基于人体碎片的特征提取算法研究  

Features Extraction Algorithm Research Based on Persons′ Patches

作　　者：程丽芳[1] 赵明昌[1] 张向文[1] 何英[1] 

机构地区：[1]桂林电子科技大学计算机与控制学院,广西桂林541004

出　　处：《现代电子技术》2008年第8期131-134,共4页Modern Electronics Technique

基　　金：广西青年基金(桂科青0542035);广西科学研究与技术开发计划项目(桂科攻0632007-1G)

摘　　要：特征提取一直是模式分类问题中的关键步骤,对基于身体碎片外形轮廓及其和人物中心相对位置特征的提取算法进行研究。此算法操作简单,这些特征是在所有位置和所有大小上提取的,使制作的分类器满足平移和缩放不变性,并且他还是基于有标注数据库的。此算法可用于基于物体碎片特征的人脸识别系统、车辆和人体的自动视频监测系统以及其他的监控系统和图像检索中。Features extraction is always the key step. This article describes the extraction algorithm based on the figure of persons＇ patches and correlational location with the center of person. This algorithm can be performed simply. And these fea- tures are extracted at all the location and scales. So the classifier made with it is unvaried in translation and scale. And it is also built on the label database. This algorithm can be used in the face recognition system,person and vehicle automated visual surveillance,other surveillance system and image searches, which based on the patches feature.

关 键 词：特征提取 滤波器 边缘检测 物体碎片特征 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]