Miniature Unmanned Aerial Vehicles Target Tracking Control Strategy Research

In recent years,with the continuous development of microelectronics technology and embedded chips,various types of autonomous mobile robots have emerged.Rotarywing UAVs have gained widespread attention among various types of mobile robots,especially small and miniature unmanned aerial vehicles,which have been applied to many fields such as military,agriculture,industry and civil use.Among them,miniature unmanned aerial vehicles have the advantages of small size,light weight,stealthiness and portability,which make them unique in military reconnaissance and aerial photography.On the other hand,due to the continuous improvement of computer vision technology,the performance of target tracking algorithms has been continuously improved.To address the problems of limited computing power and insufficient range of miniature unmanned aerial vehicles,which make them unable to deploy complex tracking algorithms,this paper proposes a miniature unmanned aerial vehicles target tracking system based on monocular vision.A complete miniature unmanned aerial vehicles hardware platform and software architecture are built,and the feasibility of the system designed in this paper is finally verified by software simulation and actual flight.The specific work is as follows.(1)According to the mission needs,the relevant vision sensors,on-board computers,autopilot and UAV frames and power kits were selected.The UAV target tracking experimental platform was designed and assembled,and the relevant parameters were debugged to enable stable fixed-point flight.In addition,the software architecture of the tracking system was designed as a whole so that each module can work closely with each other.(2)For target tracking,the KCF kernel correlation filter tracking algorithm is used to track the target,and the KCF fused Kalman filter target tracking method is proposed for the short-time occlusion problem.The KCF algorithm is ideal for deployment on micro unmanned platforms because of its small CPU resource consumption.The average time to run the KCF algorithm once on the Jetson Nano airborne computer used in this project is around 30 ms,and the average frame rate is around 30 FPS,which can fully meet the needs of the tracking task.Finally,the target position estimation algorithm is used to solve the obtained two-dimensional motion coordinates of the target to the threedimensional world coordinate system by coordinate transformation.(3)The mathematical model description of the UAV is established,and the model predictive control is used as the outer-loop position controller,and the calculated desired attitude angle is sent to the flight control for attitude control through the serial port.(4)Build the simulation environment for the designed system,verify the relevant algorithms,and design different tracking scenarios for the target position solving algorithm and control algorithm.Finally,the code is ported to the onboard computer side to verify the feasibility of the algorithm by tracking the pedestrians on the ground and the UAV targets in the air.