Automated Stereophotogrammetry.
Abstract
Using a pattern recognition approach, we have developed a fast program for finding characteristic points and calculating disparities in pairs of photographs. The procedure derives oriented edge-vector graphs from the photos, locates singular points (nodes) in the graphs, and then tests a small window about these points for match. In common with other window-matching approaches, the procedure uses a coarse-to-fine analysis, where the averaged blocks in the coarse resolutions are staggered to simulate a hexagonal array. This enables use to take advantage of the unique topological properties of such arrays. Under a slow image processing environment (GIPSY on VAX 11/785) a dense network of matched points was obtained at the rate of 5 pts/s, compared to 15 pts/s at ETL on a CYBER-730. The ETL procedure required interactive priming with a certain number of initial matches. Keywords: Stereophotogrammetry Pattern recognition, Topological method, Hexagonal scan array.
Document Details
- Document Type
- Technical Report
- Publication Date
- Dec 10, 1987
- Accession Number
- ADA189617
Entities
People
- Morton Nadler
Organizations
- Virginia Tech