Abstract: The classical W4 algorithm can not eliminate the shadow of moving object, so a moving target detection algorithm is proposed. Firstly, it gives initial frame and calculates the minimum, the maximum and the maximum frame difference gray value with the traditional W4 algorithm. Secondly, the maximum gray value and the minimum gray value of each pixel are extracted. Then it combines with LBP texture features and adopts the idea of Gaussian mixture model that uses multiple modes to describe background model. In order to reduce matching complexity and reach real-time, many kinds of LBP are cut down. Experimental results show that the algorithm can effectively reduce the impact of shadow on moving target detection, and also meet the requirements of real-time.

Key words: target detection, W4 algorithm, Gaussian mixture model, Local Binary Pattern（LBP） texture features, shadow

摘要： 针对传统W4背景模型算法无法消除运动目标阴影的问题，提出了一种有效消除阴影的运动目标检测算法。首先，给定背景初始帧，用传统W4算法计算出每一个像素点的最小灰度值、最大灰度值以及最大相邻帧间差分值；其次，对每个像素点提取的最大灰度值和最小灰度值进行线性加权；之后结合能抵抗阴影影响的改进的LBP纹理特征，采用类似混合高斯背景模型原理的思想提取多个运动目标检测背景模型。精简提取得到的LBP纹理的种类，减少计算量，以达到实时性的要求。实验结果表明，该算法与同类算法相比更有效地去除阴影对运动目标检测的影响，也满足实时性的要求。

关键词: 目标检测, W4算法, 混合高斯模型, 局部二值模式（LBP）背景模型, 阴影