您的位置:首页  > 论文页面

一种基于自组织神经元网络的背景差分算法研究与实现

发表时间:2012-04-30  浏览量:1414  下载量:465
全部作者: 吴晶,刘亮
作者单位: 北京邮电大学计算机学院
摘 要: 结合实际场景,重点分析和发现了基于自组织神经元网络的背景差分(self-organizing neural network based background subtraction,SOBS)算法的缺陷,并进行相应的改进。主要研究工作如下:1) 设计自动建立空背景算法,弥补了SOBS算法不能自动空背景建模的缺点;2) 从阴影的物理、概率统计特性及上层的经验知识进行研究,并且利用这些特性改善SOBS算法阴影检测的弱点;3) 使用简单的YCbCr色彩空间代替HSV的色彩空间,提高算法效率。最后,大量的实验结果显示所提出的改进算法有效。
关 键 词: 计算机应用技术;背景差分;自组织神经网络;混合高斯模型;空背景重建;阴影检测
Title: Research and realization of a self-organizing neural network based background subtraction algorithm
Author: WU Jing, LIU Liang
Organization: School of Computer Science, Beijing University of Posts and Telecommunications
Abstract: This paper emphasizes on analysis and disadvantages of self-organizing neural network based background subtraction (SOBS) in the special scenes, and propose some algorithms to improve it. 1) This paper proposes an algorithm to automatically and quickly empty the background modeling algorithm, which makes up the fault in SOBS and improve its usability. 2) This paper studies the physical properties, characteristics of probability and statistics of the shadow, as well as the level of knowledge on the experience of the shadow, and uses these characteristics to improve the weaknesses of shadow detection algorithm in SOBS. 3) This paper uses simple YCbCr color space instead of HSV color space for improving efficiency. Finally, a large number of experimental results show the improvement made by the algorithm is effective.
Key words: computer application technology; background subtraction; self-organized neural network; Gauss mixture model; background reconstruction; shadow detection
发表期数: 2012年4月第8期
引用格式: 吴晶,刘亮. 一种基于自组织神经元网络的背景差分算法研究与实现[J]. 中国科技论文在线精品论文,2012,5(8):714-720.
 
0 评论数 0
暂无评论
友情链接