Despite tremendous advances in machine vision techniques, the accuracy and reliability of camera-based detection systems suffer during periods of poor visibility where other technologies may offer an alternative.
Radar is one such technology. It too has seen significant development in recent years and according to
Navtech itself was set up in 1999 as a spin-off of a Sydney University project using radar for precisely locating stationary and moving objects for driverless vehicles in ports. Later the technology was developed for security and highway applications and Navtech now has installations in more than 40 countries.
“Radar cannot completely replace visual verification; what it does is overcome the limitations and provide additional capabilities and reliability,” says Navtech’s marketing director, Nigel Crisp. “Often authorities only consider radar when they are not getting what they need from their installed vision-based monitoring systems and there are three main challenges: technical; environmental; and financial,” he adds.
On the technical front, authorities want a high degree of automation often including automated alarms and camera control; they want maximum accuracy and range (to reduce the number of units required); and the minimum number of false alarms.
Environmental challenges include extreme temperatures, varying lighting conditions including glare, reflections, dust, snow and ice which Crisp says can cause problems for visual technologies and laser scanners.
tunnels and bridges,” Crisp concludes.
As radar does not depend on light it can work in total darkness and according to Crisp the technology is less susceptible than vision-based systems to vibration – for instance on a bridge being crossed by heavy loads.
On the financial side, he says authorities should consider the total cost of ownership, as with a radar installation fewer cameras are needed and, more importantly, the servicing requirements are far lower.
Luca Perletta, product manager at Navtech, explains: “Although machine vision techniques can minimise the number of operators needed to monitor the cameras, in applications like tunnels the cameras need regular cleaning otherwise the number of false alarms becomes overwhelming.”
False alarms were a key factor in the decision to use Navtech’s radar for incident detection in the Rennfast tunnels in Norway during an upgrade to meet domestic and EU legislation. Tunnel operator Vegtrafikksentralen (VTS) was suffering multiple false alarms at its centralised control centre, from a video-based incident detection system deployed in another of its tunnels which was fitted with 100 cameras. Over a six-month period there was an average of more than 30 alarms per day but typically only two or three required any action.
Perletta adds: “In tunnels the cameras often need to be cleaned every six weeks and that can mean lane closures which are very costly, disruptive for users and potentially dangerous for workers and motorists. Radar is unaffected by dust and is pretty much a case of installing it and forgetting it is there. With Navtech’s systems a minor service is needed every three years.”
With the Rennfast tunnels (which have a combined length of 10.3km, carry around 9,000 vehicles per day and are fitted with 160 cameras) due for an upgrade, VTS believed the level of false alarms from a camera-based monitoring system would be unsustainable. So it commissioned Norwegian consultants Multiconsult to consider alternative solutions including loops and radar.
Radar can provide a black and white image but not of the quality required for all monitoring purposes so in critical situations like tunnels, bridges and smart motorways, it has to be used in conjunction with normal cameras. In such dual deployments the cameras are not required for incident detection, stopped vehicles, debris or pedestrians so may not require the highest resolution and can be spaced further apart. Like cameras, the radar works on line of sight so the road/tunnel layout is a large factor in how many radar units are needed.
In Navtech’s case the equipment for highway applications can detect people over a 700m stretch of road and vehicles over 1km, and can simultaneously monitor lanes travelling in opposite directions. Despite this coverage, the energy transmitted is only 15mW – less than 10% of that emitted by a mobile phone.
In the case of tunnel installations they are often designed with a degree of overlapping to cover a ‘blind’ spot immediately beneath the radar units that are usually mounted close to the roof of the tunnel.
Furthermore, with four rotations a second, delivering 400 data samples per revolution and by transmitting a signal at 77GHz, the company says the resolution is far higher than that of traditional radar systems, provides range accuracy of 25cm, and can distinguish between vehicles with 2° of azimuth (or bearing) separation. “You cannot achieve that resolution with a 24GHz or 10GHz system,” Perletta adds.
So what are the practical implications for highway engineers?
“For incident detection in a tunnel you would typically need a camera every 60m or so - which is why, when you add in allowances for bends, the Rennfast tunnels have 160 cameras. We provided incident detection for both tunnels using only 24 radar units,” says Perletta.
In equipment costs in the tunnel installation he says the radar may have been slightly more expensive than a camera-based alternative but both the maintenance costs and rate of false alarms are much lower. “If you look on the open highway where each installation may require a post, power and communications to be installed, if you can reduce [the cost of a vision-based system] by a factor of six, the savings would be considerable.”
Navtech’s radar doesn’t use the conventional Doppler Effect, allowing it to detect stationary objects such as stopped vehicles or a lost load in the carriageway. To do this the system maps the empty carriageway, including signs, lights and other infrastructure or furniture. It then compares the received signal with the reference image to detect any obstructions.
Moving objects are detected by comparing successive scans to detect if an object such as a pedestrian is moving within the covered area. As the transmitter scans a full 360°, the system can be programmed to ignore happenings outside the road boundary such as cows moving in a field adjacent to the highway – although it can be configured to detect wildlife such as deer, heading towards the highway.
In Australia a Navtech system has been installed to detect stopped vehicles and pedestrians on Bolte Bridge in Melbourne, but a ship passing beneath the bridge would not trigger an alarm. In the UK, the Highways Authority (now Highways England) is trialling Navtech’s radar on a ‘Smart Motorway’ section where the hard shoulder has been converted into a running lane. “In such situations it is vital that the authority detect a stopped vehicle or pedestrians as soon as possible, regardless of lighting or weather conditions,” says Crisp.
Currently a10Mbit data connection transfers the raw radar data to the control centre for processing and triggering of alarms. In the near future ‘edge processing’ will be employed either adjacent to the radar unit or incorporated within it, thereby reducing the network traffic to “a few kbps”.
In many instances the radar system can be connected to the cameras in order that they automatically zoom in on any detected incident. “When the system detects something the operator will immediately be able to view the scene – they don’t have to start searching for the stopped car or pedestrian,” says Crisp. “We typically get one false alarm per radar per day – so the operators do investigate them rather than ignore them as happens with higher false alarm rates.”
According to Perletta, the radar’s rate of false alarms is 10 times better than with visual systems and is enhanced by creating modes where particular features are activated or disabled. For instance the radars covering a reversible lane can be preprogramed to expect the reversed traffic direction while still monitoring for wrong-way drivers, or pedestrian detection can be disabled when roadwork crews are deployed, while retaining the stopped vehicle monitoring.
“We don’t necessarily compete with video verification systems. Radars provide additional data and more complete situational awareness to the operators. Which all helps to improve safety on our roads, tunnels and bridges," Crisp concludes.
On trial in America
Navtech is currently are applying for FCC approval to use 77GHz in the US and has a system being trialled in a high crash zone on the Minneapolis ring road. According to Crisp the initial feedback from the seven month trial is good with the single unit detecting a high number of stopping or reversing vehicles that were previously undetected. There were only two false alarms over the first three days.
