测绘通报 ›› 2017, Vol. 0 ›› Issue (6): 17-20.doi: 10.13474/j.cnki.11-2246.2017.0181

• 学术研究 • 上一篇    下一篇

一种新的影像位姿直接估计方法

李艳焕1, 徐振亮2, 邵良杉1   

  1. 1. 辽宁工程技术大学系统工程研究所, 辽宁 阜新 123009;
    2. 中国资源卫星应用中心, 北京 100094
  • 收稿日期:2016-10-23 修回日期:2017-01-13 出版日期:2017-06-25 发布日期:2017-07-03
  • 通讯作者: 徐振亮。E-mail:xuzhenliang@pku.edu.cn E-mail:xuzhenliang@pku.edu.cn
  • 作者简介:李艳焕(1979-),女,博士,工程师,主要研究方向为计算机视觉、数据分析及处理。E-mail:lyh09523162@163.com
  • 基金资助:
    国家自然科学基金(71371091)

A Novel Direct Estimation Method for Image Pose

LI Yanhuan1, XU Zhenliang2, SHAO Liangshan1   

  1. 1. System Engineering Institute, Liaoning Technical University, Fuxin 123009, China;
    2. China Centre for Resources Satellite Data and Applications, Beijing 100094, China
  • Received:2016-10-23 Revised:2017-01-13 Online:2017-06-25 Published:2017-07-03

摘要: 针对影像位姿耦合不易直接求解的问题,建立了一种高精度分层位姿参数估计的方法。首先,通过选择3点构成面积最大的像点对应的物点作为基点,经投影方程的改化形式方程,得到估计深度及平移信息;然后利用旋转矩阵保范性获得相机外参线元素尺度参数;最后,将问题转化为由向量估计旋转关系问题,利用Procrustes理论得到外参角元素的最佳估值。该方法可有效解决该深度与耦合位姿参数难以处理的问题。试验证明,此方法确定的位姿参数估计模型形式简洁,容易收敛,参数估计精度较直接对投影矩阵分解方法质量要高。

关键词: 齐次坐标, 直接线性变换, 基点, Procrustes, 位姿估计

Abstract: In order to solve the problem of image pose coupling calculating, this paper has established a high-precision hierarchical estimated pose parameters of image. Firstly, we select corresponding three image points of 3D points which constitute the largest area in image as a base to estimate the depth and translate information. Then based on the above method, we obtain the scale parameter of camera exterior information. Finally, the topic is transformed to a problem of estimating rotation relationship by vector, using Procrustes theory to obtain the best estimation of the angle elements of external parameters. The method can effectively solve problems which depth and coupling pose parameters can not deal with. Experimental results show that this method of determining position and orientation parameter estimation model is of briefness, easy convergence and it can also achieve higher parameter estimation accuracy than the direct projection matrix factorization.

Key words: homogeneous coordinates, direct linear transformation (DLT), base point, Procrustes method, pose estimation

中图分类号: