Research On Solar Cell Surface Defect Detection Method Based On Visual Saliency

With the drying up of the non-renewable energy and the deteriorating of the issue of environmental pollution,the development and utilization of renewable clean energy have become a key planning direction of the country.The photovoltaic generation technology has gradually become one of the main means to solve such problems and has been extensively concerned and participated by the society at large.The solar cells belong to the core components of photovoltaic power generation system,and the server life and the generating efficiency are directly affected by the quality of solar cells surface.So a novel method of solar cell surface defect detection based on visual saliency is proposed to ensure the qualified rate of online production of solar cells in the factory.The main research contents of this paper are summarized as follows.First of all,the captured complete solar cell surface image is preprocessed.The median filter and the anisotropic diffusion are applied to reduce the noise.The method of locating the busbars and fingers based on the sum of gray value of the row and column is proposed to remove those white regions of busbars and fingers.Those regions are filled by using the least square method at the same time.Secondly,visual saliency detection is performed on the preprocessed images.A self-learning feature method is proposed to acquire the feature matrix of preprocessed images.The low rank matrix recovery method is adopted to further obtain the visual saliency map,which is the initial detection result of defect.Then,the defect regions in the initial detection result are precisely located.A method that combinesvisual saliency and image segmentation is proposed to precisely locate the position of the defect regions.At last,the defect regions in the precise detection result are optimized.The final detection result is obtained using the optimization method of morphological operation.In addition,a software of solar cell surface defect detection is designed based on the proposed method in this paper,which has a concise and intuitive GUI interface for users to operate.In this paper,two types of testing databases of solar cell surface defect image are established.The subjective and objective evaluations on those databases demonstrate that the proposed method has a high detection accuracy and has a significant effect for improving the qualified rate of the solar cell.