This project is funded by: DFE CAST award in collaboration with MetroCCTV
Subject: Computer Science and Informatics
Key dates
- Submission deadline: Friday 5 February 2021
- Interview Date: 25 March 2021
- Preferred student start date: mid September 2021
Summary
Established in Cookstown in 2012, Metro Surveillance Group Limited (Metro) specialises in CCTV design, installation, maintenance and monitoring across the UK and Ireland, with a focus on the petroleum station sector. Each of these fuel stations are monitored by CCTV cameras, primarily looking at two key aspects. The first is fulfilling government-driven requirements such as verifying that the correct (legal) fuel is being purchased, and that it is being purchased by an adult. The second is monitoring theft, which often includes vehicles with changed or modified number plates, or reoffending suspects. Currently, this is a very labour intensive process where humans in a control room watch several cameras simultaneously, initiating fuel pumps for use and identify potential theft.
The aim of this project is to automate this process via both vehicle and person identification and re-identification using the CCTV data currently available from fuel stations. Person and vehicle identification and re-identification have recently become major topics in the fields of computer vision and image processing and an essential component in the video surveillance system. Surveillance systems are no longer just one camera, but multi-camera systems that require significant effort to match the appearance of objects across a number of non-overlapping camera views at a range of angles.
This is a challenging problem due to occlusions, background activity, viewpoint change, weather conditions etc. Many approaches to re-identification are based on the use of images (rather than video) and focus on extracting discriminative features or metric learning. However, using only static image data is limiting due to visual ambiguity as well as the lack of spatio-temporal data. Despite recent remarkable progress, re-identification methods still suffer from weak feature representation and the ability to accuracy identify the salient object. To mitigate these issues, we build on our recent work using patch selection techniques and deep learning techniques. Many approaches focus only on person re-ID, whereas we aim to extend these to accurate person and vehicle re-ID.
The project will have two main aspects. Firstly, we will use deep networks such as a Deep Decompositional Network (DNN) to parse CCTV images into semantic regions, combined with a range of colour matching and background removal techniques. Then, building on the ongoing research of Coleman and Kerr we will develop novel approaches for spatio-temporal feature extraction algorithms for use with video data. Such techniques will underpin new approach of spatio-temporal saliency detection in video surveillance. Initially benchmarking performance evaluation will be conducted using the well-known Market- 1501, DukeMTMC-reID and CUHK03 datasets prior to real-world evaluation using MetroCCTV’s testbeds (such data are a valuable contribution to the project).
The successful candidate will have the opportunity to work at the cutting edge of video surveillance algorithmic development alongside a well-established CCTV company. Additionally, development of these novel video surveillance algorithms will result in high quality conference and journal papers. Opportunities will be provided to attend and present the work at IEEE international conferences and also attend trade shows to understand the competitive nature of this research domain.
Essential criteria
- To hold, or expect to achieve by 15 August, an Upper Second Class Honours (2:1) Degree or equivalent from a UK institution (or overseas award deemed to be equivalent via UK NARIC) in a related or cognate field.
- Research proposal of 1500 words detailing aims, objectives, milestones and methodology of the project
- A demonstrable interest in the research area associated with the studentship
Desirable Criteria
If the University receives a large number of applicants for the project, the following desirable criteria may be applied to shortlist applicants for interview.
- First Class Honours (1st) Degree
- Masters at 70%
Continue with Facebook
