An Evaluation Method for Visual Search Stability in Urban Tunnel Entrance and Exit Sections Based on Markov Chain

Zhiwen Zhou,Ting Tan,Tao Lu,Jianxiao Ma,Song Fang,Gen Li

doi:10.1109/access.2020.2986272

Abstract

Tunnel section is the throat of transportation and attracts lots of attentions. This paper proposed a method to evaluate the driver's visual search stability based on the Markov Chain properties of eye movements. Firstly, visual and physiological data about 16 participants driving through 13 urban tunnels were collected. Then, the view area was divided into six AOIs (Area of Interest) by fast clustering of the drivers' fixation points. The one-step fixation transition probability and the stable distribution of different lane changing behavior were obtained based on the division of the view area. The probability of transition from the forward windscreen to the left rearview mirror and other 6 visual parameters were selected as indexes by correlation tests. And the first four principal components which covered 96.1% of all information were extracted. Then an evaluation method for visual search stability was implemented by principal component analysis. In order to validate the method, average lane change times, average speed and SDNN (Standard Deviation of NN Intervals) of the drivers' heart rate were clustered into two categories. According to the consistency between the evaluation results and the clustering results, the evaluation method proposed in this paper has been proven to be reliable. Finally, the score threshold for judging the driver's stability was obtained as $E=0.313$ . The method could be applied to adjustment of tunnel facilities, assistance in driving training and development of auto driving system by assessing whether a driver can take over the control of the vehicle or not.

Highlights

Driving on public roads is a highly complex and responsible action, with mistakes or risks-taking possibly having fatal consequences
This paper mainly studied short urban tunnels, and a method based on Markov chain was proposed to evaluate drivers’ visual search stability which reflects the ability to observe and process information in urban tunnel entrance and exit sections
ONE-STEP FIXATION TRANSITION AND STATIONARY DISTRIBUTION The probability of the repeat fixation in 6 areas of interest (AOI) with 3 manoeuvres under high and low traffic flow density in tunnel entrance and exit sections were obtained respectively

Summary

Introduction

Driving on public roads is a highly complex and responsible action, with mistakes or risks-taking possibly having fatal consequences. As reported in [1], approximately 1.35 million people in a year lost their lives and 50 million people were injured in road accidents. As for public area, tunnels are considered as scary spaces with severe accidents which are difficult to handle. It has been proven that tunnel entrance and exit sections are accident-prone positions [2], [3] for its sharp changes of the driving environment. Because drivers’ psychological factors are directly impacted by mutual interference of vehicles and the sudden change of environment. Driving behavior is seriously affected by the bad feelings, such as fear, hesitancies and anxiety, while approaching

Methods

Results

Discussion

Conclusion