基于单目视觉多视图平面点位置的最优估计  

Optimal Estimation of Multi View Plane Point Location Based on Monocular Vision

作　　者：丁尚文 王纯贤 DING Shang-wen;WANG Chun-xian(Department of Basic Course,Xuancheng Campus,Hefei University of Technology,Xuancheng 242000,China;School of Mechanical Engineering,Xuancheng Campus,Hefei University of Technology,Xuancheng 242000,China)

机构地区：[1]合肥工业大学宣城校区基础部,宣城242000 [2]合肥工业大学宣城校区机械工程系,宣城242000

出　　处：《自动化与仪表》2018年第10期15-18,58,共5页Automation & Instrumentation

基　　金：校博士专项科研资助基金项目(JZ2015HGBZ0122);合肥工业大学教学质量与教学改革工程项目(JYQN1704);合肥工业大学宣城校区博士产学研专项基金项目(XC2015JZBZ23)

摘　　要：针对空间点的位置估计问题,提出基于单目视觉多角度测量的平面点位置最优估计。该方法通过透视成像模型获得特征点与投影点相关联的投影矩阵;通过不同视角下的投影矩阵重建特征点坐标;由于相机标定误差的影响,不同视角下重建的特征点坐标不完全相等,通过特征点坐标误差最优估计模型获得特征点坐标最优估计值。通过试验验证了本文算法的可行性。试验结果表明,该算法能够提高空间平面点的位置精度,满足视觉导航空间平面点的定位精度要求。Aiming at the location estimation of space points,this paper presents the optimal estimation of the location of plane points based on monocular vision according to multi angle measurement.The projection matrix associated with the feature point and the projection point is obtained by the perspective imaging model.The feature point coordinates are reconstructed by the projection matrix from different angles of view.The feature point coordinates of the reconstructed image are not completely equal because of the camera calibration error,and the optimal estimation mode is obtained through the coordinate error of the feature points.The optimal value of the feature point coordinates is obtained.The feasibility of the algorithm is verified by real experiment.The experimental results show that the algorithm can improve the position precision of the space plane point and meet the positioning precision requirements of space plane point for the visual navigation.

关 键 词：单目视觉:投影矩阵:位姿:像素 

分 类 号：TB921[一般工业技术—计量学]