测量误差不确定性加权的立体视觉位姿估计目标函数  

A objective function with measuring error uncertainty weighted for pose estimation in stereo vision

作　　者：霍炬 张贵阳[1] 崔家山[3] 杨明 HUO Ju;ZHANG Gui-yang;CUI Jia-shan;YANG Ming(School of Astronautics,Harbin Institute of Technology,Harbin 150001,China;School of Electrical Engineering and Automation,Harbin Institute of Technology,Harbin 150001,China;The 8357 Institute,China Aerospace Science&Industry Corp.,Tianjin 300308,China)

机构地区：[1]哈尔滨工业大学航天学院,黑龙江哈尔滨150001 [2]哈尔滨工业大学电气工程与自动化学院,黑龙江哈尔滨150001 [3]中国航天科工集团8357研究所,天津300308

出　　处：《光学精密工程》2018年第4期834-842,共9页Optics and Precision Engineering

基　　金：国家自然科学基金资助项目(No.61473100,No.61021002);重点高校基础研究经费资助项目(No.HIT.NSRIF.2014019);哈工大重点实验室建设基金资助项目(No.HIT.KLOF.2013081)

摘　　要：针对视觉测量过程中配合目标特征点在成像时灰度模式的各向异性且非独立同分布对位姿估计求解的影响,建立了基于特征点测量误差不确定性加权的位姿估计目标函数。利用协方差矩阵描述特征点的方向不确定性,分析成像特征点的不确定性对目标函数的作用权重,并将特征点测量误差的不确定性融入到空间共线性误差函数中,进而构造了基于成像特征点测量误差不确定性加权的新目标函数。该方法适应于特征点测量误差具有不同程度的方向不确定性的情况,最后通过广义正交迭代算法对该目标函数进行迭代优化求解。实验表明,当测量空间为2 300 mm×1 400 mm×1 400mm,采用新目标函数得到靶标重投影后图像坐标的最大误差不超过0.11pixel,立体视觉系统对标准杆的相对测量精度优于0.01%。所得的结果验证了新目标函数的位姿估计精度高、稳定性强,适用于实际的工程应用。In order to investigate the effect of anisotropic and correlated non-identical gray distributions of feature points on pose estimation,a novel objective function with error uncertainty weighted of feature points was proposed.In the method,the inverse covariance matrix was utilized to describe the directional uncertainty of feature points,and weighted contribution of uncertainty to the error objective function was analysed.By transforming the uncertainty into a covariance-weighted data space,the novel objective function was constructed,which was suitable for actual applications.Finally,the optimized solution to the novel objective function was obtained via generalized orthogonal iterative algorithm.The simulation and practical experiments show that the maximum error of re-projection image coordinates of the target is less than 0.11 pixels and the measurement relative error for standard gauges is superior to 0.01%within the space 2300 mm×1400 mm×1400 mm.The results verify the high accuracy and strong robustness of the proposed approach,and should therefore have potential for engineering applications.

关 键 词：立体视觉 不确定性 误差加权 目标函数 位姿估计 

分 类 号：TP394.1[自动化与计算机技术—计算机应用技术] TH691.9[自动化与计算机技术—计算机科学与技术]