Dissertations and Theses @ UNI

Availability

Open Access Thesis

Keywords

Roads--Remote sensing; Academic theses;

Abstract

The goal of this study was to extract transportation infrastructure features from hyper-spectral imagery using data from the high resolution Airborne Visible/Infrared Imaging Spectrometer (AVIRIS). A "Minimum Noise Fraction" (MNF) and a classification and regression tree (CART) were used to reduce the number of spectral dimensions to be analyzed. Then, hyperspectral classifiers such as the Spectral Angle Mapper (SAM), Mixture-tuned matched filtering (MTMF) and mixture-tuned matched filtering combined with classification and regression tree (MTMF-CART) were used to extract the transportation infrastructures from the image using different image processing software such as ENVI and ERDAS IMAGINE. Finally, WinTopo software was used to make the results GIS compatible by vectorizing the classification output. The study found that the MTMF classifier outperformed all other classification methods with an overall accuracy of 88. 92% compared to the SAM and MTMF-CART methods that produced accuracies of 81.89% and 84.32% respectively. Nevertheless, in the majority of cases all three classifiers succeeded in the difficult task of distinguishing different road classes from each other and from similar urban materials such as roofing material. On the other hand, all three classifiers extracted the city street class poorly, and certain classes such as gravel and concrete in the SAM image and highway in the MTMF-CART image were frequently confused with background material. The study also found that the WinTopo software was unable to create any useful vector data from the raw classification output. Overall, this study provides the foundation for future research regarding the use of remotely sensed imagery and transportations infrastructure. Future research should be directed at increasing the usefulness of transportation feature extraction using remote sensing data. In particular, this requires research aimed at perfecting the conversion of raster classification data into a GIS compatible vector format, as well as research aimed at enhancing the classification methods in order to extract additional information such as road quality and road condition.

Year of Submission

2005

Degree Name

Master of Arts

Department

Department of Geography

First Advisor

Ramanathan Sugumaran

Second Advisor

Tim R. Strauss

Third Advisor

James F. Fryman

Comments

If you are the rightful copyright holder of this thesis and wish to have it removed from the Open Access Collection, please submit a request to scholarworks@uni.edu and include clear identification of the work, preferably with URL.

Date Original

2005

Object Description

1 PDF file (117 leaves)

Language

en

File Format

application/pdf

Included in

Geography Commons

Share

COinS