对光照、阴影和反光具有鲁棒性的变化检测算法及实现

变化检测算法研究就是要发展一种能实现目标自动检测的算法,实现对视频图像序列的初级分析,稳定可靠地识别目标区域,实现运动图像分割,它是实现目标的识别、跟踪和报警等高层次的自动视频监视应用的重要基础,也是计算机视觉研究的一个重要领域。文章在研究了目前的多种变化检测算法的基础上,提出了一种基于线性相关的变化检测的算法。该算法以每一象素及其邻域组成的集合作为图像矢量来描述图像,用局部线性相关检测器来判断背景图像和当前图像的对应图像矢量是否线性相关,从而确定是否有变化发生。室内外的试验结果表明,基于线性相关的变化检测新算法对真实环境中的光照、阴影和反光具有较强鲁棒性,对噪声也有较强的抑制作用。该算法可准确地检测语义目标及其内部,且边缘光滑,检测目标与原图像中的目标准确地吻合,可以提高变化检测的精度,极大地增强自动视频监视系统对环境光照变化的适应能力。

A Robust Change Detection Algorithm for Varying Illumination,Shadow and Reflection Conditions

Video surveillance systems can provide varied degree of assistance to humans by providing an extended per-ception and reasoning capability about situations of interest that occur in the monitored environments.Specially,video surveillance systems can provide powerful supports to detection potential criminal activity and terror in public areas by abnormal event detection.With the increasing maturity of algorithms and techniques,they are used in various application sectors such as security,transportation,and the automotive industry.Change detection plays a very important role in video surveillance systems because object recognizing and tracking,event detection and alarm generation are based on it.This paper provides a robust linear dependence based change detection algorithm for varying illumination,shadow and reflec-tion conditions.At last,some results and conclusions are given.