Postgrad LIVE! Study Fairs

Bristol

European Molecular Biology Laboratory (Heidelberg) Featured PhD Programmes
Engineering and Physical Sciences Research Council Featured PhD Programmes
University of Glasgow Featured PhD Programmes
Anglia Ruskin University Featured PhD Programmes
Karlsruhe Institute of Technology Featured PhD Programmes

Deep learning for Segmentation and Classification of 3D Urban Surface Models

  • Full or part time
  • Application Deadline
    Friday, May 17, 2019
  • Funded PhD Project (European/UK Students Only)
    Funded PhD Project (European/UK Students Only)

Project Description

The University of Sheffield invites applications from outstanding candidates for this PhD scholarship opportunity.

Humanity is now an urban species, as such our cities are primary support mechanism for society, however the current understanding of the built environment, its energy and material use across cities is poor. This understanding is needed if we are to put in place measures to reduce our impact upon the planet.

In recent years the use of high-sampling density LiDAR data and photogrammetry data to create 3D urban surface maps to monitor urban infrastructure has grown substantially. These 3D surface maps have been shown to have potential use in a wide range of applications such as architecture, civil engineering, virtual and augmented reality, and robotics. However, while LiDAR and photogrammetry systems provide a fast and accurate method for capturing spatial data, the subsequent semantic labelling of data would require an unfeasible amount of man power to be done manually. Thus, the problem of automatic labelling of 3D urban data to associate each 3D point with a semantic class label, such as “house”, “tree” and etc, has gained momentum in the computer vision community.

Automatic segmentation and labelling of urban point cloud data is challenging due to a number of issues. First, high-end laser scanning devices output millions of data points per second, and therefore the methods need to be efficient to cope with the sheer volume of the urban scene datasets. Second, point cloud sub-regions corresponding to individual objects are imbalanced, varying from sparse representations of distant objects to dense clouds of nearby objects, and incomplete (only one side of objects is scanned by LiDAR). Third, for accurate object recognition a sufficiently large labelled training data set is needed to train these classification models.

This project will involve the use of a sensing platform mounted on a vehicle that creates high resolution, mutli-spectral 3D urban surface maps. The mobile sensing platform will collect and synchronise visual, thermal, LiDAR, GPS and inertia measurement data for the static reconstruction of 3D surface maps. These 3D surface maps will be converted into 3D voxels before rule-based segmentation methods will be used to label easily discernible objects, such as the ground surface. The remaining unlabelled points will be classified by a supervised machine learning method, such as a convolutional neural network classifier, so that automatic characterisation with minimal manual annotation can occur.

Linking to the Sheffield Urban Flows Observatory, this project will capture information about what the city is made of and ultimately improve be utilised to inform retrofit needs in the city.

REFERENCES
Babahajiani, P., Fan, L., Kämäräinen, J.-K., Gabbouj, M., 2017. Urban 3D segmentation and modelling from street view images and LiDAR point clouds. Machine Vision and Applications 28, 679–694.

Bell, S., Upchurch, P., Snavely, N., Bala, K., 2014. Material Recognition in the Wild with the Materials in Context Database. arXiv:1412.0623 [cs].

Cho, Y., Bianchi-Berthouze, N., Marquardt, N., Julier, S.J., 2018. Deep Thermal Imaging: Proximate Material Type Recognition in the Wild through Deep Learning of Spatial Surface Temperature Patterns. arXiv:1803.02310 [cond-mat] 1–13.

Kumar, S., Deshpande, A., Ho, S.S., Ku, J.S., Sarma, S.E., 2016. Urban Street Lighting Infrastructure Monitoring Using a Mobile Sensor Platform. IEEE Sensors Journal 16, 4981–4994.

REQUIREMENTS
Candidates are expected to at least hold the equivalent of a first class or upper second class degree in relevant numerate discipline, ideally computer sciences, but also mathematics or engineering. Programming skills are essential and previous experience with machine learning or machine vision methods would be beneficial. Additionally, experience with reality capture software/hardware/methods would be beneficial.

Funding Notes

Students receive fees and stipend (at the standard RCUK rate).
GBP-British Pound.

How good is research at University of Sheffield in Civil and Construction Engineering?

FTE Category A staff submitted: 34.80

Research output data provided by the Research Excellence Framework (REF)

Click here to see the results for all UK universities

Email Now

Insert previous message below for editing? 
You haven’t included a message. Providing a specific message means universities will take your enquiry more seriously and helps them provide the information you need.
Why not add a message here
* required field
Send a copy to me for my own records.

Your enquiry has been emailed successfully





FindAPhD. Copyright 2005-2019
All rights reserved.