Feature Extraction from High Resolution Satellite Imagery as an Input to the Development and Rapid Update of a METRANS Geographic Information System (GIS)

摘要

This report describes an accuracy assessment of extracted features derived from three subsets of Quickbird pan-sharpened high resolution satellite image for the area of the Port of Los Angeles, CA. Visual Learning Systems Feature Analyst and Definiens eCognition Developer software platforms were used for image classification and results were compared. A methodology was developed to determine the accuracy of the feature extraction algorithms using both software packages in the ESRI ArcView GIS interface, where the results of automatic and manual image classification methods were compared. The findings indicate that Definiens Ecognition Developer was overall more successful at identifying ground features in the areas where research was conducted.