Segmentation technique for the detection of Micro cracks in solar cell using support vector machine

Micro cracks in solar cells lower the overall performance of the solar panel. These cracks result from poor handling during transportation, fabrication, and installation. Another reason could be the harsh environmental conditions under which they are deployed. Identifying micro-cracks and their replacement is always needed to get the best performance out of available solar panels. Image processing and machine learning are two commonly used schemes for detecting the same. The former techniques cannot produce accurate results because they perform segmentation using fixed equations, whereas the latter techniques learn complex nonlinear features that are difficult for the human mind to process. This paper uses a Support Vector Machines (SVM) model for detecting micro-cracks in solar cells. An image processing technique is proposed to train the SVM model and to generate ground truth for segmentation on Electro-Luminescence Photo-Voltaic (elpv)-dataset, which was used by researchers for defect percentage classification and contains 2624 images in total. The proposed SVM model performed exceptionally well in terms of accuracy (91.079%), precision (87.289%), recall (96.314%), and F1 score (94.678%) in comparison to other available machine learning models.