智能监控系统中的运动目标分类方法研究  

Research on Targets Classification in Intelligent Video Surveillance System

作　　者：王强光[1] 伍铁军[1] 

机构地区：[1]南京航空航天大学机电学院,江苏南京210016

出　　处：《工业控制计算机》2011年第1期16-17,20,共3页Industrial Control Computer

摘　　要：运动目标的分类是智能监控系统中最重要的研究内容之一。提出了一种用于智能监控系统中的运动目标分类方法,可以较为准确地将运动目标分为汽车、人和自行车三类。首先检测出监控视频序列中的运动目标,然后定义了四种目标的形状特征、占空比、长宽比、面积与周长的平方比和惯性主轴方向,这四种特征可以有效地解决因监控的视角和远近带来的形状特征的变化。并且经过大量的实验统计出这四种特征的均值和方差。根据贝叶斯决策理论利用这四种形状特征对目标进行分类。实验结果证明了该方法具有较好的准确性、鲁棒性和实时性。Object classification is one of most important research contents in the intelligent video surveillance system.An object classification algorithm is proposed in this paper used for video surveillance,which can classify moving objects into three categories：car,human,and bicyclist.First detecting the moving objects in the monitoring video frequency sequences,then defined four kinds of shape features：occupy-space-proportion,the ratio of height to width and principal axis of inertia direction.These four kinds of features can solution the question of features changing by the angle of view and far and near.And calculated the average value and the variance of these four kinds of features by massive experiments.Using these four kinds of shape features can classify different objects according to the Bayes decision.

关 键 词：目标分类 形状特征 贝叶斯决策理论 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]