Application of remote sensing for extraction of road
Information
Manzul Kumar Hazarika, Kiyoshi
Honda, Lal Samarakoon, and Shunji Murai
Asian Center for Research on Remote Sensing (ACRoRS)
STAR Program, Asian Institute of Technology,
Km. 42, Paholyothin Highway, Klong Luang, Pathumthani 12120, Thailand.
Tel : +66-2-524-6148 Fax : +66-2-524-6147
E-mail : manzul@ait.ac.th
Keywords: Remote Sensing, information extraction.
Abstract
Road network of the Asia plays a vital role in economic development of the region by
providing access to underdeveloped areas. Keeping the importance of an efficient road network
in view, Asian Highway project has been initiated by the United Nations Economic and Social
Commission for Asia and the Pacific (UN-ESCAP) to promote and co-ordinate the development
of international road transport in the Asian region and stimulate economic growth. The major
part of Asian Highway is the existing roads and a considerable portion of these roads needs to
be upgraded to meet the Asian highway standard. A database on these roads is required and
remote sensing satellites with their synoptic view and repetitive coverage offer a possibility for
obtaining such information.
Capabilities of remote sensing data in road identification and their width estimation have
been examined by setting up several test-sites in Thailand. Widths of the roads at the test sites
are 5m, 15m, 35m and 64m. SPIN-2 (2m) data, used in this study, has the best spatial resolution;
followed by ADEOS Panchromatic (8m), SPOT Panchromatic (10m), ADEOS Multispectral
(16m) and LANDSAT TM (30m).
All the sensors can identify 35m and 64m wide road. ADEOS Multispectral and
LANDSAT TM data cannot identify a road having a width of 15m or less. A 5m wide can be
identified by SPIN-2 data only. Spatial resolution of data contributes more to the clarity of a
road than the multi-band observation capability. However, the surrounding environment along
the road also affects on its clarity. If the background of a road has a very different reflectance
characteristic from the road itself, for example, when a road passes through a paddy field,
possessing homogeneous vegetative cover, then it becomes very distinct. On the other hand, if a
road passes through an urban area, which has similar reflectance, it is difficult to identify.
Two methods have been tested for road width estimation. One is to measure the road width
on printed images and the another is to count the number of pixels on a computer display. Result
shows that in most of the cases, remote sensing data has the capability to estimate the width with
an accuracy of half of the spatial resolution or at least the accuracy better than its resolution.
Road materials like asphalt and concrete can not be discriminated using the data from any
of the five sensors, even though asphalt and concrete have different reflectance characteristics.
In ideal condition, it may be possible to distinguish one from other. But on a road, due to
wearing of rubber particles from the vehicles, which stick to the road surface, both the concrete
and asphalt surface gives similar reflectance characteristics, after a road become operational.
Introduction
Major portion of the route identified by the UN-ESCAP is existing roads which needs
upgrading to Asian Highway specifications. Accordingly, information of these routes are
required for making a plan. Many of the Asian countries are very poor and some of them are
suffering from economic and political instability. In such a situation, only a few countries
possess updated information for existing roads. Further, updating road information through
physical survey is not only an expensive task but also time consuming. Remote sensing
technology can be effectively used to overcome such problems.
Study Area and Data Used
The area in this study falls in the northern part of the Bangkok City. Study area is extended
from 100°15" to 100°45" in the East and 14°00" to 14°30" in the North. In this study, satellite
data of various spatial resolutions with different sensors have been used. Data used in this study
are shown in Table 1.
Table 1 Satellite data used in the
| Satellite/Sensor |
Spatial Resolution | Path/Row
| Acquisition Date |
| SPIN 2/KVR-1000 (Analog) |
2m | - | 1995 |
| ADEOS/AVNIR (Panchromatic) | 8m |
113/333 | 28-05-97 |
| ADEOS/AVNIR (Multispectral) |
16m | 113/333 | 25-01-97 |
| SPOT/HRV (Panchromatic) | 10m | 262/322 | 28-06-96 |
| LANDSAT/TM (Multispectral) | 30m | 129/50 | 21-05-95 |
Methodology
Data Processing
Satellite data are corrected geometrically, using Ground Control Points (GCP) taken form
topographic maps (Scale 1:50,000). A first order polynomial transformation equation has been
used to rectify the data set.
Identification of Road Characteristics in Different Backgrounds
Road sections with various backgrounds have been considered in study. False Colour
Composite of the multispectral data and intensity image panchromatic data are used for
investigating the road characteristics.
Estimation of Road Width
Road width is estimated using analog and digital methods. In analog method, each of the
geometrically corrected images from different sensors is printed in a hardcopy and widths of
roads are estimated from the actual measurements made on the hardcopies. SPIN-2 is printed at
a scale 1:8,000 whereas ADEOS Panchromatic, SPOT Panchromatic, ADEOS Multispectral and
LANDSAT TM data are printed at a scale of 1:20,000. These scales are found to be most
appropriate, because pixels of the roads are not shown individually in the hard copy. A road can
be measured up to an accuracy of 0.25 mm on a hardcopy image, using a ruler. Field survey was
conducted to find out the actual width of the road sections for verification of results. The
measured width includes both the pavements and shoulders of the road and median strip, in case
of roads having more than one lane.
In digital method, numbers of pixels are counted in the perpendicular direction of a road
and width of the road is estimated by multiplying pixel numbers with the pixel size. However,
there are certain difficulties in counting numbers of pixels perpendicular to a road. Mixed pixels
exist on the edge of a road and, in such cases, it is not an easy task to find out an accurate width
of the road. This is more critical in low-resolution data.
