Research On Unmanned Aerial Vehicle(UAV) Tracking Method Based On Visual And Inertial Sensors

Quadrotors are widely used in military and civilian fields because of its simple structure,good maneuverability,strong concealment and strong adaptability to the environment.As detecting and tracking moving objects is one of the hotspot of UAV applications,it's worthy of deeply studying how to improve the tracking accuracy and robustness.There is a problem of misdetection or loss moving targets due to image degradation caused by camera motion for vision-based UAV tracking method.To solve this problem,a relative pose solving method based on Multi-rate Kalman Filter is proposed.This method integrates the Phone IMU data and UAV IMU and image data.Meanwhile,this paper utilizes the Intel NUC as the UAV processor to solve the problem of limited processing resources,and ROS(Robot Operating System,ROS)is build on the Linux.Finally,experimental results proved that the proposed method improved the accuracy and robustness of UAV motion tracking.(1)To solve the problem of limited accuracy of image-based UAV motion tracking method,a UAV motion tracking method based on visual and inertial sensors which integrating mobile phone and UAV multi-sensor parameters is proposed.In this multi-data fusion method,mobile phone IMU data is sent through by Android application to UAV,and combined with UAV image and IMU data.(2)To solve the problem of misdetection or mismatching of moving targets happened at the UAV image acquisition process,the proposed method utilizes ORB feature descriptor to extract feature points,and RANSAC is used to remove the exclusion points and reduce the feature points.This method can reduce the processor load while ensuring accuracy.(3)To solve the case of different IMU and image sampling rates,a relative pose solving method based on Multi-Rate Extended Kalman Filter is proposed.If having a measurement data,the system will do the time updating and measurement updating at the same time,but if not,it will only do time updating.This method can improve the data update rate,reduce the waste of IMU information and improve system robustness.