几种人体运动检测算法的比较分析  

Contrast Research of Several Human Motion Detection Algorithm

作　　者：周游[1] 刘艳滢[1] 王春民[2] 陈静[1] 陈秋萍[1] 魏雅娟[1] 

机构地区：[1]中国科学院长春光学精密机械与物理研究所,长春130033 [2]吉林大学仪器科学与电气工程学院,长春130061

出　　处：《吉林大学学报（信息科学版）》2009年第6期652-657,共6页Journal of Jilin University（Information Science Edition）

摘　　要：为了提高计算机视觉运动人体检测的效果,采用了OGHMs(Orthogonal Gaussian-Hermite Moments)法应用于人体运动检测中,在实现过程中对OGHMs加以简化,对运动图像进行Gaussian平滑处理,并在模块运算后期引入腐蚀运算从而得到更精确的检测效果。分析了时间差分法、背景减除法、光流法等常用的几种运动人体检测算法。通过对现有的实验结果分析比较表明,增加腐蚀作用后的OGHMs法不仅具备常用算法各自具有的一些优点,而且抗干扰能力强,同时还较好解决了计算量大的问题,不需要进行背景重建和更新等步骤,并且检测出的运动信息较多,能较完整、较精确地检测出运动人体。In order to improve the effect of the Moving human detection in computer vision, this paper used the OGHMs method. In the implementation process, we have simplified OGHMs （ Orthogonal Gaussian-Hermite Moments）, used the Gaussian smoothing to the movement image, and added an erode algorithm to get more accurate results. Simultaneously several methods used in moving human detection are analyzed, such as temporal difference, background subtraction and optical flow. The experimental result indicated that the OGHMs which is added erode process has some advantage of common algorithms, and has strong anti-jamming capability. This algorithm has better solved the problem of large quantity of calculaion, it is not necessary to carry on the process of background rebuilding and updating. The detection result of this algorithm contains more movement information, it can finish the moving human detection more integralhy and accurately.

关 键 词：人体运动检测 正交高斯Hemite矩 时间差分法 背景减除法 光流法 

分 类 号：TP751[自动化与计算机技术—检测技术与自动化装置]