METHOD AND SYSTEM FOR RECONSTRUCTING A SCENE AS A THREE-DIMENSIONAL (3D) MODEL

摘要

One method reconstructs a scene as a three-dimensional (3D) model by first capturing an image sequence of the scene with a camera. Subsequently, dot hits or line matches of features are extracted from the image sequence; Variables for optical centers and 3D points of cameras are initialized using random values, and n copies of the variables are created depending on an availability of n conditions. The n copies are projected to meet each of the n conditions. Thereafter, the n copies are replaced with averages of the copies, and the projecting and replacement are repeated until convergence to provide the 3D model.