Polar Activities at DLR Maritime Security Lab Bremen in the Projects EISTAK and EMS Susanne Lehner, Anja Frost, Rudolf Ressel German Aerospace Center
Chart 2 Maritime Security Lab in Bremen German Aerospace Center Main activities develop and operate near real time services using SAR data Wind and sea state monitoring Oil spill detection Ship and wake detection TerraSAR-X Sentinel1 Iceberg detection Ice classification Support campaigns
Chart 3 Support campaigns Polarstern campaign in Antarctica Polarstern campaign in Antarctica Assist the Australian Maritime Safety Authority during the accident of the research vessel Akademik Shokalskiy in Antarctica Lance campaign off the Svalbard coast, Arctic Ocean Oden cruise at the marginal ice zone in Beaufort Sea, Arctic Ocean 15.08.2013-16.10.2013 20.12.2013-05.03.2014 30.10.2013-09.01.2014 14.03.2014-29.03.2014 17.07.2014 17.08.2014
DLR.de Chart 4 Maritime Security Labs Bremen and Neustrelitz - EOC s NRT Services provided by Neustrelitz - EOC is operating an international Ground Station Network for the reception of satellite data
Chart 5 Outline TerraSAR-X imaging modes and NRT iceberg detection Ice type classification Analysis of a TerraSAR-X image taken from the trapped Akademik Shokalskiy Algorithm for ice type classification Test results Conclusions source: guardianlv.com
Chart 6 First Tests on Near Real Time Ice Type Classification in Antarctica Anja Frost 18th July 2014 TerraSAR-X image acquisition in different modes Spotlight swath width: 10 km resolution: 1 m Stripmap swath width: 30 km resolution: 3 m ScanSAR swath width: 100 km resolution: 17 m Wide ScanSAR swath width: 200 km resolution: 34 m
Chart 7 First Tests on Near Real Time Ice Type Classification in Antarctica Anja Frost 18th July 2014 TerraSAR-X image acquisition in different modes Spotlight swath width: 10 km resolution: 1 m Stripmap swath width: 30 km resolution: 3 m ScanSAR swath width: 100 km resolution: 17 m Wide ScanSAR swath width: 200 km resolution: 34 m The minimum size of icebergs to be detected depends on the resolution. Stripmap mode ScanSAR WideScanSAR Bergy bits Small icebergs Medium icebergs
Iceberg detection within 20 minutes Section of a TerraSAR-X Stripmap image Disko Bay 23rd April 2013 1 km
Iceberg detection within 20 minutes Section of a TerraSAR-X ScanSAR image Grand Banks 26th June 2014 Thanks to Airbus 3 km
Chart 10 First Tests on Near Real Time Ice Type Classification in Antarctica Anja Frost 18th July 2014 Outline TerraSAR-X imaging modes and NRT iceberg detection Ice type classification Analysis of a TerraSAR-X image taken from the trapped Akademik Shokalskiy Algorithm for ice type classification Test results Conclusions source: guardianlv.com
Chart 11 First Tests on Near Real Time Ice Type Classification in Antarctica Anja Frost 18th July 2014 Analysis of a TerraSAR-X image recorded on 30 th December 2013 ScanSAR ascending incidence angle 27-36 HH polarisation 10 km
Chart 12 First Tests on Near Real Time Ice Type Classification in Antarctica Anja Frost 18th July 2014 Analysis of a TerraSAR-X image recorded on 30 th December 2013 ScanSAR ascending incidence angle 27-36 HH polarisation Xue Long Akademik Shokalskiy 10 km
Chart 13 First Tests on Near Real Time Ice Type Classification in Antarctica Anja Frost 18th July 2014 Analysis of a TerraSAR-X image recorded on 30 th December 2013 ScanSAR ascending incidence angle 27-36 HH polarisation Xue Long Akademik Shokalskiy 10 km fast ice pack ice ice edge open water
