Deep Segmenter system for recognition of micro cracks in solar cell

A solar panel is array of Photo-Voltaic modules (PVC) that are mounted together in a mechanical frame and are placed in the open fields so that sunlight impinges on those cells to produce electricity. The effectiveness of solar panels is cogently restricted by the impurities and defects present in the PVC. These imperfections bring profound energy levels in the semiconductor bandgap, depreciating the carrier lifetime and quantum efficiency of cells. It is significant to recognize the defects physics so that apposite methods may be employed to restrain the formation of severe flaws. In various past techniques, image processing, machine learning, and deep learning techniques are implemented to recognize, classify, or predict the probability of defects and their effect on PVC's overall performance. One of these approaches is an automatic recognition of micro-cracks, which is a compelling but challenging task. To achieve this, a deep learning approach based on the classification and segmentation process is proposed in this paper. This mechanism not only detects the micro-cracks but also effectively locates the area of the defected pixels. For the categorization of defects, VGG16 is used as a CNN classifier, and a Deep crack approach for the segmentation process is used. Thresholding and Decision Making are added to remove redundant pixels related to diverse types of frames present in PVC's, and finally, a decision is made. An unsharp filter is utilized because of efficient performance. This technique exhibits effective results in decision making, whether the solar cell needs to be replaced or not based on the percentage area of irregularity. The proposed model outperforms state-of-the-art methods with better performance in all aspects.