多像机非共视场的非合作飞行器位姿测量方法  

Pose measurement method for non-cooperative space vehicle using multiple non-overlapping cameras

作　　者：苗锡奎[1,2,3,4] 朱枫[1,3,4] 郝颖明[1,3,4] 

机构地区：[1]中国科学院沈阳自动化研究所,辽宁沈阳110016 [2]中国科学院大学,北京100049 [3]中国科学院光电信息处理重点实验室,辽宁沈阳110016 [4]辽宁省图像理解与视觉计算重点实验室,辽宁沈阳110016

出　　处：《红外与激光工程》2013年第3期709-715,共7页Infrared and Laser Engineering

基　　金：国家自然科学基金(51005229)

摘　　要：以空间非合作飞行器视觉位姿测量为背景,针对近距离及超近距离情况下由于成像空间小、像机视场等限制,位姿测量所用的视觉特征将不能在单像机中完整成像而无法完成定位的问题,提出一种多像机非共视场的非合作飞行器位姿测量方法。首先将多个像机配置成非共视场的形式,标定各个像机之间的位置关系;然后利用多个像机对目标上的不同特征成像,来自不同像机的底层信息既有冗余又有互补,为位姿测量提供足够的视觉特征和几何特征;最后结合像机之间的位置关系将各个像机中的特征信息进行融合,进而以闭式解法计算目标位姿。实验结果验证了该方法的有效性以及在大目标近距离位姿测量应用中的优越性。Taking pose measurement of non-cooperative space vehicle as research background, a new pose measurement method using multiple cameras with non-overlapping or slightly overlapping views was proposed to solve the problem that monocular vision-based method can not complete the pose estimation in close and ultra-close range, because of the limitation of the imaging space and the field of view, a monocular camera can not observe all the visual features used for pose estimation. Firstly, multiple cameras were configured in non-overlapping form, and the rotations and translations between cameras were calibrated. Then different cameras photographed the different features on the target, the underlying information from different cameras were both redundant and complementary, which provided adequated visual and geometric features for pose measurement. Finally, all the features information from different cameras were fused according to the relationship between cameras to calculate the pose in closed form solution. The experiment results demonstrate its effectiveness and superiority of the method for pose estimation for large target in close and ultra-close range.

关 键 词：位姿测量 计算机视觉 在轨服务 非合作目标 多像机非共视场 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]