Defect detection method with local low pass filter and adaptive threshold segmentation for photovoltaic cells

Any defects in the production process will seriously affect the safety and conversion efficiency of photovoltaic cells. This paper presents a defect detection method for photovoltaic cells using image processing technology based on photoluminescence (PL) imaging. First, the PL image of photovoltaic cell is preprocessed, and a local low-pass filter, which is different from the traditional low-pass filter, is designed to filter the high-frequency signal accurately in frequency domain for removing the grid lines. Then, the image is segmented by combining histogram bimodal method and adaptive threshold method. Finally, the defect detection is completed by using connectivity detection. To measure the performance of the proposed method, 45 photovoltaic cells sample images are used for testing experiments. The recall rate of detection is 95.28%, and the accuracy rate is 96.02%.