867 
TOWARDS AUTOMATIC ROAD MAPPING 
BY FUSING VEHICLE-BORNE MULTI-SENSOR DATA 
Y. Shi a ’\ R. Shibasaki a , Z.C. Shi b 
a CSIS, The university of Tokyo, 277-8568 Chiba, Japan - (shiyun, shiba)@iis.u-tokyo.ac.jp 
Musashi Institute of Technology, 224-0015 Yokohama, Japan - shizc@yc.musashi-tech.ac.jp 
Commission ICWG V/I 
KEY WORDS: Mobile mapping system, Integrated sensor system, Multi-sensor data fusion, Direct-georeferencing, Close-range 
Photogrammetry, Automatic object extraction and recognition. 
ABSTRACT: 
The demand of spatial data has been an explosive growth in the past 20 years. This demand has numerous sources and takes many 
forms, but it is an ever-increasing thirst for spatial data, which is more accurate and higher density (Called as High-Definition 
Spatial Data), which is produced more rapidly and acquired less expensively. This research aims to satisfy with the high demands for 
the road spatial data. We presents an automatic road mapping technology by fusing vehicle-based navigation data, stereo image and 
laser scanning data for collecting, detecting, recognizing and positioning road objects, such as road boundaries, traffic marks, road 
signs, traffic signal, road guide fences, electric power poles and many other applications important to people’s safety and welfare. In 
the hand of hardware, a hybrid inertial survey system (HISS) which combined an inertial navigation system (INS), two GPS receiver, 
and an odometer acquires the posture data was developed. Two sets of stereo camera systems are used to collect colour images and 
three laser scanners are employed to acquire range data for road and roadside objects. On the other hand, an advanced data fusion 
technology was developed for the purpose of precise/automatic road sign extraction, traffic mark extraction, and road fence 
extraction and so on, by fusing collected initial navigation data, stereo images and laser range data. The major contributions of this 
research are high-accuracy object positioning and automatic road mapping by fusion-based processing of multi-sensor data. A lot of 
experiments were performed to certify and check the accuracy and efficiency of our fusion-based automatic road mapping 
technology. From achieved results of these experiments, our developed vehicle borne multi-sensor based mobile mapping system is 
efficient system for generating high-accuracy and high-density 3D road spatial data more rapidly and less expensively. 
1. INTRODUCTION 
1.1 Background 
The demand of spatial data has been an explosive growth in the 
past 20 years. This demand has numerous sources and takes 
many forms, but it is an ever-increasing thirst for geospatial 
data, which is more accurate and higher density (we call that 
High-Definition Spatial Data), which is produced more rapidly 
and acquired less expensively. This research aims to satisfy with 
the high demands for the spatial data. We presents an automatic 
technology by fusing vehicle-based navigation data, stereo 
image and laser scanning data for collecting, detecting, 
recognizing and positioning road objects, such as road 
boundaries, traffic marks, road signs, traffic signal, road guide 
fences, electric power poles and many other applications 
important to people’s safety and welfare. 
As well know, the spatial data of road objects has traditionally 
been collected using terrestrial surveying technology and aerial 
photogrammetric technology; and as a new technique, mobile 
mapping system such as vehicle-borne laser mapping system 
begins to be used for road mapping. But these are some limits 
of the two front methods, such update frequently is difficult/not 
so accuracy in height. To solve these drawbacks, people 
developed land-based mobile mapping system. The most 
important benefit of MMT is a reduction in both the time and 
cost of data collection. 
Corresponding author. 
Land-based Mobile Mapping System (MMS) is just the 
platform system with using mobile mapping technology. Most 
of MMS integrate navigation sensors and algorithms together 
with sensors that can be used to determine the position of points 
remotely. All of the sensors are rigidly mounted together on a 
platform, such as truck. The navigation sensors determine the 
position and orientation of the platform, and the remote sensors 
determine the position of points external to the platform. The 
sensors that are used for the remote position determination are 
predominantly photographic sensors and thus they are typically 
referred to as imaging sensors. However, additional sensors 
such as laser range finders (Reed et al., 1996) are also used in 
MMS and therefore the more general terms of mapping sensors 
may also be used when referring to the remote sensors. 
The strength of MMS lays in their ability to directly 
georeference their mapping sensors. A mapping sensor is 
georeferenced when its position and orientation relative to a 
mapping co-ordinate frame is known. Once georeferenced, the 
mapping sensor can be used to determine the positions of points 
external to the platform in the same mapping co-ordinate frame. 
In the direct georeferencing done by MMS the navigation 
sensors on the platform are used to determine its position and 
orientation. This is fundamentally different from traditional 
indirect georeferencing where the position and orientation of 
the platform are determined using measurements made to 
control points. These control points are established through a