Font Size: a A A

Satellite mapping and automated feature extraction: Geographic information system-based change detection of the Antarctic coast

Posted on:2005-08-18Degree:Ph.DType:Dissertation
University:The Ohio State UniversityCandidate:Kim, Kee-TaeFull Text:PDF
GTID:1450390011950864Subject:Engineering
Abstract/Summary:
Declassified Intelligence Satellite Photograph (DISP) data are important resources for measuring the geometry of the coastline of Antarctica. By using the state-of-art digital imaging technology, bundle block triangulation based on tie points and control points derived from a RADARSAT-1 Synthetic Aperture Radar (SAR) image mosaic and Ohio State University (OSU) Antarctic digital elevation model (DEM), the individual DISP images were accurately assembled into a map quality mosaic of Antarctica as it appeared in 1963. The new map is one of important benchmarks for gauging the response of the Antarctic coastline to changing climate.; Automated coastline extraction algorithm design is the second theme of this dissertation. At the pre-processing stage, an adaptive neighborhood filtering was used to remove the film-grain noise while preserving edge features. At the segmentation stage, an adaptive Bayesian approach to image segmentation was used to split the DISP imagery into its homogenous regions, in which the fuzzy c-means clustering (FCM) technique and Gibbs random field (GRF) model were introduced to estimate the conditional and prior probability density functions. A Gaussian mixture model was used to estimate the reliable initial values for the FCM technique. At the post-processing stage, image object formation and labeling, removal of noisy image objects, and vectorization algorithms were sequentially applied to segmented images for extracting a vector representation of coastlines. Results were presented that demonstrate the effectiveness of the algorithm in segmenting the DISP data. In the cases of cloud cover and little contrast scenes, manual editing was carried out based on intermediate image processing and visual inspection in comparison of old paper maps.; Through a geographic information system (GIS), the derived DISP coastline data were integrated with earlier and later data to assess continental scale changes in the Antarctic coast. Computing the area of major Antarctic ice shelves between 1963 and 1997, we found that the net loss was approximately 0.8% and ice shelves retreated mostly between DISP and Scientific Committee Antarctic Research (SCAR) Antarctic Digital Database (ADD). In addition, over the 56-years (1947--present) observations on Pine Island Glacier, we found that the retreat rate has been approximately -10 +/- 65 m/yr.
Keywords/Search Tags:Antarctic, DISP, Data, Coastline
Related items