%0 Conference Paper
%A Ridha, S.
%A Wardaya, P.D.
%D 2014
%F scholars:4577
%I Society of Petroleum Engineers
%K Cost effectiveness; Neural networks; Oil spills; Pixels; Satellite imagery; Seawater, Artificial neural network algorithm; Classification results; Computation time; Computer memory; Neural network algorithm; Oil exploitation; Oil spill detection; Satellite images, Sustainable development
%P 378-386
%R 10.2118/170406-ms
%T Oil spill detection and characterization from satellite image using artificial neural network algorithm
%U https://khub.utp.edu.my/scholars/4577/
%X This paper describes the use of artificial neural network to identify and characterize oil spill acquired from satellite imagery. The objective of the algorithm is to classify every pixel of the image whether it is sea water or oil based on its intensity. In order to test the algorithm, several order of noise is introduced, simulating real situation caused by weather or other pollutant during the acquisition. The study shows that neural network algorithm provides efficient technique to differentiate the oil spill from water body. The neural network is capable of delivering accurate classification in less than one minute computation time. The required training data can also be suppressed indicating the cost effectiveness in term of computer memory. From the classification result, area prediction is performed by using standard image analysis. The technique counts the number of pixels classified as the oil and sum them to provide the areal extent of the spillage. This result gives important solution for environmental handling of the oil exploitation activity. For example, the method can be used in providing estimate on the thickness of the discharged oil so that environmental effect can be predicted more reliably. Copyright © 2014, Society of Petroleum Engineers.
%Z cited By 0; Conference of SPE Middle East Health, Safety, Environment and Sustainable Development Conference and Exhibition, MEHSE 2014 ; Conference Date: 22 September 2014 Through 24 September 2014; Conference Code:111486