Automotive and vision technologies are merging

This month Dr. Bern Hìfflinger, professor at the Institute for Microelectronics Stuttgart in Stuttgart, Germany, and a member of the Image Processing Europe editorial advisory board, writes about merging technologies.

Two major industries/technologies, automotive and vision, are intersecting at a critical juncture in time. Mobility is strongly motivated by safety and comfort demands in automobiles, and it is also threatened by high traffic density and roadway obstacles. Vision has made great strides with electronic CCD cameras, but it needs digital intelligence for fast and improved image processing and CMOS sensors and systems-on-chip to detect and process images faster and more precisely. Integrating and empowering automotive and vision technologies promises to spawn a huge market that will assure future mobility, increase safety, and vitalize industrial vision, all at reasonable costs.

Vision requirements for automobile safety and operation and roadside monitoring are highly demanding. In fact, they turn out to be tougher than aerospace requirements. Public and legal pressures to keep traffic moving and safe as well as worldwide competition for distinctive vehicle and traffic-control technologies are driving the market for more vision capabilities in automobiles and along the roadways.

Future requirements for vision sensors and processors to handle more than 20 distinctive automotive-vision tasks are difficult and advanced. They call for a dynamic range within each frame and from frame to frame of 106:1 or 120 dB; resolution of >320,000 pixels; random, high-speed access to each pixel of <100 ns; high frame rates on subfields of >1000 subframes/s; a high and constant contrast resolution of ~1% over four decades of gray; constant colors of more than four decades of gray; and a high sensitivity of better than 0.1 lux.

Sensors meeting these requirements can overcome problems such as over- or underexposure, white saturation, black-saturated shades, no details of objects and faces in backlit scenes, and distorted colors depending on local light level. These new sensors offer robust edge and feature detection of objects irrespective of whether they are brightly illuminated, hidden in shades, or occurring at the same time; low-light visibility, color content, and speed better than human eyes; correct colors of objects independent of local illumination and consequently robust identification; visual display of scenes and colors similar to the perception of human eyes; and high-speed acquisition and identification of objects and events, substantially improving the actions to avoid accidents.

Industry analysts expect that by 2010 electronics will incur as much as 30% of a car's value, or a world market of US$400 billion. Vision-based hardware and software installed inside and outside vehicles is anticipated to capture US$40 billion, or ten times today's industrial machine-vision market. The vision challenges and opportunities reaching those estimates are formidable, and the time scale is tight due to demanding automotive schedules. However, meeting the worldwide mobility demands of the future ensures vibrant and expanding vision and image-processing markets.

This page is available for machine-vision and image-processing professionals to present an industry or marketing perspective. Contact George Kotelly, editor in chief, for information: [email protected].