Machine Vision Safeguards Europe's Roadways

The intelligent transportation system (ITS) industry covers many technologies, including wireless and wireline communications (copper and optical), embedded inductive loops as well as many other classes of sensors, and computational power and software to make sense of it all.

In recent years, machine vision has joined the list of rising technologies in the ITS industry. After all, ITS' origins began as simple traffic monitoring, usually through loops and then CCTV cameras. Machine vision is building on this growing sensor network to significantly expand its market.

According to recent statistics, the European traffic monitoring market is approximately 200 million EUR, and is expected to grow 45 percent annually! In overall numbers, Europe lags behind North America in ITS expenditures, but new trends toward the privatizing of roads – something that's been happening in North America for many years – is opening up the European market to fast growth. The Japanese market isn't far behind either, offering a more complex ITS infrastructure, but in a smaller geographic area.

Traffic Applications
According to Klaus-Michael Buecher, Director Sales and Marketing for Vitronic (Wiesbaden, Germany), while the European traffic monitoring market is not homogenous in terms of deployment throughout European member nations (the Netherlands is one of the highest deployers), the drivers behind the application are universal. Buecher suggests three main drivers.

Improving traffic safety by automated monitoring and enforcement of traffice laws, such as stopping at red lights, and speed monitoring.

Maximizing revenue by automating toll road management and enforcement.

Increasing national security by applying automated license plate readers to locate stolen cars and protect against criminal and terrorist elements.

In general terms, machine vision systems support these drivers by automatically locating, identifying, and physically measuring a car or vehicle, and by using optical character recognition (OCR) to positively identify the vehicle by automatically reading its license plate.

Today, vision systems serve several traffic monitoring applications, including:

• Video enforcement of tolling
• Red light enforcement
• Speed enforcement
• Video tolling
• Travel time information systems
• Traffic monitoring
• Congestion charging
• Parking systems

Although system designs vary according to the application, the systems tend to have several common components: a camera, a processing unit, and data transmission capability (whether wireless or wireline, preferrably Dedicated Short Range Communication (DSRC) in Europe and Japan). Some systems will add active illumination, either in the visible or the near infrared (NIR) to improve contrast in low-light conditions and take advantage of increased sensor sensitivity near the red/NIR end of the visible spectrum.

'Auto-Optimized' Vision
Leaders in the European traffic monitoring market, such as JAI, headquartered in Denmark and with subsidiaries in USA, Japan and Europe and Vitronic, have created custom as well as OTS systems for various traffic monitoring applications.

According to JAI's Senior Vice President, Villads Sondergaard, one of JAI’s most recent traffic monitoring applications included work with Raytheon HTMS to design an automated toll enforcement system for several hundred toll plazas in Texas and Florida. The system uses high-speed megapixel camera tied to a dynamic light sensor. As light conditions change, the system can automatically change camera settings to make sure the collected images are good enough to identify toll breakers, including 'binning' of pixels to increase the contrast. Binning reduces image resolution, but improves contrast because it essentially increases the size and, therefore, light collection capabilities of each pixel.

In addition to high spatial resolution to allow license plate reading via OCR in low-light conditions, traffic monitoring cameras need to have high read out speeds. Another method would include the use of more expensive 'snapshot' sensors that elminate image skew and distortion caused by an object moving during the sensor read out. These sensors place a memory cell below each pixel in a CCD format, or next to it in a CMOS sensor, to temporarily store the charge to insure that all pixel values from a given moment can be stored and read without changing over time. JAI's sensors have been certified by the Dutch BWOM to capture images of objects moving at speeds up to 300 km/h.

Vitronic has packaged a toll enforcement system comprising a detection unit (sensor and electronics), classification unit (processing and database of vehicles), and identification unit (OCR for license plate reading, etc.), and a communication unit that uses the DSRC wireless communications standard. Japan and Europe have adopted the DSRC standard, which operates in the 5.8-5.9 GHz band typically used in WiFi, bluetooth, and RFID systems in the U.S.

By adding the capability to locate the car in motion and make physical measurements related to speed, direction, etc., Vitronic's PoliScan product can enforce traffic laws such as speeding, red light violations, and others. PoliScanspeed uses an eyesafe, pulsed infrared laser with time-of-flight (ToF) electronics to accurately measure vehicle speed and direction across multiple lanes even when weather conditions make visual measurements almost impossible. By adding an Internet connection to the system and the right security access, PoliScansurveillance can be employed on a mobile or stationary basis to detect the licence numbers of all vehicles in flowing traffic, automatically read them, and compare the acquired data with a database of wanted individuals.

In most if not all cases, the machine vision and communication technologies for automated traffic monitoring are extant – it's simply a matter of proper integration, experience, and education. Some of that integration will impact vision products, however, such as additional integration of Ethernet and GigE communications into the camera housing for longer transmission runs versus Camera Link® or proprietary analog standards. CMOS sensors will improve on the dynamic range, size, and cost of the camera sensor – all critical elements for outdoor distributed vision networks such as those needed for automated traffic monitoring. CMOS sensors also make binning easier and cheaper compared to CCD cameras.

Finally, as more smart cameras include more processing power, the computational units will migrate further into the camera housing, while opening the way for new features, such as facial recognition of the driver.

Driving Around the World
While machine vision for automated automobile guidance is still many years away, as this article shows, machine vision is already helping keep roadways safe.

Vision companies can expect the European traffic monitoring market to grow at an average of 45 percent annually during the next few years, with typical European applications falling into: red light, speed, congestion, and tolling systems, although video enforcement is still not possible in some EU countries due to legal procedures.

‘‘In all developed countries [law, toll, and border enforcement] are important, however, comparing the European situation with other important markets like North America or Asia leads to a slightly different focus: In North America Homeland Security is obviously the strongest issue, while traffic monitoring in Asia is most often used in Electronic Toll Collection applications,’‘ explains Vitronic's Buecher.