Performance prediction of point-based 3D measurement systems

Point-based three-dimensional volumetric measurement systems are defined as multi-view vision systems which reconstruct a three-dimensional scene by first identifying key points on the views and then performing the reconstruction. Examples of these are defocusing digital particle image velocimetry (DDPIV) (Pereira et al 2000 Exp. Fluids 29 S78–84) and 3D particle tracking velocimetry (3DPTV) (Papantoniou and Maas 1990 5th Int. Symp. on the Application of Laser Techniques in Fluid Mechanics) which reconstruct clouds of flow tracers in order to estimate flow velocities. The reconstruction algorithms in these systems are variations of an epipolar line search. This paper presents a generalized error analysis of such methods, both in reconstruction precision (error in the reconstructed scene) and reconstruction quality (number of ambiguities or ‘ghosts’ produced).