International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B2. Istanbul 2004 
4. CONCLUSIONS AND FUTURE WORKS 
This study developed a method for extracting feature change 
using time series analysis on multitemporal JERS-1/SAR 
images. The method consisted of two steps. In the first step, the 
stationarity of land surface change was extracted as a change 
process model, and in the second step, the stationarity of the 
newly acquired data was evaluated by comparing it with the 
change process model. Also the practicality of the method was 
examined from the view point of the "time inconsistency" 
problem. The results showed satisfactory values for change 
detection. The method is expected to provide an effective 
solution to “time inconsistency” problem. 
In addition, insufficient results were observed in the developed 
method, which mostly stemmed from the limitation of using a 
single type of data. The goal of this study was to develop a 
method for automated change detection in near real-time which 
was capable of practical use. Improvement in the accuracy of 
the change detection in multiple aspects (e.g. spatial, temporal, 
thematic) will be conducted in the future through assimilating 
additional techniques such as construction of a detailed base 
data, use of optical sensor imagery, and development of spatial 
analysis algorithms. 
Constructing the detailed base data 
Although the method developed in this study was designed to 
detect change, it is still difficult to determine thematic aspects 
of the change. This limitation mainly stems from the fact that 
the method is relying on a single type of data (i.e. SAR image). 
Therefore, integration of other spatial data will be conducted to 
improve the interpretation of the change and determine thematic 
aspects. The base data will be constructed to archive the 
thematic information of the intended area to identify the 
thematic aspects “before” the change. The base data will 
include land cover data, land use data and topographic data. The 
spatial resolution will be 1-10m. Higher resolution than SAR 
data will be required for the base data. High resolution optical 
images (e. g. IKONOS, aerial photograph) and laser profiler 
data will be used to satisfy this requirement. The construction 
of the base data may enable the estimation of stationary changes 
itself. The method of integrating this estimation into the time 
series analysis of multitemporal SAR images will also be 
considered as a future work. 
Use of the optical sensor image (not periodical) 
Additionally, the utilization of optical sensor images will be 
conducted to enable the identification of thematic aspects 
“after” the change. It is true that the periodicity of optical 
sensor images is poor because of the effect of cloud cover. Even 
so, optical images contain very rich information and they will 
be useful for understanding thematic information about the land 
surface. There are several studies of integrating multitemporal 
SAR images and optical sensor images (e.g. Michelson er al, 
2000; Le ef al., 2000; Lombardo et al., 2003). 
Integrating the result of spatial analysis into time series 
analysis 
Most of the conventional time series analysis is done by pixel 
based analysis, which does not consider the relations among 
surrounding pixels. The pixel values of SAR images are mostly 
reflecting the shape of the terrain (and some of roughness and 
material). The classification result may improve if each pixel is 
Un 
n3 
understood through the segmented features. The information of 
the relation among pixels and its corresponding segment may be 
used as a parameter to improve the precision of the time series 
analysis on multitemporal SAR images. The segmentation 
technique used for the high-resolution satellite image (Baatz 
and Schape, 1999) is also planned to be used. 
5. REFERENCES 
Baatz, M.,and Schape, A., 1999. Object-Oriented and Multi- 
Scale Image Analysis in Semantic Networks. Proc. of the 2nd 
International Symposium on Operationalizaion of Remote 
Sensing August 1 6" -20”, Netherlands, http://www definiens- 
imaging.com/documents/publications/itc 1999.pdf, (accessed 27 
Apr. 2004) 
Bruniquel, J. and Lopés, A. 1997. Multi-variate optimal 
speckle reduction in SAR imagery, International Journal of 
Remote Sensing, 18(3), pp. 603—627. 
Ciuc, M., Bolon, P., Trouve, E., Buzuloiu, V. and Rudant, J.P., 
2001. Adaptive neighborhood speckle removal in multitemporal 
SAR images. Applied Optics, 40(32). 
Coltuc, D., Trouvé, E., Bujor, F., Classeau, N. and Rudant J. P., 
2000. Time-space filtering of multitemporal SAR images. Proc. 
IGARSS, 7, Honolulu, USA, pp. 2909-2911. 
Le Hegarat-Mascle, S., Quesney, A., Vidal-Mdjar, D., Taconet, 
O., Normand, M. and Loumagne, C., 2000. Land cover 
discrimination from multitemporal ERS images and 
multispectral Landsat images: a study case in an agricultural 
area in France. International Journal of Remote Sensing, 21(3), 
pp. 435-456. 
Lombardo, P., Oliver, C. J., Macri Pellizzeri, T. and Meloni, M., 
2003. A new maximum-likelihood joint segmentation technique 
for multitemporal SAR and multiband optical images. IEEE 
Transactions on Geoscience and Remote Sensing, 41(11), pp. 
2500-2518. 
Michelson, D. B., Liljeberg, M. and Pilesjó, P., 2000. 
Comparison of algorithms for classifying Swedish landcover 
using Landsat TM and ERS-1 SAR data. Remote Sensing of 
Environment 71, pp. 1-15. 
Saito, K., Koyama, A., Yoneyama, K., Sawada, Y. and Ohtomo, 
N. (Eds.), 1994. A Recent Advance in Time Series Analysis by 
Maximum Entropy Method. Hokkaido University Press, Japan. 
DE 
KEY? 
ABST 
This p 
Orient 
The in 
Cross 
approa 
surface 
technic 
All thr 
The te 
LSM 1 
been 4 
Digita 
also in 
aerial 
Althou 
photos 
autom: 
remote 
techno 
With 
compu 
Photog 
techni 
and th 
using | 
Inner 
compu 
which 
applic: 
inner | 
will af 
In 19 
autom 
and px 
pixel. 
In in 
relatio 
coordi 
systen 
photos 
pixel