计算机工程与应用 ›› 2008, Vol. 44 ›› Issue (33): 13-17.DOI: 10.3778/j.issn.1002-8331.2008.33.004

• 博士论坛 • 上一篇    下一篇

改进型虚拟线式视频车辆检测算法

吴 骏1,肖志涛2   

  1. 1.天津工业大学 信息与通信工程学院,天津 300160
    2.天津工业大学 研究生部,天津 300160
  • 收稿日期:2008-07-09 修回日期:2008-09-02 出版日期:2008-11-21 发布日期:2008-11-21
  • 通讯作者: 吴 骏

Improved virtual-line based video vehicle detection algorithm

WU Jun1,XIAO Zhi-tao2   

  1. 1.School of Information and Communication Engineering,Tianjin Polytechnic University,Tianjin 300160,China
    2.Department of Graduate,Tianjin Polytechnic University,Tianjin 300160,China
  • Received:2008-07-09 Revised:2008-09-02 Online:2008-11-21 Published:2008-11-21
  • Contact: WU Jun

摘要: 在智能交通系统中,虚拟线式视频车辆检测算法广泛应用于交通流检测。虚拟线式视频车辆检测算法仅利用像素的亮度信息,受阴影和图像噪声的影响较大,在某些情况下认假率和拒真率比较高。为此提出一种改进型算法,采用两级检测方式,兼用了像素的亮度信息和色度信息。第一级处理利用亮度信息进行检测,第二级处理利用色度信息进行检测,根据色度信息修改亮度阈值。实验结果表明,改进型算法可有效克服阴影和图像噪声的影响,平均认假率为0.71%,平均拒真率为0.81%,与原算法相比均有明显降低,并且满足实时性要求。

关键词: 智能交通系统, 视频检测, 虚拟线, 两级检测

Abstract: In Intelligent Transportation System(ITS),virtual-line based video vehicle detection algorithm is widely utilized in traffic detection.The virtual-line based video vehicle detection algorithm only utilizes luminance information of pixels,therefore it surfers much from shadow and image noise and in some conditions its false reject rate(FRR) and false accept rate(FAR) are high.To solve this problem,an improved algorithm is proposed,which introduces two-level detection and utilizes both luminance and chrominance information.In the first level processing,it performs detection utilizing luminance information.In the second level processing,it does detection and modify the luminance threshold utilizing the chrominance information.The experiment results demonstrate that the improved algorithm can eliminate the influence of shadow and image noise effectively,and its FRR and FAR of are respectively 0.71% and 0.81%,which are obviously lower than those of original algorithm.Furthermore,it satisfies request of real-time performance.

Key words: Intelligent Transportation System, video vehicle detection, virtual line, two-level detection