基于共面特征点的单目视觉位姿测量误差分析  

Error Analysis of Monocular Visual Position Measurement Based on Coplanar Feature Points

作　　者：霍炬[1,2] 崔家山 王伟兴[2] 

机构地区：[1]哈尔滨工业大学电气工程系,哈尔滨150001 [2]哈尔滨工业大学控制与仿真中心,哈尔滨150001

出　　处：《光子学报》2014年第5期144-150,共7页Acta Photonica Sinica

基　　金：国家自然科学基金委创新研究群体科学基金(No.61021002);中央高校基本科研业务费专项资金(No.HIT.NSRIF.2014019);重点实验室开放基金(No.HIT.KLOF.2013081)资助

摘　　要：为了克服单目视觉测量中景深信息缺失的问题,运用改进的两段式迭代算法,对刚体位姿测量解算模型进行了优化.给出了特征标靶最优化布局,理论分析了标靶尺寸大小、标靶基线长度及特征圆点大小和解算准确度的关系.实验结果表明,优化的特征标靶布局、较大的特征标靶基线长度及较小的特征圆点靶标能够提高测量解算准确度.该研究对提高实际测量系统解算准确度具有一定的理论指导意义.In order to overcome the lack of depth information in monocular vision measurement, an improved two-stage iterative algorithm was proposed to optimize the posture and attitude measurement models. The characteristics of the target layout optimization were given. The relationships between the calculating accuracy with the target size, baseline length and characteristics of the target dot size were analyzed theoretically. Experimental results show that optimizing the layout of the target characteristic, the larger the length of the baseline and smaller characteristics of the target dots can improve the measurement target solver accuracy. The actual research has a certain guiding theoretical significance to improve the practical measurement system calculation accuracy.

关 键 词：位姿测量 单目视觉 景深估计 共面特征点 坐标统一化 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]