Computer Vision – Utilizing Public 5G Networks for Multi-Access Edge (MEC) Architectures

In many countries, regulations requiring worker protections such as Personal Protective Equipment (PPE) are published by organizations such as the Occupational Health and Safety Administration (OSHA) in the US. These regulations commonly require employers to do everything within their power to ensure that PPE is not only available but that it is used at all times in dangerous areas.

The following diagram outlines a solution that involves 5G-capable cameras positioned above the entryways to dangerous areas. These consist of magnetically locked doors that require a worker to scan their badge to unlock the door. The cameras are sending RTSP streams back to GPU-accelerated instances in the AWS Wavelength Zone in the same city, thus giving them a very low RTT – 1 ms in this case:

Figure 10.9 – 5G camera applying a CV ML model for PPE detection

The model makes an inference based on what the camera sees regarding whether the worker who just tried to badge in is wearing a hard hat or other PPE appropriate to that area. If this is not true, the door will not unlock even though the user’s badge has access to that area otherwise.

The vendor of the building security system owns this application and sells this as a feature to the local factory owners. Because they use 5G-capable cameras and AWS Wavelength, integration is a breeze. There is no need to install local application servers in the factories. Even better, there is no need to deal with the local Wi-Fi or firewalls – both of which are likely inadequate for the task given that these are not office campuses or data centers.

V2X

One of the primary V2X use cases being rolled out now is happening in response to a set of laws passed in both the EU and UK in 2022. As of July 6, 2022, all new cars sold must be fitted with Intelligent Speed Assist (ISA) devices, Autonomous Emergency Braking (AEB) systems, and data loggers that capture telemetry for analysis by insurers and investigators.