EAS-ASAR Integration in the Interferometric Point Target Analysis

EAS-ASAR Integration in the Interferometric Point Target Analysis
of 6
All materials on our website are shared by users. If you have any questions about copyright issues, please report us to resolve them. We are always happy to assist you.
Related Documents
    Rafael Zandona Schneider, Kostas Papathanassiou, Irena Hajnsek and Alberto Moreira German Aerospace Center  Microwaves and Radar Institute  PO BOX 1116, 82230 Wessling, Germany  Email: ABSTRACT SAR images over urban areas are characterised by strong geometrical distortions due to the side-looking acquisition geometry of SAR systems and the stepwise height variations of the scenes. This, combined with the complexity of the scattering processes in urban scenarios makes the interpretation and information extraction from SAR images over urban areas non-trivial. To simplify interpretation and information extraction, a subset of scatterers within the scene, the so-called coherent scatterers  (CSs), which are characterised by point-like scattering behaviour, have been previously  proposed [1]. The advantage of such CSs is that they have a deterministic or quasi-deterministic behaviour (no speckle) allowing – as far as possible – a direct interpretation, characterisation and information extraction. In recent works the estimation of CSs orientation angle (in relation to the Line of Sight direction) was demonstrated as well as the estimation of the dielectric properties of dihedral-like CSs. In this paper, the potential of CSs interpretation is further evaluated using data acquired by the airborne E-SAR system of the German Aerospace Center (DLR) in a Quad-pol mode at L-band and a Single-pol mode at X-band over the city of Dresden in Germany and over the Oberpfaffenhoffen region, also in Germany. In order to validate the CSs orientation angle estimation, a controlled experiment with dihedral corner reflectors deployed within the Oberpfaffenhofen test site has been set up. The achieved orientation angle estimation accuracy was about 1 degree. Finally, a comparison between CSs detected at X-band and L-band is  performed allowing additional insights in the physical nature of CSs. It was found out that by using SAR images with the same spatial resolution, more CSs are detected at L-band than at X-band and only very close to ‘ideal’ CSs (in their majority dihedrals- and surface-like scatterers) are detected at both frequencies. 1 INTRODUCTION Time phase stable scatterers have been proposed in terms of the so-called Permanent Scatterers (PSs) technique for SAR interferometric applications [2]. However, the identification of PSs is based on their temporal stability and relies on the availability of time series with a huge set of multiple images (about 30). If a wide bandwidth system is available, spectral properties can be used to detect point-like scatterers (Coherent Scatterers - CSs) by using only one SAR image. In [3] two methods to detect CSs based on spectral correlation of SAR images in the range direction have been  presented and validated using airborne SAR data acquired at L-band over the city of Dresden in Germany. It has been shown that the detected scatterers have in general high amplitude, high interferometric coherence and low polarimetric entropy, which characterise them as having point-like scatterering behaviour, as expected. It was also verified that CSs are strongly polarised, i.e., different CSs can be detected at different polarisation channels so that their number can increase significantly when using fully polarimetric data. CSs are deterministic or quasi-deterministic scatterers and therefore fully described by their scattering matrix. Hence, their characterisation and information extraction could be directly evaluated from these matrixes without the need of second order statistical parameters. Indeed, Line of Sight (LOS) orientation angle and dielectric constant of dihedral-like CSs have been estimated directly from the scattering matrix at L-band in [3] and [4]. In this work, validation of the orientation angle estimation using dihedral corner reflectors is addressed. In the second  part of the paper, the localisation of CSs at X-band is evaluated and compared to the CSs obtained at L-band. For both used test sites (Dresden and Oberpfaffenhofen), the total number of detected CSs is shown to be greater at L- band and the characteristic of the CSs common to X- and L-band are presented. 2 DETECTION The detection of the CSs is performed by evaluating the correlation coefficient between two parts of the full image spectrum (two sub-looks) and associate the pixels with high sub-looks correlation values to coherent scatterers [1], [3]. Usually, a focused SAR image is weighted by a Hamming window in order to reduce the side-lobes of the system COHERENT SCATTERERS IN URBAN AREAS: CHARACTERISATION AND INFORMATION EXTRACTION   _____________________________________________________________ Proc. Fringe 2005 Workshop, Frascati, Italy, 28 November – 2 December 2005 (ESA SP-610, February 2006)  impulse response. In this case, first a Hamming “un-weighting” process, i.e., a division of the full spectrum by the applied Hamming function has to be applied in order to recover the srcinal image spectrum. Then, the full spectrum is divided into two sub-looks that are shifted to the same central frequency in order to get a superposition of the sub-spectrums. A half-bandwidth Hamming weighting function is then applied to each of the two sub-spectrums. The inverse Fourier transform for both signals is performed and the normalised correlation