An Integrated Fuzzy Logic and Support Vector Machine–Based Framework for Automated Defect Detection in Photovoltaic from Thermal Infrared Imaging

This study aims to develop an automated and reliable framework for detecting and classifying defects in photovoltaic (PV) panels operating within solar arrays. As PV systems are increasingly deployed worldwide, early and accurate fault diagnosis has become essential to maintain efficiency, reduce energy losses, and lower maintenance costs. To address this need, the proposed approach integrates thermal infrared (IR) imaging with advanced image processing and machine learning techniques. Initially, thermographic images are preprocessed using a fuzzy logic–based edge detection method to suppress background noise and enhance defect boundaries. The Hough Transform is subsequently applied to localize panel structures and extract geometric features. Relevant attributes are then organized into a labeled feature matrix and used to train a Support Vector Machine (SVM) classifier with a fine Gaussian kernel. Model performance is validated using ten-fold cross-validation. Experimental results demonstrate that the proposed framework effectively identifies multiple PV panel defects, including hotspots, delamination, cell damage, and surface contamination, achieving a classification accuracy of 91%. The findings confirm the robustness and reliability of the integrated method. Practically, the system offers a cost-effective and intelligent tool for real-time fault monitoring, enabling improved maintenance planning, enhanced operational reliability, and optimized performance of large-scale PV installations.