Research on UAV Target Detection Algorithm Based on Improved Air-Borne-YOLOv3

While the rapid development of multi-rotor UAVs brings wide applications, it also brings harm to social life and personal safety due to problems such as “indiscriminate flying” and “black flying”. Therefore, there is an urgent need to carry out research work on anti-VAV systems. The most critical technology in the countermeasure system is the real-time detection of the target UAV. The traditional YOLOv3 detection algorithm cannot achieve a balance between real-time performance and detection accuracy on airborne embedded devices with limited computing power. Therefore, this paper first improves the end structure of MobileNetV3-Small and replaces the backbone network of the original YOLOv3 with it, which greatly reduces the amount of calculation, but brings about the loss of detection accuracy; Next, an improved receptive field module(RFB+) is added, which strengthens the detection ability of targets in complex backgrounds; Finally, the improved YOLOv3 algorithm is named Air-Borne-YOLOv3. The feasibility and superiority of the improved algorithm are proved by the actual flight test results.