Abstract: This paper presents a texture-based method for detecting moving objects from a video sequence.Texture is represented by histogram of Local Binary Patterns（LBP） with uniform patterns.Because they are not affected by shadow，the texture between shadow region and corresponding region in the background is similar.Therefore the background modeling and moving objects detecting are not affected by shadow mostly.Such the approach can detect objects exactly when the shadow accompanies them.Experimental results validate the conclusion.

摘要： 提出一种基于纹理的背景建模方法，进行视频序列中运动目标的检测。其纹理的表征采用局部二元图（Local Binary Patterns，LBP）统一模式直方图，由于LBP纹理基本不受阴影的影响，使得灰度图像中阴影区域和背景相应位置的纹理表征具有一定的相似性，因此背景建模和相应的运动物体检测受运动阴影影响很小，能够在阴影条件下较精准地检测出运动物体。实验结果证明了该结论。