Activity Report. April 2001 April 2003 RADAR REMOTE SENSING GROUP UBC

Transcription

1 Activity Report April 2001 April 2003 RADAR REMOTE SENSING GROUP UBC This report summarizes the research, teaching and administrative activities of the Radar Remote Sensing Group of the Department of Electrical and Computer Engineering at the University of British Columbia ( during the period from April 2001 to April The focus of the research work is on the processing of synthetic aperture radar (SAR) data, concentrating on data from remote sensing satellites such as ERS and RADARSAT. Various aspects of SAR processing, including Doppler centroid estimation, satellite roll angle estimation, ScanSAR data collection and calibration methods, polarimetric sea ice classification, SAR image data compression, and comparison of precision SAR processing algorithms have been studied during the reporting period. The research work has been documented in reports, conferences, journal papers, and in student theses. The work is briefly described below, first categorized by students supervised, then categorized under type of publication. 1 Graduate Students Supervised Bernd Scheuchl: Bernd passed his Ph.D. qualifying exam in 2002, and is progressing well in his research on the classification of sea ice in polarimetric radar data. The project is being conducted in cooperation with MacDonald Dettwiler, where he spends much of his time. His main work is on the development of polarimetric radar classification algorithms, which he is applying to data of the various types of sea ice found in the Canadian Arctic and North Atlantic. See [1 11] and Figure 1. 1

2 Dan Bast: Dan completed his M.A.Sc. thesis defense in May 2002, and spent some time helping to write a journal paper in the summer. He has recently joined the European Space Agency s apprenticeship program at their Space Technology Research Center, ESTEC, in Holland. In his UBC research work, he improved the estimation of RADARSAT s roll angle, by using extra data collected during the ScanSAR beam switching cycle. This involved proposing a new method of ScanSAR data collection, modifying a roll angle estimation algorithm to use the new data, and developing a simulation to show how much the roll angle estimate is improved. It is anticipated that the work will lead to more accurate joining of RADARSAT-2 images, eliminating the edge artifact that is often seen in current ScanSAR images. See [12 15] and Figure 2. Jing Wang: Jing finished her M.A.Sc. program in December She worked on the compression of SAR image data, using the approach of wavelet packets in an adaptive block encoding scheme. A new algorithm called Wavelet Packet Embedded Block (WPEB) coding was developed. Previous work had involved the discrete wavelet transform [16], and it was found that the increased attention paid to the higher frequencies in the SAR image by wavelet packets improved the compression efficiency. In addition, a block coding method was used that improved the coding by applying more bits to active areas of each scene. See [17 19] and Figure 3. Yewlam Neo: Yewlam joined us from Singapore, where he was a signal processing engineer with the Radar Systems Department in the DSO National Labs. He obtained a scholarship from his employer, and began at UBC in January He has just finished his course requirements, and is beginning research work on the comparison of precision SAR processing algorithms [20]. Flavio Wasniewski: Flavio came to us in September 2001 from Brazil, where he worked as a cartographic engineer with RADARSAT data. He is presently completing his UBC course requirements, and is working part time at RADARSAT International on the applications of polarimetric radar data. His research work will be on the extraction of man-made targets from polarimetric radar data. Millie Sikdar: Millie came to our group in September 2002 from the undergraduate program of the Govt. College of Engineering, Aurangabad, India. She is just finishing her course requirements, and will begin her research this summer on the analysis of polarimetric radar data. Kaan Ersahin: Kaan arrived at UBC in January 2003, after earning his M.Sc. at the Istanbul Technical University (ITU) in Turkey. He worked at the satellite 2

