SBMDA Briefing (October 2013)
Transcription
SBMDA Briefing (October 2013)
Commercial Satellite Electro-Optic Imagery for Space Based Maritime Domain Awareness October 2013 David L. Neyland, Associate Director Office of Naval Research – Global US Navy London, UK david.l.neyland.civ@mail.mil Dr Nigel P. Bannister, Senior Lecturer Department of Physics & Astronomy University of Leicester Leicester, UK nb101@leicester.ac.uk Any use or reuse of the material herein or distribution of such must attribute the study to the US Navy Office of Naval Research - Global and the University of Leicester, UK . Distribution A: Approved for public release; distribution is unlimited. The Problem • Small vessels at sea may not be detectable by current land and space-based approaches with AIS and/or SAR • Space Based Maritime Domain Awareness (SBMDA) is recognized Internationally as mutually beneficial • Space-Based AIS data collection is becoming a commodity • Demonstrations in 2006 and LiMES in 2010 successfully showed that EO/IR can be used with AIS and SAR to identify and track vessels at sea • There is currently no approach or incentive to provide ubiquitous commercial satellite ocean imagery for use in SBMDA Distribution A: Approved for public release; distribution is unlimited. US National Space Policy - June 28 , 2010 Use Space for Maritime Domain Awareness • • International Cooperation – Departments and agencies, in coordination with the Secretary of State, shall: • Promote appropriate cost- and risk-sharing among participating nations in international partnerships; and • Augment U.S.capabilities by leveraging existing and planned space capabilities of allies and space partners. Identify Areas for Potential International Cooperation. Departments and agencies shall identify potential areas for international cooperation that may include, but are not limited to: space science; space exploration, including human space flight activities; space nuclear power to support space science and exploration; space transportation; space surveillance for debris monitoring and awareness; missile warning; Earth science and observation; environmental monitoring; satellite communications; GNSS; geospatial information products and services; disaster mitigation and relief; search and rescue; use of space for maritime domain awareness; and long-term preservation of the space environment for human activity and use. Distribution A: Approved for public release; distribution is unlimited. Background: AIS • • • The International Maritime Organization (IMO) International Convention for the Safety of Life at Sea (SOLAS), requires Automatic Identification System (AIS) transponders to be fitted to: – All ships >300 tons gross on international voyages – Cargo ships >500 tons on any route – Passenger ships of any size built after 2002 AIS transmits identification, position, status (at anchor/underway, etc.), heading, rate of turn… AIS transmission is directly from vessel to shore, or via relay (using transponders on nearby vessels) or received using space-based AIS systems Image: Courtesy of nauticexpo.com 4 Image: Courtesy of maritimejournal.com Distribution A: Approved for public release; distribution is unlimited. Commercial space-based AIS = Solved US Coast Guard AIS Data – 24 hours – 20 November 2011 – Courtesy Navy Times, Jacqueline Klimas, 3 June 2012 One week of data collected by University of Aalborg, Denmark, AAUSAT3 – a student built and operated Cubesat launched 25 February 2013 – Courtesy Jesper Larson, Aalborg University, USN ONRG meeting 5 September 2013 • Another example, per Michael Jones, CTO, Google, Google uses two SpaceQuest satellites which capture AIS data globally • Google Earth is used for common Internet display of AIS data and there are a multitude of commercial and amateur websites displaying AIS data Distribution A: Approved for public release; distribution is unlimited. Why AIS Is Not Enough • • Only a small fraction of vessels (~200,000 of the 17,000,000 registered worldwide) are required to carry AIS. Though some others carry voluntarily, the majority of vessels are not AIS-equipped AIS can be turned off, or “spoofed” to transmit deliberately misleading information – • Image: Courtesy of NauticExpo.com E.g. Iranian tankers spoof AIS to conceal voyages into Syrian waters: http://gcaptain.com/iran-falsifying-ais-data-to-conceal-ship-movements Image: Courtesy of REUTERS / Tim Chong Space-based AIS detection still limited by available spacecraft, service levels, and limitations imposed by current data protocol – E.g. Carson-Jackson, J., “Satellite AIS – Developing Technology or Existing Capability?”, J. Navigation, 2012, 65, pp. 303-321 Images: Courtesy Jesper Larson, Aalborg University, USN ONRG meeting 5 September 2013 Distribution A: Approved for public release; distribution is unlimited. 