基于形状上下文的目标图像匹配  

Object Image Matching Based on Shape Context

作　　者：张雁腾 朱昌洪[1] 潘天赐 Zhang Yanteng;Zhu Changhong;Pan Tianci(Bowen College of Management,Guilin University of Technology,Guilin Guangxi 541006,China)

机构地区：[1]桂林理工大学博文管理学院,广西桂林541006

出　　处：《信息与电脑》2017年第5期90-92,共3页Information & Computer

基　　金：桂林理工大学博文管理学院科研项目(KY201505):

摘　　要：形状上下文描述子多用于目标识别,是一种基于形状轮廓的特征描述方法,在对数极坐标系下利用直方图描述形状特征能够很好地反映轮廓上采样点的分布情况。笔者从匹配率、匹配代价、弯曲能量最小值三方面做出改进,经实验证明,通过设定阈值判定匹配程度和识别结果,在减少匹配的迭代时间的同时取得更好的匹配结果。The shape context descriptor is used for object recognition.It uses a feature description method based on shape contour,which uses the histogram to describe the shape feature in the logarithmic polar coordinate system to reflect the contour sampling point distribution.The author from the matching rate,the matching cost,minimum bending energy value of three made improvements,proved by the experiment,by setting the threshold to determine the matching degree and the recognition result,obtain better matching results in reducing the iteration time at the same time.

关 键 词：形状上下文 TPS形变模型 形状匹配 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]