Date of Award

January 2017

Document Type

Thesis

Degree Name

Master of Science (MS)

Department

Mechanical Engineering

First Advisor

Jeremiah Neubert

Abstract

This thesis presents the development of a novel method to generate an accurate, metric spatial database of a large multi storied building during construction. The algorithm uses the 3D CAD model of the building and the video of the structure captured by an Unmanned Aircraft System (UAS). The spatial database is then used to perform several inspection procedures such as, metric data analysis, spatial query for images, visualization through 3D textured model. The video is processed using a simultaneous localization and mapping (SLAM) system. SLAM generates a sparse 3D map of the environment. Our algorithm registers the 3D map with the 3D CAD model to generate the accurate metric spatial database. The user can click on the desired part of the CAD model for inspection and the image of that part will be shown by using the spatial indexing between the CAD model and the spatially distributed images. The image returned by the spatial query can be used to extract metric information. The spatial database is also used to generate a 3D textured model which provides a visual as-built documentation. The metric data calculation and textured model reconstruction methods have been compared to the state of the art Pix4D software (Latest Release (Version 3.1)). The proposed method has a mean squared error (MSE) of 31.9 cm2 and standard deviation of 4.28 cm where Pix4D had a higher MSE of 45.6 cm2 and standard deviation of 4.91 cm. Using statistical t-test and ANOVA tests we have shown that we are statistically 99% confident that the proposed algorithm has performed better than Pix4D.

Download

Share

COinS