6 Synthetic Aperture Radar MDA • • Mainly L, C & X-band Spatial resolution versus swath trade-off – from 1-30 m for 30-100 km swath, to 100-1000 m for 30-500 km swath • 12 SAR spacecraft listed in Online Databases Kompsat-5 (S.Korea) RadarSat-2 (Canada) COSMO-SkyMed (4 spacecraft, Italy) RISAT-1 & 2 (India) HuanJing-1 (China) TerraSAR-X & TanDEM-X (Germany), Meteor-M N1 (Russia) • Images: http://www.crisp.nus.edu.sg Access to data is limited and tightly controlled by the asset operators Distribution A: Approved for public release; distribution is unlimited. The Electro-Optical SBMDA Concept • Imaging of Earth from space in the visible-IR band is common – At least 8 times as many EO sensors in Low Earth Orbit compared to SAR instruments, with EO resolutions in the 1 – 300 m range • • • A large number of VIS-IR spacecraft generate commercially available data (for free/fee) Commercial imagery providers are currently undertasked for over-water locations With coordination between operators, these satellites could form a SBMDA “Virtual Constellation” generating regularly updated information on locations of maritime vessels Google Image from Bahamas (small vessel ~15-20m) – Questions: Availability, Shareability, Affordability, Utility * ONRG / University of Leicester SBMDA Study, Dr Nigel Bannister – August 2013 Distribution A: Approved for public release; distribution is unlimited. 8 The EO SBMDA Initiative • Can we take advantage of the capabilities of commercial satellite EO imagery of the open ocean: – Whether collected today or not, – Using crowd sourced development, – In a Big-Data-Cloud, – To determine vessel locations, – To provide for Safety of Life at Sea. Distribution A: Approved for public release; distribution is unlimited. Previous / Ongoing Discussions • Dr John Mittleman – NRL, Office of the DoD Executive Agent for Maritime Domain Awareness • Greg Bader, Mike McGuinness, HQ USEUCOM/J8 S&T Advisors • Giulio Busulini, Scientific Attache, Italian Embassy (DC) • Richard (Rick) M. Williams , Counsellor, Defence R&D, Canadian Embassy (DC) • Michael Jones (CTO), Vint Cerf (CTO), Google • Dr Brian Young, Ralph Marrett, Defense Technology Agency, NZ • Doreen Dyck, Defense Counsellor, Canadian Embassy (London) • Guy Thomas – Chairman, Global Maritime Awareness Institute for Safety, Security & Stewardship, Taksha University • Delcan Kirrane - EU Science: Global Challenges, Global Collaboration • Chris Reynolds, Director, Irish Coast Guard • William Arras, VP Government Programs, Digital Globe • Dino Lorenzini, CEO, SpaceQuest Distribution A: Approved for public release; distribution is unlimited. Initial Steps • • • • • • • • Engage University of Leicester to conduct a preliminary study Define desired imaging parameters, such as “ground sample distance (GSD) Define “commercially available” as applied to electro-optic satellite imagery Identify EO-SBMDA assets, i.e., what commercially available electro-optic satellites sources exist and their capabilities Create an AGI System Toolkit (STK) model of a “virtual constellation” of commercially available electro-optic satellites Define a data collection test zone to explore with the STK model Simulate collection of data against real-world vessel tracks Determine the viability of output from the virtual constellation For the context of this study, commercially available is defined as any electrooptical satellite sensor which generates data that can be accessed by the international public, either as a free/open-source data product or as a paid-for service, regardless whether the satellite is owned and/or operated by a nation-state, non-government agency or commercial enterprise. Distribution A: Approved for public release; distribution is unlimited. Identifying EO-SBMDA Assets • • Satellites identified through regularly maintained online resources and STK catalogue For selection they must – Carry a visible wavelength sensor with GSD better than 350 m – Generate data that is available to the public “for free or for fee” • Key resources include OSCAR, Committee on EO Satellites, Space-Track, NASA-NSSDC and EOPortal 12 * ONRG / University of Leicester SBMDA Study, Dr Nigel Bannister – August 2013 Distribution A: Approved for public release; distribution is unlimited. Resolution versus Image Detail • • Care is required when considering “resolution” – angular versus spatial, Full Width Zero Maximum versus Full Width Half Maximum National Imagery Interpretability Scale (NIIRS) rates the level of information that can be extracted from an image – Applied post-facto – General Image Quality Equation can predict NIIRS score but requires detailed information on sensor performance and post-processing not available in this study • • • A useful and commonly quoted value is Ground Sample Distance (GSD): – GSD is the distance between adjacent pixel centers