Generating Reference Views of Traffic Intersection for Safe Driving Assistance - Digital Signal Processing for In-Vehicle Systems and Safety - page 194

Chapter 14
Generating Reference Views of Traffic
Intersection for Safe Driving Assistance

Jien Kato and Yu Wang

Abstract In this chapter, we address the problem of driving assistance along traffic
intersections by providing drivers with additional visual information to expand their
visual field. Our goal is to generate image stream of a virtual viewpoint which
follows the host vehicle from a higher position, using images from multiple
roadside cameras. Our approach is based on view morphing, but we extend it by
integrating robust fundamental matrix estimation and sparse key point matching.
This enables some tasks which previously rely on manual operation to be done
automatically.

Keywords Driver visual field • Driving assistance • Image-based rendering (IBR)
• Intersection assistance • Reference view • Vehicle blind spots

14.1 Introduction

Driving is becoming more and more stressful due to increasing traffic density.
The situation seems to be more serious at intersections. According to the Annual
Report 2007 from the National Police Agency of Japan, 46.3% of all traffic
accidents in Japan occurred near intersections. In addition, a very large percentage
of them happened either because of blind spots in the vehicles or inter-object
occlusion due to traffic density at the intersections. These issues limit a driver's
visual field. As a result, drivers find it more challenging to monitor their surround-
ings and forthcoming situations.
In the context of intersection assistance, Benmimnoun et al. presented a system [ 1 ]
that utilizes intervehicle communication which updates the position measurements
received from the onboard GPS and transmits all warning information to vehicles via

J. Kato ( * ) • Y. Wang
Nagoya University, Nagoya 464-8603, Japan
e-mail: [e-mail] ; [e-mail]