Research On Solar Mono-crystalline Silicon Wafer Surface Quality Inspection Based On Image Processing

Solar energy, as a new type of energy, is clean, rich reserves and it is attracting extensive attention all over the world. With the development of solar energy industry, quality control in mono- crystalline silicon wafer production is significant. Nowadays, solar industries mostly adopt the manual method for quality control. However, these methods have many drawbacks, such as bad real-time performance and defect escape etc. Therefore research and develop an automatic wafer quality inspection systems are necessary.The focus of this paper is on the following aspects:(1) Based on relevant research reports and data, review the history of the solar industry and development trends. Introduce the mono-crystalline silicon wafer surface inspection system design and outline the system structure, data flow and work principle.(2) Adopted a adaptive histogram equalization for image enhancement processing, this method not only can enhance the contrast and details but also resist the image noise. For image de-noising, designed a new type of improved median filter, the weight given to improve the formula, which can effectively reduce the noise, and as a hybrid filter combines the median and mean filter process, so it has the advantages of both, to achieve better noise reduction. Apply eight-direction edge extraction algorithm, experimental results show that the algorithm effectively enhance the effect of edge detection.(4) Study the classification principle of the SVM(support vector machine). Construct a SVM discriminant function for defects detection according to the differences of defect and non-defect images, which reduce by 95% classification amount, more suitable for online real-time requirements. Select RBF(Radial Basic Function) as kernel function and propose a orthogonal experiment method for SVM parameters optimization to obtain the best values, experiments show that accuracy over 90%, indicating that the parameter value selected by proposed method has a strong anti-interference ability and wide scope of application.