基于手眼视觉的两种目标位姿测量法仿真分析  

Simulation Analysis of Two Kinds of Algorithm of Pose Estimation Based on Hand-Eye Vision

作　　者：杨小平[1] 孙国祥[1] 

机构地区：[1]哈尔滨工业大学深圳研究生院,广东深圳518055

出　　处：《计算机仿真》2012年第6期168-170,222,共4页Computer Simulation

摘　　要：空间机器人手眼视觉系统利用合作目标物体上的合作特征点对目标物体进行位姿估计,以完成跟踪捕捉任务。针对空间机器人实时性的要求,给出了两种实时性较好的4个特征点的位姿测量算法,并且在相同仿真环境下对两种P4P位姿算法进行了分析比较。仿真结果表明:四面体体积测量的位姿测量算法对特征点的像点量化误差更加敏感,位姿测量结果偏差较大;坐标变换的位姿测量算法稳定,精度较高。后者在实际应用中,可以达到更高的精度。The relative pose parameter of target object is determined for realizing tracking and capturing coopera- tive target tasks By Hand - eye vision system of Space robot using cooperative feature points on cooperative target ob- ject. This paper gives two kind of algorithm for pose estimation, which has more real time for the real time require- ments of Space robot , and making comparison and analysis to those two kinds of algorithm in same simulation condi- tion. The simulation results show that P4P based on volume measurement is more sensitive feature points of pexal measure error, has more big pose measure error; that P4P based on coordinate transformation is more stable, more high accurate. In actual using, latter has more acctuate.

关 键 词：空间机器人 手眼视觉 位姿测量 

分 类 号：TP301.6[自动化与计算机技术—计算机系统结构]