Research On Online Inspection Technology Of Silicon Wafers Series-welding Using Machine Vision

At present,the demand of solar cells is increasing with the optimization of energy structure,the quality problems of the resulting high-speed automatic cells production process need to be solved urgently.Silicon wafers series-welding is the key link in solar cells manufacturing,which has completely realized the high-speed automatic welding,while the silicon series of automatic welding production line is unable to realize the online detection due to the welding defect types and fast production rhythm.Based on the in-depth investigation and analysis,this paper has carried out the research on the online inspection technology of silicon wafers series-welding using machine vision technology.Based on the analysis of the research status at home and abroad on vision recognition and detection of solar cells,the overall scheme design of high-speed welding online detection system and construction of experimental platform are completed.The system includes mechanical structure,visual imaging unit,industrial controller and other parts of hardware and system software.Firstly the calibration of the vision system is expounded,several methods of image smoothing filtering and morphological processing are discussed combined with the initial cell location and type checking.Using multiple color space conversion and improved Canny edge extraction algorithm,the ROI division of different image defects is finished.The multi threshold Otsu method is accelerated according to the Multi-Otsu criterion so as to achieve the image segmentation based on defect feature,which is compared with the Niblack local threshold algorithm.According to the characteristics of the welding belt offset,the corresponding model is set up and the measuring method referring to the boundary tracking idea is put forward,and the measurement methods of other dimensions are explained.On this basis,the recognition and classification method of characteristic defects of post welding battery string is studied in depth.Since the occurrence areas and preprocessing methods of stain and scratch are familiar,with an analysis for obvious feature types between two kinds of defects,the algorithm of SVM is adopted to classify them after the removal of irrelevant image noise through the same pretreatment.An improved region growing algorithm based on the detection of fine grid connectivity is proposed to locate the off-grid features one by one and get the off-grid position;The chamfering edge is determined according to the external parameters and the image reconstruction of edge defect area is completed based on the results of edge detection.Finally,the Euclidean distance between the sample and template image in YCbCr space is calculated,so that the color system of each half is judged and the color difference test is completed at the end of the series.On the basis of the above studies,this paper successfully developed a prototype of online inspection for silicon wafers series-welding based on machine vision.The field test of the production line is carried out,and the experimental results verify the feasibility and effectiveness of the online inspection technology of high-speed silicon wafers series-welding based on machine vision.