| Project Detail |
Surveillance cameras are widely used by governments, businesses, and ordinary citizens to monitor private and public areas to deter crime and identify offenders. As of 2021, it was estimated that there are over 1 billion surveillance cameras worldwide.
However, there are great concerns that continuous monitoring severely restricts peoples privacy and freedom of movement. There have been reports of abuse of surveillance footage
including by businesses and government institutions. Regulators, civil rights groups and citizens are thus rightly concerned about infringements on freedoms of speech and association, especially in this era of AI when monitoring, tracking, and profiling techniques have been made easier, cheaper and more accurate.
An initial step in video scene analysis is pose estimation which enables monitoring the flux of people and detecting criminal actions or intent. Considering the privacy concerns of continuously recording citizens and automated image analysis, this project intends to develop a novel camera system that captures just enough detail required for pose estimation but excludes identifying information. Existing software methods of deidentifying videos require a clear video to be taken first, and so do not entirely solve this problem. Proposed experimental hardware methods that avoid this issue using lenses have bulky optics and a limited set of parameters.
With MetaCam we will implement the optimized optical devices (metalenses) that allow ultimate control of all the light degrees of freedom. In this project, we will design and fabricate the metalenses and build an optical table prototype of the camera to demonstrate its secure privacy-preserving performance. Supported by business developers we will examine the
business potential, and map and engage the stakeholders later required to get from a proof- of- concept to a market-ready product. |
