A Cooperative Target 3D Tracking Method Based on EPnP and Adaptive Kalman Filter
Cooperative target 3D tracking can be treated as an extension of the pose estimation problem,since the pose tracking result is not only based on the correspondences between the control points and the image points,but also based on estimation of the target movement.Therefore,a two stages method is proposed in this paper.The first stage is the initialization stage,the initial pose estimation is acquired by using EPnP algorithm in this stage.The second stage is the pose tracking stage,where the pose varying and its varying rates are acquired by using the extended kalman filter,denoted as EKF.However,the statistic characters of the noise of the motion model and the measurement model are assumed to be fixed and predefined in EKF.Whereas,in reality,because of the uncertainty of the movement between the cooperative target and the camera,the motion noise is usually not available.Therefore,an adaptive estimation process for the motion noise is introduced in this paper.Experimental results show that a good 3D tracking result can be acquired by combining EPnP and adaptive extended kalman filter.
Measurement Pose estimation 3D tracking Kalman filter
Haodong Ding Kun Liu Peng Chen Haiyong Chen
School of Artificial Intelligence and Data Science,Hebei University of Technology,Tianjin 300130,China
国际会议
江苏镇江
英文
580-591
2019-09-20（万方平台首次上网日期，不代表论文的发表时间）