Debating the future development of ANPR

What future is there for automatic number plate recognition? Will it be supplanted by electronic vehicle identification, or will continuing development maintain the technology's relevance?

In recent years, digitisation and IP-based communication networks have allowed Automatic Number Plate Recognition (ANPR) to achieve ever-greater utility and a commensurate increase in deployments. But where does the technology go next - indeed, does it have a future in the face of the increasing use of, for instance, Dedicated Short-Range Communications to identify vehicles in free-flow conditions?

ITS International recently asked a number of the leading enforcement system manufacturers to: quantify ANPR's technological advances in recent years, and what we might expect to see over the next five-20 years; comment on digitisation's effect on the relationship between front- and back-end systems, and on operating procedures; compare ANPR's 'best effort' identification of vehicles using traditional, non-electronic methods (the licence plate) with DSRC and other electronic means - and on the continuing need (or otherwise) for visual identification; and contrast the current emphasis on systems and operators to achieve positive identification with what could be 'pushed' back from the vehicle/owner side (such as standardised licence plates and a tighter wedding of vehicle/driver identification). They were also asked whether the main impediments to technological progress reside in the laboratory or the chambers of law. What follows is a snapshot of ANPR now and into the future.

"Next step the hardest"

"In the last 10 years we have seen ANPR make stunning progress in terms of reducing human intervention but the next step, to zero human intervention, will be the hardest to take," says Paolo Sodi of 125 Sodi Scientifica.

"Digitisation and ANPR have grown alongside each other and it has been natural for clients to include ANPR when upgrading, mainly because by the time digitisation reached traffic enforcement clients had a more open mind towards computers and expected their enforcement units to do what their home PC was able to. From the manufacturer's perspective, our aim is to make the operational aspects of enforcement as easy as possible and adding ANPR software to our digital units was a natural step. Where jurisdictions still use analogue systems the reasons are to be found either in the volumes of infractions they generate or the back office rather than in a specific technology. Moving to digital is about more than buying a new unit; it is also about having a back office able to exploit in real time such capabilities as connection to a national vehicle registration database.

"DSRC use implies efforts by states and automakers. That takes matters to a whole new level and despite the undoubted benefits the time needed to implement and make it universally acceptable leaves huge room for further ANPR development. Presently, the most logical way to identify a vehicle is still its licence plate which, being mandatory, allows officers to identify without doubt the vehicle and its owner. In the near future I can see a faster and more accurate ANPR engine which gets close to zero human intervention. Further than that, things becomes hard to predict without facing the risk of being totally wrong.

"In terms of correct identification, at present the pressure is mainly on the systems and the operators and without doubt a lot could be done to change that. Standardisation of plates and characters would indeed be a great help. Back office automation is already helping but much more could be done. New technologies will improve character recognition, for instance better cameras will lead to more efficient handling of shades on the plate. Unfortunately, the greatest impediment to ANPR is obstruction, dirt or inclement weather, and I see very little possibility of improvement in this regard.

"Someone I knew very well used to say, 'The only limits to technological development are cost and imagination. If you can imagine it, you can do it.' Time has proven that to be true, it is implementation that can be an obstacle. If we do move away from ANPR many new players come into the game and having them work together is the real challenge. Changing the mentality of the officers, drivers and administrations that since the late 1800s have relied on licence plates to identify a vehicle, its use and owner is the real task. Once we have done that a new era will begin.

"In the short- to mid-term I strongly believe that the issues facing us are mainly procedural, political and administrative rather than technological. Only a political decision can make changes happen quickly. We can have the best, most innovative and convenient vehicle recognition techniques but the public sector is not so quick to make changes. What we can do as manufacturers is continue our research, make long-term investments in new technologies and rely only on our capacity to stimulate the market. That's what happened with digital systems. These are now taken for granted but were hardly seen as necessary when we first presented them in 1996."

"25-30 per cent growth"

The ANPR market will grow by around 25-30 per cent in the next five years. Enforcement solutions combining, for example, speed measurement, camera and ANPR will drive future growth, according to Erno Szucs of ARH, Inc.

"There has been a clear move toward digitisation in the past three to five years; over 90 per cent of our partners' projects are based on digital technologies and IP environments. Connectivity and networking are easier to solve and architectures are simpler. Digital cameras have solved the resolution issue and even an entry-level digital camera will cover a full highway lane. On the other hand, digitisation has given rise to new issues. High image compression rates lower OCR accuracy and are to be avoided but large, high-resolution image processing takes more time, so image sizes for ANPR have to be the smallest possible which will still cover the required areas. Increasing CPU speeds are only a partial solution.

