室外场景下目标分割和目标识别算法  

An Algorithm for Object Segmentation and Object Recognition on Outdoor Scene

作　　者：黄建新[1] 

机构地区：[1]华侨大学数学系,福建泉州362021

出　　处：《华侨大学学报（自然科学版）》2005年第4期353-356,共4页Journal of Huaqiao University(Natural Science)

摘　　要：视频监控应用场所可分为室内和室外.室外环境受到诸如光照、下雨、落叶等因素的影响,整个场景变化复杂,给视频处理带来许多困难.文中介绍一种室外场景下目标分割和目标识别的方法,使用基于像素颜色特征的混合概率模型,将当前图像中与模型匹配的像素视为背景,然后更新模型中各个参数.为了去除目标区域的阴影部分,引入一种基于阀值和区域特征的阴影消除算法.同时,采用基于支持向量机的分类方法,识别场景中新出现的目标.The site where video monitoring is applied can be divided as indoor and outdoor. The outdoor surroundings bring about difficulties to video-frequency processing due to the influence of such factors as illumination, raining and fallen leaves which make the entire scene to change and to be complicated. A method is presented here for object segmentation and object recognition. The author regards the pixels in present image matching with a model of mixed probability based on the characteristic of pixel color as background, and updates parameters in the model; and eliminates the shadow in object area by leading in an algorithm based on threshold and characteristic of area; and moreover, recognizes new-emerging object in the scene by adopting method of classification based on a supporting vector machine.

关 键 词：目标分割 混合概率模型 阴影消除 支持向量机 

分 类 号：TP394.1[自动化与计算机技术—计算机应用技术] TN914.1[自动化与计算机技术—计算机科学与技术]