3D Indoor scanning & map building

Abstract

This research project aims to develop a 3D indoor scanning and map building application (algorithm) at the Hanoi University of Science and Technology (HUST). 3D technology is often applied in: transport, navigation, robots, 3D object scanning and more. This project focuses on gathering more information from the environment with a real-time cycle time for indoor environments. The goal is to create a 3D map on a computer, with all the obstacles that are in the environment.

A major part of this master's thesis was the programming of an algorithm that makes use of PCL (Point cloud library) software and a RGD-D hardware. The algorithm creates a point cloud with the depth image received from the RGB-D camera. This point cloud correlates to everything that is in the camera's field of view. Every point cloud is taken from a specific view-point in the real world. The algorithm combines these point clouds of the scene to one point cloud (registration).

The algorithm can combine two point clouds in ±450 ms and needs ±700 ms to update the 3D map. The scene is represented well but still contains small errors, especially relating to the alignment of the small objects in the scene.