Keiji Saneyoshi

3-D Image Recognition System For Drive Assist

This paper describes a three-dimensional image recognition system developed to recognize road boundaries and obstacles. Based on stereoscopy, the system can detect two white lane marks, two fences along the road and as many as six obstacles simultaneously within 0.1 second. This system is effective during night, and bright sunny days which cause strong shadows. It can also work under counter-light, and adjust to the rapid change of brightness.

3-D image recognition system by means of stereoscopy combined with ordinary image processing

A new image recognition system using stereoscopy and ordinary image processing method has been developed for an autonomous vehicle. The information which has been lost in the distance distribution image that was yielded by stereoscopy was revived from the original image by ordinal image processing method. And it was combined to the information of the distance distribution to increase the recognition ability. As the result, this recognition system can be used not only in a simple scene such as on highways but also in a complex scene such as on streets in a town though it is not perfect. In this report, three examples treated by this method are described.