基于CPDH和三角形面积的形状匹配算法

Classification shape matching algorithm based on CPDH and triangle area

为了提高检索准确率,改进轮廓点分布直方图检索性能,提出一种将CPDH和局部三角形面积相结合的形状匹配方法.首先全局采用CPDH描述子描述全局信息;其次,局部上提出通过SURF算法得到对应特征点,并利用三角形面积描述子描述特征点的局部信息,同时进行对应特征点间的距离计算.在国际通用数据库Kimia-99上的检索效果很好,在MPEG-7图像库中也达到了90. 72%的检索精度,都远高于传统方法 CPDH,体现了算法的优异性.

In this letter,a shape matching method combining CPDH and triangle area representation was proposed,in order to improve the retrieval accuracy and the performance of the contour points distribution histogram（ CPDH）.Firstly,the global information was described by CPDH descriptors; secondly,SURF algorithm was employed to get corresponding feature points,the local information of which could next obtained through the triangle-area descriptors,with calculating the distance between the corresponding feature points simultaneously. This proposed method is much more prominent than the traditional CPDH because of its great retrieval accuracy on the international commonly used database Kimia-99,and an excellent retrieval precision reaching 90. 72% based on the MPEG-7 data sets.