טכניון מכון טכנולוגי לישראל
הטכניון מכון טכנולוגי לישראל - בית הספר ללימודי מוסמכים  
M.Sc Thesis
M.Sc StudentGanor Dov
SubjectAn Object Oriented Model for Identifying and Tracking
Marine Pollution from Satellite Images in a
Coastal Zone Information System
DepartmentDepartment of Civil and Environmental Engineering
Supervisors Dr. Ofer Zilberstain
Dr. Gideon Tibor


Abstract

This research deals with methods for detecting and tracking marine pollution analyzed from Satellite images in a Costal Zoon Information System. An innovative objected oriented model was developed and conducted on several pollution events. Special attention was given to the Israeli coastal zone and to the event of sewage failure in Tel-Aviv Region Of Interest (ROI) during January-April 2003.

The research applies a new concept that represents water parameters as objects. A Data Analysis Object Oriented Model (DAOOM) was developed that operates in a GIS environment and is based on time-series of Earth Observation (EO) products. The core of this concept is a spatial feature, which extracts, analyzes and tracks pollution event as an object from its birth (e.g. appearance) to death (e.g. disappearance), 

 The model flow consists of several phases: (I) segmentation and extraction of features from the image; (II) classification of features relations between series of images; (III) change detection; (IV) tracking events in space and time. The classification procedure is based on solving a global optimization problem by the Earth Movers Distance (EMD) algorithm. The degree of correlation between two phases of the object is calculated with Fuzzy Logic (FL) and the detection of global displacement is done with the KLT algorithm.

The use of GIS applications such as CZIS (Costal Zone Information System) together with SIS (Satellite Information System) integrates Remote Sensing data to the costal zone and improves the monitoring of water quality. The goal of this research was to create applicative knowledge from the information gathered in the GIS by designing automatic and intelligent methods using expert systems for pattern recognition. These methods offer tools for effective management of the marine environment.