多摄像机场景下的目标匹配比较研究  

Comparison of Object Matching Method in Multi-camera Scenario

作　　者：戴玥[1] 陆小锋[1] 朱民耀[1] 王晓丽[1] 

机构地区：[1]上海大学视频与空间信息学科组,上海200072

出　　处：《电视技术》2013年第19期219-223,共5页Video Engineering

基　　金：上海大学创新基金项目(A10-0107-09-902)

摘　　要：基于如颜色直方图(Color-histogram)等特征的目标匹配方法由于特征本身受光线变化、背景复杂度等因素影响而对于不同场景中目标的识别准确率存在差异。针对这一现状,选取了基于Color-histogram、SIFT和SURF特征的目标匹配方法分别应用于多摄像机场景下的目标匹配。然后,通过纵向比较同一种方法应用在不同视频集中所显示出的识别率和横向比较同一视频组中应用不同方法所得到的识别率来分析它们之间应用效果的差异性。由实验结果可知,该比较研究所得结论可以为此领域的后续研究提供一种关于特征选择的参考。Due to factors such as illumination change and background complexity,the performance of the object matching method,which is based on color-histogram or the other features,varies in different scenarios.With respect to the current situation,three kinds of classical features are selected to apply in object matching in multi-camera,and they are color-histogram,SIFT and SURF.Then,the application results are analyzed by comparing the recognition rates of each method applied in different video datasets.Experimental results show that it can distinguish the most suitable scenario for each feature matching method by a series of performance comparison,and it will provide a reference with feature selection for the subsequent researches in this field.

关 键 词：特征匹配 Color-histogram SIFT SURF 比较研究 

分 类 号：TN911.73[电子电信—通信与信息系统]