Background of the Invention Repeat-pass space-borne interferometric SAR is a powerful <BR> <BR> technique for the observation of land surFace deformation, see, e. g. , P. Rosen et al.,"Synthetic Aperture Radar Interferometry", Proc. IEEE, Vol. 88, No. 3, 333-382 (2000). Causes for deformation include tectonic, seismic, and volcanic activity, ice and rock glacier motion, slope instability, and subsidence caused by ground water pumping, mining, hydrocarbon extraction, construction activities, natural compacting, and others.

For distributed scatterers the SAR image shows the typical"speckle" behavior. One consequence of the scattering characteristics is that it is the baseline dependent geometric decorrelation of the interferometric phase which limits the use of interferometry to baselines shorter than a so-called critical baseline, see the reference cited above. Similarly, differences in the skew angle result in geometric decorrelation. This baseline and skew angle dependent geometric decorrelation does not exist for a perfect point target.

Scatterers showing a significantly reduced geometric decorrelation are termed in this document as"point-like scatterers". For interferometry, the use of point-like scatterers has the advantage that the range of baselines for which the interferometric phases of pairs can be interpreted gets significantly extended towards larger baselines as compared to what is possible using distributed targets. Similarly, the interpretation of interferometric phases of pairs having larger differences in the skew angle (or Doppler frequency)

becomes possible. A point-like scattering characteristics in the cross-track direction is required in the first case, and a point-like scattering characteristic in the along-track direction in the second case.

Several years ago, the existence of large numbers of point-like scatterers in SAR imagery was demonstrated; see, e. g. , A. Ferretti, A. Monti- Guarnieri, C. Prati, and F. Rocca, Multi-baseline InSAR DEM Reconstruction: Data Fusion, EUSAR'98,25-27 May, Friedrichshafen, Germany, VDE-Verlag, ISBN 3-8007-2359-X, pp. 369-372,1998, A. Ferretti, A. Monti-Guarnieri, C.

Prati, and F. Rocca, Multi-image DEM Reconstruction, Procs. IGARSS'98, 6-10 July, Settle, WA, USA, pp. 1367-1369, 1998, and A. Ferretti et al., Non-linear subsidence rate estimation using permanent scatterers in differential SAR interferometry, IEEE TGRS, Vol. 38, No. 5,2202-2212, 2000,. The long-term stability observed for many of the point-like scatterers is of specific interest for deformation monitoring.

Description of the Prior Art Several methods for the identification of point-like scatterers used in the past are summarized in the following.

In multi-temporal series, point-like and distributed scatterers can be distinguished based on their different variability of backscatter intensity.

Distributed scatterers show a higher variability than point-like scatterers because of their speckle noise. An important drawback of this method is that a relatively large number of repeat-pass SAR images is required.

In SAR interferometry, coherence is used as a measure of the variability of the local spatial phase and the interpretability of the interferometric phase. Using this measure in stacks of multiple interferograms permits to identify pixels which remain coherent over time. Strictly speaking, this method does not identify point-like scatterers but pixels which remain coherent over time. An important limitation of this method is that the estimation of a spatial coherence requires multiple pixels. Multiple interferometric pairs are required and at least in a first phase pairs with large baselines or skew

angle differences cannot be used. As in method (a), a relatively large number of repeat-pass SAR images is required.

A. Ferretti, A. Monti-Guarnieri, C. Prati, and F. Rocca, Multi-image DEM Reconstruction, Procs. IGARSS'98, 6-10 July, Seattle, WA, USA, pp.

1367-1369,1998, describe a method to identify point-like scatterers in interferometric data stacks. Multiple point-wise differential interferograms, obtained after subtraction of the topography related phase term, are generated.

These complex images are summed up coherently. Only those pixels that contain very stable reflectors increase their amplitude. The authors present two examples of coherent and incoherent sums which allowed to identify Permanent Scatterers (PS).

In the so-called PS method, as described in (polimi PS patent/how to reference this?), more than 20 SAR scenes taken over a multi-year period are required, and the method requires a quite extensive analysis of interferometric phases.

All these methods of the prior art which were listed so far suffer from the major drawback that a quite large number of repeat-pass SAR scenes are required and that these methods are thus not applicable for single SAR scenes and small data stacks.

J. -C. Souyris, C. Henry, and F. Adragna, On the use of complex SAR image spectral analysis for target detection: assessment of polarimetry, IEEE TGRS, Vol. 41, No. 12, pp. 2725-2734,2003, present a method to identify point-like scatterers based on their spectral characteristics. They divide the range and azimuth spectra into two spectral bands. As possible measures to be used in the target indentification a cross-products between sub-looks, an interferometric coherence between sub-looks, and an"internal hermitian product"between sub-looks are investigated. In each of these cases ensemble averages are caluculated in the spatial domain over several pixels in range and azimuth direction. Furthermore, considerations concerning the use of polarimetric signatures are presented. The main application considered is ship detection. For the investigations airborne data are used. The authors mention

the identification of permanent scatterers in interferometry as a possible further application.

This last method does not depend on multiple SAR scenes. The method presented is for airborne SAR data. The method requires an averaging step in the spatial domain which results in a reduction of the spatial resolution of the target detection capability. This reduction in the spatial resolution, away from the single look geometry, is a major draw-back for the use of the method in the context of a subsequent interferometric analysis.