"Any discussion about replacing licence plates has to consider that photographic proof is needed for an offence such as speeding. The majority of an enforcement system's cost is in the image creation infrastructure, which will be expected to be in place in the future. ANPR's contributory cost can be as low as 5 per cent; if the camera system has to be in place why add a parallel system?

I don't think traditional license plate identification will disappear in the next few decades. But I can see DSRC being used in concert. There is however a lot to do in certain parts of the world where plate standardisation is concerned. Machine-readable plate types improve OCR accuracy dramatically and the technology will probably be improved to include automated identification of vehicle model and colour.

"A next step might be more intelligent vehicles which are able to store driving information, including violations. I agree that tying a vehicle to a responsible owner or driver is an important issue but the ideal, having the driver of a specific vehicle set centrally at all times, is hard to achieve. Automated personal identification would be needed - not a popular idea. It's more realistic to have a responsible person dedicated to each vehicle at all times."

"Advances in facial recognition"

According to Ivo Paton of 65 Imagsa Technologies current levels of reliability, which are still limited for some vehicle speeds and under some environmental conditions, should improve significantly: "Value will be added in the form of vehicle classification and the addition of features such as GPS time and location, analysis of vehicle trajectory and speed, facial recognition and traffic signals status. ANPR systems have evolved from distributed solutions which are difficult to install and set up, to compact all-in-one solutions. This will continue and include smart tools to automate on-site set-up.

"Digital technologies allow significant cost and performance enhancements. For instance, processing embedded into the camera allows for reliable ANPR and reduces bandwidth requirements, allowing for easier to deploy and more cost-efficient solutions. Although some jurisdictions still require photographic proof for enforcement applications, advances in digital security techniques overcome the risk of faking.

"Electronic identification of vehicles using DSRC onboard units will always require some complementary, non-intrusive technique such as ANPR. The reason is obvious: assuring road administrations that every vehicle can be identified electronically would require complete agreement between all countries and vehicle manufacturers. That's far from being possible at the moment.

"Visual identification is still in its infancy and has a very long way to go. The number of vehicles will keep growing, and as these techniques add new value, new applications will continue to emerge. The benefits of license plate standardisation in terms of enhanced ANPR performance and reduced cost are clear but efforts like the one with European plates are far from being possible in today's world.

"Ideally, enforcement systems should chase offending drivers, and not anonymous vehicles, so I believe that in future ANPR will benefit from the advances in facial recognition. Although drivers' privacy should be preserved, advances in system security can allow only the identities of offending drivers to be used, thus addressing concerns over privacy.

"Driving from one country to another requires the integration of different administrations' data. This has been recognised in Europe, and we believe it will be in the future in other regions in the world.

"'Big Brother backlash"

Geoff Collins of 126 Speed Check Services points to improvements in both the technology and its implementation.

"This has made ANPR a much more effective tool. However, there will always be a limitation on what a visual monitoring system can achieve - it simply isn't possible to visually read a plate that is very dirty, obscured or simply not there.

"Digitisation allows data to be immediately integrated into back office systems but requires procedures to be revised and updated to account for the different media used. For example, there is no longer a negative print that can be considered the primary evidence.

"It is likely that some kind of electronic tag will be used widely in future, as a unique, fail-safe identifier but there will always be the need for a human-readable identifier. A police or enforcement officer must be able to directly identify a vehicle and the human eye will always have a place in this process.

"It would be sensible to make vehicles and their owners more readily identifiable, particularly when considering the strong links between vehicle use and crime. However, there is already a strong backlash against perceptions of 'Big Brother' control. A car offers the freedom of the open road but this freedom is limited somewhat if a log of who you are and when and where you travelled is available for scrutiny.

"A carefully considered approach will allow a link to be created between a given vehicle and its driver for a given journey. Driving licence details could be linked with the vehicle registration mark; in Sweden, roadside camera images are compared with the images from driving licenses relating to the registered address for a vehicle.

"Tighter integration will deliver a much more useful and powerful solution, which ultimately will benefit both road users and operators. Technically and procedurally, it is complex but achievable. Politically, it is far more of a challenge. To be successful, an integrated approach would only be possible if managed extremely carefully at a political level, with a long-term view on the outcomes and benefits."

