The Research And Detection Of The Solar Cells Defect

As one kind of clean energy,development and utilization of solar cells achieve a huge contribution to environmental protection.It may appear many defects in the manufacturing process of solar cells,accordingly reduce the practical utility.For guarantee the quality of solar cells,the defects of solar cells need to be detected and analyzed.It's necessary to realize the automatic classification and recognition in the real industrial because of manual inspection's low efficiency and long period.In this paper,the detection and classification algorithms of several defects which may occur in the process of solar cell production are studied.The main research work contains some aspects as following:(1)According to the gathered images may appear uneven illumination,blurring,geometric distortion and so on,because of environment,lens,nonlinear of imaging system.In order to ensure the consistency of images and facilitate the later detect defections,the image filtering,hat transform,edge detection,morphological changes and etc were studied.(2)To realize the maximum no damaged of re-cut silicon wafer,detect the defection of corner and edge breakage in the cutting stage.First,use the new line segment detector TSAP to edge line detect the silicon wafer after preprocessing and use difference image method to realize defect detection.Then use improved perspective correction algorithm correct distortion defective images to the angle of elevation view.At last,find the maximum non damaged rectangle to realize the maximum no damaged of cut silicon wafer.The accuracy of the algorithm is 98.87%.(3)Detect the gray and texture defects that appeared in the screen printing stage and achieve the classification of defects.Gray defects are mainly stains,holes and cracks.Texture defects are mainly silk screen process.Aiming at the defects in this stage,the author uses the top-hat transformation and morphological corrosion to remove the influence of illumination and noise.Combination of image segmentation algorithm and projection method to detect the target defection.The last the characteristic parameters towards the target region,majorly including defects area,aspect ratio,cohesions and coupling factors were calculated.Application of extreme learning machine to realize the classification of different defects,and the correct classification rate is 98.84%.The defects of solar cells that may appear in the silicon wafer cutting stage and screen printing stage were detected.The main defects are corner and edge breakage,holes,cracks and silk screen process.The maximum non damaged rectangle cut in the silicon wafer test stage and the classification of defects in the screen printing stage were realized.After a large number of samples training and testing,results show that the results of defect detection are fine.