coefficient (i.e. coherence) is estimated by applying for example a box-car average window. Pixels with a higher sub-looks coherence than a fixed threshold value are interpreted as CSs. 3 ORIENTATION ANGLE ESTIMATION Orientation estimation method For deterministic and quasi-deterministic scatterers (i.e. scatterers with not fully developed speckle pattern) as the CSs discussed here, Cameron’s decomposition allows the estimation of the orientation of the scatterer symmetry axis [5]. The scattering matrix of a single CS (in the reciprocal case) 40),])[sin(])[cos((][ minmax           sym symVV VH  HV  HH  S S  AbaccbaS S S S S   (where a ,  b , and c  are the corresponding Pauli components and    is related to the scatterer degree of symmetry) can be decomposed into two symmetric contributions [S maxsym ] and [S minsym ] [5]. The angle    that maximises the maximum symmetric contribution of [S] can by written as a function of the second and third Pauli components as 22** |||| )2tan( cbcbbc         is proportional to the orientation angle   a  = -    /2  of the symmetry axis of the maximum symmetric contribution relative to the basis in which the scattering matrix [S] was measured – relative to the H - V basis in this case. The ratio of the maximum symmetric power to the total power 10,|||||| |)sin()cos(||| |||||| |||| 2222222222   DoS cbacbacbaa DoS        is referred as the degree of symmetry (  DoS  ), and can be interpreted as a measure of confidence for the orientation angle   a  estimation. Experiment with dihedral corner reflectors In [3] and [4] the orientation angle of CSs was evaluated over the Dresden test site. While surface and dipole-like CSs were found to have in general a wide variety of orientation angles (different orientations over the city), dihedral-like CSs were in general oriented close to 0°, i.e., vertical oriented. In order to validate the orientation angle estimation procedure in a more quantitative way, an experiment using two dihedral corner reflectors as ground truth was performed on September 2005. Fig. 1 on the left shows one of the used dihedral reflectors. The dimensions of the vertical and horizontal surfaces are both 70x70 cm, which results in a theoretical maximal radar cross section at L-band:  max  = (8  .D 4 )/   2     20.2 dB (where D=0.7m is the dimension of the surfaces side). The dihedrals were positioned in the middle of a soccer field separated by about 40m from each other in azimuth. One dihedral was oriented in the vertical direction (   a  =0° ) while the other was about 5.0 degrees rotated in relation to the vertical (   a  =5° ). Fig. 1 on the right shows a SAR image at L-band of the region with the two dihedrals and the detected CSs. The dihedrals were detected through the CSs procedure and correct classified (red colour) as dihedral scatterers through the  polarimetric alpha angle. The two strong scatterers close to the dihedrals are two L-band transponders that were also  placed in the scene and used for another experiment. The Shh/Svv ratio was 0.978 which corresponds approximately to the theoretical ratio of 1.0 for a dihedral with ideal metallic surfaces [3], [4]. The estimated difference of the orientation angle between the two dihedrals was 4.03 degrees, i.e., an error of about 1.0 degree. The accuracy of 1.0 degree is a  satisfactory one and still a comment should be made here concerning the certainty of the dihedrals positioning. The  positioning was performed with a compass and a levelling, and consequently could also contain “man-induced” errors. If the high accuracy of this technique is confirmed, a new axis of measurement (the Line of Sight orientation angle) of deterministic or quasi-deterministic scatterers position could be obtained from full polarimetric data. Differential interferometry detects scatterer changes along the LOS direction while the scatterers orientation angle is estimated exactly in the plane perpendicular to the LOS direction, characterising full new information about the scatterer position. Hence, combined with differential interferometry, this technique could give additional information in order to achieve a two dimensional coordinate system of measurement (not projection) of scatterers changes. Fig. 1. Left: Dihedral corner reflectors. Dimensions: 70cm x 70cm. Maximal RCS at L-band:  max     20.2 dB. Right: Test site Oberpfaffenhofen. Detected CSs and classification through polarimetric alpha parameter. 4 L-BAND AND X-BAND COHERENT SCATTERERS COMPARISON In this section, a comparison between the detection of CSs at L-band and X-band is performed. The data acquired over the city of Dresden (August 2000) and over Oberpfaffenhofen (September 2003) are full polarimetric at L-band, however at X-band only the VV-channel is available. The resolution of the SLC images at the two frequencies is the same in range direction (1.5m) but different in azimuth: 1.2m at L-band and 0.6m at X-band. The comparison was done concerning the number of CSs detected at each band and the properties of the CSs common to L- and X-band. A rescaling factor in the azimuth direction of the L-band data was applied in order to correct the differences in the images dimension due to different airplane velocities in the acquisition time. Also, a corregistration procedure between L- and X-band data was necessary. A misregistration of about 1.5 pixels from near to far range was verified. In azimuth no misregistration was present, at least at the dimension of pixels. Fig. 2 shows the CSs detected in the Dresden test site at both L- and X-band using a threshold of 0.95 and a 5x5 window in the