Improving virtual representations of environments for usage in augmented reality

Abstract

In this thesis, I present a method to merge data captured from the real world in the form of a point cloud with a lowdetail virtual model of the same environment. The method first extracts the relevant structure from the captured point cloud and then uses this structure to generate a low-detail version of the point cloud. This low-detail version of the point cloud will then be fitted into the virtual model, using a combination of a custom global registration and ICP. The fitted point clouds will then be merged into a single point cloud, resulting in a version of the virtual model with added data from the real world.