Single View Computer Vision in Polyhedral World: Geometric Inference and Performance Characterization

 Mingzhou Song, Aiwen Guo and Robert M. Haralick

Abstract

An algorithm for making consistent 2-D to 3-D geometric inference in polyhedral world using one perspective line drawing is described.  Hypotheses are made on the internal angles of visible faces.  The normals to the face planes are then determined.  Valid normals lead to the reconstruction of the 3-D polyhedral world up to a scale factor.  The performance of the algorithm is verified by using covariance matrix propagation.  The experimental results show satisfactory performance.

The general propagation formulae for the covariance matrix of both input and output variables are also derived.