Object of the Invention The object of the present invention is to overcome the drawbacks of the known identification methods and to provide a new and useful method that does, before all, not require a great number of SAR images in order to identify point-like scatterers at the full spatial resolution of the SAR imagery.

Furthermore, the method to be provided should be able to be implemented with standard synthetic aperture radar sensors without need for special adaptations.

This object and still others are attained by the method of this invention that is defined in independent claim 1. Special or preferred embodiments of the method are the subject of dependent claims.

The method of the invention will now be explained in more detail by an example of execution.

For a SAR data set multiple images are processed with sub- bandwidths in the range and/or azimuth spectrum. This can either be done in the so-called SAR processing from RAW data by using appropriate processing spectra or from already processed single-look complex (SLC) imagery (see Figure 1), by using band-pass filtering. In the example presented the second method was used. The azimuth and range spectra were divided into 4 sub- bandwidths, resulting in a total of 16"spectral looks"images. A"spectral coherence", ys, is then calculated from these"spectral looks"images by:

in which Si denotes a the complex value of the reference spectral look, a spectrally centered reference is used, and s2, denotes the complex value of a different spectral look. The brackets () represent the ensemble average, which is estimated by averaging over the i different second looks. In the specific example described the azimuth and range spectra were divided into 4 sub-bandwidths, resulting in a total of 16"spectral looks"images. One was selected as reference and the other 15 correspond to the S2, i images.

Consequently, the ensemble averages were calculated based on 15 samples.

The resulting"spectral coherence"image is shown in Figure 2.

This"spectral coherence"can then be used to identify point-like scatterers. In the specific example described this was done using a simple thresholding with a threshold value of 0.5. Figure 3 shows the points with "spectral coherence"values above a certain threshold. One important aspect is that this"spectral coherence"has a distinct maximum which permits identification of the point-like scatterers at the full spatial resolution of the full- bandwidth SAR imagery. Figure 4 shows the resulting point-like scatterers identified from the single SLC.

Another important aspect is that the method is applicable with a single or multiple complex SAR images. In the case of multiple images the identification of the point-like scatterers is done based on the"spectral coherence"values calculated for each complex SAR image.

In the specific example described this was done for 5 repeat-pass SAR images over the same site. The average of the 5 spectral coherence images were then averaged and a thresholding with a threshold value of 0.4 was applied to identify the point-like scatterers. Figure 5 shows the"average spectral coherence"and Figure 6 the points with"average spectral coherence" values above a 0.4. The"average spectral coherence"has a distinct maximum which permits identification of the point-like scatterers at the full spatial

resolution of the full-bandwidth SAR imagery. Figure 7 shows the resulting point-like scatterers identified from the 5 SLCs.

The advantage of using multiple SLCs is that the identification gets more robust in the sense that a higher fraction of the points identified are useful for interferometric analysis. Further criteria based on the backscatter intensity and/or the mean-to-sigma ratio of the sub-bandwidth looks images can also be used to enhance the robustness of the methodology.

A high fraction of the identified points were found useful for the interferometric analysis. This was confirmed by the fact that interpretable interferometric phases were obtained for these points even for long interferometric baselines.

Brief Description of the Drawing Fig. 1 represents a small section of a Single Look Complex (SLC) image (or more precisely speaking the detected single look intensity), using power-law gray scale for display. The area shown corresponds to 100 x 200 SLC pixels or approximately 2 km x 0.800 km on the ground. ERS raw data copyright ESA, processing by GAMMA.

Fig. 2 represents the"spectral coherence"calculated for the SLC of Figure 1, using linear gray-scale for display. A total of 16 looks were processed by dividing the range and azimuth spectra in four parts each.

Fig. 3 represents the"spectral coherence"shown in Figure 2, but using white for pixels with values below a threshold of 0. 5 and a linear gray- scale for the display of values above the threshold.

Fig. 4 shows the locations of the point-like scatterers identified based on the spectral coherence calculated for the single SLC of Figure 1. The

area shown corresponds to 100 x 200 SLC pixels or approximately 2 km x 0.800 km on the ground.

Fig. 5 represents the"average spectral coherence"calculated from 5 SLCs, using linear gray-scale for display. For each of SLC 16 looks were processed by dividing the range and azimuth spectra in four parts each. For each pixel the 5 resulting spectral coherence images were averaged.

Fig. 6 represents the"average spectral coherence", calculated from 5 SLCs, using white for pixels with values below a threshold of 0.4 and a linear gray-scale for the display of values above the threshold.

Fig. 7 shows the locations of the point-like scatterers identified based on the"average spectral coherence"calculated from 5 SLCs. The area shown corresponds to 100 x 200 SLC pixels or approximately 2 km x 0.800 km on the ground.

Detailed Description of the Invention The method according to the invention, an Example of its implementation having just been described, is applicable with single or multiple complex SAR images. In the case of multiple images, the identification of the point-like scatterers is done based on the"spectral coherence"values calculated for each complex SAR image.

It was shown that the point-like scatterers identified with the method of this invention are suited for interferometric analysis of multi-temporal data stacks (see for example Wegmüller U. , C. Werner, T. Strozzi, and A.

Wiesmann,"Multi-temporal interferometric point target analysis", Proceedings Multi-temp 2003 conference, Ispra, Italy, 16. -18. July 2003) The method provides interpretable interferometric phases for point-like scatterers even for long interferometric baselines. One major advantage of the proposed method is that it permits identification of point-like scatterers in small data stacks while previous methods required large data stacks.