Research on virtual intelligent reconnaissance training system based on YOLOv3 tiny

DOI:	10.11809/bqzbgcxb2023.08.027
Keywords:	virtual training; UAV; intelligent reconnaissance; Unity3D; YOLOv3
Abstract:	Virtual training system is widely used in medical, aerospace, military and other fields, which can effectively reduce training cost, improve training efficiency and ensure training safety. Specific to the harsh conditions of flight control, high delay of onboard algorithm, high potential security risks and high training cost of the UAV reconnaissance training, this paper proposes an overall research and development scheme of virtual training system applying target detection algorithm in virtual scenes. Based on YOLOv3 tiny algorithm, this system uses Unity3D camera component instead of drone video stream for data transmission by using YOLOv3 tiny algorithm to detect vehicles and personnel through OpenCV For Unity plug in, and returning the detection screen in real time. Through system test experiments, the intelligent UAV reconnaissance function is realized, the target detection speed is stable at 25 fps, the recognition confidence is over 80%, and the proportion of good indicators is over 88%, which meets the flight control simulation and algorithm detection training.
Published:	2023-08-28
PDF HTML