5 questions on IT implementation of a Safety CV system

You want to get the benefits of an AI-powered computer vision (CV) safety system. You’ve defined some problems you want to solve. You’re excited about the prospect of working with AI and CV. But now you have to get your IT team (and the team managing your CCTV if that’s done separately) on board. Here are five questions you should be prepared for before you have that conversation.

Will we need extra bandwidth?

We’ve become so familiar with the idea of “cloud computing” that we might have forgotten that in the past, computers were “stand alone”, usually desktop PCs, processing and storing data locally. The software and information on that computer could only be processed and used on that computer. The cloud made it possible to access applications and information from any computer, including mobile devices. But streaming all of your CCTV streams across the Internet would be a headache for your IT team due to bandwidth and security constraints. “Edge computing” helps to overcome both problems.

Some security cameras can consume as little as 5 Kbps in “steady-state”. The average bandwidth consumption of an IP camera is 1-2 Mbps. Protex does not require the highest quality stream or the highest frame rate available, so its bandwidth consumption is on the lower end. Live streaming only occurs between the cctv network and the edge device so this is where that bandwidth would be consumed. The edge device reviews the video and identifies selected clips (according to the instructions you’ve provided) and only these are forwarded to the management system in the cloud. Protex does not require a significant amount of bandwidth on the internet gateway network as a result.

What about the security implications?

If the camera network is connected to the edge device, and the edge device is connected to the Internet, isn’t there a possibility of a security breach? Scare stories in the media about hackers being able to view video from inside hospitals, schools and police departments add to this impression. 

A good CCTV supplier will have considered security before the introduction of CV, but this is a good time to review arrangements. Careful configuration of the camera network, and of the edge device, will help ensure that your CCTV network is secure.

The internet gateway should have a firewall with inbound rules to prevent unauthorised access and outbound rules to tell the Protex edge device what it’s allowed to communicate with in the internet (tell your IT team that Protex can provide these rules for you). Best practice is for the camera network is physically separate from the corporate network via 2 Network Interface Cards (NICs), but in some cases your CCTV network can also have an internet gateway.

 Access to live camera streams should be controlled – both physically (with locked rooms for any viewing screens and network recorders) and through software (with user authentication). The link between the edge device and the camera network will be a managed ethernet connection. The CCTV system should be set up with a read-only account to be used by the edge device. The edge device then selects relevant portions of video, blurs the faces and encrypts the clips before they are forwarded to the management system in the cloud.

What about the data protection implications?

The data protection implications of your existing CCTV network should already have been considered, balancing the legitimate interests of the organisation with the rights to privacy of employees. Some countries require a specific data protection impact assessment (DPIA) for CCTV networks, but these are valuable tools even where not mandatory. Review the changes that the CV will make to the use of the cameras, and inform anyone affected. See our article on data protection and computer vision for more advice.

Even with blurred faces in the images, the management system that reviews and catalogues the video clips selected by the edge device should have suitable access arrangements. Be clear about who needs access to the system, and insist on a Single Sign On (SSO). This will ensure that when someone leaves an organisation, they lose access to the CV management system.

Do we need extra cameras?

If you have a clear idea of the problems you want to investigate using CV, you will know whether the existing camera network covers the areas required. Minimising the number of extra cameras to be installed will help meet the data protection principle of minimisation as well as keeping costs down. It might be possible to reposition existing cameras, so have a clear picture of what you need before you approach IT or your CCTV contractors. 

A helpful approach is to start small, and build from there. If you are running a smaller pilot first, make sure the IT and CCTV teams have a view of where the project might go eventually. How easy is it to add cameras and new rules to the system you want to buy? 

Anything else we need to know?

What training and ongoing support will the EHS team members using the system need? Make sure that your CV partner is committed to supporting you in the long term, as expecting IT to run a support line for a new product used by a small number of people in EHS could be a deal breaker.

‍

‍