Ship Detection in ENVISAT ASAR Alternating Polarisation Images

• Main Authors: VAN WIMERSMA GREIDANUS Herman, THOORENS Francois-Xavier, BAUNA Tony, INDREGARD Marte, ARNESEN T.N.
• Language: English
• Published: Canadian Space Agency 2008
• Subjects: Barents Sea; Indian; Asar; North Atlantic
• Online Access: https://publications.jrc.ec.europa.eu/repository/handle/JRC37317; http://www.space.gc.ca/asc/eng/events/2007/asar.asp

RADARSAT has been used with much success for ship detection, and considerable experience has been built up in the past. ENVISAT ASAR does not yet have such a long history. It has long been known that VV polarisation is not very accommodating to ship detection, although it is for ship wake imaging. More recent work with ENVISAT has made it clear that HH is best with shallower incidence and HV with steeper incidence. But there is still a need to have better quantitative performance figures for ship detection in the different polarisation channels and beam incidence angles. In the framework of ship detection campaigns, JRC has acquired close to 100 ENVISAT ASAR Alternating Polarisation (AP) images, most between 2004 and 2006. Many of them contain a number of known ships, from VMS (fisheries transponder), AIS (merchant ship transponder) or patrol aircraft. Different regions are covered, notably the Baltic Sea, the North Atlantic, the Barents Sea and the Indian Ocean, and different vessel types: fishing vessels and merchant vessels. This study will present detection and false alarm statistics, look at main causes of false alarms and compare HH, VV and HV performance. Although maybe generally not so much appreciated, wakes of the larger and faster moving vessels show up also in HH and HV images. Frequency of occurrence of these is also derived. At time of writing of this abstract, the detailed statistical analysis has progressed through 10 AP HH/HV images, all in the Baltic Sea on a marine traffic route. These images contain a total of 296 targets that are interpreted, by visual analysis, as most likely being ships. Their statistics show e.g. that in IS3 (steep) beam, 90 % of the ships are seen both in HH and HV, 2 % only in HV and 8 % only in HH. In IS7 (shallow) beam, 74 % are seen both in HH and HV, 2 % only in HV and 25 % only in HH. In IS3 beam 30 % of the ships display a (short) wake in HH and the same percentage do in HV, whereas in IS7 beam almost no HV wakes are seen anymore while the HH wakes are shorter, although still occurring with 30 % of the ships. The wakes, in fact, are sometimes so strong that they cause 30 % of all false alarms in IS3. Other important causes of false alarms are azimuth ambiguities of strong ships or strong reflectors on land; nearly exclusively in HH. Whereas such false alarms caused by strong ships may be corrected for, because the ships that cause it are detected as well, those from land are harder to recognise and pose a real problem in coastal waters. JRC.G.4 - Maritime affairs