Design considerations in the generation of digital terrain models by image matching for GIS databases
Abstract
There is much software in use today which generate DTMs by image matching, but the resulting DTMs usually need to be inspected and erroneous posts corrected manually. The manual correction process is very time-consuming. This research explored an Object-Space-Driven (OSD) method of generating DTMs by image matching. An OSD algorithm was designed, written, and tested on real and simulated images. The output DTMs were filtered using a 3 × 3 median filter. Experiments on selected windows of a model as well as on whole models, were conducted to demonstrate how match-window sizing, progressive shaping and image contrast affects the match function and the height accuracy. A terrain profile with a break-point was simulated and its images generated. The OSD algorithm was then used to determine the height at the break-point, firstly using a horizontal match-window and secondly an edge-shaped match-window. The results indicate that the shape of the match-window is a strong accuracy influencing factor. In another experiment, images of a simulated building profile were generated, and then the OSD algorithm used to reconstruct the building. The results led to the formulation of a multi-aggregation level DTM generation and processing method using either the Ll-Norm, or L2-Norm, or a median filter. Also a novel idea of using a small area as a training sample was hypothesized and tested. It was established that the use of a median filter is the most viable option because it gives better accuracy, and its implementation does not require manually measured posts. The present state of the OSD algorithm enhanced with progressive window shaping and multi-aggregation level processing by a median filter, are promising approaches. Considering the promising potential shown, it is recommended to take the current OSD algorithm and explore all possibilities of refining it further to achieve higher efficiency and accuracy.
Degree
Ph.D.
Advisors
Bethel, Purdue University.
Subject Area
Civil engineering|Geographic information science|Physical geography|Geography
Off-Campus Purdue Users:
To access this dissertation, please log in to our
proxy server.