JSTE Journal of Traffic Engineering
Online ISSN : 2187-2929
ISSN-L : 2187-2929
Special Edition A (Research Paper)
Research on anomaly driving detection applying deep learning using driving behavior at intersections
Kota MIYAUCHIKazuyuki TAKADAMoeko SHINOHARAMakoto FUJIU
Author information
JOURNAL FREE ACCESS

2021 Volume 7 Issue 2 Pages A_19-A_28

Details
Abstract

In recent years, with the development of intelligent transportation systems (ITS), the performance of automobiles has improved. On the other hand, in Japan, intentional dangerous driving by drivers and accidents by elderly drivers have become serious social problems. As a method to solve these problems, it is necessary to have a technology that detects the factors that lead to the occurrence of accidents and allows the vehicle to take appropriate control.

In this study, we focus on driving behavior at intersections and propose a method to detect factors that lead to accidents appear. Applying the Long Short Term Memory Auto Encoder, we assumed that anomaly driving occurred when driving behavior that deviated from normal driving was observed and verified the effectiveness of the method. Comparing the proposed method with previous studies, it was confirmed that the detection accuracy was the highest and the detection timing could be detected relatively quickly.

Content from these authors
© 2021 Japan Society of Traffic Engineers
Previous article Next article
feedback
Top