International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B2. Istanbul 2004 
  
  
according to the maximum change magnitude value. The 
growing criteria are taking into consideration the radiometric 
parameters and the gradient magnitude of the neighboring 
pixels. The gradient magnitude is computed using (wo 
directional 3x3 Sobel operators. Figure 4 shows an example of 
changed-pixels cluster which was detected as part of large-area- 
object. 
4.4 Change Identification 
After the process of labeling objects which were changed, a 
process of identifying the type of the change is implemented. 
The change identification process is performed using the rule- 
based system. This process consists of two tests for each object. 
The first test takes into consideration the radiometric and the 
textural parameters. For each type (code), the radiometric and 
textural parameters are checked whether they meet the specified 
rules. The result of this test could be zero, one or multi 
matching types. If no matching type is found then the object 
could not be identified. This could occur due to small area 
objects or untested types, such as building shadows or lack of 
statistically valid samples in the tested spatial database. On the 
other hand, if one matching type was found then the object will 
be identified by this step. Otherwise, the next test is 
implemented. This test includes the geometrical parameters and 
the topological relationships. This test is performed in a 
hierarchical order to find the best fitting type for the object. For 
example, the two buildings in figure (3b) were identified as 
"red" roof building (code=702) using only the radiometric and 
textural test. On the other hand, the road in figure (3d) was 
identified, in the first test, as dirt road (code=111) and also as 
"cultivated area" (code=608). However, by using the 
geometrical and the topological rules, in the second test, the 
system was able to define this object as a dirt road, due to the 
high value of the elongation and the connection to the road 
network. 
5. SUMMARY 
[n each stage of this study, different methods were implemented 
and tested. Also, optimal thresholds were searched for, in order 
to enable automatic execution of these methods. In summary, 
the research has a number of contributions: 
i. Automatic detecting and labeling of "changed" 
objects. The detection was implemented in "object-wise" 
manner rather than single pixel treatment, as used in many 
traditional methods. The labeling process is done by using the 
region growing segmentation method, which takes into 
consideration four different data sources: change intensity 
index, edge gradient, radiometric data and the objects from the 
existing spatial database. 
2. Improvement of methods of classification, by using 
rule based system, both in terms of accuracy and efficient 
operation. The rule base system is based on four data sources: 
radiometric data, geometric parameters, texture parameters and 
topological relationships between different objects. 
3 Implementing quality control process of the spatial 
database, according to the remotely sensed data. This process in 
itself is an important enhancement toward automatic updating of 
the spatial databases. 
6. REFERENCES 
Bellacicco, A., 1996. Fuzzy Logic in the Analysis of Spatial 
Data. In: Sistema Terra: Remote Sensing and the Earth, 
December, 1996, Rome, pp. 82-83. 
Jha, C.S. & Unni, N.V.M., 1994. Digital Change Detection of 
Forest Conversion of a Dry Tropical Indian Forest Region. In: 
International Journal of Remote Sensing, 1994, Vol. 15, No. 13, 
pp. 2543-2552. 
Metternicht, G., 1999. Change Detection Assessment Using 
Fuzzy Sets and Remotely Sensed Data: An application of 
topographic map revision. In: /SPRS Journal of 
Photogrammetry and Remote Sensing, Vol. 54, No. 4, pp. 221- 
233. 
Mouat, D.A., Mahin, G.G. & Lancaster, J., 1993. Remote 
Sensing Techniques in the Analysis of Change Detetction. In: 
Geocarto International (2) 1993. pp. 39-50. 
Muchoney, D.M. & Haack, B.N., 1994. Change Detection for 
Monitoring Forest Defoliation. In: Photogrammetric 
Engineering & Remote Sensing, October 1994, ASPRS, Vol. 
60, No. 10, pp. 1243-1251. 
Peled, A., 1993, Change Detection: First Step toward Automatic 
Updating, ACSM-ASPRS Annual Convention & Exposition 
Technical Papers, Vol. 3, pp. 281- 286, USA 
Peled, A., Gilichinski, M., 2004. GIS-Driven Analyses of 
Remotely Sensed Data for Quality Assessment of Existing Land 
Cover Classification, 20" ISPRS Congress, 12-23 July, The 
International Archives of Photogrammetry and Remote Sensing, 
Vol. IV, Istanbul. 
Tilton, J.C., 1998. Image Segmentation by Region Growing and 
Spectral Clustering with a Natural Convergence Criterion. In: 
Proceedings of the 1998 International Geoscience and Remote 
Sensing Symposium, Seattle, WA, July 6-10, 1998. 
Warner, T.A., & Shank, M.. 1997. An Evaluation of the 
Potential for Fuzzy Classification of Multispectral Data Using 
Artificial Neural Networks. In: Photogrammetric Engineering 
& Remote Sensing, November 1997, Vol. 63, No. 11, pp. 1285- 
1294. 
500 
Interna 
  
_ 
+ 
pe 
  
  
  
— 
Figure