混合高斯模型与CamShift算法结合的台球跟踪方法  

Tracking of Billiard Based on CamShift Algorithm Combined with Gaussian Mixture Model

作　　者：袁伟[1] 王朝立[1] 张旭飞[1] 

机构地区：[1]上海理工大学光电信息与计算机工程学院

出　　处：《微型电脑应用》2013年第2期10-12,共3页Microcomputer Applications

基　　金：国家自然科学基金项目(60874002);上海市教委重点科技创新项目

摘　　要：传统的CamShift跟踪算法是基于颜色信息的算法,这种算法在背景与目标颜色相近时会出现跟踪不准确的情况,而且不能自动跟踪。针对这两个缺点,本文结合混合高斯模型来对这一跟踪算法进行改进,通过混合高斯模型来实现对运动目标的检测得到目标的轮廓,将图像处理中的形态学变换应用到目标轮廓检测过程中,可以得到更好的检测结果和目标轮廓。对检测的目标再利用CamShift算法进行跟踪,在程序中引入Kalman预测算法,对目标的位置进行估计,提高跟踪的准确性,实验结果表明了这种方法的有效性,实现了对多目标的自动跟踪。Traditional CamShifi algorithm is an algorithm based on color information, it will appear to inaccurate tracking in the algorithm if the background and the target have similar color, and the automatic tracking also can＇t be achieved. To be aimed at the two shortcomings, in this paper, the tracking algorithm is improved by using Gaussian mixture model which is used to detect moving targets and find the contours. Then much better results and contours with morphological transformation in image processing are ob- tained. After that, CamShift algorithm is used to track the detected targets, what more, this paper uses Kalman prediction algorithm in the program so that the position of the targets can be estimated and tracking accuracy is improved. The experimental results prove that the method is effective and accesses to automatic tracking of the multi-targets.

关 键 词：混合高斯模型 CAMSHIFT算法 轮廓检测 多目标跟踪 

分 类 号：TH133[机械工程—机械制造及自动化] TP183[自动化与计算机技术—控制理论与控制工程]