Can AI cameras provide robust pedestrian detection for your site?

The safety of your staff is your top priority, but you also have plenty of operational challenges on your plate. That’s where SiteZone comes in. We’re proven to eliminate plant pedestrian collisions on all types of waste and recycling sites.

So, when it comes to comparing the different options available for your plant pedestrian safety, we want to make sure you have the facts to compare and understand the pros and cons of all options. Here we unpick some of the differences between human form detection cameras (or AI cameras as they’re often wrongly called!) and tag-based proximity warning systems.

Not seeing what’s right under your nose (or bucket!)

It’s time to be frank when we discuss this point. There’s a fundamental flaw in the ability of cameras to provide 360 detection. A human form detection camera is unable to see in front of a loading shovel bucket and when a load is being carried on telehandlers and forklifts this can often obscure the AI camera’s view. It’s also difficult to detect anything in close proximity (under 0.5m). So why would you buy a system that doesn’t give you total protection around a vehicle?

The front bucket of a loading shovel is where many of the fatalities on waste and recycling sites have occurred, and was highlighted by HSE in a safety notice in 2021. Just look at the image below to the scale of the blind spot:

Your SiteZone tag will be detected anywhere within the zone – this includes in front of the bucket and close to the machine.

The obstacle is in the way

Look round your site. I bet you can see obstructions everywhere. Piles of waste, stacks of pallets, cabins, walls, entrances, other vehicles. The list goes on and on. On waste and recycling sites in particular this can be one of the greatest risks. Not seeing someone before it’s too late. And a camera relying on field of vision will also not detect a person until it’s too late.

Your SiteZone tag will be detected anywhere within the zone – this includes around blind corners, waste piles, etc.

Who owns the data?

The use of human form detection cameras raises privacy concerns among operators, which can lead to resistance or misuse of the technology. Ensuring compliance with data protection regulations is crucial to avoid legal issues. Do you have permission to film everyone on your site, including visitors?

SiteZone data is anonymised. The tag ID is the unique identifier, and you can control your data (i.e. who has which tag). The data protection remains within your organisation and control.

Cyber security issues

Do you know where a camera’s data is stored? A large percentage of the AI cameras on the market come from China, and this is often where the data is stored. Plus, is the system protected against the possible risk of a hack?

SiteZone data is accessed via Microsoft PowerBI, taking advantage of the Microsoft security protocols that we all use every day (Outlook, Teams, etc). And we’ve recently achieved Cyber Essentials Plus accreditation, a Government back scheme for cyber security.

Screen fatigue

If you’re an operator of a vehicle, you’d think your job would have been made easier with advancement in technologies. But this isn’t what they tell us. As well as completing their job efficiently, they now have to monitor a number of different video screens inside the cab. This requires their attention and intuition to act on what is being presented.

We’ve always kept the SiteZone System simple. It’s a flashing display, no video screens. When alerted you stop your vehicle and look for the hazard. Once you have ensured the risk is removed and the screen stops flashing you can continue with your work. Simple as that.

And once again we’re putting all the emphasis on to the operator with a camera. Where is the pedestrian alert? With a tag vibration being personal to the wearer it reminds them of their responsibility for personal safety.

1 in 10 detections not made

Human form detection cameras can sometimes misinterpret safe actions as hazardous or fail to detect actual hazards. This can lead to unnecessary stoppages or missed dangers, affecting both safety and productivity. It can also cause distrust in the system with people switching off to alerts.

One supplier of these ‘AI’ cameras has said: “The cameras are extremely accurate, about 92%.” So nearly 1 in 10 times the camera detects wrongly.

In a recent YouGov survey, when asked about AI in the workplace, 26% of employers and only 13% of employees thought it would make their workplace ‘more safe’.

With an RFID tag-based proximity warning system you will only detect the tag. The tag is a part of your PPE and staff put it on with their hard hat or hi-visibility clothing. Tags are tested at the start of the shift, and you can monitor this in the data platform for compliance. There is no delay in the alert to the driver.

Changing behaviours

Ultimately the way to eliminate vehicle to pedestrian fatalities is to change behaviours. And to change behaviours we need to identify what are the wrong behaviours, by who, and where on site.

SiteZone customers are already doing this:

“The analytical data platform is an exceptional risk management tool from which improvements can be tracked. The information contained within the platform is anonymised (based on tag ID) and supported development of a strong, positive, safety culture changing behaviours to make a safer site.”

Chris Beddoes MCGI CMIOSH AIEMA, Business Manager for Stop'N'Go