STYRELSEN FÖR VINTERSJÖFARTSFORSKNING WINTER NAVIGATION RESEARCH BOARD Research Report No 89 Juha Karvonen, Markku Similä and Alexandru Gegiuc ESAR - ENHANCED SAR IMAGERY FOR THE BALTIC SEA WINTER NAVIGATION WNRB Research Project, Short Summary Finnish Transport Safety Agency Finnish Transport Agency Finland Swedish Maritime Administration Swedish Transport Agency Sweden
Talvimerenkulun tutkimusraportit Winter Navigation Research Reports ISSN 2342-4303 ISBN 978-952-311-038-0
FOREWORD In its report no 89, the Winter Navigation Research Board presents the outcome of the project: ESAR - enhanced SAR imagery for the Baltic Sea winter navigation. Utilization and dissemination of the dual-polarized Radarsat-2 SAR data was studied in the project. The dual-polarized data includes two channels instead of the single channel present in the earlier SAR imagery. Three different false color image alternatives were generated. In the false color images the two channels and values computed based on the two channels are placed into the RGB-channels of a three-channel color image. These color images do not present an interpretation of the SAR image but only try to merge the information of the two channels into one color image. Also graytone representations better describing the two channels and two alternatives of PC images have been studied. The incidence angle dependency at Radarsat-2 HV channel was studied and it was noticed that there is a regular noise pattern present in the HV-images. The processing and delivery chain for the dual-polarized Radarsat-2 data was established and tested. The Finnish and Swedish ice breaker officers and maritime administrations have been contacted and feedback has been requested. The Winter Navigation Research Board warmly thanks Mr. Juha Karvonen, Mr. Markku Similä and Mr. Alexandru Gegiuc for this report. Helsinki and Norrköping June 2014 Jorma Kämäräinen Finnish Transport Safety Agency Peter Fyrby Swedish Maritime Administration Tiina Tuurnala Finnish Transport Agency Stefan Eriksson Swedish Transport Agency
ESAR (Enhanced SAR imagery for the Baltic Sea winter navigation), WNRB Research Project, Short Summary Juha Karvonen, Markku Similä, Alexandru Gegiuc Finnish Meteorological Institute (FMI) PB 503, FI 00101, Helsinki, Finland Ph. +358 29 539 1000 email : Juha.Karvonen@fmi.fi, Markku.Simila@fmi.fi, Alexandru.Gegiuc@fmi.fi June 30, 2014
1 Summary In the project utilization and dissemination of the dual-polarized Radarsat-2 SAR data was studied. The dualpolarized data includes two channels instead of the single channel present in the earlier SAR imagery. The channels are formed by changing the polarization of the transmitted and received signal. The single channel images have used the polarization combination HH (horizontally polarized transmission and horizontally polarized reception), and the dual-polarized images also include the HV channel (horizontally polarized transmission and vertically polarized reception). In the project three different false color image alternatives were generated. In the false color images the two channels and values computed based on the two channels are placed into the RGB-channels of a three-channel color image. These color images do not present an interpretation of the SAR image but only try to merge the information of the two channels into one color image. Also gray-tone representations better describing the two channels have been studied. The most promising were the principal component (PC) approach where the gray-tone image is a weighted linear combination of the two channels, and the weights are defined by the principal component analysis. Two alternatives of PC images were studied: The principal components defined for the whole image and Principal components defined locally (i.e. different linear combination in different parts of the image depending on the local statistics). The studies related to PC images were performed after the Baltic Sea ice season, and therefore samples of these images were not delivered to ships. The capability of delivery exist through the usual delivery channel (because they are gray tone images they can be delivered as the HH-channel images). Also the incidence angle dependency at Radarsat-2 HV channel was studied and it was noticed that there is a regular noise pattern present in the HV-images. This noise pattern is clearly visible in the SAR imagery and makes visual interpretation difficult. It can be significantly reduced by filtering, because the pattern remains constant as a function of the incidence angle (or range from the instrument). The filtering was implemented and included in the FMI Radarsat-2 processing chain. Also a presentation on this topic was given at the URSI (International Union of Radio Science) Commission F Microwave Signatures 2013 Specialist Symposium on Microwave Remote Sensing of the Earth, Oceans, and Atmosphere, organized in Espoo in October 2013. The processing and delivery chain for the dual-polarized Radarsat-2 data was established and tested. Dualpolarized false color images were delivered to ice breakers and some feedback received. In the winter 2013-2014 probably the PC images will be delivered instead. The processing and delivery chain for the PC images now exists at FMI. The end-users i.e. Finnish and Swedish ice breaker officers and maritime administrations have been contacted and feedback has been requested. Only little feedback has been received, but based on this feedback it seems that full utilization of the color imagery is not very easy and would require some training. The interpretation of the PC imagery would be easier because the users are familiar to use gray-tone SAR imagery. In short the essential outcomes of the project are: Incidence angle correction of the HV-channel, improving the visual appearance of the HV channel significantly. False color image studies and tests. Principal component image studies. Establishment and testing of the color and PC image processing and delivery chain. 2
2 HV-Channel Incidence Angle Correction Figure 1: Incidence angle dependence for HH and HV channels. The incidence angle dependent pattern for HV is clearly visible, for HH the incidence angle dependence is linear. Figure 2: A SAR HH-channel image before and after incidence angle correction. 3
3 Principal Component Imagery Figure 3: A detail of a HH-channel, HV-channel, PC image, and local PC image. 4
4 False Color Imagery Figure 4: HH and HV channels, PC image and three different false color image of the same SAR scene, March 15, 2013. 5
5 Data Processing and Dissemination Figure 5: The SAR processing and dissemination chain. 6