Real time fault detection in photovoltaic cells by cameras on drones

Abstract

Hot spots are among the defects of photovoltaic panels which may cause the most destructive effects. In this paper we propose a method able to automatically detect the hot spots in photovoltaic panels by analyzing the sequence of thermal images acquired by a camera mounted on board of a drone flighting over the plant. The main novelty of the proposed approach lies in the fact that color based information, typically adopted in the literature, are combined with model based one, so as to strongly reduce the number of detected false positive. The experimentation, both in terms of accuracy and processing time, confirms the effectiveness and the efficiency of the proposed approach.