Finding geometric and structural information from 2D image frames

摘要

This paper describes the process of obtaining geometric and structural information, in the form of relational tables, from two-dimensional intensity image frames. The process is divided into two subprocesses: the heuristic edge follower and the structure finding algorith. The first subprocess searches an intensity image to locate and extract potential edges satisfying some criteria and fits linear equations to them using linear regressional analysis. The second subprocess finds valid intersections between these linear equations and hence obtains the structural relationship between these intersections. The outcome of the process is a relational table, containing all vertices found and the relationships between them, for the given intensity image. The process has been tested on a number of rigitd, planar-faced three-dimensional objects, producing accurate relational tables from the given views of the objects. The process is described here with reference to one such object and the performance in terms of processing time is reported.