一种基于曲率法的曲线特征点选取方法

An algorithm of extracting feature points in curve based on the curvature

本文提出使用弯曲树来组织曲线的弯曲形态,根据弯曲特征以一定范围计算曲线弯曲程度所得到的值度量点的宏观弯曲量,作为特征点的选取标准;并根据微观弯曲量微调领域内特征点的位置,来确定特征点。实验对比发现,本算法可以在一定程度上克服传统曲率算法的不足,选取的特征点能够控制曲线的宏观形态,并且位置准确。

This paper proposed to organize the curves by bend tree, select bends through the level of the bend tree, and detect the feature points in the corresponding selected bends. On the one hand, the macro-curvature which is calculated by the macroscopic cam- ber of the curve was used as the criterion to evaluate the feature points roughly. On the other hand, the micro-curvature was used to de- termine the exact feature points around the rough ones. Comparison experiments showed that some of the feature points extracted by Douglas-Peucker algorithm were not the real feature points, and a number of the feature points extracted by traditional curvature algo- rithms were unwanted, and may cause unnecessary trembles in the curve, while the feature points extracted by the proposed algorithm could avoid those above defects.