基于码本和运行期均值法的双层背景建模方法

背景建模是视频处理的重要部分,是后续运动目标检测、识别和跟踪的基础。针对现有的背景建模方法无法兼顾抗干扰性、适应光照、背景更新速度和遮挡等问题,提出结合码本和运行期均值法对视频进行双层背景建模的方法。首先在第一层提取亮度和颜色特征,使用聚类的方法进行码本建模,接着在第二层建立运行期均值法模型,通过两种背景模型的有效结合快速准确地实现运动目标分割。实验结果表明,该背景建模方法计算简单、背景更新快、抗噪声能力强并且能较好地适应光照变化,适应于复杂环境下的目标检测。

A two layer background modeling method based on codebook and running average

Background modeling is a critical task in video processing, and it is the foundation of subsequent detection, recognition and tracking of moving objects. Considering the existing problems in background modeling methods, we propose a two layer background model method based on codebook and running average. Firstly, we extract brightness and color features in the first layer and build the codebook model using the method of clustering. Then in the second layer we construct the running average model, and through the effective combination of the two background models we achieve a fast and accurate moving object segmentation. Experimental results show that the background modeling method has advantages of low computation complexity, strong anti noise ability and fast background update speed, and it can also adapt to light change, thus applicable to moving object detection in complex environments.