detection procedure (sub-looks coherence). The number of CSs detected at L-band is greater than at X- band of about 2.4 times for this specific case. Fig. 3 shows the results obtained over the Oberpfaffenhofen test site. Also in this case, the number of CSs was higher at L-band than at X-band, but now of a factor of almost 4.8 times. Fig. 4 shows the common CSs to X- and L-band, classified with the alpha angle evaluated using the polarimetric L-band data. One can see that the majority of the common CSs are dihedrals. More common CSs may be expected when a more effective corregistration procedure is applied. However, a quantitative analysis of the properties of the common CSs to L- and X-band is possible. Fig. 5 shows some characteristics of the common CSs detected at L- and X-band in the Dresden and in the Oberpfaffenhofen test sites. On the left are presented the amplitude, polarimetric entropy and polarimetric alpha angle normalised histograms of the Dresden region while on the right the corresponding histograms are presented for the Oberpfaffenhofen data. The dotted red line represents the histograms of the CSs detected only at L-band while the blue continuous line represents the histograms of the CSs detected at both L- and X-band data. The histograms show that the common CSs are in general the CSs with higher amplitude and lower polarimetric entropy, indicating that in general the common CSs are closer to ‘ideal CSs’. It can also be seen that they are in their majority dihedral-like and surface-like scatterers. This allow us to conclude that the resolution cells where CSs were detected at L-band but not at X-band may contain (smaller) secondary scatterers that are not sensed at L-band (due to the longer wavelength) but are sensed at X-band. In Surface Dihedral Dipol RGB-Coding: Range Dihedrals Transponders  other words, when the geometric resolution is the same in both frequencies, it may be more difficult to find one resolution cell with a point-like characteristic (no multi-scattering effects) at X-band than at L-band, as smaller wavelengths are more sensible to smaller scattering structures. Physical differences in the scattering structures inside a resolution cell will induce differences in the path of the electromagnetic wave leading to phase oscillations (variances) that will be greater at X-band than at L-band. These phase variances consequently induce decorrelation between sub-looks and the resolution cell in this case does not characterise point scattering behaviour and will not be detected as containing a CS. Fig. 2. Dresden test site. Left: L-band (184673 CSs detected). Right: X-band (78581 CSs detected). Fig. 3. Oberpfaffenhofen test site. Left: L-band (19513 CSs detected). Right: X-band (4040 CSs detected). L-band X-band L-band X-band   Fig. 4. Region of the Dresden test site highlighted in Fig. 2 with the common CSs detected at L-band and X-band. The majority of the common CSs are of type dihedral (red). 5 CONCLUSION & DISCUSSION High bandwidth systems allow the detection of point-like scatterers by using only one SAR image. If more SAR images are available, the common CSs can be selected, or each other merge procedure implemented, in order to obtain a temporal character for the set of CSs. The detected CSs are characterized by a strong polarimetric-dependent behavior. If full polarimetric data are available, the number of CSs can be significantly increased (up to 10, depending on the used thresholds and by applying  polarimetric optimisation techniques) [3], the detected CSs can be characterised in the sense of their scattering type (at least in canonical forms: surface, dipole, dihedral-like scatterers), and additional information can be extracted (as the CSs orientation angle and dielectric properties). It was also verified that by similar resolution systems more CSs can be detected at L-band than at X-band. The characteristics of the common CSs to L- and X-band indicate that these scatterers are the most close to ‘ideal CSs’, having (in general) higher amplitudes, lower polarimetric entropy, and that they are in their majority dihedral- and surface-like scatterers. 6 REFERENCES [1]   R. Z. Schneider, K. Papathanassiou, I. Hajnsek, and A. Moreira, “Analysis of coherent scatterers over urban areas,” Workshop on Applications of SAR Polarimetry and Polarimetric Interferometry (POLinSAR’2005) , Frascati, Italia, 17 - 21 January 2005. Available on [2]   A. Ferreti, C. Prati, and F. Rocca, “Permanent scatterers in SAR interferometry,”  IEEE Transactions on Geoscience and Remote Sensing  , vol. 39, no. 1, pp. 8-20, January 2001. [3]   R. Z. Schneider, K. P. Papathanassiou, I. Hajnsek and A. Moreira, “Polarimetric and interferometric characterisation of CSs in urban areas,”  IEEE Transactions on Geoscience and Remote Sensing  , in press. [4]   R. Z. Schneider, K. P. Papathanassiou, I. Hajnsek and A. Moreira, “Polarimetric interferometry over urban areas: information extraction using coherent scatterers”  Proceedings of Geoscience and Remote Sensing Symposium (IGARSS’2005) , Seoul, Korea, 25-29 July 2005. [5]   W. L. Cameron, N.N. Youssef and L.K. Leing, “Simulated polarimetric signatures of primitive geometrical shapes”  IEEE Trans. Geoscience and Remote Sensing, vol. 34, pp. 793-803, May 1996. Surface Dihedral Dipol RGB-Coding: Range
Related Search
We Need Your Support
Thank you for visiting our website and your interest in our free products and services. We are nonprofit website to share and download documents. To the running of this website, we need your help to support us.

Thanks to everyone for your continued support.

No, Thanks