3 receiving station in Istanbul, including the processing of RADARSAT data, and is keen to pursue his studies in radar remote sensing. Vincent Yip and Vincent Siu: These two students completed their Master of Engineering program at UBC in April They worked on the MPEG-1, Level-3 audio data compression algorithm, known as MP3. The algorithm was simulated in MATLAB, and audio examples were generated that illustrated the effect of various bit rates from 64 to 256 Kbits/s [21]. 2 Presentations at Conferences During the reporting period, the following conferences were attended by Ian Cumming and/or his students. IGARSS 01 IEEE International Geoscience and Remote Sensing Symposium, Sydney, Australia, July 9 13, 2001 [1, 22]. IGC was a session chairman. CRSS 01 23rd Canadian Symposium on Remote Sensing, Quebec City, August 21 24, 2001 [2]. ASAR 01 4th Workshop on Advanced SAR Technology, Canadian Space Agency, St. Hubert, Quebec, October 1 3, 2001 [3]. IGARSS 02 IEEE International Geoscience and Remote Sensing Symposium, Toronto, June 24 28, 2002 [4 6, 12, 17]. IGC was on the technical organizing committee and was a session chairman. PolInSAR 03 Workshop on Applications of SAR Polarimetry and Polarimetric Interferometry, ESA/ESRIN, Frascati, Italy, January 14 16, 2003 [7, 8]. DCC 03 Data Compression Conference, IEEE Computer Society, Snowbird, Utah, March 25-27, 2003 [19]. Future Conferences Papers has been submitted to the following future conferences: IGARSS 03 IEEE International Geoscience and Remote Sensing Symposium, Toulouse, July 21 25, 2003 [9, 20]. IGC is a session chairman. 3

4 ASAR 03 Advanced SAR Workshop, Canadian Space Agency, Montreal, June 25 27, 2003 [10, 15]. IGC is on the technical organizing committee and is a session chairman. sectors/technology/development/csa labs/ expertise/asar2003.asp 3 Research Book on SAR Processing Frank Wong and Ian Cumming have been writing a research book on SAR signal processing over the last three years. The book brings together all the major developments in satellite SAR processing in the last 20 years, based on their research and experience at MacDonald Dettwiler. Chapter titles are: 1. Introduction 2. Signal Processing Fundamentals 3. Pulse Compression 4. Synthetic Aperture Concepts 5. SAR Signal Properties 6. The Range Doppler Algorithm 7. The Chirp Scaling Algorithm 8. The Omega-K Algorithm 9. The SPECAN Algorithm 10. ScanSAR Processing Algorithms 11. Doppler Parameter Estimation 12. Comparison of Algorithms 13. Summary The book is now over 400 pages long, and it is expected that about 100 more pages will be written before it is completed. Arrangements are underway for Artech House to publish the book in late Papers Submitted and/or Published Because of the time and effort devoted to writing the book, there were fewer journal papers written than usual during the reporting period. 4

5 Jonathan Zeng s work on SAR speckle reduction and data compression using wavelets was completed in the last reporting period, and appeared in the IEEE Transactions of Geoscience and Remote Sensing in March 2001 [16]. It shows that SAR speckle reduction and data compression can be efficiently combined, and that wavelets can provide better data compression than traditional methods. The work with Dan Bast produced a number of new ideas, including a new method of ScanSAR data collection on SAR satellites, and the use of the new data for improving the estimation of the satellite s roll angle. When the roll angle is known more accurately, the radiometry of the received data can be corrected with greater precision, with the result that radiometric artifacts in the ScanSAR images can be reduced. The new data comes from pulses transmitted on one beam, and received on another beam, after the beams have been switched in range during the normal ScanSAR cycles. The new data is referred to as hybrid beam data, and Figure 2(a) illustrates how the new data is received within the normal inter-pulse periods. Figure 2(b) shows that the roll angle accuracy improves by about 25% with the additional hybrid beam data, when three burst periods are used in the estimator. The concepts and results have been documented in a technical paper, which has been reviewed by the IEEE. The paper was re-submitted in April 2003 with editing changes requested by the editor [14]. When Prof. Malek Hussain was an academic visitor in the Radar Remote Sensing Group at UBC, he worked on the analysis of radar systems using very large bandwidths, referred to as impulse radars. One significant advantage they have is the lack of side lobes in the radiation patterns. Impulse radars can be conveniently analyzed using Space-Time Processing, as Prof. Hussain explains in a paper produced from his work at UBC [23]. Ian Cumming has refined his work on Doppler centroid estimation, by adding an Earth/satellite geometry model that obtains the Doppler centroid from satellite roll and pitch measurements. The earlier work used spatial diversity and quality criteria to fit a Doppler surface over a whole frame of received data, without using roll and pitch measurements [22]. The addition of the geometry model reduces the dimensionality of the estimation problem, and places constraints on the solution to ensure a physically plausible estimate. A draft of a technical paper has been written, and it is planned to submit it for publication in the near future [24]. 5 Educational Material An interesting activity in the last year has been writing material for the educational web site of the Canada Centre for Remote Sensing. 5

