Ship Detection : from Processing to Instrument Characterisation
Guillaume Hajduch(1), Vincent Kerbaol(1) and Romain de Joux(1)
(1) BOOST Technologies, 115, rue Claude Chappe, 29280 Plouzané, France
Ship detection is a crucial application for global monitoring for environment and security.
It permits to monitor traffic, fisheries, and to associate ships with oil discharge.
Nevertheless, each of those applications has specific constraints in terms of spatial coverage and resolution, radiometric quality, and contrast between vessels and ocean.
By choosing adapted modes, polarisation and processing level ENVISAT/ASAR gives access to a wide range of data which can be optimally processed for specific applications.
This paper addresses performance assesment of specific ship detection algorithms which have been developped for various demanding applications and extensively tested on ASAR data.
It will be shown that proper instrument characterisation is necessary to achieve low probability of false alarm which is highly desirable for operational exploitation of ship detection reports.
First, typical ship detection schemes and strategies are presented.
Past research efforts on automatic target detection in SAR imagery have clearly demonstrated that no single detection algorithm will produce satisfactory results with sufficient detection sensitivity and small false alarm rate.
Thus, ship detection systems generally consist of four stages:
(1) preprocessing: generally adrressed by calibration process.
(2) (land) masking: enabling to focus on area of interest.
(3) prescreening: in order to highlight candidate targets.
(4) discrimination: in order to reject false alarms.
Second, specific processing for false alarm rejection are addressed.
We will focus on false alarm induced by instrument artefacts and meteo-oceanic phenomena.
Some dedicating processing are proposed including :
(1) rejection of azimuth ambiguities and short life time events on detected and SLC products.
(2) rejection of range/nadir ambiguities
(3) rejection of false alarm near noise equivalent sigma nought
The interest of using such processing is illustrated on some examples, including coastal regions.
Eventually, a statistic study of performance assesment for ship detection algorithms is performed, with respect to local meteo-oceanic conditions, target parameters and recording geometry.
It will be shown that it is necessary to discriminate as much artefacts as possible to perform such an analysis.
Major result of this paper is a characterization of noise equivalent sigma nought performed by rejection of false alarms.
This result illustrates that precise radiometric calibration must be achieved in order to define operational services of ship detection.
Keywords: ESA European
Space Agency - Agence spatiale europeenne,
observation de la terre, earth observation,
satellite remote sensing,
teledetection, geophysique, altimetrie, radar,
chimique atmospherique, geophysics, altimetry, radar,