Polarimetric UHF Calibration for SETHI

Transcription

1 PIERS ONLINE, VOL. 6, NO. 5, Polarimetric UHF Calibration for SETHI H. Oriot 1, C. Coulombeix 1, and P. Dubois-Fernandez 2 1 ONERA, Chemin de la Hunière, F Palaiseau cedex, France 2 ONERA, Base aérienne 71, Salon AIR, France Abstract The SETHI sensor from ONERA is an airborne radar equipped with different bands (UHF-VHF, L, X) on a Falcon 2. The UHF-VHF band is fully polarimetric and can operate between 225 MHz up to 46 MHz. In August 29, this system was used in French Guiana for the TROPISAR experiment. Data were acquired from incidence angle up to 65 incidence angle and a swath of about 6 m. The antenna pattern being quite large, this configuration induced a secondary reflection of the incident wave with the wing leading to an antenna pattern deformation (in intensity, phase and frequency). This paper explains the calibration procedure that was performed on the data. First the crosstalk is estimated and removed, then, the antenna pattern is compensated. 1. INTRODUCTION The SETHI sensor from ONERA (Figure 1) is an airborne radar equipped with different bands (UHF-VHF, L, X) on a Falcon XX [1]. The UHF-VHF band is fully polarimetric and can operate between 225 MHz up to 46 MHz. In August 29, this system was used in French Guiana for the TROPISAR experiment. This experiment was conducted to provide ESA and CNES with measurements of temporal coherence at P- and L-band over tropical forests for time intervals compatible with spaceborne missions (typically 2 3 days). In this paper we present the methodology used to calibrate the UHF data. Data were acquired from incidence angle up to 65 incidence angle and a swath of about 6 m. The UHF antenna is located under the left wing of the Falcon 2. For this experiment, it was pointing with an incident angle of 45 and the bandwidth used was from 335 MHz up to 46 MHz. The antenna pattern being quite large, this configuration induced a secondary reflection of the incident wave with the wing leading to an antenna pattern deformation (in intensity, phase and frequency). This paper explains the calibration procedure that was performed on the data. First the cross-talk is estimated and removed, then, the antenna pattern is compensated. 2. PROCESSING The full polarimetric data were processed using the ONERA SAR processor [2]. This processor uses a modified version of the RMA algorithm. Data were processed with a 1 m meter resolution in range and 1.5 m resolution in azimuth. The analysis of the four channels show some defaults: Figure 1: Falcon XX equipped with the ONERA Sethi system.

2 PIERS ONLINE, VOL. 6, NO. 5, Figure 2: Color composite of the calibration test-site (Rochambeau airport, French Guiana). Hh channel appears in red, Hv, in green, V v in blue. The aircraft is flying from the left-bottom of the image to the right bottom of the image. - Hv and V h appear to be unbalanced. - Radiometry varies across the swath and these variations differ from one channel to the other (this variation can be observed on Figure 2: the main color of the image varies with range which indicates that the radiometric variations are different from channel to channel). - The polarimetric phase Hh-V v is also affected. The data acquisition geometry is suspected to be responsible for this behavior: The pod is located under the wing and the antenna pattern is large so a secondary reflection of the incident wave with the wing leads to an antenna pattern deformation (in intensity, phase and frequency). This deformation has to be compensated by the calibration procedure. This procedure is divided into two steps: the first one aims at removing the cross-talk that can be observed on the images; the second one aims at improving the radiometric and phase calibration of the four channels. 3. CALIBRATION PROCEDURE 3.1. Crosstalk Removal The crosstalk removal procedure was inspired by the Quegan s polarimetric calibration procedure described in [3]. S. Quegan models the cross-talk by a Transmission and Reception matrices. SmHh SmV h RHh RV h SHh SV h THh TV h = => SmHv SmV v RHv RV v SHv SV v THv TV v 1 wk SHh SV h αk αkz SmHh SmV h = SmHv SmV v v 1 u k1 SHv SV v RHh TV v RV v with α =, k= RV v THh RHh where Smij stands for the measured signal for the polarization channel ij, Sij the polarized signal on the ground, Rij, reception matrix Tij Transmission matrix. The R and T matrix coefficients are rewritten as a set of parameters (u, v, w, z, α, k). The complete set of parameters can be found in [3]. Under the hypothesis that the co-polar and cross-polar channels are uncorrelated over natural clutter (under the assumption of reflection symmetry), and that the Hv channel and V h channel are equal (if one neglects the noise level). The covariance matrix hs Si can be written as: * SHh + σhh Hh σhh V v σhv Hv σhv V h S hs Si = Hv [SHh SHv SV h SV v ] = σhv V h σv h V h SV v σhh V σv v V v SV v v S. Quegan shows that the parameters (u, v, w, z, α) only depend on hsm Smi. Therefore, they can be computed from the polarimetric covariance matrix estimated form the data.