6 GlobeSAR-2 Nine chapters have been written for the GlobeSAR-2 Educational Resources for Radar Remote Sensing. The material is now available at: e.html The chapters that Ian wrote have the following titles and page numbers: 1. Introduction to RADAR Remote Sensing, pages RADARSAT-1, pages RADARSAT-2, pages RADAR Systems, pages SAR Image Formation, pages SAR Image Characteristics, pages SAR Systems and Digital Signal Processing, pages Radar Polarimetry, pages Radar Interferometry, pages Glossary CCRS has a large glossary of remote sensing terms on their web site. Ian has been updating the glossary entries pertaining to polarimetric radar. About 62 current terms have been updated and 52 new terms added. These terms are currently being incorporated into the CCRS Glossary on Remote Sensing at: e.html Tutorial In addition, there is a tutorial on Remote Sensing on the CCRS web site, which includes a module on SAR remote sensing. Ian has written three new chapters for the tutorial: 1. Understanding the Basics of SAR Sensors, 23 pages 2. Fundamentals of Polarimetry, 8 pages 3. Advanced Polarimetric Tutorial, 44 pages These new chapters are in the process of being added to the web site Fundamentals of Remote Sensing : e.html 6

7 6 Teaching at UBC In September 2001 and again in September 2002, the course EECE 591 Applied Digital Signal Processing was given to about 20 postgraduate students. After covering the main theory and tools of DSP, a number of applications are explained. The main application is processing of SAR data, but other applications include image interpolation and compression, GPS navigation, and wavelets. Over 400 pages of electronic notes have been prepared and made available to the students via the web. In January 2002, the course EECE 466 Digital Signal Processing was taught to about 30 4th year undergraduates. A new book was used by Ifeachor and Jervis, which combines theory and practice in a very readable format. In January 2003, the lecture/lab course, EECE 467, was taught to 33 4th year students. Four experiments were designed in which DSP algorithms were programmed on a TI DSP board to perform functions such as an audio graphic equalizer, an IIR bandpass filter in fixed point arithmetic, an audio filter implemented with FFTs, and a detector for dual-tone multi-frequency telephone dialing. The lectures covered the DSP theory needed to understand the experiments, and the technology of DSP hardware. Everyone passed and we all had fun! One novel aspect of the 467 course was using WebCT for web-based learning. Lecture notes, lab instructions, and quizzes were posted on the web. We actually ran the exam on WebCT, which worked well after we got over the embarrassment of prematurely ing the exam to all the students by mistake! In September 2002, a 6-person ECE 285 group was supervised on the topic of Radar Remote Sensing, concentrating on the new technologies being developed for RADARSAT- 2. In September 2001 and January 2002, groups of 3rd and 4th year students were supervised in an ECE 496 project on GPS receivers. After learning how signal processing is used in the GPS system, they used a hand-held receiver to assess the position and velocity accuracy in dynamic experiments. PC-based mapping software was used to plot the GPS measurements on topographic maps. In September 2002, January 2003 and again in May 2003, ECE 496 groups were supervised on a new project building an ultrasound model of the GPS system. The students have to understand how the GPS system works, and then build a model in the lab to illustrate the principles and to show how signal processing is used. Ultrasound signals were used so that the transmitted signal would not be noticeable (i.e. not offensive to the ears), and to obtain resolutions in the order of a centimeter using a bandwidth as low as 4 KHz. The first group experimented with various ultrasound transducers, and 7

