一种基于曲率提取轮廓特征点的方法

对边缘轮廓线进行了自适应平滑，在以轮廓点为中心的支撑区域内定义了曲率角，并根据定义的曲率角及曲率符号，采用简单的模板匹配技术，提取出候选的轮廓特征点，对候选点附加抑制条件得到真正的特征点．此方法减少了数字化误差及噪声带来的影响，抗干扰性好，且运算量较小．实验表明，该方法具有很好的轮廓特征点检测及定位能力

Approach to Extract Feature Points on Boundary Based on Curvature

Feature points were detected by first applying adaptive smoothing to coordinate function along the curve so that the quantization error and noise can be reduced. The curvature angle was defined based on supporting area of which the boundary pixel is the center. The candidates of feature point were detected based on curvature orientation and its sign pattern with template matching technique. Certain constraints were applied to select the actual feature points. The experimental results show good capability of detection and localization.