This report aims to describe the technical details behind the development of a 3D reconstruction system for face recognition. Initially the methods used for constructing a 3D model of the face from several cameras are discussed followed by some consideration of the pitfalls and problems behind some of the techniques.
The reconstruction system uses the Direct Linear Transform (DLT) method for both calibration and triangulation along with the Gabor Jet correspondence measure and Voronoi cell propagation matching strategy for stereo correlation. We also discuss the accuracy of the DLT method and assess its suitability for use in model construction for face recognition.
The final section of the report shows some example reconstructions using the described methods in combination with a six camera capture rig with a random light pattern projector and suggests how some of the problems currently being encountered could be solved.