Applications of artificial intelligence to digital photogrammetry
Abstract
A great deal of interest has arisen lately in the application of expert systems to problems in which computer solutions were previously inapplicable. The aim of this research was to explore the application of expert systems to digital photogrammetry, specifically to photogrammetric triangulation, feature extraction, and photogrammetric problem solving. In 1987, prototype expert systems were developed for doing system startup, interior orientation, and relative orientation in the mensuration stage. The system explored means of performing diagnostics during the process. In the area of feature extraction, the relationship of metric uncertainty to symbolic uncertainty was the topic of research. Error propagation through the Dempster-Shafer formalism for representing evidence was performed in order to find the variance in the calculated belief values due to errors in measurements made to gather the initial evidence needed to begin labeling of observed image features with features in an object model. In Photogrammetric problem solving, an expert system is under continuous development which seeks to solve photogrammetric problems using mathematical reasoning. The key to the approach used is the representation of knowledge directly in the form of equations, rather than in the form of if-then rules. Then each variable in the equations is treated as a goal to be solved. Once the solution set of equations able to solve the problem is determined, the set is submitted to the Vaxima (MACSYMA) expert system for solution. Development of a rule-based part to deal with the heuristic knowledge needed in solving the equations, such as choosing optimum equations, and selecting suitable approximations, has been described.
Degree
Ph.D.
Advisors
Mikhail, Purdue University.
Subject Area
Civil engineering|Artificial intelligence
Off-Campus Purdue Users:
To access this dissertation, please log in to our
proxy server.