"At-the-edge processing"

"Advances in high-resolution digital cameras, LED illumination, more powerful DSPs, faster, lower-power, general-purpose processors, and increased network bandwidth have improved ANPR system performance. As of now, improvements in vehicle detection and identification have resulted in even higher accuracy than previously possible, including full identification of country, state or province of issue, plate colour and so on," says Tom Hayes of 1919 Perceptics.

"Over the next few years, more cameras may start using onboard DSPs to implement video analytics tailored to improve ANPR performance - from smart exposure control to eliminate motion blur, to real-time skew correction and vehicle detection.

"High-resolution digital cameras have pushed processing to the edge of the network, mainly to avoid having to transmit large digital images. The impact that the large volumes of image data have on transmission latency, network bandwidth and storage requirements mean analogue cameras won't be completely displaced overnight.

"Without a mandate to use the same technology and/or standard to identify vehicles across multiple jurisdictions, there will always be a need for visual identification. Given the current lack of uniform licence plate standards, each jurisdiction seems to give itself the freedom to dictate the layout and looks of its plates. The use of technologies such as DSRC to allow the electronic capture of license plates might create privacy concerns. Future use of DSRC technologies for vehicle identification will probably need to allow the user some control over which non-governmental third-party agents have access to their information.

"Visual identification will always be necessary to help prevent cloning, spoofing and other electronic methods of tampering. Given that vehicles will continue to have a visually discernible, unique registration number, the role of ANPR systems may actually grow from exception-based violation enforcement. ANPR could become the primary method used to identify vehicles, as licence plates are commonly viewed as public documents and any agency needing to perform an enforcement action based on the information it holds would have access to a database holding the associated private information.

"The onus will always be on the enforcement systems and their operators to continue to develop robust systems able to correctly and uniquely identify every vehicle in all conditions. Most standardisation attempts seem to have been limited to relatively small geographical regions; more and more local governments use vanity and/or special plates as a source of revenue.

"In most jurisdictions, different plate colours and styles are used to distinguish between different types of vehicle uses - private or commercial, emergency, military, government and so on. Linking owner/driver ID to vehicle registrations at the front end creates privacy concerns as it would be very easy for third parties to capture and store individuals' travel histories without their knowledge. The link needs to be made on the back end, where access to databases can be properly controlled.

ANPR systems have already been integrated with other technologies such as RFID to capture ID tokens in real-time. Such tokens, by themselves, carry no private information, but can be used by the agencies that issue them as an index to a properly secured database containing personal information about the driver and/or other passengers inside the vehicle.

"No front/back-end distinctions"

According to Nicola Bartesaghi of 592 Tattile, miniaturisation of the relevant technologies and improvements in abilities to read license plates that already exceed 96 to 98 per cent in good operating conditions will lead to new applications.

"ANPR solutions have revolutionised accessibility to information, real-time processing, process automation, security and safety. The distributed intelligence of these systems has also changed the characteristics of the networks used. Having separate front- and back-end technologies could become an outdated concept as analogue systems are replaced by a single element: the smart camera. Cumbersome illuminators are also often being abandoned in favour of LED technology.

"Smaller microprocessor technology with greater computing capabilities has drastically reduced systems' visual impact, while low power consumptions allow installation wherever there is access to sunlight. One of the world's most sophisticated ANPR systems, which is fully contained in the roof lights of patrol vehicles and so is virtually invisible, is supplied to Italian police and Carabinieri teams.

"The system is capable of reading and recognising up to 50 plates per second on vehicles moving at high speeds, comparing in real time with a huge database of stolen or suspect vehicles. Arrest or other notification alerts can be passed to a remote central location without implications for the roadside team.

"Visual identification will retain an important role but there will be complementing technologies, some of which are still emerging. Next-generation ANPR solutions will include a whole range of classification capabilities in one unit. Information supplied by vehicles' onboard GPS or mobile phones will bring some surprises and contributions to the accuracy of location and identification of vehicles, their habits and their possible infringements.

"But if GPS and cellular technologies will help tie a vehicle and driver together, much could still be done to standardise plates' sizes, fonts, colours and so on. The important thing is to apply thought and do this well; doing it badly could be very costly in all sorts of ways.

Existing technologies already allow high levels of identification but could be seen to prejudice privacy. The challenge is to balance cost and the public and private interests."