Abstract
This dissertation explores an alternative to traditional fiducial markers where geometric
information is inferred from the observed position of 3D points seen in an image. We offer an alternative approach which enables geometric inference based on the relative orientation
of markers in an image. We present markers fabricated from microlenses whose appearance
changes depending on the marker's orientation relative to the camera. First, we show how
to manufacture and calibrate chromo-coding lenticular arrays to create a known relationship
between the observed hue and orientation of the array. Second, we use 2 small chromo-coding lenticular arrays to estimate the pose of an object. Third, we use 3 large chromo-coding lenticular arrays to calibrate a camera with a single image. Finally, we create another type of fiducial marker from lenslet arrays that encode orientation with discrete black and white appearances. Collectively, these approaches oer new opportunities for pose estimation and camera calibration that are relevant for robotics, virtual reality, and augmented reality.
Committee Chair
Robert Pless
Committee Members
Bryan Clair, Yasutaka Furukawa, Viktor Gruev, Tao Ju, Joseph O'Sullivan
Degree
Doctor of Philosophy (PhD)
Author's Department
Computer Science & Engineering
Document Type
Dissertation
Date of Award
Summer 8-15-2016
Language
English (en)
DOI
https://doi.org/10.7936/K74F1P4D
Recommended Citation
Schillebeeckx, Ian, "Geometric Inference with Microlens Arrays" (2016). McKelvey School of Engineering Theses & Dissertations. 192.
The definitive version is available at https://doi.org/10.7936/K74F1P4D
Comments
Permanent URL: https://doi.org/10.7936/K74F1P4D