You must have a javascript-enabled browser and javacript and stylesheets must be enabled to use some of the functions on this site.
 Earthnet Home  
Sessions and Session Summaries
First Annoucement
Scientific Committee
ESA Specific Links
Organising Committee
Round Table Discussion Questions
Conference Photos
List of Participants
All papers


Multitemporal and/or Polarimetric SAR Characterization of Urban Areas

Fabio Dell'Acqua (1)and Prof. Paolo Gamba(1)

(1) Università di Pavia, Via Ferrata, 1, I-27100 Pavia, Italy



A few studies have shown that polarimetric SAR data sets may be useful for discriminating different urban environments. In particular, in [1] the AIRSAR data over Sydney have been considered, and the potentials for characterizing different building clusters by extracting single, double and triple bounce effects have been demonstrated. Similarly, in [2] the different statistical properties of built aggregates in a urban area using SIR-C measurements have been discussed in order to provide a basis for an efficient segmentation of polarimetric urban SAR data sets. Finally, the availability of SRL-1, 2 and possibly SRTM mission allows considering multitemporal and multiparametric characterization of urban areas discussing the mutual effects of temporal/polarimetric redundancy [3]. To provide a quantitative discussion on the usefulness and the limits of SAR data sets at the spatial resolution that satellite sensor are currently able to provide, and to further validate these results, a comprehensive study is proposed in this paper using ERS, RADARSAT and SIR-C/X-SAR data over the town of Pavia, northern Italy.


Using nine ERS images, 5 SIR-C/X-SAR data sets and one RADARSAT image over the same urban area, many investigations may be produced. The precise knowledge of the site and the availability of a well tested neural classifier allows a quantitative comparison on both the possibility to discriminate among different urban environments (city center vs. residential areas vs. sparse buildings) and among some of the most important land cover/land use classes (built up areas, streets and railways, vegetation, water...). In particular, the neural kernel-based classifier is based on a two-step approach, performing first the image segmentation on a pixel-by-pixel basis. Then, the spatial analysis is carried out by means of a second classification, but using as input vector the percentages of pixels in a window around the current pixel position assigned to each class by the first classifier. These steps are implemented either by an unsupervised neural network (ART-2), followed by a fuzzy clustering using a standard Fuzzy-C-Means (FCM) algorithm [4], or a supervised fuzzy ARTMAP structure [5]. Results and conclusions We compare classification results obtained by using polarimetric, multitemporal or multitemporal/polarimetric data sets, aggregating SAR polarizations or dates and exploring also the interactions among SAR data recorded by different sensors. Since SIR-C/X-SAR data were recorded in April 1994 , they partially overlap with the ERS images, providing a unique set for this kind of analysis. Classifications are obtained also by exploiting texture measures for a better discrimination of urban environments. We want to discuss which are the advantages of a polarimetric data sets, and to evaluate the usefulness, for our purposes, of using more polarizations with respect to a combination of multitemporal single polarization images. This could also be used to determine which of the many functioning modes of the ASAR sensor would be most useful for urban mapping and/or monitoring applications.


[1] Y. Dong, B. Forster, and C. Ticehurst, 'A new decomposition of radar polarization signatures,' IEEE Trans. Geosci. Remote Sensing, vol. 36, pp. 933-939, May 1998.

[2] E. Costamagna, P. Gamba, P. Lombardo, G. Chinino: "Statistical analysis and neuro-fuzzy classification of polarimetric SAR images of urban areas", Proc. of the ERS/ENVISAT Symposium, Gotheborg, Sweden, Oct. 2000.

[3] F. Dell"Acqua, P. Gamba, P. Lombardo, T. Macrì Pellizzeri, D. Mazzola: "Multiband SAR classification using contextual analysis: annealing segmentation vs. a neural kernel-based approach", Proc. of IGARSS"02, Toronto (CAN), June 2002, Vol. V, pp. 2593-2595.

[4] P. Gamba, B. Houshmand: "An efficient neural classification chain for optical and SAR urban images", International Journal of Remote Sensing, Vol. 22, n. 8, pp. 1535-1553, May 2001. [5] G. Amici, F. Dell"Acqua, P. Gamba, G. Pulina: "Fuzzy, neural and neuro-fuzzy classification of pre- and post-event SAR images for flood monitoring and disaster mitigation", Proc. of the First International Workshop on Multi-temporal Analysis of Remote Sensing Images, Trento, Italy, 13-14 Sept. 2001.


Full paper


  Higher level                 Last modified: