93
ISPRS, Vol.34, Part 2W2, “Dynamic and Multi-Dimensional GIS”, Bangkok, May 23-25, 2001
USE DSM/DTM TO SUPPORT CHANGE DETECTION OF BUILDING IN URBAN AREA
Hong FAN Jianqing ZHANG Zuxun ZHANG Zhifang LIU
LIESMARS of Wuhan Univ., 129 Luoyu Road,Wuhan,China,430079
Tel: 86-027-87881292, Fax:86-027-87643969,Email:fh@hp01 .wtusm.edu.cn
Key Words: DSM , change detection, gradient direction analysis
Abstract
Timely change detection of urban area is very useful for the city's management, development and update of the urban geography
information as well. Aerial Image has proven to be a valuable data source for these kind of application. As we know, the buildings are
typically 3-dimendional real objects whose change will cause the change of height of Digital Surface Model (DSM for short). If we could
make use of the "height" information to assist the detection of building change, it would be helpful for improvement of efficiency and
performance of the automatic detection.
This paper presented an new approach using DSM/DTM to support the change detection of man-made objects especially building in
urban regions, which use both the height information and gray and texture information of building as one kind of data fusion technology
to detect the change. In this paper, the corresponding methods and experiment results would be presented and analyzed in detail.
1. PREFACE
Rapid development of urban makes the data update more often
than ever. On the other hand, manual handling of data update is
a formidable task. An automatic or even semi-automatic way of
data update will increase the speed of data update greatly. One
of the most available and feasible approaches for detection
automation is to utilize aerial images to explore changes of
man-made features in urban area.
Techniques of change detection have been widely used in
change analysis of land use, monitoring of shifting cultivation,
study of seasonal changes in pasture production, crop stress
detection and other environmental change detection (Singh,
1989), in the meantime, some methods were proposed by the
previous references (Fung T, 1987) such as image difference,
image regression, principal components analysis and
background subtraction, most of which just use single image
analysis such as gray information analysis to detect the
changes.
Up to now, for the change detection of man-made objects, such
as buildings and roads, little research had been done in the field
(Cushnie, 1989). Due to the limit of the spatial resolution satellite
images are difficult to be applied in detection of building of urban
area, aerial Image has proven to be an valuable data source for
these kind of application. In this paper, a new approach of
change detection of building based on aerial images was
explored and introduced.
As well known, changes of man-made objects of urban area are
certainly 3-dimendional real objects' changes. Buildings are
typically 3-dimendional real object whose change will cause the
change of height of DSM. If we could make use of the height
information to assist the detection of building change, it would
contribute to raise the efficiency and performance of the
detection.
This paper presented a new approach to combine the height
information and gray information of building as one kind of data
fusion technology to detect the change of building in urban area.
The corresponding theory and experiments were introduced and
analyzed below.
2. THE FUSION APPROACH OF DSM/DTM AND SINGLE
IMAGE ANALYSIS
The fusion approach of DSM/DTM and single image analysis
synthetically make use of both the technique of stereo image
analysis and of single image analysis such as gray and feature
analysis, exactly speaking, it was a method that single image
analysis was applied on the result of stereo image. It included
the following steps, new and old DSMs were created
automatically respectively by image matching firstly. By
comparing the new and old DSMs the changed regions were
extracted. After features were extracted from these regions, the
straight lines were detected. Finally, the changed buildings and
roads could be detected based on gradient direction analysis
and recognition of line pattern of building between registered
images.
