结合边缘纹理和抽样推断的自适应阴影检测算法  

An Adaptive Shadow Detection Algorithm Using Edge Texture and Sampling Deduction

作　　者：姜柯[1] 李艾华[1] 苏延召[1] 

机构地区：[1]第二炮兵工程大学502教研室,西安710025

出　　处：《西安交通大学学报》2013年第2期39-46,共8页Journal of Xi'an Jiaotong University

基　　金：国家自然科学基金资助项目(61132008)

摘　　要：为改进阴影检测的准确度和场景自适应能力,提高运动目标检测精度,设计了一种自适应的阴影检测算法。该算法利用候选前景与原始背景的Y、U、V分量变化比率来检测阴影像素,并结合全局边缘纹理特征及抽样推断方法来估计检测阈值。算法能自动完成阈值估计及阴影判别过程而无需人工干预,并可自动适应各种光线条件,具有较强的鲁棒性。对不同光线环境的标准视频检测实验表明,该算法在精度和实时性上均有所提升,阴影检测综合性能指标达到了94%以上。An adaptive shadow detection algorithm is proposed to improve the accuracy and scene adaptive capacity of the shadow detection and to raise the effect of moving object detection.The change ratios of YUV components between candidate foreground and original background are used to detect shadow pixels,and the global edge texture and sampling deduction methods are employed to estimate the detection threshold values.The algorithm automatically complete the processes of both thresholds estimation and shadow discriminant without any manual intervention,so the algorithm is adaptive to different light conditions and has a strong robustness.Experiment results on standard videos with different lighting conditions show that both the accuracy and stability are raised by the proposed algorithm and the average comprehensive index of the proposed algorithm can reach more than 94%.

关 键 词：运动目标检测 阴影检测 YUV色彩空间 边缘纹理 自适应阈值 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]