Attitude Estimation Based On Stereo Vision For Unmanned Aerial Vehicle

Unmanned aerial vehicle(UAV)has always been a cutting edge issue and research hotspot today,and UAV’s position and attitude measurement plays a crucial role for its flight stability.Therefore,the accuracy,real-time performance and robustness of UAV position and attitude measurement must be satisfied.Among the traditional UAV position and attitude measurement methods,global navigation satellite system(GNSS)signals will be blocked indoors,while inertial measurement unit(IMU)has high sampling frequency,high short-term accuracy,but also unavoidable cumulative error.In view of the above shortcomings,the binocular vision measurement method is discussed in this paper,which is helpful to improve the indoor position and attitude accuracy of UAV.However,since the binocular-vision-based technology involves image processing,it often has poor real-time performance.In order to obtain reliable and real-time UAV position and attitude data,this paper combines the advantages of binocular vision and IMU based on Kalman filter algorithm,and also proposes a binocular vision/IMU UAV position and attitude measurement system.The main content of this paper is as follows.The position and attitude measurement principles of binocular vision and IMU have been analyzed.Both binocular vision and IMU completed the pose estimation based on the cumulative state change in the motion analysis process of UAV,and the attitude and position change can be expressed in a unified matrix form.On the above basis,a binocular vision/IMU combined measurement model has been proposed in this paper.After the problem of inconsistent sampling frequency is solved by preprocessing,the Kalman filter is used to complete the attitude data fusion.Additionally,based on the principle of the complementary filtering,binocular vision measurement results are used to correct the velocity estimated by IMU.The proposed data fusion method has lower computational burden and can further improve the precision of position and attitude measurement while retaining the advantages of high short-term precision and fast response of IMU.The measurement effect of binocular vision,IMU and composite system were verified respectively using the public EuRoC dataset of.The results prove that the proposed method has higher accuracy than pure binocular vision or IMU measurement,and can meet the requirements of real-time,robustness and high precision for UAV position and attitude measurement.