POLARBEAR - Pattern Of Life ANPR Behaviour Extraction Analysis and Recognition
Start date: 1 October 2014
End date: 31 March 2016
The organised nature of some crimes including drug trafficking and terrorism can be difficult to identify. One approach is to identify offenders travelling in 'convoy'. UK police forces have been using ANPR (automatic number plate recognition) data collected by traffic cameras to perform convoy analysis; however, this is typically done manually with prior information of one known vehicle. This project aims to develop a distributed processing system for large-scale ANPR convoy analysis and other criminal behaviours that currently go undetected. Our solution will enable criminal investigations to be more effective, accurate and resource-efficient. By applying automated data mining techniques to ANPR data, our solution will enable criminal investigations to focus on medium and high priority issues such organised crime as opposed to minor traffic infringements, and thus can help to justify the growing use of ANPR to the public at large.
The project will be conducted by a consortium formed by researchers from the University of Surrey and two industrial partners, with input from ANPR stakeholders such as UK police forces and ANPR solution providers. The University of Surrey team will be contributing mainly to the ANPR convoy analysis part which includes visual analytic approaches used in HMI (human-machine interface) to assist ANPR operators to interpret the data and results more easily and more effectively. Human behaviour analysis and machine learning will be two key elements of the developed ANPR convoy analysis techniques. The possibility of making use of other data sources other than ANPR will be investigated to improve both the accuracy and efficiency of the convoy analysis results. The University of Surrey team will also work with the two industrial partners to generalise the convoy analysis work to other types of suspicious behaviours and activities that can be detected from ANPR data, and to verify the scalability of the developed algorithms to nationwide data sets.
The project will involve use of anonymised police data and other public data sources such as Crime Map data on police.uk, public data from the Highways Agency, road topology, routing and labelling data on Google Maps and OpenStreetMap, which allows more effective analysis of the ANPR data without violating privacy of vehicles and their drivers.