华南理工大学学报(自然科学版) ›› 2016, Vol. 44 ›› Issue (9): 94-99.doi: 10.3969/j.issn.1000-565X.2016.09.014

• 电子、通信与自动控制 • 上一篇    下一篇

基于梯度特征的隧道场景车辆灯光干扰抑制方法

赵敏 石雨新 孙棣华   

  1. 重庆大学 自动化学院,重庆 400044
  • 收稿日期:2015-12-31 修回日期:2016-03-03 出版日期:2016-09-25 发布日期:2016-08-21
  • 通信作者: 赵敏(1980-),女,副教授,主要从事智能交通系统、数字图像处理研究. E-mail:min992215@163.com
  • 作者简介:赵敏(1980-),女,副教授,主要从事智能交通系统、数字图像处理研究.
  • 基金资助:
    国家自然科学基金资助项目(61573075);重庆市重点产业共性关键技术创新专项(cstc2015zdcy-ztzx60002);重庆市自然科学基金资助项目(cstc2016jcyjA0565);重庆市教委科学技术项目(KJ1503301);重庆大学中央高校基本科研业务费专项基金资助项目(106112014CDJZR178801)

Gradient-Based Suppression Method of Vehicle Light Interference in Tunnel Scenes

ZHAO Min SHI Yu-xin SUN Di-hua   

  1. School of Automation,Chongqing University,Chongqing 400044,China
  • Received:2015-12-31 Revised:2016-03-03 Online:2016-09-25 Published:2016-08-21
  • Contact: 赵敏(1980-),女,副教授,主要从事智能交通系统、数字图像处理研究. E-mail:min992215@163.com
  • About author:赵敏(1980-),女,副教授,主要从事智能交通系统、数字图像处理研究.
  • Supported by:
    Supported by the National Natural Science Foundation of China(61573075)and the Natural Science Foundation of Chongqi(cstc2016jcyjA0565)

摘要: 在隧道场景下,车辆灯光的干扰对车辆目标的准确提取产生严重的影响,而现有方法对此尚缺乏针对性的解决方案. 为此,文中提出了一种基于灯光梯度特征的灯光干扰抑制方法,根据光照辐射特性以及隧道空间位置关系,建立了车辆光照区域的光强模型并构造其梯度函数,进而利用光照区域梯度方向不变特性,筛选出非车辆灯光区域并构造前景掩膜,最后与运动目标前景叠加,实现光照干扰的抑制. 实验结果表明,该方法能有效地抑制车辆灯光干扰,提高车辆目标识别的准确性.

关键词: 灯光干扰, 光照模型, 梯度特征, 目标提取

Abstract: In tunnel scenes,the interference from vehicle light has a strong impact on the accurate extraction of vehicle targets.However,in the existing methods,there is no solution to this problem.Therefore,a suppression method of vehicle light interference is proposed on the basis of light gradient features,and according to the optical radiation and space structure features,a light intensity model of vehicle light illumination on roads and its vehicle light gradient function are constructed.Furthermore,by utilizing the invariance of gradient directions of vehicle light illumination on roads,the non-vehicle light illumination area is screened and a foreground mask is made.Fi- nally,the foreground mask is overlaid with the moving target foreground,thus suppressing the vehicle light Interfe- rence.Experimental results show that the proposed method can effectively suppress the vehicle light interference,and thus improve the extraction accuracy of vehicle targets.

Key words: light interference, illumination model, gradient feature, target extraction

中图分类号: