|
Abstract
|
A nighttime pedestrian recognition assistance system that is expected to help reduce nighttime pedestrian accidents has been developed.
This system is fitted with far-infrared stereo cameras that can make an image of what is ahead, even when it is dark. The system displays the image on a heads-up display directly in front of the driver. At the same time, through the use of image processing, pedestrians crossing the road ahead, as well as those on the side of the road are detected, and the driver is informed of their presence, with a warning sound and via framed highlight of the pedestrian in the display image.
This system's pedestrian detection algorithm is based on stereoscopic distance data. The system identifies pedestrians through their position and movement as well as size and shape. Results of evaluations on suburban and city roads show good detection performance.
In addition, the results of monitor driving tests using this system, carried out on public roads using test subjects, have confirmed that this system is effective in detecting pedestrians at night.
https://doi.org/10.4271/2006-01-0461
Language: ja |