Silicon Crystallinity Detection Based On The Digital Image Processing Technology

Polysilicon is widely used as substrate in photovoltaic industry, the crystallinity of it is one of thekey factors which determines the performance and the product rate of solar cell. The crystallinity isdetected by visual assessment at present. It is urgent need for enterprise to realize the samplingdetection to full detection by using machine automatic detection instead of the traditional artificialdetection. Machine automatic detection can improve the quality of product and reduce the cost. Localthreshold segmentation was utilized to extract the polysilicon edge automatically count and theinfromations of crystal region, an important basis was provided for the crystallinity detection by thoseinformations. The main research contents of this thesis include polysilicon image collection, imagepreprocessing, crystal regional contour extraction and area statistics. The aim of image collection is totake polysilicon images with uniform luminance and distinct crystal region contour. Imagepreprocessing is the preparation for the later image processing, including image gray processing, noiseabatement and background correction, etc. Threshold segmentation and the contour of crystal regionextraction, which are the most important parts in our research. A technology based on thecharacteristics of image is chosen lastly, which is to extract the contour of the local segmentationpolysilicon image firstly and then synthesize the whole image contour. In the design of thresholdsegmentation algorithm, the bimodal method and Otsu method are combined to realize adaptivedetermination of the numbers of thresholds and the corresponding precise threshold values. Theresults had showed that this method in the thesis provided good detection results and improved theefficiency obviously. Our basic research lays a firm foundation for the industrial online automaticdetection.