How do you think the new GigE standards will influence the machine vision industry?
Respond or ask your question now!
By Lee J. Nelson
An increasing number of vehicles have some type of sensor or video system for the rear view. With today’s high beltline and trunk configurations, it is sometimes impossible to see someone bending down behind a sedan (see Figure 2). And in some SUVs, the driver cannot see the ground behind the vehicle for 50 ft. due to sightlines.
But, forward vision would seem to save more lives because of forward speed versus reverse speed. Installed in a Lincoln Navigator, Ford Motor Company’s (Dearborn, MI) TrafficView uses two forward-facing cameras, one mounted out-board on each side. That set-up allows the driver to see around other large vehicles and to look ahead for hidden cars or pedestrians. Ford’s SensorCar technology, mounted in a Mazda 626, represents breakthrough work conducted by the Mazda Technical Research Center (Hiroshima, Japan). Pre-impact warning systems are key to upcoming improvements in collision avoidance; and, the SensorCar test vehicle aims to reduce accidents and rear-end impacts. It does so via grille-mounted LIDAR that monitors pedestrian movements ahead of the vehicle, and activates a warning light, a beeping tone and even blows the horn if someone steps into the projected travel path. In spite of those forewarnings, if the driver collides with a pedestrian, a wide over-the-hood airbag—triggered by a pre-crash sensor—unfurls just above the bumper and conforms to the hood’s shape. The airbag affords abdominal and hip protection to medium and large persons and head and chest shielding for children and smaller-stature adults. A separate cowl airbag deploys from in front of the windshield base. This airbag covers “hard points” including wipers and lower A-pillars to offer head defense.
Hardware + Software = Safety
Micron Technology, Inc.’s (Boise, ID) specialty image sensors are inspiring new approaches to building safer, smarter automobiles, relying on the low-light sensitivity and image quality of their MT9V022 and MT9V111 chips. The detectors feature a global shutter (simultaneous total-pixel exposure) system, enhanced near-infrared sensitivity, wide VGA format and exceptional dynamic range. The MT9V022 provides automotive scene understanding for lane tracking and departure warning, collision and pedestrian avoidance, smart airbags and accident reconstruction.
Advanced Imaging readers will be quick to note that any imaging sensor mounted in or on a moving vehicle is confronted by two major challenges: motion and vibration. Sensors also must contend with dirt, humidity, salt, fuel additives and extremes of temperature, all without significant deviation in accuracy over the vehicle’s entire lifetime. To that end, realtime, full mechanical computer simulations are permitting engineering teams to review even very complex models using virtual reality interfaces. To help identify and resolve such implementation issues earlier—thus reducing the number of physical prototypes—many automotive manufacturers are conducting project reviews in SGI Reality Center facilities, powered by Onyx2 and Onyx 3000 series systems from Silicon Graphics, Inc. (Mountain View, CA).
As image-based active and passive driver-assist safety features continue to emerge, the software content to support them will grow exponentially. Under increasing pressure to accelerate timelines, augment quality and reduce costs, engineers look for alternatives to traditional approaches that are neither terribly effective nor efficient, hampered by pricey or incomplete testbeds and requirements for nearly continuous re-programming. To bridge traditional gaps in software development, many major automotive manufacturers turn to The MathWorks, Inc. (Natick, MA). MATLAB and Simulink users enjoy significant advantages; designing, modeling, simulating, testing and programming control strategies in a unified environment. Further, “executable specifications” can be begun in the intuitive and self-documenting background of Simulink and Stateflow, both buttressed by MATLAB’s computational, analytical and visualization capabilities.