Intelligent Surveillance Trigger Challenge

About NSTIx OpTech Co-Creation

Context

Unmanned Aerial Vehicles (UAVs) offer National Security organisations the capability to provide surveillance from a stand-off position. These are typically mounted with electro optic sensors to track a person’s movements. This requires an operator to pilot the UAV and a further person to process the live imagery ensuring full surveillance coverage.

There are “follow me drones”, but they don’t fulfil the use case in our complicated scenario where staying discreet is necessary, so altitudes may be high, and the angle of imagery is not always favourable so a person’s face may not always be in view. Coupled with use cases in city location so buildings may obscure line of sight, movement from other people and vehicles increasing complexity, and the person under surveillance entering and exiting buildings. Currently, processing the live imagery / information is resource intensive and is open to human error, therefore it is highly desirable to integrate a level of intelligence or machine learning into a system that can provide a trigger when a person of interest comes into view.

The resulting surveillance trigger might be used to notify a human in the loop to verify this information, enable the UAV to continue its own surveillance task or to inform other teams to initiate other activities.

To inform future capability development of novel approaches to overcome the challenge of providing an intelligent surveillance trigger of a positive identification of a person of interest. Ideally this would lead to significantly reduced human interaction, noting human machine teaming will always be required for key decision making.

The surveillance trigger could be based off a database of information, for example facial recognition, a person’s gait, or other novel identification methods. The benefits of intelligent and unsupervised learning could also be utilised to improve reliability.

The Challenge

This Challenge will last for 12 weeks with an indicative budget of £60k (ex VAT) per solution provider. We are keen to encourage consortiums to form and provide joint proposal, as such a higher budget of up to £180k (ex VAT) will be available for consortia, with a total portfolio ceiling budget up to £540k (ex VAT).

The focus of the proposal should be on the unsupervised processing of surveillance information for a machine to intelligently provide a trigger or notification that a human in the loop or a UAV can act upon. The vehicle of delivery (i.e., UAV) is not the focus, though the technology developed will likely but not necessarily be used with a commercial UAV, so interoperability should be considered for future phases of work.

To further steer, several options could be considered:

• Positive identification could come from traditional image processing or deep learning techniques.
• Broader person identification techniques beyond facial recognition or other datasets beyond imagery.
• Utilising AI/ML algorithms to positively identify a person from available data sources.
• Edge processing to provide real-time information without the need to be cloud connected.
• The proposal should consider what data sets would be necessary to activate the trigger (be it captured imagery footage, open-source, synthetic etc). As this project is open to innovative methods.
• Thought should be given to future ethical and legal compliance implications.
• We will not be able to provide datasets, please consider in your proposal what training data could be used.

The outcome of this challenge should be a demonstrator, which could be a lab-based demo, a video showing the capability or a field-based demo, showing the concept of an intelligent surveillance trigger. This should not be UAV mounted at this stage, but consideration should be given to how this may be integrated in the future phases of work. Alongside this we expect a report detailing the developments, findings and future recommendations.

The proposal should also consider future phases of work, which could include a timeline, possible costs, and routes to exploitation leading to operational deployment.

Pre-competition Phase

We wish to have the greatest collaboration between various sectors and disciplines. We think that the following non-exhaustive list of expertise will be vital and key to achieve the best outcome collaboration between these disciplines is vital; AI/ML, deep learning, data science, computer vision, neural nets, novel sensors, future semiconductors, FPGAs, modelling, decision making, engineering.

To encourage cross collaboration between various sectors and disciplines, a pre-competition phase facilitating consortium building will be run across the dates below.

As part of this consortium building, there will also be an online forum held throughout this period to encourage enduring conversations for those unable to attend any of the events.

Failure to attend the Briefing Call or pre-competition events does not exclude solution providers from submitting a proposal by the deadline stated below.

Commercial Consideration

If this initial funded project demonstrates the concept to be feasible, and the OCCS team agree, there is the potential for multi-year, multi-million-pound funding leading to operational deployment. Therefore, it is encouraged that proposals view to de-risk the hardest components of this Challenge in the first phase with a view to a long-term collaboration.

This Challenge is open to sole innovators and to organisations of all types and sizes. This includes UK registered businesses of any size, research organisations, and research and technology organisations. Consortiums are encouraged where a sole lead organisation contracts with the Authority & flows the terms, conditions, IP & payments, to the consortium members.

Recording of the briefing call