Video Surveillance Architecture at the Edge

EasyChair Preprint 9362

Abstract

This paper examines the use of digital video in public safety and surveillance systems; traditionally these video recordings are used by law enforcement to review events retrospectively, and for evidential purposes in the pursuance of criminal prosecution. We also examine how, due to the proliferation of cameras around cities, human operators are challenged to monitor these data feeds in real-time, and how the emergence of AI and computer vision solutions can process this data. Computer vision can enable the move from a purely reactive to a predictive, real-time analysis platform. As camera numbers and the resolution and framerate of cameras grow, existing network infrastructure frequently causes challenges to the provision of low latency, high bandwidth networking to private or public cloud infrastructure for evidential storage. These technical challenges can provide issues for law enforcement providing a data chain of custody to ensure its admissibility during court proceedings. Emerging technologies offer solutions to overcome these challenges: the use of emerging edge compute capabilities, including the use of on-camera and mobile edge compute nodes to provide compute capabilities closer to the data source and new software paradigms, including CI/CD methodologies, and the use of micro-services and containerization to manage and deliver applications across the portfolio of devices, at the edge of the network.

Keyphrases: ANPR, Automatic Number Plate Recognition, CCTV, Cloud, Edge Computing, IoT, computer vision, edge, edge compute, image processing, surveillance, video analysis

@booklet{EasyChair:9362,
  author    = {Jeff McCann and Liam Quinn and Sean McGrath and Colin Flanagan},
  title     = {Video Surveillance Architecture at the Edge},
  howpublished = {EasyChair Preprint 9362},
  year      = {EasyChair, 2022}}