Research On Method Of Solar Cell Silicon Defects Automatic Detection Classification

The quality of solar cell silicon as a key factor is affecting efficiency of solar cells conversion and battery components power generation, so the quality assurance of the solar cell silicon has become a important part of the solar cells production and experiments. There are monocrystalline silicon solar cells and poly crystal line silicon solar cells for using generally, for some reasons, affected by many factors, silicon slices always have some defects more or less in slices during the production process. Common defects in polysilicon slice are edge impure, high impurity and dislocation defects, common defects in monocrystalline slice are whirlpool defects. The existence of silicon defects will greatly reduce the cell’s power generation efficiency and reduce the life of the battery pack, and even affect the stability of the PV system.Currently, in the actual production process, the methods of defects detection for silicon slices are mostly based on electroluminescent defect detection and detected by human eyes observation or automatically detected. Because the human eye observation method has a strong subjective and eye fatigue, it greatly reduces the detection reliability and efficiency. Further, since the electroluminescent defect detection is performed for solar cell detection, can not detect the silicon slice, diffusion sheet and other objects in the production process. Can not detect the defects in production process will improve the cost of production, reduce the efficiency of production. And electroluminescence detection is contact detection, may bring varying degrees of damage to solar battery. Thus a efficient, accurate and non-contact automatic defects detection method for silicon slices in the production process is very valuable. This paper, based on digital image processing theory, did some researches for silicon slices photoluminescence defect detection and classification method, and proposed a kind of new method for silicon slices defects detection and classification.This paper includes the following contents:1. Perform preprocessing for photoluminescence image, including image denoising, enhancement, edge detection, line detection, image rotation, and segment the image of target silicon slice from the photoluminescence image automatically in order to facilitate subsequent processing.2. Base on the analysis of photoluminescence images and defects characteristics, segment defects by the threshold which is calculated by gray curve Gaussian fitting for polycrystalline silicon slice gray curve, and extract area ratio and distribution characteristics of defects; base on enhanced monocrystalline silicon slice image, by Gaussian curve fitting for sum value curve of sampling gray curves in image to extract fitting standard deviation; by the method of frequency domain filtering combined with binary to extract areas ratio of high grayvalue in high frequency image; based on the thinned binary image of high frequency image, extract the result of circle detection by hough transforming method; get three defects characteristics.3. Based on decision tree classification method, constructed classification tree model, to design an open-threshold classification method to achieve defects detection and classification. For three defects in polycrystalline silicon slice, using exclusion method to detect sequentially. And design a system software, the system software based on C#programming to complete the various functional modules coding and integration. And test the system software in practical application and experiments, the results show the accuracy of defects detection and classification can reach95%, it proves that the methods and design of software are reasonable.This paper proposes a new method, this method can achieve non-contact automatic defects detection classification for solar cell silicon in the process of solar cell production, and design the automatic detection software. The experimental results show this method has high detection accuracy and efficiency, and has a great application prospects for production and research.