On the accuracy of point correspondence methods in three-dimensional measurement systems using fringe projection

Abstract

Different concepts of correspondence findings in contactless optical three-dimensional (3-D) measurement systems using fringe projection are analyzed concerning the accuracy of the 3-D point calculation. These concepts are different concerning the kind of performance of the triangulation procedure in order to calculate the resulting 3-D points and the use of geometric constraints versus second projection sequence. Triangulation may be alternatively performed between camera pixels and the phase origin of the projection, between one camera pixel in the prior camera and the image of the corresponding measured phase value in the second camera, or between the image points of certain raster phase values in the two observation cameras. Additionally, triangulation procedures can be distinguished concerning the use of two perpendicular projection directions of the fringes versus the use of geometric constraints, i.e., epipolar geometry instead of the second projection direction. Advantages and disadvantages of the different techniques are discussed. In addition, a theoretical analysis of the application of synthetic data has been simulated as well as experiments performed on real measurement data. Both simulations and real data experiments confirm the theoretical assumptions concerning the magnitudes of the random errors occurring in 3-D point determination.