06 Jan 2005
Driving in the dark could become a lot safer if affordable night-vision systems being developed by several European projects prove popular. James Tyrrell gets the industry view at SPIE's Photonics in the Automobile event in Geneva.
From Opto & Laser Europe January 2005
The application of night-vision systems in vehicles is set to step up a gear. According to research presented at SPIE's recent Photonics in the Automobile event in Switzerland, scientists and engineers are driving towards low-cost, mass-production solutions that could match consumer expectations with industry price-points.
A collaboration between the European Photonics Industry Consortium (EPIC) and SPIE Europe, the meeting brought together key industry players from firms including Audi, Lumileds, Centro Ricerche Fiat, Hella KG Hueck, DaimlerChrysler and Osram. The three-day workshop saw more than 90 delegates assemble in Geneva.
Discussions covered a number of photonics innovations for cars, such as intelligent lighting systems that respond to fog and tunnels, lane-departure warning devices and assisted cruise-control. However, along with LED-based headlamps, it was night vision systems in particular that were grabbing a lot of attention at the event. The interest in night vision among car makers is partly a response to the European Commission's call to cut the number of road-traffic fatalities; around half of all fatal traffic accidents in Europe and the US occur at night, even though road traffic is only around 20% of the daytime figure.
Presenting at the event, Jörg Moisel from DaimlerChrysler Research and Technology pointed out that only high beams provide enough light to illuminate the road sufficiently for safe driving at speed. However, drivers often have limited opportunities to use a high beam during their journey: for example, a low beam must be used in an urban setting or when sharing the road with other vehicles. To improve safety, the automotive industry is looking into other ways of providing the driver with a clear picture of the road ahead, such as radar or night-vision tools.
Night-vision systems are based on infrared (IR) technology and can be classified as either passive or active. Passive systems rely on a far-infrared (FIR) detector to sense thermal radiation from the scene in front of the car. By contrast, active systems operate in the near infrared (NIR) and use an infrared source to illuminate the road ahead.
As Moisel explains, one advantage of passive systems is their ability to differentiate warm objects, such as pedestrians and animals, from a cold background. For this reason, passive systems are particularly effective during the dark winter months. Additionally, passive systems are capable of penetrating haze, fog and smoke, betraying their origins as a battlefield technology.
Military research has driven the development of infrared sensors that can detect thermal radiation in the 8-14 μm range. And over the past few years, commercial IR camera systems have slowly begun to appear in some luxury cars. For example, US car-maker Cadillac offered a night-vision option developed by US defence supplier Raytheon with its 2000 model DeVille sedan.
However, customer reaction to the units has been mixed. In fact, Cadillac announced recently that it has decided to drop the option on its 2005 model. "If we bring [out] a night-vision system again, it would be an active system," Jay Spenchian, Cadillac's marketing director, told US car magazine AutoWeek. "But there are no plans for that at this point."
In contrast, Honda says that it plans to offer a passive system to its Japanese customers this year. Clearly, the automotive industry is still settling on the best way to deliver a night-vision package to its customers. Key to the technology's success is the need to balance customer expectations with price. Fortunately, low-cost, high-volume solutions may be on the horizon, thanks to several major European projects.
The ICAR (infrared camera for car) project is a three-year European initiative, now nearing completion, that aims to develop a low-cost automotive FIR camera for driver-vision applications. Sensing experts from CEA-LETI in France and ULIS (formed in June 2002 by Sofradir and CEA) developed the detector, while the electronic circuit boards were built by French IR systems specialist CEDIP. German optoelectronic developer Zeiss Optronik assembled the cameras and performed the necessary electro-optical testing.
With an estimated budget of €5.7 m, one of ICAR's key targets was to create an FIR microbolometer detector. These uncooled focal-plane arrays, which in the case of the ICAR project contain 160 x 120 pixels at a pitch of 35 μm, could be an affordable solution for high-volume applications such as use in vehicles. The low-cost strategy also benefits from moulded infrared glass lenses supplied by project partner Umicore. The international materials specialist has developed a process using its GASIR material - a chalcogenide glass - that allows direct moulding into the final lens shape. According to the team, lens performance is comparable to that of optics made using germanium aspherics, which are more expensive to manufacture as they require cutting, shaping and polishing.
