cells is becoming more and more common.However,due to the variety of materials,complex manufacturing process and improper operation of operators,many solar cells defects are produced during production and transportation,which highly reduces the performance life of solar cells.At the same time,the existence of complex anisotropic random texture on its surface makes the existing machine vision detection methods to have poor anti-interference and low detection accuracy.Thus,in order to suppress the influence of complex lattice background on detection and recognition,and to highlight the characteristics of weak surface defects,three related algorithms and some evaluation indexes are proposed from the perspectives of multispectral image acquisition,multispectral image fusion,and image quality evaluation of solar cells,respectively.The specific research contents are as follows:(1)Firstly,an automatic multispectral image acquisition system for solar cells is designed based on CIE 1931 chromaticity theory.According to the calculated RGB three-channel ratio,RGB three-channel source is controlled by the digital controller to obtain the visible light of the desired bands.And high speed industrial camera is controlled by hard trigger to automatically acquire multispectral image when inputting rising edge signal of the desired bands.The system with stability,rapidity,accuracy and band adjustability provides convenience for image fusion of solar cells.(2)Secondly,a single-channel multispectral gray image saliency fusion algorithm is proposed.First,the spectral reflectance of the solar cell images is analyzed,and a set of multispectral images of a specific wavelength are acquired.Then the structural texture decomposition method is used to suppress the complex background.The multiscale decomposition is performed by the two-dimensional tensor empirical wavelet with empirical information,and the detail layers containing the defect information are obtained.The saliency analysis is performed to enhance the contrast of the larger weight,and resulting in a strong defect feature is generated.The experimental results show that the method can effectively suppress the interference effects of complex lattice background,highlight surface defects and obtain clear solar cell images.(3)Thirdly,two three-channel multispectral image fusion algorithms are proposed.Firstly,the differences of spectral information of various defect information is analyzed.A group of multispectral images of specific wavelengths are acquired,and the optimal fusion bands and the optimal fusion space are selected.Then,according to the different representation forms of the three-channel optimal fusion space of the multispectral defect image,each channel is processed separately.According to the hybrid weight and the gradient weight,empirical wavelet with multi-resolution characteristic is used for image fusion.The experimental results show that the method can effectively suppress the interference effects of complex lattice background,highlight surface defects and obtain clear solar cell images.(4)Finally,the validity of the proposed method is verified by a large number of experiments.According to the analysis of the results of multispectral fusion of solar cells,some objective quality evaluation indexes for multispectral fusion results of complex surface images of solar cells are selected,which better verifies the performance of the results,and provides better help for the subsequent defect detection and classification.