Abstract

In this dissertation, we propose data-driven approaches to reconstruct 3D models for indoor scenes which are represented in a structured way (e.g., a wall is represented by a planar surface and two rooms are connected via the wall). The structured representation of models is more application ready than dense representations (e.g., a point cloud), but poses additional challenges for reconstruction since extracting structures requires high-level understanding about geometries. To address this challenging problem, we explore two common structural regularities of indoor scenes: 1) most indoor structures consist of planar surfaces (planarity), and 2) structural surfaces (e.g., walls and floor) can be represented by a 2D floorplan as a top-down view projection (orthogonality). With breakthroughs in data capturing techniques, we develop automated systems to tackle structured modeling problems, namely piece-wise planar reconstruction and floorplan reconstruction, by learning shape priors (i.e., planarity and orthogonality) from data. With structured representations and production-level quality, the reconstructed models have an immediate impact on many industrial applications.

Committee Chair

Tao Ju

Committee Members

Ayan Chakrabarti, Yasutaka Furukawa, Brendan Juba, Ulugbek Kamilov,

Comments

Permanent URL: https://doi.org/7936/mxpa-n449

Degree

Doctor of Philosophy (PhD)

Author's Department

Computer Science & Engineering

Author's School

McKelvey School of Engineering

Document Type

Dissertation

Date of Award

Spring 5-15-2019

Language

English (en)

Author's ORCID

http://orcid.org/0000-0002-1628-3792

Download

Share

COinS