Polysilicon Solar Wafer Surface Defect Detection And The Design And Development Of Software Inspection System

Solar energy is a clean and pollution-free high-quality energy,compared to fossil fuels which causes the environmental problems, it hasbecome an important part of the new energy. Solar power technology is animportant form of solar energy. Polysilicon solar cells have been widelyused in photovoltaic manufacturing with its relatively simple process, lowproduction cost advantages.Quality Management is an critical process in manufacturing. Toensure the quality and reduce production costs, defects inspection is animport process after the antireflection film plating process. As non-uniformcolor and complex texture exist on the polysilicon solar cell, manualsurface inspection adopted by most domestic factories suffers from lowefficiency and poor repetitive detection ability.From the industrial engineering quality management perspective,based on machine vision and SVM, an automation and high classificationaccuracy polycrystalline silicon wafer surface defect detection andclassification system has been developed in this paper, which has thetheoretical and value in use.This mainly work in this paper can be seen as follows:(1) According to the practical problems of a polycrystalline siliconsolar cell manufacturer, common solar wafers surface defects have been studied and classified, with the defects causes analyzed. According to theactual needs, an overall design based on machine vision and patternrecognition has been proposed, with the inspection framework, theinspection process and the working principle of the inspection software,realizing the automation and information in polysilicon solar cells’ qualitymanagement.(2) An edge fitting method has been proposed to conduct precisewafer positioning to overcome the random feeding, which produces thereliability of data sources to improve the follow-up classification accuracy.A color space transformation has been proposed to reduce the color patternspace dimension. In the spot detection, the saturation channel has beenselected to reduce the influence of the texture on the wafer surface.(3) For each common defect, the feature extraction method and thedesign of the SVM classifier have been proposed to classification, and theclassification parameters are optimized to improve the inspection result. Tospeed up the inspection process, a classifier scheduler has been proposed,which combined with the frequency of the defection occurrence and theinspection classifier speed.Experimental results show that the classification accuracy rate of94.5%and the cycle time less than1s, which meets the company’s actualproduction needs. The polysilicon wafer surface defect detection andclassification method based on computer vision has fairly anti-interferenceability and can be widely used of applications, and thus becomes a fairlyimportant quality management tool.