8 built a system that obtained 2 cm accuracy over one meter. The second group built an improved modulator using FSK, and obtained 1.5 cm accuracy over 15 meters. The third group is now building a system to measure position in two dimensions. The students learn how spread spectrum and pulse compression techniques can obtain accurate position and velocity measurements in the face of low signal/noise ratios. The following ECE 571 reading courses were designed and conducted for individual students: 1. Polarimetric SAR Interferometry (PolInSAR), Bernd Scheuchl, April SAR Systems and Signal Processing, Yewlam Neo, July Fundamentals of Polarimetry, Flavio Wasniewski, April Principles and Applications of Imaging Radar, Millie Sikdar, May SAR Signal Processing, David Alton, Univ. of Calgary Approximately two coop students have been supervised per term. 7 Theses Supervised Being on sabbatical in 2000, new students were not accepted until This meant that the number of theses in were below average. The two M.A.Sc. students who completed their theses during the reporting period were Dan Bast [13] and Jing Wang [18]. In addition, two M.Eng. students were supervised, Vincent Yip and Vincent Siu, who wrote a report almost as long as a masters thesis [21] (80 pages). 8 Educational Service Ian Cumming was invited to be the external Ph.D. examiner of Juergen Holzner at the University of Edinburgh in November As the topic of ScanSAR Radar Interferometry was very pertinent to the group s research, Ian travelled to Edinburgh, and conducted a 6-hour exam. A very good thesis and a worthwhile trip. Other educational activities included: serving on the committee of five Ph.D. examinations held at the Faculty of Graduate Studies at UBC, serving on the committee of 11 ECE Master s and Ph.D. examinations in the ECE department, 8

9 presenting two seminars on Radar Polarimetry at Pre-conference Workshop at IGARSS 02 (Ian Cumming and Bernd Scheuchl), running a reading course for a student at the University of Calgary (see Section 6), reviewing 9 papers for several IEEE Transactions and the Canadian Journal of Remote Sensing, reviewing 3 NSERC applications. 9 UBC Administration ECE Building committee, 2001 ECE Faculty Recruiting committee, Industrial Liaison Ian and his students visit MacDonald Dettwiler often, and usually make presentations at the MDA Quarterly R&D reviews. Their current work with MacDonald Dettwiler concentrates on developing applications for RADARSAT-2. One of the main applications of RADARSAT-1 is ice mapping, and the main innovation on RADARSAT-2 is polarimetry, so it is important to develop algorithms for ice classification with polarimetric data. This is the focus of Bernd Scheuchl s research, and present results indicate that polarimetric SAR data will provide substantially better automatic ice classification than data from existing single-channel radars (see Figure 1) [9]. Bernd was given the technical lead on the EOADP project Multi-Polarimetric SAR Product for Operational Sea Ice Monitoring at MDA. His work in 2002 included: Analysis of AIRSAR three-frequency fully polarimetric data of sea ice Analysis of Convair-580 data of sea ice (incidence angle analysis, dual-pol and quad-pol classification) Preparation of a sample data set including polarimetric data, polarimetric parameters, data description, and auxiliary data Four reports have been written to date on the work at MDA [25 28]. Bernd s work is continuing with an investigation of ENVISAT alternating polarization data with respect to the separation of sea ice and open water, and its data visualization. 9

10 We also have connections with RADARSAT International Inc. (RSI) of Richmond, BC. In particular, Flavio Wasniewski has spent about a quarter of his time at RSI, working on the interpretation of Convair-580 polarimetric SAR data from the Gagetown base, in order to explore the potential of RADARSAT-2 for various applications. The analysis results have been used for customer training and marketing. The work involved: Exploration of the polarimetric signatures of targets; Addition of noise to Convair-580 multi-looked data to emulate RADARSAT-2 data. Extraction of visual examples that illustrate the advantage of multi-polarized data; Comparisons between RADARSAT-1 and Convair-580 data to predict RADARSAT- 2 feature extraction capabilities; Identification of features in the Gagetown images to verify via fieldwork. fieldwork has been done and there is now more data to work on. The Acknowledgments We would like to express our sincere appreciation to MacDonald Dettwiler, NSERC, the Canadian Space Agency and the BC Advanced Systems Institute for providing financial support for the research activities of the UBC Radar Remote Sensing Group throughout the period of this report. We continue to enjoy access to radar data from the RADARSAT, ERS and X-SAR/SRTM programs. Ian Cumming May 8, 2003 References [1] B. Scheuchl, R. Caves, I. Cumming, and G. Staples. Automated Sea Ice Classification Using Spaceborne Polarimetric SAR Data. In Proceedings of the IEEE International Geoscience and Remote Sensing Symposium, IGARSS 01, pages , Sydney, Australia, July 9 13, [2] B. Scheuchl, R. G. Caves, I. G. Cumming, and G. Staples. H/A/alpha-based Classification of Sea Ice Using SAR Polarimetry. In Proceedings of the 23rd Canadian Symposium on Remote Sensing, Quebec City, August 21 24,

