Photogrammetry for Fine-Scale 3D Model Acquisition
Research Institute Reference Number: EMRC-TC-2018-1-PhD
The focus of this project is on the photogrammetric acquisition of 3D virtual objects. Potential for the exploitation of a-priori knowledge of lighting conditions and the use of calibration procedures will be explored to enhance acquisition algorithms. The project ambitions support the digital preservation of cultural heritage from museum collections.
Specific Requirements of the Project
• Degree in Electronic Engineering, Computer Science or equivalent.
• Experience in signal processing and/or image processing techniques.
• Computer programming expertise.
Project Aims and Objectives
This focus of this project is on the photogrammetric acquisition of 3D virtual for cultural heritage.
This project will extend the collaborative efforts of an international team of researchers, summarised at http://virtualcuneiform.org, that have an ambition to support virtual access to cuneiform tablets and to reconstruct cuneiform tablets by joining virtual fragments together. The techniques developed in this project will have relevance to the virtual cuneiform project and also to the acquisition of cultural artefacts more generally. The research proposed will have significant impact by supporting worldwide access to virtual 3D cultural artefacts which would otherwise be unavailable or at-risk.
A low-cost 3D acquisition system using photogrammetric reconstruction has been developed. This prototype system works well for the purpose of matching fragments of broken cuneiform tablets. However, the third-party open-source photogrammetric reconstruction algorithms currently used are designed for general-purpose applications and do not take advantage of the characteristics of inscribed or engraved artefacts such as cuneiform tablets. Also, they do not cater for situations where conditions such as lighting and camera positioning can be, at least partially, controlled and where calibration processes are possible. As a consequence, the accuracy of the 3D models produced is not as precise as the visual perception created by the photographic texturing suggests. A closer estimate to the actual shape of an object could be obtained by exploiting a-priori knowledge of the characteristics of the impressions and of the homogenous surface texture regions. Also, lighting conditions and material properties could be more accurately modelled in order to differentiate features that are more likely to be caused by the object geometry as opposed to those caused by surface colouring.
Approaches to improve the 3D model accuracy will be investigated. For example, the modification of the point-cloud reconstruction and surface meshing algorithms by exploiting a-priori knowledge of the properties of clay cuneiform tablets to prevent smoothing in key areas where abrupt colouring change is apparent (indicating a likely sharp edge). Another possible approach is the introduction of a post-processing stage that modifies the model based on characteristic features in the texturing image, i.e. a texture-informed 3D sharpening filter. Both would be informed by lighting and calibration information. Novel approaches will be explored and new 3D processing and reconstruction algorithms developed.
Practical development, demonstration and validation of the algorithms will be key activities in this project. Algorithm development will require practical general programming skills and will provide an opportunity to learn GPU processing, for example using OpenGL or CUDA.
Project is open to: Home/EU and overseas
Informal enquiries can be made to
Dr Tim Collins, Tel: +44 (0) 161 247 1658, E-mail: [Email Address Removed]