Contents I. Introduction
Machine vision (MV) is a comprehensive technology. Its frame for defect inspection mainly includes three parts: image acquisition, object location, and defect detection [1]. It has been widely used for product quality monitoring and production process optimization in the automatic manufacturing industry, particularly the photoelectric semiconductor industry [2]. Over the past decade, polycrystalline silicon products have played critical roles in the photovoltaic application market due to their cost advantages. To improve their photoelectric conversion efficiency, many researchers pay increasing attention to detecting defects inside polycrystalline solar cells (PSCs). Fourier transform [3], independent component analysis (ICA) model [4], mean shift filter [5], and vessel algorithm [6] are exploited to enhance the defects by removing the silicon grains in the background. The OR gate is utilized to compare the test image with the standard template to acquire the salient defects [7]. Nevertheless, these methods are merely developed for crack-type defects in the cropped region of cells. With the continuing decrease of cost, the high-efficiency monocrystalline solar cell (MSC) will replace the dominant position of PSC in the photovoltaic market [8], but the studies about defect inspection for MSC interior are fewer. We thus concentrate on the studies of positioning cells and detecting defects inside MSC in this article.