projected onto the target Need to understand what this means in terms of detail levels and the size of vessel which can be recovered Example is a high-resolution aerial image of a vessel with known dimensions and image properties, and resampled / rescaled to illustrate the detail that can be obtained, under favorable conditions, for a range of vessel sizes and sensor GSDs * ONRG / University of Leicester SBMDA Study, Dr Nigel Bannister – August 2013 From Zheng & Tidrow, Infrared Physics & Technology, 52, 2009, pp.408-411 Distribution A: Approved for public release; distribution is unlimited. Resolution versus Image Detail Example Resampled • Original aerial view of HMS Protector used as a test image, with background area increased using cloning. Original at https://navynews.co.uk/archive/news/item/1271 • Initial results indicate that a GSD less than 33% of vessel length required for detection • Sensors with GSD up to 350 m (largest tanker) included; detection may be possible in “ideal” conditions * ONRG / University of Leicester SBMDA Study, Dr Nigel Bannister – August 2013 Distribution A: Approved for public release; distribution is unlimited. Identifying EO-SBMDA Assets • • • • • Current asset list includes 55 satellites from 19 countries and agencies These spacecraft carry a total of 85 sensors falling within the GSD / information availability requirements Satellites include research EO spacecraft, commercial mapping spacecraft, disaster monitoring, meteorological and engineering prototype spacecraft All are in Low Earth, near-polar orbits All but one are Sun-synchronous * ONRG / University of Leicester SBMDA Study, Dr Nigel Bannister – August 2013 Distribution A: Approved for public release; distribution is unlimited. GSD of EO-SBMDA Constellation Sensors • • • 85 sensors will be proposed in the virtual constellation All sensors operate in the visible part of the spectrum Most sensors also include NIRIR channels – IR may be effective for night-time tracking but the limitations are not yet understood in this study – For this study no IR observations are assumed; all detections take place during times when the Sun is above the horizon for the target location Note: 350 m GSD is marginal as a SBMDA asset – i.e., it provides ~one pixel detection of a big tanker * ONRG / University of Leicester SBMDA Study, Dr Nigel Bannister – August 2013 Distribution A: Approved for public release; distribution is unlimited. Suggested SBMDA Study Box 8 deg by 10 deg = ~250k sqnm Suggested collection box from Dr Brian Young, NZ Defense Technology Agency, 3/15/2013, as “everything has to go around Cape Reinga (the northern-most tip of NZ) on the way to Australia…this would also capture merchant vessels crossing the Tasman and cruising boats arriving in the Bay of Islands (Oct to Dec timeframe), and would have a high probability of capturing Navy inshore patrol vessels.” Distribution A: Approved for public release; distribution is unlimited. Study Test Zone • Study area is a box of approximately 500 x 500 nautical miles and includes part of the NZ Exclusive Economic Zone (NZEEZ) • Bounded by coordinates 27S 169E 27S 179E 35S 169E 35S 179E • Current model* based on 3 km (~1.9 mile) granularity within test zone (significantly smaller than sensor footprints) * ONRG / University of Leicester SBMDA Study, Dr Nigel Bannister – August 2013 Image: courtesy mfe.govt.nz Distribution A: Approved for public release; distribution is unlimited. Example Satellite Pass Example Swath ~9 nm Wide * by 500nm traverse in <2 min * GeoEye-1 Instrument/Product Description (Public Domain 2009) lists 15.2 km (~9 nm) swath width Distribution A: Approved for public release; distribution is unlimited. Another Satellite Follows Distribution A: Approved for public release; distribution is unlimited. And Another Pass Distribution A: Approved for public release; distribution is unlimited. Time Geo-temporal Coverage Image strips from different satellites can be fused in a time-phased mosaic Longitude Buried in the geo-temporal mosaic are the vessels tracks This is what we need to store in the Cloud and let the crowds figure out how to process! Distribution A: Approved for public release; distribution is unlimited. EO-SBMDA Virtual Constellation 85 electrooptical sources from 19 countries* * ONRG / University of Leicester SBMDA Study, Dr Nigel Bannister – August 2013 Distribution A: Approved for public release; distribution is unlimited. One Day Coverage, Daylight Hours • • • Model run over one period of daylight, from 18:49 UT (Aug 1) – 05:17 UT (Aug 2) Results show the form of coverage over a typical day (since equator crossing times of satellites are fixed) Gap length estimations for 1-day simulations describe the frequency of coverage within a single period of daylight (i.e. are not dominated by night time outage) * ONRG / University