Hotspot detection of photovoltaic modules in infrared thermal image based on saliency analysis

With the increasing awareness of environmental protection, the solar photovoltaic (PV) industry has been developing dramatically in recent years. PV module is an important part of PV power generation system. In all the defect types of PV modules, hotspots occur more frequently and cause more serious damage. Therefore, it is necessary to detect hotspots. An approach is proposed using saliency analysis method for hotspot detection, combing the cognitive property about visual saliency and the temperature property of hotspots. Not only does this approach not require manual threshold setting, but also it does not require a large number of learning samples. This approach can also achieve the defect diagnosis and hotspot location when compared with the electrical characterization method. The infrared thermal (IRT) images trained in experiments are obtained using the thermographic camera FLIR Vue Pro with the unmanned aerial vehicle from a PV plant in Jiangsu, China. Altogether 135 IRT images are collected and 1020 PV modules are extracted from these images. Experiments have proved great accuracy and robustness when using the proposed method to detect hotspots.