3 PIERS ONLINE, VOL. 6, NO. 5, In this experiment, we modified the Quegan s methodology by (1) removing the geometrical cross-talk due to attitude variation of the plane over the flight (2) estimating the Quegan s parameters both in range and frequency. Figure 3 shows an example of the α parameter estimated over frequencies (vertical axis) and range (horizontal axis) on our data. This figure shows the dependency of the parameter with both range and frequency. Figure 4 is a plot of the v parameter according to the incidence angle before removing the aircraft attitude (left figure) and after removing the aircraft attitude. The different curves correspond to different acquisitions and attitude conditions. The v parameters differ if the aircraft attitude is not removed whereas, they display the same behavior when the attitude is compensated Image Quality Improvement The secondary bounce of the wave on the wing results in an antenna pattern deformation. This deformation depends on (1) the frequency, (2) the incidence angle of wave and (3) the polarization state. In order to remove this effect, we used the corner reflectors that have been placed across the swath on the calibration area. 5 corner reflectors were installed and several acquisitions were gathered using different flight configurations so that we have the frequency response of the corner reflectors for 11 incidences. By measuring the frequency response of each corner reflector on the images and interpolating these frequency responses in between, the residual radar transfer function was estimated and removed from the different channels. 4. RESULTS In order to assess the quality of the calibration we used several measurements: - we verified that the ratio Hv/V h was equal to 1 in intensity and in phase. - we measured the system noise by computing the Hv Hv - Hv V h. This measurement should only be noise and should not depend on the nature of the ground. It is very useful measurement to assess the quality of the cross-talk removal. Results are shown on Figure Figure 3: Quegan s parameter α estimated for different range and different frequencies on the TROPISAR data. Figure 4: Quegan s parameter v estimated over range before aircraft-attitude compensation (left image) and after aircraft-attitude compensation (right images). The different curves correspond to different data with different aircraft attitudes.

4 PIERS ONLINE, VOL. 6, NO. 5, The left image corresponds to the noise image before cross-talk removal, the right image corresponds to the one after noise removal. The ground is less visible on the right image than on the left image showing that the cross-talk removal procedure was effective. - We measured the radar cross section of the different corner reflectors across the swath to verify that the radiometry was correctly calibrated. Results are presenting in Table 1. The final color composite after calibration is presented in Figure 6. The radiometric artifacts that were observed on Figure 2 have disappeared. Table 1: Radar cross section of 3 corner reflectors placed across the swath for the different polarization states. Polarization Corner reflector #1 Corner reflector #2 Corner reflector #3 V v 22.6 dbm 2 23 dbm 2 23 dbm 2 V h.5 dbm dbm dbm 2 Hv.7 dbm dbm dbm 2 Hh 22.1 dbm dbm dbm 2 Figure 5: Noise image computed using the cross-polarized channels before (left image) and after (right image) cross-talk removal. Figure 6: Color composite of the calibration test-site (Rochambeau airport, French Guiana) after calibration (right). Hh channel appears in red, Hv, in green, V v in blue. The aircraft is flying from the left-bottom of the image to the right bottom of the image.

5 PIERS ONLINE, VOL. 6, NO. 5, CONCLUSION In this paper, a calibration procedure for large band UHF full polarimetric data was developed. This procedure is based on a modified version of the Quegan s algorithm and the estimation/correction of a the residual radar transfer function across the swath, which was found to depend on range due to a secondary bounce of the incident wave on the aircraft wing. ACKNOWLEDGMENT The TropiSAR campaign was conducted by ONERA, CESBIO, CIRAD and EDB under financing from ESA and CNES. The SAR data was acquired by ONERA and the research work described in this paper was covered by ONERA internal ressources. We want to thank Hubert Cantalloube for SAR processor, the RIM team for acquiring the data Cécile Titin-Schneider for fruitful discussions. REFERENCES 1. Bonin, G. and P. Dreuillet, The airborne SAR-system: SETHI Airborne microwave remote sensing imaging system, Proceedings of EUSAR, Cantalloube, H. and P. Dubois-Fernandez, Airborne X-band SAR imaging with 1 cm resolution: technical challenge and preliminary results, IEE Proc. Radar Sonar Navig., Vol. 153, No. 2, April Quegan, S., A unified algorithm for phase and cross-talk calibration of polarimetric data Theory and observations, IEEE Trans. Geosci. Remote Sensing, Vol. 32, 89 99, 1994.