of Leicester SBMDA Study, Dr Nigel Bannister – August 2013 Distribution A: Approved for public release; distribution is unlimited. One Day Coverage, Daylight Hours Number of accesses Total observing time Max gap between observations All Sensors Only sensors with GSD of 30 m or better * ONRG / University of Leicester SBMDA Study, Dr Nigel Bannister – August 2013 Distribution A: Approved for public release; distribution is unlimited. Modeling Specific Vessel Contacts • • • • Using YotReps & AIS track histories from sailwx.info, routes & movements of 9 “real” vessels are included in the model Tracks time-shifted so that all vessels enter the test zone during the 9 day model period All instances in which a vessel was within a sensor field of view, inside the test zone, when the Sun was above the horizon, identified Weather outages are not included at this stage – “clear sky” conditions assumed * ONRG / University of Leicester SBMDA Study, Dr Nigel Bannister – August 2013 Distribution A: Approved for public release; distribution is unlimited. Vessel Contact Results No night-time IR data used Vessel enters zone at night All Sensors GSD 150 GSD 50 GSD 25 * ONRG / University of Leicester SBMDA Study, Dr Nigel Bannister – August 2013 Distribution A: Approved for public release; distribution is unlimited. Example Case: Niña • • • • • • • 85 year old, 21 metre-long US schooner Niña left Opua, NZ North Island, May 29th 2013 for Newcastle, Australia (1500 mile journey) Estimated voyage duration 10 days (avg. 5.4 knots) Last contact June 4th, 370 miles WNW Cape Reinga (northern tip of the North Island) in 26’ seas, travelling at 4 knots, bearing 310º Reported overdue 12 June, search efforts ended after overflying 737,000 square nautical mile area No further contact; vessel with 7 onboard presumed sunk Carrying only manually activated emergency beacons Ship5 in the SBMDA model represents a Niña-like voyage leaving Opua for Newcastle at 12:00 UTC, August 1st 2013 at a steady 4 knots http://www.sailmagazine.com/nina Nina modeled as Ship5 track in SBMDA STK model * ONRG / University of Leicester SBMDA Study, Dr Nigel Bannister – August 2013 Distribution A: Approved for public release; distribution is unlimited. Ship5 Contacts (one day of voyage) Nina modeled as Ship5 track in SBMDA STK model Potential contact for Beijing-1 of Ship5 track in SBMDA STK model 500 km Approx 26 km OceanSat-2 Direction of motion NigeriaSat-X Terra Deimos-1 FengYun 3A & Landsat-8 Terra Meteor M N1 Beijing-1 Rapideye-4 Numbers indicate GSD of observing satellite * ONRG / University of Leicester SBMDA Study, Dr Nigel Bannister – August 2013 NZEEZ SBMDA Test Area Distribution A: Approved for public release; distribution is unlimited. Ship5 Contacts (as a function of GSD) Approx 560 km Max distance without contact: 132 km All Sensors Max distance without contact: 172 km GSD < 100 m Max distance without contact: 192 km GSD < 25 m * ONRG / University of Leicester SBMDA Study, Dr Nigel Bannister – August 2013 Max distance without contact: 172 km GSD < 200 m Max distance without contact: 174 km GSD < 50 m Max distance without contact: 537 km GSD < 7 m Distribution A: Approved for public release; distribution is unlimited. How Much Data? • The University of Leicester study computed the volume of data (bytes of imagery) for the original study area – Simplified approach based on RapidEye Specification (available online) – scaled for others using their specs – RapidEye imager generates a footprint with a cross-track dimension of approximately 77 km, and an along-track unit of length 25 km with a Ground Sample Distance (GSD) of 6.5 m. – The number of pixels in a 77 x 25 km image is (77 x 25) km2 / (6.5 x 6.5) m2 = 53,472,222 – The image product has a 16 bit depth, so a single 77 x 25 km image is predicted to have a size of 16 x 53,472,222 = 855,555,556 bits or approximately 102 Mbytes – Rapideye quotes a single frame file size of 462 Mbytes for 5 wavebands, perband size of 462/5 = 92.4 Mbytes, or 91% of the value estimated above * ONRG / University of Leicester SBMDA Study, Dr Nigel Bannister – August 2013 Distribution A: Approved for public release; distribution is unlimited. How Much Data (continued) • • • • • Along track and crosstrack footprint of all 85 sensors estimated The swath length is divided by the sensor GSD to determine the number of pixels along-track The number of pixels along-track is multiplied by the number of cross-track pixels = the total number of pixels within the observed swath Multiplying by the bit depth then provides an estimate of the size of the data product for that pass Total data storage requirement is simply the sum of these data product sizes for each access of each sensor ...etc. Note: If at least one pixel is inside the test zone and the illumination conditions are met for that pixel, then