中文摘要：

提出一种基于ViBe算法的改进运动目标检测方法。改进的方法利用20帧图像进行背景模型的初始化，通过比较SILTP特征值来对前后景像素进行分类，模型更新机制采用替代异常值的方式。在不同场景下的实验结果表明，改进后的算法对光照变化、鬼影产生和噪声具有强鲁棒性。

英文摘要：

An improved method of moving objects detection is proposed, which is based on the ViBe algorithm. The improved algo- rithm uses background model initialization of twenty images and elassifys the foreground and background pixels by compariing the SILTP characteristic values. Also, the model updating mechanism of the improved algorithm is the method of replacing the abnor- mal values. Under different eonditions, the results of experiments show that the improved algorithm has stronger robustness to illu- mination changes, the emergence of ghost and noise than ViBe algorithm.