Optimizing Thermal Image Segmentation Using K-Means Clustering and Morphological Operations for Photovoltaic

Thermal imaging has become a critical tool in the diagnosis and maintenance of photovoltaic (PV) panels, particularly in detecting localized hotspots that indicate underlying faults. We introduce an optimized segmentation framework an optimized segmentation framework that integrates K-means clustering with adaptive histogram equalization and morphological operations. The aim is to improve hotspot detection accuracy in thermal images of PV panels. Evaluations under various conditions, including the presence or absence of filters and data augmentation, reveal that the proposed method significantly enhances segmentation performance. The highest precision and F1-score are achieved achieving up to 0.87 precision and 0.71 F1-score with filtering and augmentation.