Hot-Spot Detection for Thermographic Images of Solar Panels

Hot-spot detection facilitates the discovery of damaged solar panels, which plays a critical role in the solar energy utilization. Since most hot-spots are not visibly distinguishable in ordinary optic images, it is necessary to take thermographic images for hot-spot detection. This paper proposes a method to detect hot-spots for thermographic images of solar panels. Firstly, a thermographic image is transformed from the RGB color space to the HSV color space and all cells of solar panels are segmented based on the H channel. Secondly, the edges of solar panels are extracted using the Canny edge detection algorithm. Then, cells of solar panels are segmented based on the extracted edges, and the mean and standard deviation of each segmented cell in the B channel are computed to construct features. Finally, a support vector machine(SVM) decides whether there is any hot-spots in each cell based on the extracted features of that cell. To further enhance the robustness against overexposure of thermographic images, the mean and standard deviation of a cell in the gray space are also computed and combined with the mean and standard deviation of that cell in the B channel to construct composite features, which are taken in the decision of SVM. Experimental results confirm the effectiveness of the proposed hot-spot detection method.