基于三视图几何的树木点位图重构  

Reconstruction of tree position map based on three-view geometry.

作　　者：张青[1] 唐守正[2] 

机构地区：[1]北京林业大学理学院 [2]中国林业科学研究院资源信息研究所

出　　处：《北京林业大学学报》2008年第S1期183-186,共4页Journal of Beijing Forestry University

基　　金：国家自然科学基金项目(30271079)

摘　　要：为了探求计算机视觉在森林调查中的应用,该文运用计算机视觉的三维未标定的重构技术,对树木点位图的测量进行了尝试。首先运用未标定CCD相机对样地获取3张像片,根据多视图几何计算3张图像的几何关系,求算三焦点张量,进行图像匹配,然后由仿射重构的矩阵分解算法恢复树木特征点的仿射空间坐标,最后将仿射坐标点校正到度量坐标下,完成点位图测量。To study the application of computer vision in forest inventory,a new method to plot the map of treeposition by uncalibrated 3D reconstruction is presented in this paper.First,three images of one stand wereobtained with an uncalibrated CCD camera,and then geometrical relationship of the three images was caculatedby multi-view geometry.The three images were matched based on the trifocal tensor.Then some feature pointsof trees were reconstructed in an affine coordinates frame through the matrix factorization algorithm of affinereconstruction.Finally,after the affine reconstruction was transformed to a metric reconstruction,the map oftree position was obtained.

关 键 词：计算机视觉 重构 三视图几何 三焦点张量 点位图 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]