Sample batches of the GASIR lenses are being made at Umicore's plant in Quapaw, US, and full production start-up is scheduled for June 2005. According to Umicore's Marleen Van den Bergh, the facility is dedicated to the production of GASIR assemblies for a long-term contract with Autoliv, a major automotive component supplier.
A defining requirement for ICAR is that the set-up must be able to detect a pedestrian at a distance of at least 190 m, and other vehicles at 500 m. In terms of field-of-view, the system must cover the width of the road plus any pavement on either side.
As OLE went to press, ICAR's prototype was being test-driven at Centro Ricerche Fiat near Turin, Italy, with input from global automotive equipment supplier and project partner Valeo. "The [prototype] unit is fitted in Fiat's Multipla vehicle," Arnaud Crastes of French infrared-detector manufacturer ULIS told OLE. "Because the car's instruments are in the middle of the cockpit, there is space for us to mount the display behind the steering wheel."
Along with cost, functionality remains an issue for car night vision. This is especially true in the case of passive FIR systems, which often output a "negative image" - hot objects appear white against a cold black background - the unfamiliarity of which reduces the driver's reaction time.
French firm Sagem, another attendee at the SPIE event, is tackling this issue by developing an advanced driver-assistance system that interprets IR images and provides a warning of potential obstacles. Its work formed part of PAROTO (radar and optoelectronic anticollision project for the car), which concluded in 2004 and investigated obstacle detection using IR and radar information.
Jacques Lonnoy of Sagem reports that hot objects in an FIR image such as wheels, exhaust pipes, pedestrians and cyclists provide effective cues for the detection software. Using software algorithms, the idea is to track obstacles detected in the camera image and alert the driver if he or she appears to be unaware of the oncoming danger.
The alternative to FIR systems is the active night-vision set-up. Active systems employ standard silicon cameras (CCD or CMOS) to image a scene illuminated with an NIR radiation source in the range 780-1,000 nm. EDEL (enhanced driver's perception in poor visibility) is a European project investigating active technology that is due to deliver its results next month. The project involves industry heavyweights Osram, Jaguar, Bosch, Hella and Centro Ricerche Fiat, as well as the University of Genoa, the University of Karlsruhe and the University of Siena.
As partner firm Bosch points out, because road boundaries and lane markings are visible in the NIR image, it is easy for the user to differentiate between those obstacles that are in their driving lane and those that are not. It also means that the system could be used for other vision-based safety functions such as lane-departure warning. By contrast, FIR images from passive systems can mask lane information, as signs and markings generally adopt the temperature of their surroundings.
Active systems have already made their commercial debut. Toyota released an active system back in 2002 based on a halogen IR bulb. Ideas such as a modified headlamp that combines an IR high beam with a visible low beam, sound like a perfect solution. Unfortunately, because the 970 nm emission-peak of a halogen lamp is located at the edge of a silicon camera chip's spectral range, most of the IR radiation emitted remains undetected.
A custom-designed solid-state IR source, such as a semiconductor laser or LED, could turn out to be a much more effective IR illuminator - an idea shared by both the EDEL project and DaimlerChrysler. The car-maker has developed a prototype based on a high-power 808 nm laser diode and a CCD camera with an interference filter centred at 808 nm with a 20 nm FWHM bandwidth.
To help manage the heat load, a thermoelectric cooler is attached to the laser's copper subassembly. Temperature remains a concern with solid-state lighting, because operating enclosures can reach up to 100 °C. However, one benefit of using a laser is that the IR source could perform other roles, such as forming part of a LIDAR scheme for rangefinding and environmental sensing.
Rather than continuous illumination, the industry may eventually decide upon pulsed emission. Pulses of IR radiation delivered in synchronization with the unit's camera could help to suppress noise from stray light-sources such as streetlamps. Additionally, pulsed schemes could help prevent the camera from being "blinded" by 808 nm beams from oncoming vehicles.
Ultimately, LED illumination could prove to be the most cost-effective technique. Moisel's team at DaimlerChrysler has prototyped a 5 x 15 array of 0.3 mm LED chips glued directly to a copper heat-sink. According to market data presented by Moisel, the cost (excluding optics) of 1 W of laser power is around €25, compared with just €5 for IR LEDs. However, as he explains, a key parameter is the size of the IR illuminator's exit pupil, which is often determined by the vehicle's design and styling. Essentially, the smaller the headlight, the more likely it is that a laser system will have to be used. Moisel concludes that if size was not an issue, infrared LEDs could bring about a low-cost solution to night-vision illumination.