Ceiling visual odometry for SLAM

• Main Author: Chang, Poo Hee.
• Other Authors: Yong Chuen-Tze, Mark
• Format: Final Year Project
• Language: English
• Published: 2012
• Subjects: DRNTU::Engineering::Computer science and engineering::Computing methodologies::Image processing and computer vision
• Online Access: http://hdl.handle.net/10356/49159
• Institution: Nanyang Technological University

SLAM implementation in an indoor environment relies on odometry reading. The motivation of the project is to explore the approach on Ceiling Visual Odometry. This report describes the process and methodology to obtain practical methods for Ceiling Visual Odometry. A consumer grade wide angle camera and depth sensor were used. Compared to conventional forward facing cameras common in visual odometry methods, the ceiling approach was less explored on. Ceiling visual odometry offers its advantages as it is less disturbed by dynamic environment, e.g. movement of people. It also simplify the scan matching dimension to one that only deal with rotation, affine transform without scale changes. Fourier-Mellin Transform, Optical flow and SURF were implemented and tested on the data sets. Accurate and fast computations were emphasized. The results obtained were shown to be encouraging.