the entire field of view is recorded. * ONRG / University of Leicester SBMDA Study, Dr Nigel Bannister – August 2013 Distribution A: Approved for public release; distribution is unlimited. Data from Each Senor for Study Case • Histogram showing the size of each access in terms of the number of bits which represent the observed swath. * ONRG / University of Leicester SBMDA Study, Dr Nigel Bannister – August 2013 Distribution A: Approved for public release; distribution is unlimited. Daily data volumes for the simulation • • 3.9 TB/ 9.5 days = 0.43 TB data per day 10-day data volume is estimated to be 4.33 TB for the ~500nm x ~500nm study area * ONRG / University of Leicester SBMDA Study, Dr Nigel Bannister – August 2013 Distribution A: Approved for public release; distribution is unlimited. EO-SBMDA Study - Preliminary Results • • • • • • • • 55+ satellites from 19 countries carrying a total of 85 imaging sensors generate images which are available commercially / internationally / publically for free or for fee When combined, the high resolution (GSD ≤ 350 m) sensors on these satellites give complete coverage of the test zone (and by inference most regions of Earth) in one day Excepting vessels which enter the test zone only briefly (e.g. transit across a corner), at least one position fix per day is observed In several cases, two or more fixes are available, separated by an hour or more In favorable situations 10+ fixes are available at useful resolution Using a coordinated approach of a virtual constellation, the availability of geo-temporal imagery over water could be significant for improving safety of life at sea Significant cross-processing of the imagery is crucial to extract meaningful time-phased information from the data This study shows that there may be great benefit in building an international, open collaboration between imagery providers, cloud storage and open source algorithm development to translate the potential collection of open ocean imagery into a meaningful maritime domain awareness information * ONRG / University of Leicester SBMDA Study, Dr Nigel Bannister – August 2013 Distribution A: Approved for public release; distribution is unlimited. Where to Now? • • • • • • The study suggests that Commercial Satellite EO may provide sufficient data (vessels contacts) to hypothesize vessels tracks – if the right algorithms existed The intellectual issue is “how to derive tracks from sparse data of unrelated detections from unrelated sensors spread across time” The study is only a model of what satellites could collect – the challenge set needs to be populated with real world imagery Real world imagery would enable development of the algorithms to hypothesize vessel tracks from geo-temporal, independent detections The amount of imagery for ten days of collection from 85 sensors would be ~10-20 TB Cloud Storage and Processing of the imagery is the only possible path to developing algorithms Distribution A: Approved for public release; distribution is unlimited. Data Collection Opportunity • • • • • Sail to Fiji - June, 2014 – 20+ “small” sailboats – ~1100 nm from Auckland, NZ to Port Denarau, Fiji Each sailboat will be equipped with “Yellow Brick™ Iridiumbased Global Satellite Tracking Characteristics and imagery of all vessels documented a-priori Provides a unique “truth case” for EO imagery collection and algorithm development NZ DTA actively interested in facilitating this as an experiment Distribution A: Approved for public release; distribution is unlimited. Data Collection Area for Sail to Fiji • • Original “Study Area” was ~500nm x ~500nm = ~250k square nm To accommodate Sail to Fiji, Data Collection Area proposed to be ~500 nm by ~1260nm = ~668k square nm Sail to Fiji Data Collection Area Original Study Area Sail to Fiji Data Collection proposed to be bounded by coordinates 17 S 170 E 17 S 180 E 38 S 170 E 38 S 180 E Distribution A: Approved for public release; distribution is unlimited. Next Steps for ONRG – Possible Path • • • • • • • Distribute the results from University of Leicester model of the time-phased satellite coverage for the NZ collection zone – October-December 2013 Pool resources for a S/W Development Prize – November-December 2013 – Suggest $100k based on experience Convince Owner/Operators to provide imagery for 8-14 days for “Sail to Fiji” collection area – January-April 2014 (data collect June 2014) Convince a “Cloud” owner/operator to provide data access, processing power and connectivity – January-April 2014 (storage and processing June-October 2014) Conduct Virtual Workshop via web and international location to present problem and challenge to the research community – February-April 2014 Run the Algorithm Challenge in August-October 2014 Prize Award ~December 2014 Distribution A: Approved for public release; distribution is unlimited.