多摄像机在无重叠区域的目标匹配算法  

Target Matching Algorithm in Multiple Non-overlapping Cameras

作　　者：王晓丽[1] 陆小锋[1] 陆亨立[1] 

机构地区：[1]上海大学通信与信息工程学院,上海200444

出　　处：《电视技术》2015年第10期11-16,共6页Video Engineering

摘　　要：基于颜色特征的目标匹配方法具有很好的区分性和直观性,但是特征本身受光线变化、背景复杂度等影响而对于目标识别准确率存在差异。针对这一现状,提出了一种新的基于时空局部特征和颜色特征的目标匹配方法。利用时空局部特征3-D LSK对待测视频中的运动目标进行行为识别,识别出具有相同行为目标。再结合目标的颜色特征,通过基于区域权重(Weighted Region Matching)的颜色匹配方法来识别出具有相同颜色的目标。对无重叠区域的多个测试视频进行的相关实验表明,该算法能够有效识别出普通场景下的不同目标,对于衣着颜色相近的不同目标也有很高的识别率。The target matching method of color features has good distinction and intuition, but small dependence on the shape and attitude of the target. Due to factors such as illumination change and background complexity, the performance of the object matching method varies in different scenarios. With respect to the current situation, this paper proposes a new target matching algorithm, combining with space-time local features and color features. Firstly, 3-D LSK （space-time local steering kernel） features are used to identify the behavior of moving targets in video. Then color features and Weighted Region Matching method are used to identify targets with the same color. Experimental results on several test videos in non-overlapping areas show that the new algorithm can effectively identify the different targets in common scenarios, especially for different targets dressed similar color.

关 键 词：颜色特征 时空局部特征 无重叠区域 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]