首页期刊视频编委会征稿启事出版道德声明审稿流程读者订阅论文查重联系我们English
引用本文
  • 丁祖春.基于机动车车灯轮廓特征矢量的机动车款式识别[J].电讯技术,2025,(12):2124 - 2131.    [点击复制]
  • DING Zuchun.Vehicle Style Recognition Based on Contour Feature Vector of Vehicle Lamps[J].,2025,(12):2124 - 2131.   [点击复制]
【HTML】 【打印本页】 【下载PDF全文】 查看/发表评论下载PDF阅读器关闭

←前一篇|后一篇→

过刊浏览    高级检索

本文已被:浏览 244次   下载 114 本文二维码信息
码上扫一扫!
基于机动车车灯轮廓特征矢量的机动车款式识别
丁祖春
0
(广州工商学院 产业学院,广州510850)
摘要:
为了实施对机动车的管控,需要查找特定机动车或分析其行驶轨迹。因道路监控图像数量巨大,识别机动车的款式可以大幅度压缩查找范围。由于监控摄像头与被拍摄的车辆之间具有一定的夹角,监控图像中的车辆会有一定的形变,影响识别结果的准确性。考虑监控拍摄夹角及车灯表面的非平面性的影响,构建车灯的特征矢量用于车辆款式识别。提取监控图像的车灯区域,通过形态学方法去除背景噪声,提取机动车车灯轮廓并进行编码,将车灯的轮廓分割为多个扇区提取多维特征,通过分区加权方法计算机动车款式特征矢量的相似度实现款式识别。实验中使用了监控设备拍摄的15 300幅机动车图像作为测试数据集,覆盖不同地理位置、不同时段、不同光照条件、不同天气下的机动车,所提出的算法对机动车款式的识别准确率达到了98.95%,高于现有主流的先进方法。该方法根据车辆的不可改变的车灯结构的轮廓特征识别车辆款式,在交通管控上具有实践意义。
关键词:  交通管控  机动车款式识别  车灯轮廓  特征矢量
DOI:10.20079/j.issn.1001-893x:250808001
基金项目:广东省教育厅新一代信息技术重点领域专项(2020ZDZX3036)
Vehicle Style Recognition Based on Contour Feature Vector of Vehicle Lamps
DING Zuchun
(School of Industry,Guangzhou College of Technology and Business,Guangzhou 510850,China)
Abstract:
To implement the regulation of vehicles,it is necessary to identify vehicles and analyze driving trajectories.Because of the huge quantity of road surveillance images,recognizing vehicle types can significantly reduce search scope.Due to the angle between surveillance cameras and vehicles being taken pictures,there is deformation in surveillance images,which affects the accuracy of recognition.Considering the shooting angle and the non-planar surface of vehicle lamps,feature vectors are constructed to identify vehicle types.The lamp is extracted,and then noise is removed using morphological methods.The contours of the vehicle lamps are extracted and encoded,then the contours are segmented into sectors to extract multidimensional lamp features,and the similarity of lamp feature vectors are calculated by weighted sector feature for vehicle type recognition.In the experiment,15 300 vehicle images captured by surveillance monitors are used,covering vehicles from different geographical locations,periods,illumination and weather.The proposed algorithm achieves an accuracy of 98.95%,surpassing the existing state-of-the-art methods.This approach recognizes the vehicle type based on the unalterable contour features of lamps and it is meaningful in traffic management with practical significance.
Key words:  traffic management  vehicle style recognition  vehicle lamp contour  feature vector
安全联盟站长平台