PhD studentship Automated Asset Recognition using Machine Learning from Photogrammetry and LiDAR data

Duration – 3 years

Funding –The scholarship covers UCL registration fees and provides a stipend of between £16,000 and £18,000 per annum tax free. Travel expenses and research equipment will also be paid for.

Funding Body: Department of Civil, Environmental and Geomatic Engineering and Bentley Systems, Inc.

Vacancy Information

Applicants are sought for an exciting PhD studentship with University College London (UCL) and Bentley Systems, Inc. The project aims to enhance civil engineering industry capability in in the field of infrastructure asset management and construction monitoring as well as drive machine learning technology. The research will challenge the way construction projects are conducted in the UK and possibly world-wide. It can potentially alter the way progress monitoring, compliance checking and even payments are conducted in industry.

Studentship Description

Today photogrammetry and LiDAR both provide very effective ways to capture the geometry of existing assets. Both technologies can create datasets with hundreds of millions of points and triangles. While those datasets describe the overall geometry of a scene they are not capable of identifying individual objects in the scene, such as the engineering assets.

This research work is aimed at exploring the use of machine learning techniques to automatically identify assets in captured geometry data. The data can be captured using photogrammetry or LiDAR or a combination thereof. The data can be particularly easily captured from UAVs or even from handheld cameras. The costs of such a data acquisition are comparatively low, which makes it possible to repeat them in regular intervals, e.g. bi-weekly. This research project aims at automating the asset extraction from these data sets. The current manual process of delineating and tagging assets is time-consuming and prohibitively costly. As data gets captured more regularly and at increasing resolution the data volume is ever going to increase. This makes manual extraction even less desirable and demands automation. Machine learning techniques in combination with geometric processing provide a high potential for an automated solution.

This research work will also explore the additional potential and advantage that photogrammetry might offer over LiDAR, in that it combines both 3D geometry data and colour imagery. The additional information should provide a better level of identification.

Person Specification
Applicants should have a strong computing science, mathematics or engineering degree (first/upper second). Experience in machine learning, CNN, programming, data analytics and reality capture will be advantageous. Candidates should have excellent communication and presentation skills and will ideally have some relevant previous research and/or industry experience. Candidates should be highly motivated individuals with a want to apply academic research to engineering practice and a desire to gain industry experience as part of the studentship.

Eligibility
The award covers home/EU fees, stipend (minimum UK Research Council with London weighting) and may be used to support UK and EU nationals only.

Start Date

The post will be available immediately with the latest start date in September 2017