11 [3] I. Hajnsek, B. Scheuchl, R. G. Caves, and I. G. Cumming. Surface Parameters Inversion from Sea Ice Using Spaceborne Polarimetric SAR Data. In 4th Workshop on Advanced SAR Technology, Canadian Space Agency, St. Hubert, Quebec, October 1 3, [4] B. Scheuchl, I. Hajnsek, and I. G. Cumming. Model-Based Classification of Polarimetric SAR Sea Ice Data. In Proceedings of the IEEE International Geoscience and Remote Sensing Symposium, IGARSS 02, pages , Toronto, June 24 28, [5] B. Scheuchl, I. Hajnsek, and I. G. Cumming. Sea Ice Classification Using Multi- Frequency Polarimetric SAR Data. In Proceedings of the IEEE International Geoscience and Remote Sensing Symposium, IGARSS 02, pages , Toronto, June 24 28, [6] B. Scheuchl and I. G. Cumming. Potential of RADARSAT-2 for Sea Ice Classification. In Proceedings of the IEEE International Geoscience and Remote Sensing Symposium, IGARSS 02, pages , Toronto, June 24 28, [7] Bernd Scheuchl, Irena Hajnsek, and Ian Cumming. Classification strategies for polarimetric SAR sea ice data. In Workshop on Applications of SAR Polarimetry and Polarimetric Interferometry, ESA/ESRIN, Frascati, Italy, January 14 16, [8] Bernd Scheuchl, Dean Flett, Gordon Staples, Gordon Davidson, and Ian Cumming. Preliminary classification results of simulated RADARSAT-2 polarimetric sea ice data. In Workshop on Applications of SAR Polarimetry and Polarimetric Interferometry, ESA/ESRIN, Frascati, Italy, January 14 16, [9] B. Scheuchl, I. G. Cumming, and I. Hajnsek. Consolidation of a Pixel-Based Classification Using Neighbourhood Information. In Proceedings of the IEEE International Geoscience and Remote Sensing Symposium, IGARSS 03, Toulouse, July 21 25, [10] B. Scheuchl, I. Hajnsek, and I. G. Cumming. Potential of SAR Polarimetry for Sea Ice Discrimination. In Proceedings of the Advanced SAR Workshop, ASAR 03, Canadian Space Agency, Montreal, June 25 27, [11] B. Scheuchl, D. Flett, R. Caves, and I. Cumming. Potential of RADARSAT- 2 multiple polarization and polarimetric data for operational sea ice monitoring. Canadian J. of Remote Sensing, submitted February [12] D. Bast and I. Cumming. RADARSAT ScanSAR Roll Angle Estimation. In Proceedings of the IEEE International Geoscience and Remote Sensing Symposium, IGARSS 02, pages , Toronto, June 24 28,

12 [13] Daniel C. Bast. ScanSAR radiometric calibration based on roll angle estimation. Master s thesis, Dept. of Electrical and Computer Eng., The University of Britsh Columbia, 2356 Main Mall, Vancouver, BC, Canada V6T 1Z4, May [14] I. G. Cumming and D. C. Bast. A New Hybrid-Beam Data Collection Strategy to Support ScanSAR Radiometric Calibration. IEEE Trans. on Geoscience and Remote Sensing, To be published, final editorial revisions have been completed. [15] I. G. Cumming and D. C. Bast. A Method of Improving Roll Angle Estimation in Future RADARSATs. In Proceedings of the Advanced SAR Workshop, ASAR 03, Canadian Space Agency, Montreal, June 25 27, [16] Z. Zeng and I. G. Cumming. Modified SPIHT Encoding for SAR Image Data. IEEE Trans. on Geoscience and Remote Sensing, 39(3): pp , March [17] I. Cumming and J. Wang. Polarimetric SAR Data Compression Using Wavelet Packets in a Block Coding Scheme. In Proceedings of the IEEE International Geoscience and Remote Sensing Symposium, IGARSS 02, Toronto, June 24 28, [18] Jing Wang. Investigations on single and multipolarization SAR image compression. Master s thesis, Dept. of Electrical and Computer Eng., The University of Britsh Columbia, 2356 Main Mall, Vancouver, BC, Canada V6T 1Z4, December [19] I. Cumming and J. Wang. Compression of RADARSAT Data with Block Adaptive Wavelets. In Proc. Data Compression Conference, IEEE Computer Society, Snowbird, Utah, March 25-27, [20] I. G. Cumming, Y. L. Neo, and F. H. Wong. Interpretations of the Omega-K Algorithm and Comparisons with other Algorithms. In Proceedings of the IEEE International Geoscience and Remote Sensing Symposium, IGARSS 03, Toulouse, July 21 25, [21] Vincent Yip and Vincent Siu. MATLAB Implementation of the MP3 Encoder. Master of Engineering Report, Department of Electrical and Computer Engineering, The University of British Columbia, 80 pages, April [22] Ian Cumming. Model-Based Doppler Estimation for Frame-Based SAR Processing. In Proceedings of the IEEE International Geoscience and Remote Sensing Symposium, IGARSS 01, pages , Sydney, July 9 13, [23] Malek G. M. Hussain. Principles of Space-Time Array Processing for Ultrawide- Band Impulse Radar and Radio Communications. IEEE Transactions on Vehicular Technology, 51(3): pp , May

