Authors

MONICA MAGDY, OMAR SALEH, MONICA HANY, MADONNA SAID

SUPERVISED BY : DR. AYMAN NABIL

ASSISTANT SUPERVISED BY: ENG. YOUMNA IBRAHIM, ENG. MAHA SAYED


Publishing Date

27-DEC-2021

Abstract

THE HIGH DEMAND OF OIL CAUSED A HUGE LOSS IN FORM OF OIL SPILLS DURING THE PROCESS OF EXPORTING THE PRODUCT, WHICH LEAD TO AN INCREASE IN THE POLLUTION SPECIALLY THE MARINE ENVIRONMENT. OUR PROJECT MAINTAIN THIS PROBLEM USING MODERN TECHNOLOGY BY DETECTING OIL SPILLS USING SATELLITE IMAGERY. THIS DETECTION IS ACHIEVED WITH THE ASSISTANCE OF VARIOUS MACHINE LEARNING TECHNOLOGIES AND IMAGE PROCESSING METHODS USING SATELLITE HYPERSPECTRAL IMAGES. THE GEOLOGIST TEAM HAD DEFINED HYPERSPECTRAL IMAGES AS THE PRINCIPLE FEATURE WHICH FOCUS ON DETECTING THE OIL SPILLS DURING ALL THE PROCESS DURING THE EXPORTING PROCESS. THE SYSTEM USES  THIS SPECTRUM AND SPECIAL FEATURE THAT WILL BE EXTRACTED FROM IMAGES USING  SOME FILTERS AND CLASSIFIERS. WE PERFORM PREPROCESSING ON THE IMAGE, GOING THROUGH  THE PROCESS THEN USING VARIOUS CLASSIFIERS TO ACCOMPLISH THE BEST ACCURACY. THE STUDY AREA WAS DECIDED BASED ON THE HUGE AMOUNT OF OIL  DETECTED IN SUCH AS GULF OF  MEXICO WHICH IS OBTAINED USING AVIRIS DATASET. 

1.1 Purpose of this document

The purpose of this Software Description Document is to provide a brief overview of the various functions of our system and the reasons for our development, mentioning the scope and the references used. This document explains how the output of the system is built and meets the technical requirements. The main purpose of our system is to detect the oil spills using remote sensing data using hyperspectral images in order to detect the oil spills. The target audience of the system is the petroleum firms especially the exploration department (geologist team). The system will be working by locating certain locations in the hyperspectral images to be labeled into three main classes according to reflectance spectra of oil emulation then reclassifying raster to extract the three priority colored image to get thickness of oil spill.

1.2 Scope of this document

This document, will be explaining more how this software is used to obtain a better understanding of the whole system and define other functional and non-functional requirements that we may develop along the way as the project progresses.

1.3 System Overview

The proposed system consists of some linked stages: The first stage is data input. The Satellite hyperspectral camera starts to capture the earth’s ground with a wide spectrum of light. The second stage is the feature extractions. Those images are taken as a 3D array for each R-G-B to find new features by selecting existing features to reduce the feature space without loss of information by using Principal Component Analysis (PCA) to reduce the excess of dimensions of the HIS. Then the machine learning stage. The system has some algorithms used to choose the highest accuracy. First; The Support Vector Machine (SVM) is a machine learning algorithm based on statistical supervised learning theory used to analyze and classify. Second, the three-band ratio to enhance the spectral differences between bands and to reduce the effects of the atmosphere on camera. Then the Convolutional Neural Network is used to analyzing visual imagery. Finally, the accuracy report will decide to classify the ground truth classes for the current dataset.

1.4 System Scope

The proposed system aims to automatically detect new oil spills by using different algorithms and classifiers by satellite imagery. The main purpose is increasing the accuracy of the Classifiers to experiment with various techniques by ensemble different frame works. Ensemble classifiers have been developed to integrate several classifiers as supervised or unsupervised learning techniques to enhance the accuracy and consistency of a single classifier. The system will be working by locating certain locations in the hyperspectral images those images will be classified according to hydrocarbon emissions that are emitted from the oil reservoirs then Inform the results detected from our system to competent authorities.