基于多特征融合的前方车辆检测方法研究倡.pdf

第 28 卷第 9 .9
2011 年 9 月 Vol .2011No
Application Research puters Sep
基于多特征融合的前方车辆检测方法研究倡
慕永云, 王荣本, 赵一兵, 郭烈
(大连理工大学汽车工程学院, 辽宁大连 116024)
摘要: 为了有效地避免交通事故的发生,提高车辆检测的准确率和效率,提出一种基于车辆阴影、边缘、纹理
和对称性等多种特征融合的前方车辆检测方法。利用多阈值分割得到车辆阴影,选取边缘确定出车辆可能存在
的感兴趣区域,通过计算感兴趣区域内的分形盒子维数进一步确定车辆的候选区域,利用对称性测度验证定位
车辆。实验结果表明,该方法能够实现不同环境下的车辆检测,相比较单个特征的检测方法具有较高的准确率
和可靠性。
关键词: 车辆检测; 多阈值; 阴影; 边缘; 分形盒子维数; 对称性测度
中图分类号: 391畅41 文献标志码: 文章编号: 1001唱3695(2011)09唱3572唱04
TP A
: /. .
doi j issn
唱
Study on detecting method of preceding vehicle based on multi feature fusion
唱, 唱, 唱,
MU Yong yun WANG Rong ben ZHAO Yi bing GUO Lie
(School of Automotive Engineering, Dalian University of Technology, Dalian Liaoning 116024, China)
Abstract: 唱: , ,
This paper proposed a method to detect preceding vehicle based on multi feature fusion .shadows, texture edges唱
and symmetry to唱 prevent traffic accidents and improve. the accuracy and efficiency on vehicle detection First, used the segmen
tation. with multi threshold for shadow detection Used shadows to mark areas in which a vehicle may exist namely interesting
areas Then determined potential vehicle region by calculation of the fractal box dimension to. judge, whether there exist vehicles唱
in the areas and to locate the vehicles with edges feature was achieved in region of interest. At last propos