单目-无扫描3D激光雷达融合的非合作目标相对位姿估计

Relative Pose Estimation of Noncooperative Target Based on Fusion of Monocular Vision and Scannerless 3D LIDAR

针对传统的利用单一视觉传感器难以实现复杂非合作空间操控导航的问题,提出一种基于单目相机与无扫描三维激光雷达融合的非合作目标相对位姿估计方法。首先,设计了基于成像几何关系的单目纹理-非扫描距离图像的快速配准与融合方法;之后,在构建目标同步定位与建图(SLAM)贝叶斯滤波模型基础上,提出一种扩展卡尔曼滤波-无损卡尔曼滤波-粒子滤波联合的滤波估计算法,可实现尺度模糊下相对位姿的快速鲁棒估计;其次,针对估计中的尺度模糊问题,提出基于融合图像的全局尺度系数确定方法,将尺度系数估计问题转化为简单线性滤波问题。基于Open GL生成的2D/3D图像实验表明:所提出的方法具有较优的精度和鲁棒性;相对位置估计误差与尺度估计误差相关,二者近似成线性正比关系。

Only one vision sensor in the traditional way is incompetent for the navigation of complicated noncooperative operation. In order to solve this problem,a scale-ambiguous relative pose estimation method based on the fusion of monocular vision and scannerless 3D LIDAR is proposed. Firstly,the imaging geometrical relationship of two cameras is used to map the scannerless range measurement onto the monocular texture image. Secondly,on the basis of establishing the SLAM（ Simultaneous Localization and Mapping,SLAM） Bayes filter estimation model,the scale-ambiguous relative pose estimation algorithm based on EKF（ Extended Kalman Filter）-UKF（ Unscented Kalman Filter）-PF（ Particle Filter）combination filter is presented. Thirdly,a global scale factor estimation algorithm based on the fusion image is proposed to estimate the scale factor,which can be solved by a simple linear filter algorithm. Some simulations based on 2D /3D images generated in Open GL not only demonstrate both good accuracy and robustness of the proposed approach but also show that the position estimation error is approximately proportional to the scale estimation error.