Science Service System

Summary of Proposal COA2312

TitleToward an integrated system for ship identification and oil spill movement using high-resolution SAR data
Investigator Yang, Chan-Su - Korea Institute of Ocean Science and Technology (KIOST), Korea Ocean Satellite Center
Team Members
Professor Ouchi, Kazuo - Korea Institute of Ocean Science & Technology (KIOST), Korea Ocean Satellite Center
SummaryThe purpose of the study is to develop (1) ship detection and classification algorithms (2) estimation methods of oil spill area (3) forward and inverse models of spilled oil transport using high-resolution and multi-polarization SAR data (4) an operational GUI-based module consisting of above three algorithms For ship detection by SAR, the detection accuracy will be examined quantitatively by applying CFAR to amplitude, dual-polarization coherence, and entropy images, and estimating signal-to-noise ratio (SNR), FAR, and receiver operating characteristics (ROC). In particular, accuracy dependencies will be investigated on polarization, incidence angle, the ratio of ship size to resolution, and sea states such as wind speed and waveheight. For CFAR processing, setting of suitable initial FAR value for required purposes is crucial, and hence it is important to quantify the relation between the initial and post-processed resultant FAR values. The algorithms for polarimetric analyses are based on the quad and dual polarization data. Coherence is high on the sea surface due to the dominant surface scattering, while the coherence of ships (and land) is low because of multiple (volume) scattering. Due to the random nature of multiple scattering, dual polarization entropy of ships is high, and low over the sea surface. In the theme of ship classification, we are not certain at this stage how much progress can be made to develop algorithms based on the parametric vectors using empirical data because of limited number of available images. However, a study is in progress to simulate the shapes of ships in SAR image by ray-tracing. We intend to extend the current simple ray-tracing algorithm by including scattering theories such as small perturbation approximation (SPM) for the sea surface and edge diffraction. In the theme of oil slick detection and forecast, we shall examine the abilities of slick area extraction by the methods using CFAR applied to amplitude images, difference in statistical distribution of polarimetric data, dual polarization eigenvalue analysis, and coherence also used in ship detection. The method of eigenvalue analysis has been used to extract underwater laver cultivation nets by our team, and the method can also be used for oil slick detection. In order to compute RCS, waveheight profiles are first simulated by the Monte Carlo method for a given waveheight spectrum, and RCS from the simulated waveheight variation is computed using scattering models. Extraction of oil-covered sea surface is made from the difference in RCS of clean surface. The study on the oil transport forecast will be made using an adaptive wind drift factor Q developed by our group instead of the conventional fixed Q. An attempt will be made to improve the accuracy by including the fate process such as evaporation and dissolution. We shall evaluate the accuracies of the algorithms in the above themes, clarify the advantages and disadvantages, integrate the algorithms into a single module, and develop a ship detection and identification - oil spill transport system that can be operated on the GUI base. The TerraSAR-X data required for the proposed project are those on the waters around the Korean Peninsula, including the Gyeonggi Bay in the Yellow Sea where traffic is very busy, and AIS and in-situ meteorological data can easily be accessed. For the validation of small ships and those without AIS signals, visual observation simultaneous with TerraSAR-X data acquisitions will be made. Although it is very unlikely, if TerraSAR-X acquires data containing oil spills by, for example, tanker accidents, as much meteorological and ground-truth data will be collected. We have several projects on SAR applications supported by Ministry of Oceans and Fisheries and KIOST.

Back to list of proposals

DLR 2004-2016