Smart vehicles
Smart vehicles use a vision system to detect other vehicles, respond to traffic signals, and avoid
pedestrians and obstacles.
By C. G. Masi,Contributing Editor
Actor Arnold Schwarzenegger searches for his clone in the motion picture The Sixth Day and is driven
to an airport by an intelligent car. Detecting close traffic, pedestrians, and surrounding highways, the
car appears to drive itself as the driver sits behind a steering wheel.
Such a smart-vehicle system appears more plausible to the general public than it does to systems
integrators who know the complexities involved in a vision system imaging a target-rich environment.
Yet, researchers at DaimlerChrysler Research (Ulm, Germany) have built a smart-vehicle-theory
demonstrator called the Urban Traffic Assistant (UTA) that uses brute-force calculations performed
off-line from images stored on video tape.
As yet, the UTA cannot make complex decisions. However, the DaimlerChrysler research team has
demonstrated that by using a multitude of algorithms it can maintain separation from the vehicle ahead,
see and respond to traffic signals, and avoid pedestrians and obstacles while moving through urban
traffic situations at normal driving speeds under the supervision of a human driver.
Hardware architecture
A pair of video cameras from JAI AS (Copenhagen,
Denmark) serve as the UTA's eyes. Arranged in a stereo
configuration, the cameras are mounted on a horizontal bar
near the rear-view mirror. The bar separates the two cameras
by approximately 30 cm and maintains their orientation. The
horizontal scan lines of both cameras must be tightly aligned
and strictly parallel to the line joining their optical axes.
Furthermore, the cameras' optical axes must be coplanar, as
well. These specifications are needed because the vision
system must infer a correspondence between the scan lines
of one camera and those of the other. These video cameras
provide progressive-scan images with 720 x 576-pixel
resolution.
The UTA uses an Imaging Technology Inc. (now Coreco Imaging; St Laurent, Que., Canada) IMPCI
frame grabber and a commercial monitor with 1024 x 786-pixel resolution. Its central processing unit
comprises several Ethernet-interconnected general-purpose PCs with dual-Pentium 600-MHz
processors, MMX-based technology from Intel Corp. (Santa Clara, CA) for image processing, and one
Motorola (Tempe, AZ) Lynx/604e PowerPC for controlling the vehicle's sensors and actuators (see Fig.
1).
Click here to enlarge image
object to
System connections are established by so-called administrators. An administrator controls a set of
Connecting a group of modules forms an application. One version of pedestrian recognition, for
Click here to enlarge image
Click here to enlarge image from that pixel to the nearest feature pixel. The
image-feature pixel. A DT value of zero equates
Click here to enlarge image

版权声明：本站内容均来自互联网，仅供演示用，请勿用于商业和其他非法用途。如果侵犯了您的权益请与我们联系QQ:729038198，我们将在24小时内删除。