Extraction of photogrammetric linear features using active contours
Abstract
The focus of modern photogrammetric research is to interpret and extract information from digital imagery. A major activity in mapping and database construction is the delineation of linear features (one dimensional features, not necessarily straight lines). Examples of such features are roads, trails, rivers, coast lines, shore lines, area feature perimeters etc. This activity is very labor intensive and proposed machine algorithms are, to date, imperfect and generally usable only in research environments, not in production. Although automated methods of extraction may appear faster, the requisite editing and fixing causes the manual method to remain in effect faster and more accurate. A semi-automated assist is needed to reduce the manual efforts of image analysts. This research proposes a method for semi-automated reconstruction of 3D photogrammetric linear features, from overlapping aerial photographs, with the option of exploiting redundant DEM data. The following procedures are applied to extract the 3D linear features from the aerial images semi-automatically. (1) Using the active contour model with iterative variational approach, a sub-pixel accuracy result of 2D linear feature extraction can be obtained with simultaneous processing of two images. (2) The normalized photo coordinate system makes efficient the corresponding points matching process. (3) Space intersection enables the extracting of the 3D linear feature. (4) The weighted mean is a simple way to adjust the extracted data to an existing DEM data set.
Degree
Ph.D.
Advisors
Bethel, Purdue University.
Subject Area
Civil engineering
Off-Campus Purdue Users:
To access this dissertation, please log in to our
proxy server.