Method and system for reconstructing a scene as a three-dimensional (3D) model

摘要

The method first reconstructs the scene as a three-dimensional (3D) model by acquiring a sequence of images of the scene using a camera. Next, a feature point match or feature line match is extracted from the sequence of images, the camera's optical center and 3D point variables are initialized using random values, and according to the availability of n constraints, the variable n Copies are generated. The n copies are projected to satisfy each of the n constraints. The n copies are then replaced with the average of these copies, and projection and replacement are repeated until convergence to provide a 3D model.