一种新的基于实时视频流的车型特征提取算法  

New Vehicle Recognition Algorithm Based on Real-Time Videos

作　　者：张全元[1] 戴光明[1] 

机构地区：[1]中国地质大学计算机学院,武汉430074

出　　处：《工程地球物理学报》2008年第3期346-350,共5页Chinese Journal of Engineering Geophysics

摘　　要：背景的实时建模、车辆存在判定及车型特征提取是基于实时视频流的车型识别系统中的关键性问题。本文首先从建模质量和建模速度两个方面对比分析了几种常用的背景建模算法。然后提出了一种新的车辆存在判定、特征提取算法——"弹性松弛算法",最后给出了实验结果。结果表明:该算法可以有效的进行车辆存在判定,并能够实时的提取车型长度特征,从而完成车型识别系统中的车辆计数工作,为车型识别奠定了良好的基础。In the vehicle recognition systems based on the real--time videos, the construction of the background, the judgment of vehicles existence and features extraction are important. Firstly, this paper compares several common background update algorithms, and points out the advantages and disadvantages. And then a new algorithm method to extract the vehicle features is proposed. The experiments show that the algorithm is very adaptive to the real--time environment, very sensitive to the existence of the iudgment of vehicles has it has very good vehicle feature extraction and classification function.

关 键 词：车辆存在判定 弹性松弛算法 特征提取 

分 类 号：P63[天文地球—地质矿产勘探] TP3[天文地球—地质学]