摘 要：针对目前运动目标检测算法中常将阴影误检为前景目标的问题，提出一种基于Phong物体光照模型的阴影检测算法。基于Phong物体光照模型，我们对场景中象素的亮度值进行分析，通过定义一个亮度相对变化量，推导出他在整个阴影区域是比较稳定的，所以在一个(5×5)的模板上用协方差来衡量这种稳定性，从而得到第一个阴影判决式。又推导出阴影区域亮度相对变化量随时间的变化保持相对稳定，设计一个滤波模板来增大目标区域的不稳定性，从而得到第二个阴影判决式。最后结合以上二个阴影判决式进行阴影检测，并对实验结果进行定性和定量的评估。与前人提出算法比较，本文提出的算法在阴影检测率和区分率等方面都得到了提高，具有较强的鲁棒性。
关键词：Phong光照模型;阴影检测;运动目标检测;智能监控
中图分类号：TP391 文献标识码：B
文章编号：1004373X(2008)0512404
A Shadow Detection Algorithm Based on Phong Lighting and Radiosity Model
WU Liang,ZHOU Dongxiang,LIANG Hua，CAI Xuanping
(School of Electronic Science and Engineering,National University of Defense Technology,Changsha,410073,China)
Abstract：Focusing on the problem that shadows cast by moving objects are detected incorrectly as foreground targets by most of the current moving objects detection algorithm,a method of shadow detection based on the Phong lighting and radiosity model is proposed.Based on the Phong model,we analyze the brightness of pixels in image sequences,the Relative Change of Brightness (RCB) in shadowed regions is proved to be more stable than moving objects regions,it is measured by the covariance of RCB of pixels on a template (5*5) so as to acquire the first discriminant.As the RCB in shadowed regions is stable in image sequences,a filter template is designed to make the RCB more unstable in regions of moving objects,so the second discriminant is presented.Shadow detection is carried out by fusing the two discriminant formulas described above,experimental results are evaluated quantitatively and qualitatively,and show that our method is robust and offers more advantage over other algorithms presented previously on detection rate and discrimination rate.
Keywords：Phong lighting and radiosity model;shadow detection;moving object detection;intelligent monitoring
1 引 言
智能视频监控中一个重要技术是实现对运动目标检测，但是运动目标及其投影阴影[1]经常同时被检测为运动前景，因此阴影检测对运动目标准确检测是至关重要的，也对后续的跟踪、识别和分类产生重要影响。（剩余1149字）