一种基于三焦点张量的度量重建方法  

Approach to metric reconstruction based on trifocal tensor

作　　者：黄以君[1,2] 刘伟军[1] 赵吉宾[1] 

机构地区：[1]中国科学院沈阳自动化研究所先进制造技术重点实验室,沈阳110016 [2]中国科学院研究生院,北京100039

出　　处：《仪器仪表学报》2009年第6期1307-1312,共6页Chinese Journal of Scientific Instrument

摘　　要：从二维图像恢复物体的三维结构是计算视觉的一个重要研究方向。通过研究三视图之间的对极几何,提出了一种基于三线性关系的度量重建方法。不需要对相机的运动或者景物结构施加任何约束,使用直接重建方法,利用三焦点张量恢复场景的度量重构。仿真实验和真实图像实验表明,该度量重建方法具有很高的准确性和实用性。Reconstructing the structure of a scene from 2D views has been an important research field in computer vision. By investigating the epipolar geometry between three views, this paper develops a novel method for metric reconstruction of a scene based on the trilinear relations. Our algorithm does not put any restriction on the camera movement or the scene structure; the algorithm adopts direct reconstruction method and applies the trifocal tensor to metric reconstruction. Experiments with both simulation data and real image data show that the proposed method of metric reconstruction is very accurate and practical.

关 键 词：度量重建 三线性关系 三焦点张量 计算视觉 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]