基于实时视频流的车型识别系统设计  

Vehicle Recognition System Design Based on Real-time Videos

作　　者：张全元[1] 戴光明[1] 陈良[1] 

机构地区：[1]中国地质大学计算机学院,湖北武汉430074

出　　处：《微计算机信息》2008年第31期288-290,共3页Control & Automation

基　　金：国家863高技术研究发展计划项目<混沌密码系统的理论与实现技术>资助(NO:2006AA01Z426)

摘　　要：基于视频的运动车辆检测及车型识别系统是智能交通系统的重要组成部分。本文提出了一种在固定摄像头拍摄的车辆图像序列中检测出运动车辆并识别其车辆类型的系统结构,提出了基于投影检测提取车型特征的新方法,达到了统计每种类型车辆数目和收费总数的目的。本算法处理过程分为数字图像获取、图像处理和车辆外观长、宽、高特征数据提取三个主要部分。现场实验表明本方法实用可行,尤其满足实时性较高的车行检测系统,并且得到了较好的结果。Video-based movement of vehicles and models to identify detection system is an important component of ITS. This paper suggested a system architecture which based on serial videos and a fixed camera and a new algorithm to detect the traits of cars based on image projections. The algorithm process into digital image acquisition, image processing and vehicle exterior length, width, and height data extraction features three main sections. Field experiments indicate that this method is practical, particularly to meet the real-time dealers higher detection system, and a better result.

关 键 词：车型识别 背景差分 投影检测 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]