Texture-based terrain classification and optimal sampling in support of digital elevation model extraction
Abstract
The purpose of this research is to enhance the efficiency of the task of extracting elevation data from digital (images captured by a digital imaging system) and digitized (standard aerial images scanned into a digital format) aerial imagery on a digital photogrammetric workstation. This efficiency was gained in several areas. First, image and terrain texture features were used together with maximum likelihood classification to classify elevation points into terrain classes that are useful in the elevation extraction and editing process. These classes were bare ground, obscured ground, and water body. Image and terrain texture information was found to be adequate to perform this classification to a high level of accuracy. The information from the terrain labels was used to drive automatic editing functions. These functions, when successful, removed the effects of the ground obscurations from the Digital Elevation Model. The terrain classification information was also used to provide input to an iterative refinement algorithm that aided the extraction process in collecting elevations in the presence of terrain discontinuities (such as buildings and trees) through the process of terrain masking. Optimal sampling algorithms were developed to select the sparsest set of elevation posts that could be used to define the elevation model. The techniques developed in this research provide the basis for the automation of much of the digital elevation model extraction and editing process.
Degree
Ph.D.
Advisors
Bethel, Purdue University.
Subject Area
Civil engineering|Earth|Geography
Off-Campus Purdue Users:
To access this dissertation, please log in to our
proxy server.