基于ViBe的运动目标检测改进方法

发布时间：2018-02-02 15:01

本文关键词： 视觉背景提取算法光线突变帧间差分背景差分运动目标检测背景模型重置　出处：《计算机工程与设计》2017年02期 　论文类型：期刊论文

【摘要】：针对运动目标检测算法视觉背景提取(visual background extractor,ViBe)对光线突变适应缓慢和易产生目标丢失等问题,提出一种ViBe的改进算法。充分利用ViBe算法初始化特点,在ViBe中加入光线判定方法并设计生成ViBe新背景模型的方法,改变原始ViBe算法结构,使ViBe算法在光线突变时找到合适的帧重置背景模型,将光线突变时模型恢复问题变为模型重置问题,达到迅速适应光照突变的目的。实验结果表明,使用该方法可以使ViBe有两方面的改进:在光线突变时不会产生目标丢失;使ViBe算法更加快速地适应光照变化。
[Abstract]:Visual background extractor is extracted from the visual background of moving target detection algorithm. In order to solve the problems of slow adaptation of light mutation and easy loss of target, a modified ViBe algorithm is proposed, which makes full use of the initialization characteristics of ViBe algorithm. In order to change the structure of the original ViBe algorithm, a new background model of ViBe is designed by adding the light judgment method into ViBe. The ViBe algorithm can find the appropriate frame reset background model when the light changes, and change the model restoration problem into the model reset problem, so as to adapt to the illumination mutation quickly. The experimental results show that. Using this method, ViBe can be improved in two aspects: when the light is abrupt, the target will not be lost; The ViBe algorithm can adapt to the illumination change more quickly.
【作者单位】：中北大学计算机与控制工程学院;
【基金】：国家科技支撑计划基金项目(2013BAH45F02)
【分类号】：TP391.41
【正文快照】： 0引言常用的运动目标检测方法[1,2]主要有帧间差分法、背景差分法以及光流法。背景差分法是目前最为流行的检测方法,它结果的好坏取决于背景模型。背景模型分为两类:1参数化模型,利用含参模型对每个像素点建模。2非参数化模型,使用已观察的像素值对该像素点建模[3]。其中参数

【参考文献】