View Item 

Creation of 3D models from large unstructured image and video datasets

Access Status
Open access
Authors
Hollick, J.
Moncrieff, Simon
Belton, David
Woods, Andrew J.
Hutchison, Andrew
Helmholz, Petra
Date
2013
Type
Conference Paper

Metadata
Show full item record
Citation
Hollick, J. and Moncrieff, S. and Belton, D. and Woods, A.J. and Hutchison, A. and Helmholz, P. 2013. Creation of 3D models from large unstructured image and video datasets, in Heipke, C. and Jacobsen, K. and Rottensteiner, F. and Sörgel, U. (ed), Proceedings of International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, XL-1/W1, May 21-24 2013, pp. 133-137. Hannover, Germany: ISPRS Archives.
Source Title
ISPRS Archives – Volume XL-1/W1, 2013
Source Conference
ISPRS Hannover Workshop 2013 High-Resolution Earth Imaging for Geospatial Information
DOI
Remarks

This article is published under the Open Access publishing model and distributed under the terms of the Creative Commons Attribution License http://creativecommons.org/licenses/by/3.0/. Please refer to the licence to obtain terms for any further reuse or distribution of this work.

URI
Collection
Abstract

Exploration of various places using low-cost camera solutions over decades without having a photogrammetric application in mind has resulted in large collections of images and videos that may have significant cultural value. The purpose of collecting this data is often to provide a log of events and therefore the data is often unstructured and of varying quality. Depending on the equipment used there may be approximate location data available for the images but the accuracy of this data may also be of varying quality. In this paper we present an approach that can deal with these conditions and process datasets of this type to produce 3D models. Results from processing the dataset collected during the discovery and subsequent exploration of the HMAS Sydney and HSK Kormoran wreck sites shows the potential of our approach. The results are promising and show that there is potential to retrieve significantly more information from many of these datasets than previously thought possible.