Video analytics enhances urban rail safety
First publishedin ITS International
An intruder detected alongside a moving train on the Montréal metro computer
David Crawford explores some promising innovations for North American commuters.
North America is experiencing a surge in commuter rail and metro development. The US now has 75 light rail and metro networks in operation; and California, in particular, is actively exploring ways of developing the state’s existing passenger rail operations into a fully integrated system.
But global consultant Steer Davies Gleave, which led what it claims is the first comprehensive review of the state’s passenger rail infrastructure, published in 2015, has highlighted a key problem.
Especially in California, it says, ‘geographic and topographic constraints are major factors impacting the development of rail infrastructure and services, leaving little room for expanding local passenger rail systems to a state-wide level.’
One such constraint is being overcome using modern safety detection technology on a US$2.3bn, 25.5km southern extension of the San Francisco-centred Bay Area Rapid Transit (BART) network into the Silicon Valley technopolis, in Santa Clara County. This has been a regional aspiration since the late 1990s.
The first phase is due to become operational in late 2017 and earlier this year the US Federal Transit Authority approved the final phase to Santa Clara University which will include an 8km (5 miles) tunnel under the centre of the city of San Jose.
The Apple HQ in Silicon Valley, which will enjoy better BART connections
For one 9km stretch, new automated passenger trains will run alongside a Union Pacific heavy freight line in an existing rail corridor. Normally, this would mean a 20m (65ft) safety separation gap between the alignments, to guard against such risks as a fast-moving freight train hitting an obstacle and jackknifing wagons into the commuter tracks; or shedding part of its load.
One alternative, in a highly-developed urban corridor where such a gap was unfeasible on the grounds of land availability and dislocation, was a reinforced concrete crash-wall barrier, which would have had to be around 1m thick and 6m high and cost around US$6 million.
The other was the innovative video analytics solution that is now being installed - a camera-based railroad intrusion detection system (RIDS) developed from one originally introduced to protect commercial harbours from unwelcome visitors climbing over perimeter security fences. There, said Matt Shipman, a systems engineering manager for the prime consultant, PGH Wong Engineering, based in San Francisco, “it’s a lot more complicated, because the cameras are tracking people through the port, so they can map a perpetrator.
“Here, they’re simply detecting an object coming over or through a fence and determining if it is large enough to be a threat”. The system is being designed to distinguish between, for example, a child’s stray helium balloon and a dislodged length of pipe being transported for utility mains laying; assess the risk; and only halt the commuter train when necessary. It is costing around US$500,000.
The protective fence alongside the Santa Clara BART extension.
If it receives a high-level alert – for example, indicating that the fence has been significantly damaged; or that the cameras have detected a large object going over or through it – the system issues a zero speed command and the train automatically stops. A low-level intrusion will alert the BART control centre which will review the video and decide whether to halt trains or issue a ‘slow’ command.
The ‘slow’ command allows an oncoming service to continue at low speed until the control centre operator can see the object and decide if the train can brush it away or if maintenance staff need to go and remove it.
Pole-mounted thermal imaging cameras are positioned by a standard wire fence every 300m or so along the route, acting as sensors for the GPS-equipped analysis array, which picks up any danger spot on the detection boundary. Alternating poles have tilt sensors to detect if an incident has deflected the fence far enough into the commuter route’s right of way to give grounds for concern.
The team has already deployed the system over an 800m track section as part of a six-month prototype test of, among other factors, the effects of weather. Field testing of the completed system began in 2016.
Construction of the extension, which is being overseen by the Santa Clara Valley Transportation Authority, is due for substantial completion in early spring 2017, when there will be full-scale integrated testing of all the control systems. It will relieve congestion on two crowded roadways, interstates I-680 and I-880.
The solution could benefit other North American cities where the scope for maximising existing rail rights of way is limited by land prices, development pressures or safety standards.
Meanwhile, the nearby city of Palo Alto is trialling a ‘suicide watch’ video-camera based intrusion detection system along the tracks of Caltrain commuter services running north into San Francisco, where a significant number of teenagers have committed suicide since 2009. The Suicide And Accident Intervention System (SAAIS), launched in 2015 by locally-based CSC Integrations, will, the city hopes, eventually replace the current US$1.6 million a year Track Watch programme, using contracted, private security guards, which is experiencing a number of concerns.
The unfortunate aftermath of a suicide.
One is the practical limit on how far the guards can see along the tracks, especially in the dark, so the new system will use night vision cameras. In addition, regional media has reported that some guards have been found neglecting their duties in favour of texting or even criminal activity. As a temporary measure, the city has demanded guards carry GPS-enabled smartphones, so that their locations can be monitored, and installed temporary lavatories. In contrast to the use of guards, SAAIS uses video analytics to detect potential suicides’ behavioural patterns and movements and triggers alerts to stop trains in time.
The system is the brainchild of company founder Jason Jenkins, whose earlier career in law enforcement confronted him with a number of attempted and successful suicides – the most distressing of which he found involved trains and their drivers. (A 2011 national survey of commuter rail drivers for the US Federal Transit Administration found that more than 12% of drivers who had been involved in a critical incident reported post-traumatic stress disorder symptoms.)
In Canada, the Montréal metro is the country’s second, and North America’s fourth, busiest network. Its operator, the Société de Transport de Montréal (STM), wanted to enhance its existing surveillance system, which deploys some 300 cameras covering tracks and tunnels, to enable the detection of, and automated response to, people entering a controlled danger zone.
A major technical issue was that the view from these cameras was frequently largely filled with moving trains and their lights, making motion-based video analytics temporarily unusable. AgilityVideo, based in Tennessee, US, had already developed its SmartCatch software to detect railway intrusion, and worked with SMT on an enhanced version that could identify a human on a tunnel catwalk with a moving train occupying most of the camera’s field. In 2015, the company incorporated an automatic audio alert to deter intruders.
Software programming for SmartCatch is subcontracted to Indian specialist TechnoYen Software. AgilityVideo has recently placed the technology on the market, while maintaining a current warranty agreement with STM.
In June 2015 the International Association of Public Transport (UITP), jointly with Sweden-HQ'd surveillance specialist Axis Communications, ran a global survey of video analytics uptake levels and adoption plans among public transport operations. This showed wide scale appreciation of the technology among staff as well as passengers, with half of the respondents reporting commitments to future investments.
At the same time, the pace of commuter rail and metro development in the US and Canada has prompted the association to create an International Urban Rail Platform for North America. This is due to hold its first regional forum in New York from 8 to 9 December, with safety high on the agenda.
One promising avenue being explored is the introduction of driver assistance systems, along the lines of those common in the automotive sector. With third-party collisions emerging as the most common incidents in light rail transit operations, the hope is that smart sensors detecting and communicating the presence of pedestrians and cyclists can bring added safety to users of the public spaces through which light rail services travel.