第
33
卷第
5
期
2
014
年
10
月
红外与毫米波学报
J. Infrared Millim. Waves
Vol. 33,No. 5
October,
2014
文章
编号
:1001 - 9014(2014)05 - 0465 - 07 DOI:
10. 3724 / SP. J. 1010. 2014. 00465
Received date: 2013 -10 -17,revised date: 2014 -07 -03
收稿
日期
: 2013 -10 -17,
修回日期
: 2014 -07 -03
Foundation items: Supported by the Nation Science Foundation of China (91120003)
; N
ation Science Foundation of China (61105092); Beijing Natu-
ral Science Foundation(4101001)
Biography: JIN Lu(1982-),male,Changzhi Shanxi,Ph D,Research area involves intelligent navigation and information fusion. E-mail: 10906009@ bit. edu. cn
*
Corresponding author: E-mail: fumy@ bit. edu. cn
Vehicle detection based on vision and millimeter wave radar
JIN Lu
1,2
, FU Meng-Yin
1,2*
, WANG Mei-Ling
1,2
, YANG Yi
1,2
(1. School of Automation,Beijing institute of technology,Beijing 100081,China;
2. Key Laboratory of Intelligent Control and Decision of Complex Systems,Beijing 100081,China)
Abstract: With the importance of automotive drive assistance system of intelligent vehicle,vehicle detection fusing
millimeter wave (MMW) radar data and vision multi -features is presented. The vehicle detection algorithm can be
divided into three steps. Firstly,a space alignment algorithm between MMW radar and vision was proposed to get
space alignment point according to the space transformation matrix of image coordinate and radar coordinate. The
second step obtains region of interest (ROI) according to the space aligned point and search strategy. At last,vehi-
cle detection was realized through features of vehicle including bottom shadow ,symmetry,left and right edges; in
this step,an improved segmentation algorithm of bottom shadow of vehicle was described in order to obtain accu-
rate vehicle width. The performance of the algorithm was verified under different scenarios. The results show the
vehicle detection algorithm is effective and feasible.
Key words: automotive driver assistance system,vehicle detection,space alignment,ROI ,bottom shadow
PACS: 07. 57. -c;42. 30. -d
基于视觉和毫米波雷达的车辆检测
靳 璐
1,2
,
付
梦
印
1,2*
,
王
美
玲
1,2
,
杨 毅
1,2
(1.
北京
理工大学 自动化学院
,
北京
100081;
2.
复杂系统智能控制与决策教育部重点实验室
,
北京
100081)
摘要
:
根据智能车辆主动驾驶辅助系统中的重要性
,
提出了一种融合毫米波雷达数据和视觉多特征的车辆检
测算法
。
车辆检测算法通过三个步骤实现
,
首先
,
提出一种空间对准算法实现毫米波雷达和视觉的空间对准
;
其次
,
根据空间对准结果和搜索策略提取目标车辆的感兴趣区域
;
最后
,
融合车底阴影
、
对称轴
、
左右边缘等车
辆特征实现车辆检测
,
其中
,
为了准确得到目标车辆的车底阴影
,
提出一种改进的车底阴影分割算法
。
算法的
性能在不同的场景下得到证实
,
实验结果表明该车辆检测算法是有效和可靠的
。
关 键 词
:
主动驾驶辅助系统
;
车辆检测
;
空间对准
;
感兴趣区域
;
车底阴影
中图分类号
:TP242
文献标识码
:A
Introduction
With the sharply growing of vehicle,the traffic acci-
dents happen frequently and cause heavy casualties and
financial losses. In order to improve the traffic security
level,developing the automotive driver assistance sys-
tems has attracted a large amount of attention lately. The
reliable vehicle detection becomes a very important as-
pect of automotive system
[1]
.
The major advantage of radar-vision fusion benefits
from the best performance of each sensor by fusing results
from different sensors together
[2]
. MMW radar can detect
moving vehicles fast and provide the long-range detection
and exact velocity measurement. Nevertheless,vision
can obtain the contour of vehicles in the short-range sens-
ing region
[3]
. So radar-vision fusion based on comple-
mentary devices can provide an improved performance:
improved reliability from multiple detections and the mer-
ging of position measurements with good longitudinal and
lateral accuracy
[4]
.
The radar-vision fusion approach mainly focuses on
the space alignment of radar-vision sensors and the verifi-
cation of radar target. The general space alignment algo-