13 [24] I. G. Cumming. A Spatially-Selective Approach to Doppler Estimation for Frame- Based SAR Processing. IEEE Trans. on Geoscience and Remote Sensing, In preparation. [25] Ron Caves, Gordon Staples, and Bernd Scheuchl. Multi-Polarimetric SAR Products for Operational Sea Ice Monitoring. Technical Report RX-RP , Mac- Donald Dettwiler, Richmond, BC, November [26] Bernd Scheuchl. Analysis of CV-580 Fully Polarimetric Data of Sea Ice Data Calibration Test. Technical Report RX-TN , MacDonald Dettwiler, Richmond, BC, September [27] Bernd Scheuchl, Gordon Staples, and Gordon Davidson. Analysis of CV-580 Fully Polarimetric Data of Sea Ice Final Report. Technical Report RX-RP , MacDonald Dettwiler, Richmond, BC, March [28] Bernd Scheuchl. Analysis of CV-580 Fully Polarimetric Data of Sea Ice Executive Summary. Technical Report RX-RP , MacDonald Dettwiler, Richmond, BC, March Many of the above references are available on the group s web site: 13

14 (a) Convair-580 SAR scene of Northumberland Strait, PEI, acquired March 8, Red VV ; Green HV ; Blue HH. (b) Classification result: Grey Young ice; White rough FYI (strong HV); Green ridged FYI; Blue Leads; Orange FYI (near range); Magenta FYI (far range); Figure 1: Classification of C-band CV-580 polarimetric sea ice data using a Bayesian algorithm with a Wishart distribution. Image size: 6.4 km (range) by 8 km (azimuth); Pixel size: 16 m; Incidence angle: degrees; Number of looks: 4x40. 14

15 beam switchover point PRF 1 = 1230 m 1 = 8 PRF 2 = 1050 m 2 = 7 Case 1: PRF DOWN 2 Delta PRF = 180 Hz PRF 1 = 1230 m 1 = 8 PRF 2 = 1170 m 2 = 8 Case 2: PRF DOWN 1 Delta PRF = 60 Hz PRF 1 = 1230 m 1 = 8 PRF 2 = 1310 m 2 = 9 Case 3: PRF UP 1 Delta PRF = 80 Hz PRF 1 = 1230 m 1 = 8 PRF 2 = 1440 m 2 = 10 Case 4: PRF UP 2 Delta PRF = 210 Hz (a) Timing of hybrid beam data collection for 4 values of PRF change. The PRF and the beam elevation angle change at the beam switchover point. Roll error magnitude (deg) Noise equivalent sigma nought = 23 db Correct beam patterns used Without hybrid beam data Recommended accuracy With hybrid beam data Scene brightness (sigma nought) (b) Improvement in roll angle accuracy with the hybrid data Figure 2: Illustrating the improvement of satellite roll angle estimation when hybrid beam ScanSAR data are used [14]. 15

16 (a) Original RADARSAT scene of Vancouver Airport with 16 bits/pixel (b) The scene encoded with the WPEB algorithm using 1 bit/pixel Figure 3: Illustrating image interpretability when the SAR image is encoded with wavelet packets by the WPEB algorithm using 1 bit/pixel [19]. 16