基于空间特征的阴影检测  

Shadow Detection Method Based on Spatial Feature

作　　者：徐韶华 赵勇[2] 

机构地区：[1]安防科技(中国)有限公司,深圳518034 [2]北京大学深圳研究生院,深圳518055

出　　处：《现代计算机》2009年第1期70-72,共3页Modern Computer

摘　　要：针对城市交通监控系统,提出了基于空间特征的运动目标阴影检测算法。首先,用混合高斯模型及边界提取算子获得所有前景区域,然后计算每个前景区域的水平位置-垂直方向上前景点个数的直方图,通过最优阈值分割获得合适的阈值,去除阴影区域,并且每隔一段时间,自动学习并确定阴影与目标的方位关系。对不同光照条件的视频测试表明,该方法具有良好的自适应性、鲁棒性和实时性。Proposes a Shadow detection algorithm base on spatial features, in the case of focusing on traffic vehicle detection system. First of all, multi-foreground rectangles are extracted using Gaussian Mixture Model （GMM） and edge detection operator of mathematical morphology. Then, histogram of horizontal location-foreground points number of vertical direction is computed, combined with optimum threshold segmentation, shadow areas are deleted. To enhance the adaptability, the system learns direction relations between target and its shadow, automatically. Results are presented for several video sequences representing a variety of illumination conditions. The results of experiment on the video of different traffic conditions show the robust, self-adaptive, and real- time of our technique.

关 键 词：混合高斯模型 最优阈值分割 阴影分离 空间特征 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术] TN911.25[自动化与计算机技术—计算机科学与技术]