Unmanned Helicopter Position And Pose Estimation Comparative Research Based On Inertial Sensors And Visual Method

Unmanned aerial vehicle has flexible control characteristic and relatively low cost,so it is widely used in reconnaissance, forest fire protection, intelligent power network,etc. Position and pose estimation is very important for the small UAV navigation andcontrol. The traditional position and pose estimation methods use airborne IMU andGPS data fusion to get position and pose of the UAV. However, the IMU’ error willaccumulate along with the time. When there are clouds in the sky or when the UAV isrolling, the GPS maybe cannot work.With the development of computer vision technology, position and pose estimationbased on the computer vision attracts more and more attentions. Compared to thetraditional position and pose estimation methods, computer vision methods use thecameras as its sensors, so it can make full use of the surrounding environmentinformation to work.In allusion to the pose estimation comparative research of unmanned helicopterbased on inertial sensors and visual method, first, in this thesis, we design aposition and pose acquisition circuit board. On this board, we make use of theacceleration and angular velocity information of the UAV to get position and poseinformation of the UAV based on the quaternion method. Then we design a position andpose estimate method based on one airborne camera. At last, in this thesis, weconduct a comprehensive experiment. In this experiment, we compare and analyze theresults of the traditional method and visual method to analyze the feasibility of the poseand position estimation based on the computer vision.A large number of results of tests show that the pose and position estimationmethod based on the computer vision in this thesis is simple and reliable. The errorof the pose estimation is small. The results of the pose estimation are up to the standard;errors of roll angle estimation and pitch angle estimation are all so far as to less than2°.The results of position estimation based on the computer vision have betterperformance than the estimation output based on the traditional method. Obviously, thismethod has certain feasibility in the UAV autonomous flight control system and thereal-time pose estimation system of the flight vehicles.