多特征点拓扑确定位姿测量算法研究

发布时间：2018-03-07 14:14

本文选题：位姿测量　切入点：拓扑确定　出处：《红外与激光工程》2017年05期 　论文类型：期刊论文

【摘要】：为解决单目视觉位姿测量时,由目标特征点较多导致图像点与目标点拓扑关系未知的问题,提出了一种多特征点拓扑确定位姿测量算法。较多特征点可在目标进行大角度运动时保证足够的特征点进行位姿解算,与较少特征点相比提高测量精度。该算法将拓扑确定的过程和位姿求解的迭代过程进行嵌套,同时进行拓扑确定和位姿计算。位姿计算的迭代过程基于平行透视投影模型,不需要目标重心投影点坐标作为迭代初始值。拓扑确定的过程转化为分配问题的求解过程。每次位姿迭代的过程中进行一次拓扑确定,拓扑确定的结果可以计算更优的位姿估计。通过多位姿测量实验和精度对比实验结果证明:该算法适合大范围、高精度的位姿测量,在-120°~120°范围内,位姿测量均方根误差为0.272°。
[Abstract]:In order to solve the problem that the topological relationship between image points and target points is unknown due to the large number of target feature points in monocular vision pose measurement. In this paper, a multi-feature point topologically accurate orientation and attitude measurement algorithm is proposed. More feature points can guarantee sufficient feature points for pose calculation when the target is moving at large angles. Compared with fewer feature points, the algorithm nests the process of topology determination and the iterative process of position and pose solution, and carries out topology determination and pose calculation. The iterative process of position and pose calculation is based on parallel perspective projection model. There is no need for the coordinate of the target barycenter projection point as the initial value of the iteration. The process of topology determination is transformed into the solving process of the assignment problem. The results of topology determination show that the algorithm is suitable for large range and high precision pose measurement, and the root-mean-square error is 0.272 掳in the range of -120 掳~ 120 掳.
【作者单位】：天津大学精密测试技术及仪器国家重点实验室;光电控制技术重点实验室;
【基金】：国家自然科学基金(51375339) 航空科学基金(20145148009)
【分类号】：TP391.41

【参考文献】