ISPRS, Vol.34, Part 2W2, "Dynamic and Multi-Dimensional GIS", Bangkok, May 23-25, 2001 
238 
AUTOMATIC RECOGNITION AND LOCATION OF ROAD SIGNS FROM TERRESTERIAL COLOR IMAGERY 
Sompoch PUNTAVUNGKOUR 
Map Information Center, Royal Thai Survey Department, Pranakorn, Bangkok 10200, Thailand 
Spuntavungkour@hotmail.com 
Xiaoyang CHEN * \ Michiro KUSANAGI 2 
Space Technology Application and Research Program, Asian Institute of Technology, P.O. box 4, Klong Lung, Pathumthani, and 12120, 
Thailand 
Xychen 1 ,Kusanagi 2 @ait.ac.th 
Keywords: Road Sign Recognition; Mobile Mapping Systems, Sign Library 
Abstract 
Mobile Mapping Systems (MMS) are an emerging technology in Geomatics. They have developed from a concept to a standard mapping 
tool. They differ from traditional mapping methods mainly by their fast data collection speed and low cost which makes them ideal for data 
collection in support of the intelligent transportation systems (ITS). The automatic recognition and location of road signs using a ground 
image algorithm, an important part of a MMS, has been developed but is still complex and not found in commercial systems. 
In this paper an automatic road sign positioning and recognition algorithm is presented. For the data collection, the two calibrated sensors, 
RTK-GPS and digital CCD camera, integrated on a van are used. Road Sign Recognition consists of two parts: Sign detection and Sign 
Identification. Sign Detection is based on Semi-geometric correction. Sign Identification is based on Image Matching with a Sign Library. The 
Sign positioning requires of data extraction and Sign position computation. Data extraction is based on Data Synchronization. The Sign 
Position is based on a new simple algorithm based on surveying principles. However, the result is accurate to realize within a real application 
in the near future. 
1 Introduction 
Road Sign Recognition integrated with a Mobile Mapping 
Systems is a significant challenge. Currently two topics are 
researched and developed widely in Developed counties such as 
German, Japan, and others. Specially, Road Sign Recognition is 
complex and expensive in an experimental step, not in the 
market. Transforming from the complex sensor-separated 
system to a simple sensor-integrated system, it is a problem 
worthy of solution in this region. 
Present Mobile Mapping Systems that have incorporated 
recognition systems are few in number (See Table 1). 
Table 1 Mobile Mapping Systems available in the world 
System 
Developer/Research 
Navigation Sensors 
Mapping System 
GeoVAN 
Geospan Crop., USA 
G PS/DR 
10 VHS, voice recorder 
GPS Van 
The Ohio State University, USA 
GPS/Gyro/Wheel 
counter 
2 CCD, Voice recorder 
GPS Vision** 
Lambda Tech. Int. Inc, USA 
GPS/INS 
2 CCD Optical disc 
Kiss** 
Univ. of Bundeswger Munich and 
GeoDigital, Germany 
GPS/IMU?lnclination 
Odometer/Barometer 
IHVHS, 2BW CCD, Voice recorder 
ON-SIGHT 
TransMap Crop., USA 
GPS/INS 
4 color CCD 
RGIAS 
Rowe Surveying and Engg, Inc, USA 
GPS 
Video/Laser 
T ruckMAP 
John E. Chance and Engg Inc., USA 
GPS/Gryo//WA-GPS 
Laser ranqe finder, 1 video camera 
VISAr* 
The Univ. of Calgary and Geofit., 
Cadana 
GPS/INS/ABS 
8 BW CCD 1 color SVHS 
WUMMS 
Wuhan Technical Uni. Of surveying 
and Mapping, China 
GPS 
3 Color CCD, Laser Range, Finder 
** MMS have road sign recognition systems 
Source: Sompoch P., Special Study: Mobile Mapping, AIT, 1999 
Mobile Mapping has been researched and developed since 1980 
when the GPS was transferred from military to civilian users. 
There are three groups of methods. The first research specified 
the photogrammetry section such as photogrammetric 
aerotriangulation by GPS (van der Vegt 1988, Colonia 1989, 
Jacobsen 1993, Ackermann and Schade 1993, Merchant 1994). 
The second group emphasized mapping land vehicle by 
combining navigation sensors such as GPS, INS, CCD camera 
etc. for high-resolution mapping. (Hein 1989, Krabill 1989,chwarz 
1993, Scherzinger 1995, Da and Dedes 1995, Schwarz 1996). 
The last one effects to research the automation rocessing such 
as the automatic matching of the ground bjects, and traction of 
the central line of a road (He and ovak 1992, Li 1994, Xin 995, 
Tao1996) 
* Supported by Visiting Scholar Foundation of Keb Lab. In Wuhan University, P. R. China