Vision-Based Target Detection and Tracking System for a Quadcopter

Tracking a maneuvering target with quadcopter is challenging problem, that involves variety of fields such as visual tracking, state estimation, and control algorithms. Most existing unmanned aerial vehicle (UAV) systems fail to track targets accurately in the long term cannot relocate after loss. This paper aims design implement vision-based tracking system for quadcopters can steadily ground well air without any prior information. We employ vision detection algorithm select quickly precisely. To fit complex practical conditions, developed based on correlation filters, which capable large-scale variation fast motion. In addition, an efficient redetection support vector machine (SVM) designed handle occlusions The states are estimated from information by improved Lucas-Kanade (LK) optical flow method extended Kalman filter (EKF). Moreover, double closed-loop Proportion Integral Differential (PID) controller using follow target. By implementing main algorithms onboard NUC computer, extensive outdoor flight evaluated platform equipped stereo camera. experimental results validate feasibility practicability system.