北京大学学报(自然科学版) ›› 2015, Vol. 51 ›› Issue (6): 1029-1034.DOI: 10.13209/j.0479-8023.2015.119

上一篇    下一篇

基于视觉认知的禁令交通标志检测

胡晓光1,2, 程承旗2, 李德仁3   

  1. 1. 中国人民公安大学刑事科学技术学院, 北京 100038
    2. 北京大学工学院, 北京 100871
    3. 武汉大学测绘遥感信息工程国家重点实验室, 武汉 430079
  • 收稿日期:2014-03-24 出版日期:2015-11-20 发布日期:2015-11-20
  • 通讯作者: 胡晓光, E-mail: hxgnol(at)126.com
  • 基金资助:
    国家重点基础研究发展计划(61399)资助

Prohibition Traffic Signs Detection Based on Visual Cognition

HU Xiaoguang1,2, CHENG Chengqi2, LI Deren3   

  1. 1. School of Criminal Science and Technology, People’s Public Security University of China, Beijing 100038
    2. College of Engineering,Peking University, Beijing 100871
    3. State Key Laboratory of Information Engineering in Surveying, Mapping and Remote Sensing, Wuhan University, Wuhan 430079
  • Received:2014-03-24 Online:2015-11-20 Published:2015-11-20
  • Contact: HU Xiaoguang, E-mail: hxgnol(at)126.com

摘要:

根据交通标志的反差特性会强烈吸引人类视觉注意的设计原则, 结合生物视网膜会强烈响应场景中大反差视觉刺激的机理, 将基于视觉反差的层次结构的显著性分析框架引入交通标志的检测问题, 提出一种适合现实街景中交通标志检测的多线索视觉注意模型, 将对交通标志的检测定位转变为对显著目标的发现与分割问题。实验表明, 所提方法优于典型的显著性方法在面对现实街景时的目标检测性能。

关键词: 交通标志检测, 街景, 显著性分析, 视觉反差

Abstract:

Considering the design principles that traffic signs is to strongly attract the human visual attention, combining the phenomenon that the retina strongly responds to large contrast visual stimulation, a hierarchy saliency analytic framework based on visual contrast is introduced. The authors propose a multi-cue visual attention model for traffic sign detection in street scene, so traffic sign detection and segmentation problem is converted to saliency object discovery and location problem. Experimental results show that the proposed method is better than typical saliency methods.

Key words: traffic sign detection, street scene, saliency analysis, visual contrast

中图分类号: