Vol.28 No.1 JOURNAL OF ELECTRONICS(CHINA) January 2011
USING COVARIANCE INTERSECTION FOR CHANGE DETECTION IN
REMOTE SENSING IMAGES
1
Yang Meng Zhang Gong
(Nanjing University of Aeronautics and Astronautics, Nanjing 210016, China)
Abstract In this paper, an unsupervised change detection technique for remote sensing images ac-
quired on the same geographical area but at different time instances is proposed by conducting Co-
variance Intersection (CI) to perform unsupervised fusion of the final fuzzy partition matrices from the
Fuzzy C-Means (FCM) clustering for the feature space by applying compressed sampling to the given
remote sensing images. The proposed approach exploits a CI-based data fusion of the membership
function matrices, which are obtained by taking the Fuzzy C-Means (FCM) clustering of the fre-
quency-domain feature vectors and spatial-domain feature vectors, aimed at enhancing the unsuper-
vised change detection performance. Compressed sampling is performed to realize the image local
feature sampling, which is a signal acquisition framework based on the revelation that a small collection
of linear projections of a sparse signal contains enough information for stable recovery. The experi-
mental results demonstrate that the proposed algorithm has a good change detection results and also
performs quite well on denoising purpose.
Key words Change detection; Covariance Intersection (CI); Fusion; SAR image; Multi-spectral
image
CLC index TP751.1
DOI 10.1007/s11767-011-0532-x
I. Introduction
Change detection is one of the most important
applications of the remote sensing technology and
it plays a more and more important role in a variety
of fields, like studies on land-use or land-cover
dynamics, monitoring shifting cultivations , burned
area assessment, analysis of deforestation processes,
identification of vegetation changes, monitoring of
urban growth, etc.. Usually, change detection aims
at discerning areas of change on two registered
remote sensing images acquired in the same geo-
graphical area at two different times. This is useful
to identify vegetation changes, monitoring shifting
cultivations, studies on land-use, etc.. Particularly,
in disaster management case, the fast and accurate
detection of affected regions in multi-temporal
images acquired at two different time instances.
1
Manuscript received date: July 13, 2010; revised date:
December 22, 2010.
Supported by the National Natural Science Foundation of
China (No. 61071163).
Communication author: Yang Meng, born in 1980, male,
pursuing the Ph.D., College of Information Science &
Technology, Nanjing University of Aeronautics and As-
tronautics, Nanjing 210016, China.
Email: yangmeng372901@163.com.
Change detection may be done either in su-
pervised or in unsupervised manner. Relevance of
unsupervised techniques is more than supervised
ones for this problem as in most of cases we do not
have the ground truth information. We can view
unsupervised change detection problem as a clus-
tering problem where the task is differencing
[1]
or
rationing
[2]
, which accomplish to discriminate the
data into two groups changed and unchanged.
Unsupervised change detection techniques mainly
use the automatic analysis of the change data
which are constructed using the remote sensing
images. Most of the unsupervised methods are
developed based on the image the change detection
by subtracting or dividing, on a pixel basis, the
images acquired at two instances to produce new
image called difference image. In Ref. [3], an
automatic technique based on the Principal Com-
ponent Analysis (PCA) for the analysis of the dif-
ference image is proposed. The method does not
use complicated data modeling and parameter
estimate, which produces promising results with a
low computational cost. It employs PCA for di-
mension reduction and feature extraction. PCA can
only separate pair-wise linear dependences between