摘要
为有效检测三角网格曲面上的角点特征,提出一种基于最小主曲率的角点检测算法.首先通过计算网格顶点处的最小主曲率,利用加权最小主曲率定义角点特征函数,并计算角点特征值;然后利用迭代阈值法自动产生检测阈值,以去除噪声和特征不明显的角点;最后采用非极大值抑制法消除局部邻域内的角点聚簇获取特征明显的角点.在此基础上,在多个尺度下分别计算每个网格顶点处的角点特征值,并通过加权将其合并成多尺度角点特征值,新的角点特征值使得角点检测算法具有较高的稳定性和鲁棒性.通过重复检测率实验和部分重叠曲面的配准实验,验证了文中算法的有效性与实用性.
A new the minimum princi algorithm is proposed to detect corners on the triangular mesh surfaces. Based on pal curvature, a corner feature function at each vertex is evaluated, which accounts for the variance of minimum principal curvature within a local area. Then an iteratively determined threshold of the corner feature function is applied to remove noisy or faint corners. Further, the nonmaxima suppression method is employed to extract the distinct corners from local clusters of candidates. To make the corner detection algorithm more robust, the above process is conducted on mesh vertices under different scales to form a multi-scale feature representation at each corner. Experiments on repeated corner detection and registration of partially overlapping surfaces demonstrate the effectiveness and robustness of the proposed algorithm.
出处
《计算机辅助设计与图形学学报》
EI
CSCD
北大核心
2009年第11期1545-1550,共6页
Journal of Computer-Aided Design & Computer Graphics
基金
国家自然科学基金(50875130)
博士点基金(200802870016)
江苏省支撑计划(BE2008136)
关键词
网格曲面
角点检测
最小主曲率
多尺度特征表示
Mesh surfaces
corner detection
minimum principal curvature
multi-scale featurerepresentation
