InfrastructurAL(PR): technology for the smart road
Author : Aneet Chopra | Vice President, Product Management, Marketing & Business Development | XMOS
01 November 2022
The smart home is no longer a lofty ideal. Voice-activated lights, remote thermostats & video doorbells are flashy, but not unattainable; you probably know several people who own smart speakers. By the end of 2022, according to market research firm, Statista, the smart home market will break $125bn in the US alone.
This article was originally featured as the cover story in the November 2022 issue of EPDT magazine [read the digital issue]. And sign up to receive your own copy each month.
Each smart home is a brick in the wall that is the smart city. As our homes become more intelligent, filled with devices increasingly capable of communicating with one another, we will start to see that model connecting homes. But what next, Aneet Chopra, Vice President, Product Management, Marketing & Business Development at British smart IoT chip firm, XMOS considers here…
The cement for the wall in this analogy – the physical space that connects each home – is the road network. Just as we’ve seen smart IoT (internet of things) technologies enter our homes and make them more intuitive to interact with, we will start to see them become a core component of road infrastructure. In time, it’ll be as normal as speed cameras or parking barriers.
One of the most advanced technologies in this field is automatic license plate recognition (ALPR – also known as automatic number plate recognition or ANPR). Already deployed across some motorways and car parks, ALPR systems can read the numberplate of a vehicle and then contextualise its behaviour. They can track speed, recognise payment statuses, open parking barriers and, in some cases, even remind you where you parked.
However, advanced is a relative term. Just as smart speakers were a decade ago, ALPR is currently on the periphery of our roads. Most car parks and motorways still rely on outdated and expensive systems or remain entirely manual. Why?
Changing gears
The difficulty is twofold: cost, and system complexity.
Current ALPR systems, generally, rely upon high-resolution cameras. These come at a premium, and capture data for extremely complex machine learning models – which in turn demand huge power draw, and can charge licensing fees for use.
These models are also unhealthily reliant upon cloud connectivity to process the images that the cameras capture. Without a constant connection to the internet, the devices can’t understand what they’re capturing – and the hardware required to remain connected will, again, drive up costs.
Interestingly, this reliance upon the cloud is one of the same obstacles that has plagued the smart home’s progress for some time. XMOS CEO, Mark Lippett has previously written for Forbes Technology Council about the need to mitigate reliance upon the cloud, imbuing individual electronics with the intelligence they need to make sense of data instead.
Just as the smart home can be enhanced and expanded with on-device processing, so can the smart road. Moving from an always-connected, high-resolution system to one that prioritises on-device intelligence and efficiency could offer the potential to ramp up adoption across huge stretches of the road network. As a leading technology in the field, ALPR is one of the best solutions to accelerate progress.
Due process
Every design is going to look different. But there are also some non-negotiables: no design should be sacrificing accuracy, and it should be using as little power as possible. Efficiency is key.
Firstly, that means assessing the camera. Not all ALPR applications demand high-resolution images; in a car park, for example, being able to read a slow-moving vehicle’s plate from 3-5 metres away is all you need. Designers need to identify the camera that ticks all their boxes, while keeping power draw and the bill of materials to a minimum.
You then need to consider processor requirements. The silicon within needs to be able to interpret the images that are captured, while operating comfortably in a low-power scenario. That package needs to be delivered at a price point which makes mass production a realistic option.
Cameras that require less power, and that don’t need to be always on, offer designers more room to manoeuvre here. The change in hardware allows designers to consider processors better suited to supporting artificial intelligence (AI), and the machine learning algorithms that ALPR relies upon.
There’s also the sensor array to consider. An ambient sensor array has the potential to allow devices to power down until triggered, removing the need to connect to the mains, opening up battery- and solar-powered environments for ALPR.
Get the wheels turning
As in the smart home, the balance of these qualities differs between designs. ALPR solutions deployed on a motorway will differentiate hugely from ones in a car park.
However, the overarching objectives remain the same. Accuracy is paramount: the challenge for designers is minimising both cost and power consumption, without sacrificing performance.
The solutions capable of this will turn to on-board processing, freeing ALPR electronics from their tether to the cloud. When combined with low-resolution image capture and an appropriate sensor array, such a processor opens up powerful possibilities for lower power consumption, while guaranteeing a far cheaper bill of materials.
At XMOS, our iteration of this idea allows service providers to take parking management from a complex, resource-intensive system to a streamlined and simplified on-device AI. Based on the xcore®.ai, chip and a light neural network model, it’s designed to minimise cost and thereby catalyse the mass adoption of ALPR.
That saving on cost can drive ALPR forward as a technology, connecting each smart home – setting the cement in the wall of the smart city.
More information...
Contact Details and Archive...
