计算机工程与应用 ›› 2009, Vol. 45 ›› Issue (17): 211-214.DOI: 10.3778/j.issn.1002-8331.2009.17.064

• 工程与应用 • 上一篇    下一篇

新车型识别方法及其在套牌车辆鉴别中的应用

王 枚1,2,王国宏1,于元港2,谢洪森1   

  1. 1.海军航空工程学院 电子信息工程系,山东 烟台 264001
    2.烟台职业学院 图像处理与模式识别研究所,山东 烟台 264000
  • 收稿日期:2008-04-07 修回日期:2008-06-19 出版日期:2009-06-11 发布日期:2009-06-11
  • 通讯作者: 王 枚

Method of new vehicle-type recognition and its application in set of sign vehicles distinction

WANG Mei1,2,WANG Guo-hong1,YU Yuan-gang2,XIE Hong-sen1   

  1. 1.Department of Electronic Information,Naval Aeronautical Engineering Institute,Yantai,Shandong 264001,China
    2.Institute of Image Processing and Pattern Recognition,Yantai Vocational College,Yantai,Shandong 264000,China
  • Received:2008-04-07 Revised:2008-06-19 Online:2009-06-11 Published:2009-06-11
  • Contact: WANG Mei

摘要: 提出一种用于判决车辆类型的车灯定位识别方法,可解决套牌车辆鉴别问题。在车牌和车标识别基础上,以车牌中心点为准,选取相对完整的单个车灯,从靠近车标一方开始,截取车灯垂直幅值倍数的水平宽度作为有效区域,然后对其进行大小和方向归一化处理;接着使用图像的不变矩距离分类器在车标确定的车系中进行车灯识别,根据车灯识别结果确定车辆具体车型;最后结合车型识别结果和牌照信息在车辆信息库中比对,判断是否为套牌车辆。经交通卡口获取的实测图像进行测试结果表明,车型识别准确度为95.5%。

关键词: 车灯定位识别, 车型识别, 归一化, 套牌车辆, 不变矩

Abstract: It gives a method of vehicle-light location and recognition for vehicle type recognition to solve the problem of set of sign vehicle identification.Based on the recognition of vehicle-plate and vehicle-logo,depending on the central position of the vehicle-plate,signal more complete vehicle-light is selected.The area is regarded as effect area by cutting times horizontal length of the vertical wide of vehicle-light from the side near vehicle-logo.Next its size and direction normalization processing is executed.Then invariant moment minimum distance achieves the vehicle-light recognition.The vehicle type is given by the fusing the vehicle-light recognition result and vehicle-logo class.At last the set of sign vehicles are distinguished by the recognition result both the vehicle-plate and vehicle-type in the vehicle database.The experimental result on actual vehicle images that are taken from traffic station shows that the vehicle-type recognition accuracy is 95.5 percent.

Key words: vehicle-light location and recognition, vehicle-type recognition, normalization, set of sign vehicle, invariant moment minimum