A Segmentation Method for PV Modules in Infrared Thermography Images

In recent years, the installation of photovoltaic (PV) plants has increased rapidly worldwide. The installation in China account for a large proportion in global market. The maintenance of PV modules is very important for their efficiency and stability. Simultaneously, fault detection becomes an important issue. Nowadays, the infrared thermography (IRT) technology is widely used for hotspot detection. Compared with manual inspection, the use of unmanned aerial vehicles (UAVs) can improved work efficiency greatly in large-scale PV plants. The IRT images processing of PV modules is important for hotspot detection. Without the segmentation of PV modules, the hotspot location cannot be determined. This paper proposed a method to analyze IRT images to acquire segmentation. The MATLAB function findpeaks is used to find the points on the boundaries and density-based spatial clustering of applications with noise (DBSCAN) algorithm is used to cluster these points. This paper shows experimental results for 1211 IRT images collected by the UAVs. The proposed method is more accurate and robust than Sobel and Canny algorithms in edge detection. Moreover, quantitative evaluation is used to assess our method. The average quality is 0.9206, which indicates the method performs well in segmentation.