Chart 14 First Tests on Near Real Time Ice Type Classification in Antarctica Anja Frost 18th July 2014 Outline TerraSAR-X imaging modes and NRT iceberg detection Ice type classification Analysis of a TerraSAR-X image taken from the trapped Akademik Shokalskiy Algorithm for ice type classification Test results Conclusions source: guardianlv.com
Chart 15 First Tests on Near Real Time Ice Type Classification in Antarctica Anja Frost 18th July 2014 Ice type classification TerraSAR-X image (ScanSAR) Aim: Classify the ice in near real time Main steps of the algorithm: 1. Select texture features describing the most significant properties of different structures in ice e.g. GLCM features, Markov random fields, wavelet based features [1][2][3] [5] 2. Feed the features into a neural network [4] texture extraction feature a feature b feature c neural network 20 km ice chart [1] Bogdanov et al., 2005 [2] Clausi, 2001 [3] Zakhvatkina et al., 2013 [4] Nissen, 2003 [5] Lehner et al., IAHR 2014
Chart 16 First Tests on Near Real Time Ice Type Classification in Antarctica Anja Frost 18th July 2014 Ice type classification TerraSAR-X image (ScanSAR) Aim: Classify the ice in near real time Main steps of the algorithm: 1. Select texture features mean value, entropy and homogeneity texture features 20 km based on σ 0 -values of all the pixels in a limited window based on the frequency distribution of σ 0 -values of two neigboured pixels mean value entropy homogeneity
Chart 17 First Tests on Near Real Time Ice Type Classification in Antarctica Anja Frost 18th July 2014 Ice type classification TerraSAR-X image (ScanSAR) Aim: Classify the ice in near real time Main steps of the algorithm: 1. Select texture features mean value, entropy and homogeneity 2. Feed the features into a neural network 10 input neurons two hidden layers with 8 and 9 neurons feature a texture extraction feature b neural network 20 km feature c fast ice First year thick ice First year ice open water with <1/10 ice of unspecified SoD open water unknown
Chart 18 First Tests on Near Real Time Ice Type Classification in Antarctica Anja Frost 18th July 2014 30 th December 2013 Xue Long 9 th January 2014 5 km 5 km Akademik Shokalskiy
Chart 19 First Tests on Near Real Time Ice Type Classification in Antarctica Anja Frost 18th July 2014 Lance campaign TSX acquisitions during Lance campaign off the Svalbard coast 14.03.2014-29.03.2014
Chart 20 Lance campaign 20 km 20.03.2014 05:09 UTC Photo: 05:10 UTC RV Lance (course: W)
Chart 21 Lance campaign 20 km 20.03.2014 05:09 UTC 3:25 3:30 Photo: 03:33 UTC 4:00
Chart 22 Lance campaign 20 km 20.03.2014 05:09 UTC Photo: 04:34 UTC 5:10 4:36
Chart 23 Classification Result for 2014/03/20, 05:10 Ice classification Open Water (<1/10 ice of unspecified SoD) Young ice First year ice Land
Chart 24 Comparison with modelling result Ice classification Open Water (<1/10 ice of unspecified SoD) Young ice First year ice Land
Chart 25 Comparison with AMSR2 based ice concentration charts of 2014/03/20 Ice classification Open Water (<1/10 ice of unspecified SoD) Young ice First year ice Land Source: University of Bremen, IUP
DLR.de Chart 26 > Björn Tings Oceanography on Sentinel-1 > 13.11.2014 SENTINEL 1, ice types, icebergs Antarktis W110 S75 S1A_IW_GRDH_1SDH_20140504T092719_20140504T092744_000445_000535_4DCF 20km
DLR.de Chart 27 > Björn Tings Oceanography on Sentinel-1 > 13.11.2014 ice types, icebergs S1A_IW_GRDH_1SDH_20140504T092719_20140504T092744_000445_000535_4DCF
DLR.de Chart 28 > Björn Tings Oceanography on Sentinel-1 > 13.11.2014 ice types, icebergs Anja Frost, DLR Bremen S1A_IW_GRDH_1SDH_20140504T092719_20140504T092744_000445_000535_4DCF
