A modified structure from motion framework based on smartphone imaging for 3D reconstruction of arbitrary shapes

Open Access
Author:
Vidvans, Amey Narayan
Graduate Program:
Industrial Engineering
Degree:
Master of Science
Document Type:
Master Thesis
Date of Defense:
July 24, 2017
Committee Members:
  • Prof Saurabh Basu, Thesis Advisor
Keywords:
  • Reconstruction
  • Smartphone
  • Digital Image Correlation
  • Reverse Engineering
Abstract:
There is a need for scalable open source 3D reconstruction systems for reverse engineering and in other applications involving additive manufacturing. This is because most commercially available 3D reconstruction systems are capital and resource intensive. To address this, a novel reconstruction technique is presented in this thesis. The technique involves digital image correlation based characterization of surface speeds during rigid body rotational motion of the object of interest. Proof of concept of the same is shown using simulation and empirical characterization. Towards this, a stage is designed and fabricated in-house using poly-lactic acid (PLA) polymer with a Makerbot 3D printer. The stage is computer controlled using an Arduino controller. Dimensional errors are characterized between the reconstructed model and the original geometry. Herein, some commonly occurring shapes are characterized. Effect of Gaussian curvature on the accuracy of reconstruction are briefly explored. It is shown that highly accurate 3D reconstructions can be performed using scalable hardware involving smart phone based optical photography.