张　峰　熊庆国

(武汉科技大学信息科学与工程学院　湖北 武汉 430081)



基于分层码本模型的阴影消除方法

张峰熊庆国

(武汉科技大学信息科学与工程学院湖北 武汉 430081)

摘要为了解决复杂环境，特别是背景改变带来的背景模型变化，对运动目标的阴影检测与消除效果产生影响的问题，给出一种在YCbCr颜色空间下基于分层码本模型的阴影消除方法。目标检测时采用分层建模与检测技术检测出运动区域并更新背景模型，从永久背景模型中提取背景图像并采用基于色度的阴影检测方法消除阴影，避免在阴影检测与消除中使用固定背景的问题。在公开测试集上的实验结果表明，该方法能有效地适应背景的动态变化，对复杂环境中的前景目标检测与阴影消除有很好的效果。

关键词阴影检测与消除背景建模分层码本目标检测

SHADOW ELIMINATION APPROACH BASED ON LAYERED CODEBOOK MODEL

Zhang FengXiong Qingguo

(School of Information Science and Technology,Wuhan University of Science and Technology,Wuhan 430081,Hubei,China)

AbstractThe background model changes brought about by the complex environment, and in particular the variation of background have the impact on the shadow detection and elimination effects for moving objects. In order to solve this problem, we proposed a shadow elimination method which is based on layered codebook model in YCbCr colour space. In objects detection phase, the method uses the stratified modelling and detection technique to detect the moving regions and to update the background model, and extracts from the permanent background model the background images, and adopts the chroma-based shadow detection method to remove shadows, thus avoids the problem of using the fixed background when detecting and removing shadows. Results of experiment on public test set demonstrated that the method was able to effectively adapt to the dynamic changes of background, and had excellent effect on foreground objects detection and shadow elimination in complex environment.

KeywordsShadow detection and eliminationBackground modellingLayered codebookObject detection

0引言

一种常用的运动目标检测方法是背景减除法，即将当前图像与参考背景模型相减或找出不同点，从而标识非静止或新出现的物体[1]。但在检测运动目标时，因目标的阴影具有和前景目标相似的运动模式及色彩变化，会很大程度上影响目标的精确提取，从而带来错误的目标检测及分类结果。针对背景无明显变化的场景现有技术能够稳定地检测出运动目标并消除阴影。但是在一些背景动态变化的场景中，即背景在连续的图像序列中不断地发生变化，此时运动目标检测、阴影检测及消除变成了一个难点，对此应采取相应的背景更新策略。……