Abstract: Shape recognition is an important research content in the field of computer vision and pattern recognition. Shape feature selection and description is a hot topic in shape recognition. In view of the shortcomings of the existing recognition methods, a method of feature extraction is proposed by describing different length contour segments, in which the feature parameters from six contour segments of the shape are achieved with five parameters for each contour segment. In the shape recognition stage, the Dynamic Time Warping（DTW） algorithm is used to measure the matching distance between shape descriptors to perform shape recognition. The method is verified in the Kimia99, Kimia216 and MPEG-7 databases. The results show that the shape feature descriptors based on multi-scale contour segments have invariance of rotation, scaling, translation and local occlusion, and the recognition rate is better than the existing methods.

Key words: shape description, feature extraction, shape recognition, multi-scale

摘要： 形状识别是计算机视觉与模式识别领域的重要研究内容。形状的特征选取与描述是形状识别的研究热点。针对现有识别方法的不足，提出一种通过对不同长度轮廓段进行描述，进行特征提取的方法。对每个形状均在6种尺度下进行特征提取，每种尺度选取5种轮廓段特征参数，实现了对形状的特征描述。在形状识别阶段，使用动态时间规整（DTW）算法度量形状描述子之间的匹配距离，实现形状识别。分别在Kimia99、Kimia216和MPEG-7数据库中进行算法验证，结果表明基于多尺度轮廓段的形状特征描述子具有旋转、缩放、平移和局部遮挡不变性，识别率优于现有算法。

关键词: 形状描述, 特征提取, 形状识别, 多尺度