Description

Digital depth cameras can be used to construct 3D point clouds of building interiors. That is, point clouds are computed from image data and depth image data. The feasibility of these point clouds as being raw construction material for 2D/3D models is of interest. At minimum, 2D floor plans are extracted, at maximum, the feasibility towards creating 3D models in standard industry formats is reviewed.

Objectives and Methodology

Intel® RealSense™ depth camera D435i, i.e., the combination of an IMU, RGB stereo camera, and a depth camera, is connected onto a laptop
and utilized by the student to scan an indoor environment. ROS libs, ORB-SLAM2, or other suitable open source code is utilized for data registration. The results may be compared against a terrestrial laser scanning point clouds to obtain rigorous accuracy estimates. The student can also explore drone-mounting options, if desired. A list of different visual SLAM open source programs:  https://github.com/tzutalin/awesome-visual-slam

Further reading