Date of Completion

12-31-2013

Embargo Period

12-31-2013

Keywords

Humanitarian crisis, Rapid mapping, Geographic object-based image analysis, Data fusion, Image segmentation, Multiresolution segmentation

Major Advisor

Thomas Meyer

Associate Advisor

Thomas Blaschke

Associate Advisor

Marco Neubert

Associate Advisor

Jeffry Osleeb

Associate Advisor

Daniel Civco

Field of Study

Geological Sciences

Degree

Doctor of Philosophy

Open Access

Open Access

Abstract

Advanced Earth observation (EO) is increasingly recognized as an indispensible tool to support rigorous and robust decision making in crisis management. In crisis scenarios, EO data need to be streamed through time-critical workflows for delivering reliable and effective information to civil protection authorities. In this context, the over arching goal of this research is to closely examine the integral segments of routing EO-based rapid mapping workflows to understand prevailing shortfalls and to devise novel approaches to catalyze conditioned geoinformation delivery to cater increasing user demand. This study envisions three interconnected objectives, which are primarily fuelled by very high spatial resolution (VHSR) imagery, data fusion, image segmentation, and geographic object-based analysis (GEOBIA) framework. Focal study areas encapsulate natural and anthropogenic crises having occurred in the recent past: the 2010 earthquake-damaged areas in Haiti, the 2010 flood-impacted sites in Pakistan, and armed-conflicted areas and internally displaced persons (IDP) camps in Sri Lanka

The first objective investigated how different data fusion algorithms perform when applied to VHSR satellite images that encompass ongoing- and post-crises scenes. The evaluation entailed twelve fusion algorithms. The spatial and spectral fidelities were assessed subjectively using fourteen quality indices. Ehlers, Wavelet, and High-pass filtering (HPF) fusion algorithms had the best scores for the majority of spectral quality indices. The University of New Brunswick and Gram-Schmidt fusion algorithms had the best scores for spatial metrics. The HPF algorithm emerged as the overall best performing fusion algorithm.

The second objective aimed to unravel the synergies of data fusion and image segmentation in the context of EO-based rapid mapping workflows. We statistically compared the quality image object candidates among twelve fused products and their original MS and PAN images. We have shown that the GEOBIA framework has the ability to create meaningful image objects during the segmentation process by compensating the fused image’s spectral distortions with the high-frequency information content that has been injected during fusion. We further questioned the necessity of the data fusion step in rapid mapping context. Bypassing time-intense data fusion steps helps to intensify EO-based rapid mapping workflows.

The third objective explored the efficacy of supervised empirical discrepancy measures for optimizing multiresolution segmentation (MRS) algorithm. I selected the Euclidean distance 2 (ED2) metric, a recently proposed supervised metric that measures dissimilarity between a reference polygon and an image object candidate, as a candidate to investigate the validity and efficacy of empirical discrepancy measures for finding the optimal scale parameter setting of the MRS algorithm. The discriminative capacity of the ED2 metric across different scales groups was tested using non-parametric statistical methods. My results showed that the ED2 metric significantly discriminates the quality of image object candidates at smaller scale values but it loses the sensitivity at larger scale values. This questions the meaningfulness of the ED2 metric in the MRS algorithms parameter optimization.

Share

COinS