A correspondence matching technique of dense checkerboard pattern for one-shot geometry acquisition

Abstract

This paper presents a correspondence matching technique for a dense checkerboard pattern displayed by a projector-camera system for one-shot geometry acquisition purposes. It does not require color coding or complicated spatial encoding techniques to encode the corresponding positions of corners, and can find corresponding positions for an incomplete checkerboard pattern. We introduce a combination of epipolar geometry and topology constraints in the checkerboard pattern to solve correspondence ambiguities. To verify the feasibility of the technique, we have created a prototype scanning system that can construct the 3D geometry of a scenario for each image frame. The results of our experiments show that the technique can identify correspondences for a checkerboard pattern displayed on discontinuous surfaces and can reconstruct 3D geometry structures in real time.