Accuracy comparision of 2D-3D point correspondence algorithms for pose estimation

• Main Author: Srinath, Hanumantha Rao
• Other Authors: Cheah Chien Chern
• Format: Theses and Dissertations
• Language: English
• Published: 2019
• Subjects: DRNTU::Engineering::Computer science and engineering::Computing methodologies::Image processing and computer vision
• Online Access: http://hdl.handle.net/10356/78644
• Institution: Nanyang Technological University

Pose estimation is one of the most basic steps when interfacing a camera with respect to the world. It has seen an increase in usage with the development of autonomy across various fields, mainly robotics. This thesis aims to compare the various 2d-3d point correspondence methods for pose estimation available to us using three different conditions: distance from camera, resolution and pattern used. There is no documentation available today that describes which pattern, distance and algorithm configuration gives us the best results possible. After getting various correspondences, a hand-eye calibration will be performed and then later used for testing which correspondence environment produces the most accurate results. Accuracy would be the metric used to calculate the results produced from the correspondence algorithms