自适应尺度的形状表示与匹配方法

Adaptive scale method for representing and matching shapes

提出了一种基于轮廓的形状特征描述方法,以及基于此形状特征的相似度度量方法.将轮廓形状的质心作为极点,在极坐标系中对轮廓等角间距采样获得轮廓点,并将相邻两采样轮廓点的极径所围面积比值作为轮廓特征向量.将改进后的归一化欧式距离作为轮廓特征向量之间的相似度距离.实验表明,基于轮廓的形状特征描述方法,仅用少量数据来表示原始轮廓信息而保留了原始轮廓信息的主要特征,且对于形状的缩放、旋转、翻转等形变操作具有尺度不变性;同时,也可以获得轮廓向量之间的相似度距离和旋转角度.实验证明了此方法可行、有效.

A novel method for shape representation and matching is presented. Firstly, the coordinate of contour centroid is calculated as the pole according to Green＇s theorem. Then contour points are sampled at equal angle interval in polar coordinates system. The ratios of area that are enclosed by the radius of contour points are regard as feature vectors of contour. Finally, an improved normalized Euclidean distance is adopted to calculate the similarity distance between two feature vectors. The new method of shape representation needs less data to represent the contour feature and preserve the main characteristic. The new method of similarity measurement can obtain angle of rotation, scale factor and flip sign effectively, which is not affected by zoom, flip, and rotation. Experimental results show that the proposed method is feasible and accurate.