Real-time UAV (Unmanned vehicle) Tracking with Autonomous Drone in Flight

Abstract In recent years, the use of unmanned aerial vehicle (UAV) platforms in civil and military fields has significantly increased and become a critical technology. The artificial intelligence (AI) embedded UAV system has great importance and potential in the future. Research and development, search and rescue operations, aerial photography, videography, border security, traffic control, forest fire prevention, working in toxic chemical gas environments, preventing poaching, natural resource exploration, agriculture, extraction, and UAV detection while in the air are all applications for AI embedded UAVs. In this project, Autonomous Drone (Vechür-SIHA) was developed for detecting and tracking other UAVs while in flight sequence. The system was simulated in Robot Operating System (ROS). Inside the simulated environment drone model was used for tracking UAV. After simulations and calculations, the Fighter drone was designed and developed. The real-time UAV detection was made with object detection algorithms inside the simulation and in real-time. The system was embedded with a tracking algorithm based on LSTM (Long-short term memory). The system achieves high-accuracy drone detection in changing backgrounds and weather conditions (%80 accuracy), LSTM drone tracking can be accomplished even with multiple drones inside the field of view, and the system can fly for 35 minutes (Take off and landing included). The system also can be locked other UAVs and track them while in the air. The system can detect and track other UAV's 4-10 seconds without losing contact.