摘要
基于三维重建中物体截面轮廓边缘序列中相邻轮廓基本相似的事实 ,提出了一种物体截面轮廓线点集之间的三角划分的快速算法。该算法首先寻找轮廓线的控制点 ,对每两条相邻轮廓线的控制点进行配对 ,其次将这样两条轮廓线之间的三角划分的任务分配在多个小的曲线段对之间进行 ,并且由于事先的控制点匹配 ,在每个小的区域中就能够进行简化快速的三角划分。通过试验验证了本算法的有效性。
We proposed an approach to fit both global request and speed of recreation.That is,considering the fact that two adjacency cross_section contours of object are basically similar to each other,we first segment the area into several sections coarsely,and then triangulate within each section.What needs to be done first is detecting corner points.Many methods for corner detection have been proposed,such as corner detection based on multi_scale wavelet transformation,direction of edge chain code,CSS(curvature scale space) and polygon approximation about curvature,etc.Since some noises exist in curve,applied method about corner detection with de_noise ability is needed.We use common global optimization method.Here the advanced corner detection based on CSS is adopted.By using multi_scale processing,corner points are extracted from contours automatically.The second step is searching counterpart corner points from the two sets of corner dots.This step is to judge whether the two contours are similar with each other and to segment the space between them.For all similar contours,the quick triangulation algorithm introduced in this paper is applied.But for dissimilar contours,any global optimization method can be applied in the triangulation between them.We judge whether two corner dots in different contours are pairs based on the distance between them.A dynamic threshold is used and it depends on the distance between the planes in which the two contours lay and the curvatures at the dots.After the procession for corners,the similarity of contours should be confirmed.We think two contours to be similar if the number of corner pairs in them exceeds 70% that of corners in the contour with fewer corners on it.The final step is triangulation between two contours.For similar adjacency contours,we segment the space between them into smaller sections based on the matched corners.For each smaller section,any triangulation algorithms can be used but here we use a simple and quick method.During triangulating,the algorithm decides the next triangle on the basis of keeping the proportion of the numbers of the dots not used on two contours in accordance with that of total numbers.This algorithm distributed the task of triangulation between two adjacency contours into several smaller sections of them.Because of the corners' matching,a simple and quick triangulation could be used.Experiments show that the visual effect of our method is basically accord with that of normal global optimizing methods,but much less time is needed.So our method is effective.
出处
《武汉大学学报（信息科学版）》
EI
CSCD
北大核心
2003年第2期227-232,共6页
Geomatics and Information Science of Wuhan University
基金
国家自然科学基金资助项目 ( 40 10 10 19与 40 0 2 30 0 )
测绘遥感信息工程国家重点实验室开放研究基金资助项目 (WLK( 0 0 )0 10 2 )
