Characterizing and Correcting Residual RFI Signatures in Operationally Processed ALOS-PALSAR Imagery

By Franz J. Meyer, Geophysical Institute, University of Alaska Fairbanks (UAF), U.S.A.; Jeremy Nicoll, Alaska Satellite Facility (ASF), UAF, U.S.A.; Anthony P. Doulgeris, University of Tromsø, Norway

Radio-frequency interference (RFI) has long been identified as a problem in L-band Synthetic Aperture Radar (SAR), as it is limiting the application, performance, and data quality of SAR at many areas around the globe. Several algorithms for RFI mitigation have been developed throughout the recent decades (Le, et al., 1998; Reigber and Ferro-Famil, 2005; Rosen, et al., 2008) in response to this problem.

Especially in the American Arctic, RFI distortions in L-band SAR data are a widespread problem (Doulgeris and Meyer, 2011), (Meyer, et al., 2011). Areas affected by severe RFI are shown in Figure 1. Here RFI levels in Advanced Land Observing Satellite-Phased Array L-band SAR (ALOS-PALSAR) data were quantified using a data screening method (Meyer, et al., 2011) and color coded according to their magnitude. This screening method was greatly facilitated by the ASF datapool and the new datapool Application Programming Interface (API), allowing easy and rapid download of thousands of images. Strong RFI presence can be identified particularly along the Arctic coast. In these areas, highquality RFI filtering is necessary to provide consistently well calibrated L-band radar data to the user community.

It was also shown (Meyer, et al., 2011) that in the American Arctic, many operationally processed ALOS-PALSAR images show distortions that can be attributed to residual RFI that was not completely removed by the Japan Aerospace Exploration Agency’s operational filtering algorithms. Figure 2 shows two examples of RFI-induced artifacts for images acquired over Barrow, Alaska, in April 2009. Pauli decomposition images derived from operationally processed

Level-1.1 ALOS-PALSAR scenes are shown. Significant polarimetric signature variations can be identified over regions that appear to have similar surface characteristics. The strong color changes in the polarimetric decomposition would indicate strong changes in scattering properties, which appear unrealistic. Additional small-scale artifacts are visible. Both signal patterns are typical for RFI-affected data (Reigber and Ferro-Famil, 2005).

From an analysis of ALOS-PALSAR Level-1.0 data, it was found that the RFI signal in this area is characterized by high-power, wide bandwidth, temporarily narrow signatures, where center frequency is randomly changing with time. In such complex RFI environments, simple notch-filtering algorithms, such as the one implemented in the ALOS-PALSAR processor, fail, resulting in residual artifacts in the SAR image, polarimetric signature, and phase (Rosen, et al., 2008).

In order to obtain consistent data quality, a new RFI filter method was developed that is optimized to remove residual RFI signatures from ALOS-PALSAR imagery that were not corrected by the notch-filtering algorithm implemented in the operational processor. The new approach analyzes SAR data in azimuth time rather than range frequency and uses statistical methods to detect and remove RFI signatures.

After range compression, sections of ~2,000 azimuth lines are transformed into the range-frequency azimuth-time domain. A specific range frequency slice is extracted from the two-dimensional data. The extracted frequency slice is log transformed to generate data of approximate Gaussian distribution and a statistical outlier test is used to identify interference signals using a Z-test approach. This new RFI filter was embedded in a customized SAR processor, where it is used in combination with a traditional notch filter to capture a wide range of RFI signatures.

To demonstrate the effectiveness of the new RFI filter, RFI-affected, full-polarimetric ALOS-PALSAR data acquired near Barrow, Alaska, were processed using both the operational and the customized processor. To quantify performance, processed data were analyzed for image quality, polarimetric signature, and Interferometric SAR (InSAR) coherence.

The improvement of SAR image quality is exemplified in Figure 3, where both the operationally (top panel) and custom focused (bottom panel) HV channel of an ALOS-PALSAR image near Barrow, Alaska, are shown. The image quality improvement achieved by the customized algorithm is clearly evident.

To visualize improvements of polarimetric signature, full-polarimetric ALOS-PALSAR scenes were processed to Pauli RGB images using both the operational (Figure 4a) and customized processor (Figure 4b). The RFI-induced color distortions evident in Figure 4a are corrected in Figure 4b to virtually flat Pauli RGB images. Classification results based on these data are shown in Figure 4c and d. They provide additional evidence that the custom processor (Figure 4d) produces more distinct and realistic results.

Probability density functions (PDF) of the coherence of a SAR interferogram acquired near Barrow, Alaska, are shown in Figure 5 for both operationally and custom-processed data. A comparison of the coherence distribution indicates clearly higher coherence in the custom-processed data. In conclusion, RFI is a severe and growing issue in L-band radar remote sensing that affects data quality in many areas around the globe. It was shown that ALOS-PALSAR data over the American Arctic coast is consistently affected by complex RFI signatures, whose effects cannot be sufficiently removed using traditional notch-filter algorithms. A novel processing scheme was presented that is capable of effectively removing RFI artifacts. Examples have shown that the developed technique leads to improved image quality, polarimetric integrity, and InSAR coherence.

References:
Doulgeris, A.P. and Meyer, F.J., 2011. Severe Radio Frequency
Interference in ALOS-PALSAR Imagery, 5th International Workshop on Science and Applications of SAR Polarimetry and Polarimetric Interferometry (PolInSAR 2011), Frascati, Italy.
Le, C., Hensley, S., and Chapin, E., 1998. Adaptive filtering of RFI in wideband SAR signals, Seventh Annual JPL Airborne Earth
Science Workshop, Jet Propulsion Laboratory, Passadena,
California.
Meyer, F.J., Nicoll, J., and Doulgeris, A.P., 2011. Characterization and Extent of Randomly-Changing Radio-Frequency Interference in ALOS-PALSAR Data, Geoscience and Remote-Sensing Symposium, 2011. IGARSS 2011. IEEE International, Vancouver, Canada.
Reigber, A., and Ferro-Famil, L., 2005. Interference Suppression
in Synthesized SAR Images. IEEE Geoscience and Remote
Sensing Letters, 2(1): 45-49.
Rosen, P.A., Hensley, S., and Le, C., 2008. Observations and mitigation of RFI in ALOS-PALSAR SAR data: Implications for the DESDynI mission, Radar Conference, 2008. RADAR ,08. IEEE, pp. 1-6.