摘要
文中提出了一种针对多视角未定标图像序列的三维测量算法,该算法无须先验信息只需利用多视角图像序列.首先对图像序列进行特征点提取进而得到基础矩阵,然后利用简化的Kruppa方法计算各个视角对应摄像机的内参数,再次通过奇异值分解得到视角间的平移和旋转运动,最后利用已知距离信息进行三维点重建完成三维测量.实验室情形下的实验证明了算法的效果.
In this paper, a novel algorithm aiming at solving the problem of uncalibrated image based 3D t is presented. Firstly, feature extraction is done to generate the matching pair and further the fundamental matrix is calculated. Secondly, simplified Kruppa equation is used to obtain the intrinsic parameters corresponding to each view. Thirdly, through using singular value decomposition the translations vector and rotation matrixes between views are acquired. Finally, 3D point reconstruction and t is done by making use of the known distance. Results of experiments in lab environment show the effect of the proposed algorithm.
出处
《微电子学与计算机》
CSCD
北大核心
2010年第9期181-185,189,共6页
Microelectronics & Computer
基金
国家自然科学基金项目(60736007)
关键词
三维测量
多角度
自定标
基础矩阵
3D measurement
multi-view
self-calibration
fundamental matrix
