Experiment of Safety Drive in an Intersection by Visual Assistances based on HIR System

摘要

In resent years, AHS(Advanced Highway Systems) has been studied to realize the automated driving. In AHS, images from cameras are used for sensing, not given to humans, and cars are controlled by using these images, GPS, VICS, etc.. However, machines make errors in sensing and can't always control cars certainly. This sometimes causes serious traffic accidents. It is very important to solve this problem in AHS. To solve the problem in AHS, we have proposed a new concept system "HIR (Human-Oriented Information Restructuring)". In HIR, images from cameras are used as "images", not for sensing in AHS. We reported an approach of applying HIR system to an intersection. However, the experiment was made on only a diorama and there was also a prerequisite condition that we know the position and direction of all cameras at real time. In this paper, we make experiment on a real intersection. And we remove the conventional prerequisite condition by using some sensors. In addition, we use the difference between three frame images in algorithms to extract an opposite going straight car. Finally, we generate visual assist images and the images displayed on the car monitor are evaluated.