An Intelligent System for Automatic Detection and Cleaning of Bird Droppings on Solar Panels

In solar power plants, the accumulation of bird droppings may decrease power generation efficiency and necessitate manual cleaning, which can be difficult in large facilities. This paper proposes a solution to this problem through an automatic unmanned aerial vehicle (UAV)-based method for detecting, localizing, and cleaning bird droppings on a large solar power plant. An automated flight route is generated for a UAV to fly over the solar farm and capture images of the solar panels, which are then stitched together to create a high-resolution orthomosaic image of the farm. A new YOLOv7-based model is proposed, utilizing an attention module to suppress complex backgrounds and highlight bird dropping features, and an extra head is used to detect small droppings. To determine the GPS coordinates of bird droppings, an algorithm is developed that uses the azimuth angle and ground level distance (GLD) between detected bird droppings and a ground control point (GCP). Finally, a system for cleaning bird droppings is established using the travelling salesman problem (TSP) to guide the drone in visiting each detected dropping once. The experimental results have shown that our approach is successful in detecting and cleaning bird droppings on solar farms. Our system has achieved a high mAP of 93.91% in detecting the accumulated bird droppings on solar panels, and has been able to accurately locate the coordinates of the detected bird droppings with a mean error of 0.14 meters.