Autonomous Following Algorithm for UAV Based on Multi-Scale KCF and KF

This paper presents an autonomous following approach for UAV, called AF, for the issue of target's motion state is unknown, complex and variable. This approach relies on the integration of Object Tracking algorithm and Flight Control algorithm. The proposed method aims to solve the problems of scale change, occlusion, speed and direction mutation during target's movement, so as to achieve accurate following flight of UA V. The method is divided into two main parts: firstly, the proposed mKCF-KF algorithm is used to track the target's position in the image sequence, which can solve the issues of target's scale change and occlusion. Secondly, based on the tracked target's position, a flight control algorithm for 3D following is designed, which can respond the issues of target's speed and direction mutation. The effectiveness of the proposed method is demonstrated by built semi-physical simulation platform based on ROS, Pixhawk, CopterSim and RflySim3D software. The results show that the proposed method achieves superior following performance.