ISPRS, Vol.34, Part 2W2, “Dynamic and Multi-Dimensional GIS”, Bangkok, May 23-25, 2001 
278 
A FRAMEWORK FOR AUTOMATED CHANGE DETECTION SYSTEM 
Haigang SUI Deren LI Jianya GONG 
Lab for Information Engineering in Surveying, Mapping And Remote Sensing (LIESMARS), Wuhan University, 129 Luoyu Road, Wuhan, 
Hubei,P.R.China,430079 
e-mail: shg@rcgis.wtusm.edu.cn 
Keywords: remote sensing, change detection, change feature extraction 
Abstract 
To enhance the ability of remote sensing system to provide accurate, timely, and complete geo-spatial information at regional or global 
scale ,an automated change detection system has been and will continue to be one of the important yet challenging problem in remote 
sensing. In this paper we designed a framework for automated change detection system at landscape level using various geo-spatial 
data sources including multi-sensor remotely sensed imagery and ancillary data layers. In this framework database is the central part 
and some associated techniques are discussed. These techniques includes five subsystems: automated feature-based image 
registration, automated change identification, automated change feature extraction, intelligent change recognition, change accuracy 
assessment and database updating and visualization. 
1.Current change detection techniques 
Automatic change detection in images of a given scene 
acquired at different times is one of the most interesting topics 
of image processing. It finds important applications within 
different contexts, ranging from visual surveillance and video 
coding to tracking of moving objects, from map updating to 
environment monitoring[Rosin,1999].The basic assumption 
behind change detection is that any changes on the ground 
must result in changes in radiance values, so the changes 
must be detected from noise by other factors ,such as 
differences in atmosphere conditions .differences in 
illumination condition, differences in relief condition 
.differences in soil moisture and registration noise. 
Usually, change detection involves the analysis of two 
registered multi-spectral remote sensing images acquired in 
the same geographical area at two different times. In the 
remote sensing literature, two main approaches to the change- 
detection problem have been proposed: the supervised and 
the unsupervised approach. The former is based on 
supervised classification methods, which require the 
availability of a multi-temporal ground truth in order to derive a 
suitable training set for the learning process of the classifiers. 
The latter performs change detection by making a direct 
comparison of the two images considered without relying on 
any additional information. 
Post Classification Comparison and Direct Multi-date 
Classification are the commonly used techniques in the 
supervised approach. Post Classification Comparison is the 
most intuitive techniques in practice change detection, which 
simply classifies the images of two times separately and 
compares the classified maps on a pixel-by-pixel basis to 
identify the changes. In contrast, the Direct Multi-date 
Classification deals with the multi-spectral images of the two 
times simultaneously. Each change combination between two 
times is represented as a output class and the whole change 
detection process is treated as one classification. The 
supervised approach exhibits some advantages over the 
unsupervised one(e.g., capability to explicitly recognize the 
kinds of land cover transitions that have occurred .robustness 
to the different atmospheric and light conditions at the two 
acquisition times, ability to process multi-temporal and/or multi 
sensor imagesj.The disadvantages include greater 
computational and labeling requirements, severe difficulty in 
obtaining individual classification accuracy and difficulties 
inherent in performing adequate accuracy assessment on 
historical data sets. Another drawback is the generation of an 
appropriate multi-temporal ground truth is usually a difficult 
and expensive task [Bruzzone et al.,2000].Consequently ,the 
use of effective unsupervised change detection methods is 
fundamental in many applications in which a ground truth is 
not available. 
The most widely used types of the unsupervised change 
detection techniques is the so-called “difference image”. These 
techniques involve image difference, image ratio, change 
vector analysis(CVA) .principle component analysis(PCA) and 
other methods such as neutral network .morphological