Detection and Positioning of Keypoints in Small-scale Photovoltaic System Based on Object Detection Network and Aerial Sequence Images

With the vigorous development of photovoltaic (PV) industry, the pressure on maintenance of photovoltaic stations is increasing rapidly. The automated system inspection via unmanned aerial vehicles (UAVs) becomes a popular approach these days, which possessed high efficiency. However, UAV automated inspection requires a global geographic information about the PV system area. This paper proposed a method based on object detection network that accurately detects the keypoints of small-scale photovoltaic system in sequence images taken by UAV at low altitude. By combining the keypoints and the UAV flight parameters, the coordinates of the PV system area keypoints can be converted to world coordinates. Experiments show that the keypoints detection method can accurately detect the keypoints and the positioning information is reliable for UAV automated inspection.