RPEOD: A Real-Time Pose Estimation and Object Detection System for Aerial Robot Target Tracking

Pose estimation and environmental perception are the fundamental capabilities of autonomous robots. In this paper, a novel real-time pose object detection (RPEOD) strategy for aerial robot target tracking is presented. The equipped with binocular fisheye camera depth to capture spatial position tracked target. RPEOD system uses sparse optical flow algorithm track image corner features, local bundle adjustment restricted in sliding window. Ulteriorly, we proposed YZNet, lightweight neural inference structure, took it as backbone YOLOV5 (the state-of-the-art detector). can dramatically reduce computational complexity reprojection error minimization network process; Thus, calculate on onboard computer carried by robot. evaluated using both simulated real-world experiments, demonstrating clear advantages over approaches, significantly more fast.