Research On UAV Target Tracking Algorithm Based On Correlation Filter And Control Method

With the advent of the intelligent era,the UAV equipped with a variety of sensors and its technology has been developed rapidly,and has brought great changes to people’s lives.The UAV equipped with camera and other visual sensors can play a greater role and perform more advanced tasks.In recent years,target tracking technology has developed rapidly.The combination of UAV and target tracking technology makes life more intelligent,and has been widely used in intelligent transportation,human-computer interaction,military and other fields.However,when the UAV is performing the target tracking task,it will encounter the problems of target occlusion and rapid movement.In order to solve these problems and make the UAV track the target accurately,we study the relevant technology of UAV target tracking based on correlation filter in this paper,including target tracking algorithm,UAV tracking control method,target tracking system construction,experiment and algorithm evaluation,The specific research contents are as follows:First of all,aiming at the real-time requirement of UAV target tracking,this paper studies the correlation filtering target tracking algorithm with fast computing speed.Aiming at the appearance feature extraction method of the target,a multi feature fusion method is designed.Aiming at the problem of boundary effect caused by the rapid movement of UAV tracking target,an adaptive spatiotemporal regularization method is designed to improve the objective function,so that the optimal filter template has higher accuracy and success rate.At the same time,in order to solve the problem efficiently and quickly,ADM M-Newton method is used to solve the optimal filtering template and the position of the target in the image.Secondly,aiming at the problem of target occlusion in UAV tracking,a multi detection template updating strategy is proposed.The template paramet er updating status of the tracker is determined by designing a variety of confidence functions and threshold conditions.Aiming at the problem of targ et scale change,a scale estimation method based on the idea of block and pyramid scale pool is designed to solve the problem of target scale change in the image due to the distance from UAV.Then,the target tracking control system of UAV is introduced,in cluding the conversion principle between image coordinate system and ground coordinate system,mathematical model of UAV kinematics and dynamics and design of cascade PID controller.Finally,the simulation system of UAV target tracking is built based on ROS/Gazebo simulation platform and the experimental analysis is carried out.Finally,the performance of the algorithm is evaluated in the data set OTB100 and UAV aerial photographing data set UAV132 and DTB70.The experimental results show that the algorithm has higher accuracy and success rate than other relevant filtering algorithms.Meanwhile,the tracking track and position change of UAV are analyzed in the target tracking simulation system,The experimental results show that the UAV has a small position and attitude error and meets the requirements of target tracking task.