Chart 29 First Tests on Near Real Time Ice Type Classification in Antarctica Anja Frost 18th July 2014 Outline TerraSAR-X imaging modes and NRT iceberg detection Ice type classification Analysis of a TerraSAR-X image taken from the trapped Akademik Shokalskiy Algorithm for ice type classification Test results Conclusions source: guardianlv.com
Chart 30 Summary and Conclusions TerraSAR-X based ice type products have been delivered to ships By the selection of significant texture features (particularly GLCM features) different structures in ice are revealed. A neural network performs the classification. Iceberg Detection Service Established Product Delivery in NRT about 15 min - provide a new NRT service for ship routing in ice infested areas using ice products - participate in validation campaigns - Use SAR multisensor products - Use Meteo Marine Parameters, e.g., sea state as well in ice generated ice chart observed ice situation (manual interpretation) 20 km 20 km
DLR.de Chart 31 > Björn Tings Oceanography on Sentinel-1 > 13.11.2014 ice types glacier Greenland S1A_EW_GRDH_1SDH_20141003T085100_20141003T085200_002663_002F72_97D0 50km
DLR.de Folie 32 Kathrin Höppner > 15th IICWG Meeting > October 20-25, 2014 > Punta Arenas, Chile Campaign support for Chile and Brazil Example: Support of Chilean vessel Aquiles in January 09-17, 2014 Course from Punta Arenas to O Higgins Station and Mendel Station TSX sequence shows changes of the ice situation at the tip of the Antarctic Peninsula 2014-01-09 / 0831 UTC O Higgins Station Mendel Station TerraSAR-X Akquisition Low Resolution Quicklook ScanSAR mode
Chart 33 First Tests on Near Real Time Ice Type Classification in Antarctica Anja Frost 18th July 2014 Ice type classification TerraSAR-X image (ScanSAR) Aim: Classify the ice in near real time Main steps of the algorithm: 1. Select texture features mean value, entropy and homogeneity based on σ 0 -values of all the pixels in a limited window based on the frequency distribution of σ 0 -values of two neigboured pixels 5 8 20 km frequency h window size 11 x 11 pixels σ 0,left
Chart 34 First Tests on Near Real Time Ice Type Classification in Antarctica Anja Frost 18th July 2014 Ice type classification TerraSAR-X image (ScanSAR) Aim: Classify the ice in near real time Main steps of the algorithm: 1. Select texture features mean value, entropy and homogeneity based on σ 0 -values of all the pixels in a limited window based on the frequency distribution of σ 0 -values of two neigboured pixels entropy = homogeneity = h(a,b) log(h(a,b)) a,b a,b h(a,b) 1+ a-b ² 20 km frequency h frequency h σ 0,left σ 0,left
> Lecture > Author Document > Date DLR.de Chart 35 MIZ recovery cruise (Norseman, Sep/Oct 2014) Sep, 26, 2014,16:58, TSX Stripmap VV
> Lecture > Author Document > Date DLR.de Chart 36 Oden Cruise (summer 2014) Sep, 20, 2014, TSX Stripmap, VV Aug, 23, 2014, TSX Stripmap, VV
> Lecture > Author Document > Date DLR.de Chart 37 Support campaign Ice classification Pack ice Open water Precampaign analysis: Time series Svalbard, April 2013
StripMap 22-08-2014 Point Barrow 50m 62m
Section of a TerraSAR-X Stripmap image Disko Bay 23rd April 2013 Iceberg detection within 20 minutes 2 km
Section of a TerraSAR-X Stripmap image Disko Bay 23rd April 2013 Statistical analysis on individual icebergs 2 km
Section of a TerraSAR-X Stripmap image Disko Bay 23rd April 2013 Statistical analysis on iceberg frequency throughout Disko Bay frequency iceberg size [m] 2 km