Full text: Proceedings of the Symposium "From Analytical to Digital" (Part 2)

  
INVESTIGATIONS OF DIFFERENT INTEREST OPERATORS FOR DTM-GENERATION 
BY EPIPOLAR LINE CORRELATION 
Johannes Piechel 
Universität Karlsruhe, Inst. f. Photogrammetrie u. Topographie 
7500 Karlsruhe, W.-Germany 
ABSTRACT 
A regular grid pattern for DTM-generation may not produce the most 
ideal points for manual or automatic measurement. The paper 
describes, and compares, techniques which produce more suitable points 
In the region of the original grid points. Even simple methods select 
DTM-points which are better defined and correlated than the original 
ones. It has been found that the best results are produced by 
maximizing grey level differences in the search area. 
1. INTRODUCTION 
The objective of the work described in this paper is the production of 
high-quality digital terrain models (DTM's) from stereo pairs of 
aerial photographs. A fast method of calculation is sought which 
produces reliable and accurate heights. This will be achieved by an 
off-line program which is currently under development. The program 
uses techniques of automatic correlation and is based on the 
DISC-system of Claus /4/. This takes full advantage of epipolar 
geometry, but has been greatly extended and modified. These 
extensions include: two-dimensional windows, hierarchical point 
densification, least squares adjustment, visual controls, and interest 
operators. 
The interest operators are functions used for finding suitable 
DTM-points, and it is these which are the subject of the current 
paper. Starting from a regular grid in the image space, many points 
are not suitable for measurement, due to a lack of structure within 
the window. Consequently, the window is shifted until the interest 
operator indicates that suitable detail is located near the center of 
the window. 
The following points should be noted: 
1) It would be possible to use the interest operator to accept or 
reject a raster point, but the current proposal uses the optimizing 
method. 
2) In contrast to other methods /2,3,6,9/ the position of the 
feature is not determined within the window, but is arranged to be 
close at the center. 
3) No attempt is made to determine if the feature has topographical 
significance. 
4) In using the interest operator to find a new feature close to the 
initial raster point, a search is made in only one direction. This 
direction is the same as that used to find the corresponding point 
in the stereo partner (the epipolar direction as explained below). 
AS a result the final set of DTM-points remains semiregular. 
- 564 -
	        
Waiting...

Note to user

Dear user,

In response to current developments in the web technology used by the Goobi viewer, the software no longer supports your browser.

Please use one of the following browsers to display this page correctly.

Thank you.