Abstract
This paper reports on an artificial eye of intelligent vehicle. The author had presented a method to recognize obstacles in the path of the vehicle and materialized it as a road pattern recognition system. The system employed a stereo TV camera assembly and a hard wired logic unit. It was capable of recognizing obstacles in the recognition region which was approximately 6.4m wide, 1m deep and 9m ahead the vehicle.
However, the depth of the region was not enough to make the vehicle run smooth. Then, this time, the author made an attempt to expand the recognition region. The improved method makes use of a fact that the recognizable area changes its location when either of the signals from 2 sets of TV camera is delayed. The TV signals are delayed in different 32 ways and all the delayed signals are processed in parallel to detect presence of obstacles in continuous 32 recognizable areas. The whole recognition region covers an area of about 16m wide, 20m deep and 5m ahead the vehicle. It is divided into sections, 128 horizontally and 16 deeply. Presence of obstacles on each section is checked, and 2 dimensional distribution of obstacles is figured out. Then, the clear path for the vehicle is computed from the 2 dimensional distribution of obstacles, and output to a problem solving device. The total time to process a vision is 36ms.
The road pattern recognition system was equipped on the intelligent vehicle and put to experiments on a test course. The system functioned successfully to recognize 2 dimensional distribution of obstacles such like other vehicles and roadside guardrails.
