Active Computing & AI Engineering

Project Haystack

In plain English

AI plain-English summary

Police and security teams will be able to locate a specific person in a crowd without ever analysing their face. Currently, finding a Person of Interest (PoI) relies on operators manually scanning multiple video feeds, cross-referencing sightings, and using local knowledge—a slow, labour-intensive process with no standard system. Project Haystack aims to replace that manual trawl with technology that processes video from many cameras while deliberately avoiding facial recognition, sidestepping the privacy concerns that have made such surveillance controversial. If the prototype succeeds, it could give authorities an ethically compliant tool for public-space searches. The same approach could also help theme park staff find a lost child without broadcasting their image or requiring a face scan. The project is not about fundamental science; it is a practical engineering challenge to build and test a detection system that works in real-world conditions while respecting privacy. Success would mean faster, less intrusive searches in settings where every minute counts.

View original technical description
Police and security teams in public spaces need an ethically compliant solution to locate Persons of Interest. Finding PoIs presents a significant challenge for authorities due to the resource-intensive and unstructured systems and processes currently used. Locating PoIs is a manual task carried out by operators trawling different video feeds, assessing sightings and reports, and factoring in as much other information as practical, such as local knowledge. The aim of the Project Haystack approach is to use technology that can process video feeds from a wide range of camera sources without any use of facial analysis and the inherent privacy concerns. This also creates potential for use outside of law enforcement (e.g. to help staff find lost children in theme parks). Project Haystack's objective is to develop, deliver and test a revolutionary Person of Interest detection prototype which can ethically and compliantly identify a Person of Interest while maintaining privacy.

View the original record at the funder ↗

Related Research

Grants with similar aims, by meaning.

Protecting public-facing professionals and their dependents online (3PO)
Horizon Scanning Through Automated Information Prioritisation
FinePoint - a study into the use of phyiscal optical correlation applied to live video to generate camera tracking information and the provision of high precision localised augmented reality.
Robust Methods for Monitoring & Understanding People in Public Spaces (REASON)
Proof of Claim (PoC): A human-in-the-loop AI based solution for semantic information integrity

Original classification

Collaborative R&D

Plain English summaries and category classifications on this site are generated by AI and may not perfectly reflect the original research.