基于改进的混合高斯模型的目标检测方法  

Target Detection Method Based on Improved Gaussian Mixture Model

作　　者：曙光 张超[1] 蔡则苏[2] 

机构地区：[1]哈尔滨理工大学计算机学院,黑龙江哈尔滨150080 [2]哈尔滨工业大学计算机科学与技术学院,黑龙江哈尔滨150001

出　　处：《计算机技术与发展》2012年第7期60-63,共4页Computer Technology and Development

基　　金：国家自然科学基金(61075076)

摘　　要：混合高斯模型是环境监控的一种有效方法,能及时侦测环境的异常变化,其基本思想为当环境与建立的模型相匹配时为背景,否则为前景。利用混合高斯模型在图像内进行目标检测,在模型替换的时候,前景容易误替换进模型内,影响背景差分效果。同时采用像素为单位做前景判断,信息量往往不足,不能有效地表示环境特征。高斯模型也存在对光照影响消除能力不足的问题,光照发生时,背景会发生变化,如果不及时检测出来,排除光照,就会造成误检,影响系统的准确性。文中针对这些问题提出了一种解决方案,并用实验证明了其有效性。Gauss mixture model is an effective method for environmental monitoring, which can detect the abnormal changes in the envi- ronment timely, the basic thought of Gauss mixture model is when the environment and the established model matches, it is as the back- ground,or the foreground. By using Gaussian mixture model for target detection in the image,foreground is mistakenly substituted into the model easily when model replaces, effecting background difference results. At the same time using pixels to prospect judgment, the in- formation is often insufficient, which can not effectively express environmental characteristics. Gauss mixture model also exists the prob- lem that the ability of illumination effect removing is not ideal, when light occurs, the background will be changed, if not promptly detec- ted the changes, excluding the light, it will cause false detection, affecting the accuracy of the system. Aiming at these problems,put for- ward a solution and use experiment to prove the validity in the paper.

关 键 词：混合高斯模型 目标检测 背景差分 光照影响 

分 类 号：TP301[自动化与计算机技术—计算机系统结构]