1、Vol119 No14 公 路 交 通 科 技 2002 M8JOURNAL OF HIGHWAY AND TRANSPORTATION RESEARCH AND DEVELOPMENTcI|: 1002O0268 ( 2002) 04O0126O04l : 2001O08O06 “: S/ “(00037)Te: ( 1973- ) , 3, “ , vYp V 3,1V Y P Ra. Z . j $ ? p_ZE李 斌, 王荣本, 郭克友( vY, = 130025)K1: 着重阐述基于机器视觉的前方车辆障碍物检测方法b首先根据公路上前方车辆的先验特征模型,建立障碍物探测的感兴趣区,以缩
2、小搜索区域; 随后提出一种新的对称变换算子,用于检测障碍物车辆的对称轴,并确定障碍物车辆的矩形轮廓b为进一步提高障碍物检测的实时性, 采用递归模板匹配法对障碍物进行跟踪b试验表明上述方法是有效的b1oM: 智能车辆; 机器视觉; 对称变换ms |: U49116 DS M : AStudy on Machine Vision Based Obstacle Detection and Recognition Methodfor Intelligent VehicleLI Bin, WANG RongOben, GUOKeOyou( Transportation College of Jilin
3、University, Jilin Changchun 130025, China)Abstract: A leading vehicle detection and recognition method based on machine vision is mainly described in this paper. Firstly, in or-der to reduce the searching area, an area of interest ( AOI) for the obstacle is gotten based on the image contour detectio
4、n and the prioriknowledge of the leading vehicle on the road. Secondly, a new symmetry transform operator used to search for the symmetry axis of theleading vehicle in the image is proposed, and its the rectangle contour is obtained using the hough transform. Then the concept of sym-metry distance i
5、s introduced to validate the leading vehicle. Further more, in order to improve real time obstacle detection, a recursivetemplate matching method is established to track the leading vehiclecs location in the image. The experiment results indicate its validity.Key words: Intelligent vehicle; Machine
6、vision; Symmetry transformpB ? V 1 ?b -Z L H1 MY a? 3 Y# Es1il,9 R -4Hqbj $ 4 Wa M W1 ? m ,P RP R V 190% 1j $ 1b7Vm) T MY?Z 9 j $(9 j $) , ? mm MY W, P9 LC j $“dt ?byN j $X “ - ? # P R | m 1ob j $/ ,4 B1 “dp L H MYZEb k rb1 5 M + Y, MY “S 5M, j $“d ? MYsO $Y “S,9 5 M Tb R -Z m1C/ j+:11+bv8 ,7 O + y1
7、 , zB1 2mW,B1 3mWb21H+b LaP BbL ! - SB = R,5-piY - R byNK - = -Z b9 , p u Vm l HLMF u, m2PH Ub - L R,5 X f /, L %FOE ( Focus of Ex-pend) , FOEm Mb l u p_ HWbL_ZE3 -T X8C 2 ,N b212 p , V ?*?C -Zp R dA1, p -bN sO qCCD25mmhFbp H, mFOE %,+ V AT , 4_, ? p + b p 5 l, z2m, 80mP Hvl40pixels, f /,+9 |, H, m
8、ibp “SB , +vM1byN T M1,p 6, A1 BQ1 1 8 HW,V7 V4p_ L H V Lbm7 U6 E BB E,NZE,y AOI H1 Q_ ,7 -B mXp, - msAOIb E VpHa a+bHa v G+4 | E,9 HW9 byN+ bM1“ ,n T6bm7 B 6 E m Mb9 -B mX uT I ,/B m,I BS = M,i9 M uM1“ ,“ l,M1vbM1Kv u L !i ubNy AOIbQk(Dx, Dy) = EXAx= 0 EYAy= 0Ik- 1(x,y)- Ik(x+ Dx, y+ Dy) 2(6), x, y
9、 MUS;Ik (x, y)k m;XA, YA vl;Dx, Dy /B m MbkTV ,B Ep6,p_ ( HW42ms 31msb5 4 pMa a p MYZE,iB Ep6,B4 EL Hb kV 4 E r Vb ID: 1 co.m 9 j $. bv, 2000O08. 2 , . “ j $ -Z ZE. Y S/, 2001 (6) . 3 4.9 m) . bv, 2000O02. 4 Thomas Kalinke, et al. A TextureObased Object Detection and an Adap-tive ModelObased Classification. Proceedings of the IEEE Intelligent Ve-hicles Symposium 1998, 1998O10: 143- 148. 5 , O. m.v, 1999 ( 1) . j $ ? p_ZE 129
