AI-powered video analytics firm, Ipsotek has today launched a new tool that detects whether or not employees are adhering to COVID-related social distancing measures.
Proximity Detection analyses shape, appearance and trajectory to identify individuals anonymously -- rather than face recognition or mobile phone signal tracking, deploying tailored algorithms to analyse camera footage in real-time.
The solution is being rolled out as part of Ipsotek’s VISuite AI platform, which is powered by a GPU-enabled, deep learning engine. It supports highly customised object classification, detection and tracking capabilities, offering what it says is a reliable method for accurately determining the distance between moving objects or people in real-time.
The company stated: "The technology works by actively monitoring large networks of cameras and tracking the GPS coordinates of an object or person within the field of view in real-time using a unique and patented geospatial algorithm. This AI algorithm automatically learns the perspective of the scene and tracks an object’s shape, appearance, colours, speed, and trajectory, meaning it is able to precisely calculate the distance between objects and/or people, without using face recognition or an individual’s mobile phone signal."
The tool monitors the proximity of any individual within the range of the camera -- staff and or members of the public, and if social distancing rules are broken an alarm is raised and a dynamic dashboard updated.
Ipsotek says that if a spike in social distancing alerts is registered, an email is automatically sent to a designated person for further investigation, and can therefore be valuable in the case of contact tracing.
The company says the technology has been successfully trialled in the UK, Europe, the Middle East, and Singapore.
Printed Copy:
Would you also like to receive CIR Magazine in print?
Data Use:
We will also send you our free daily email newsletters and other relevant communications, which you can opt out of at any time. Thank you.
YOU MIGHT ALSO LIKE