Quantifying visual road environment to establish a speeding prediction model: An examination using naturalistic...

Area of Science:

• Traffic Safety Engineering
• Driver Behavior Analysis
• Machine Learning Applications

Background:

• Speeding is a primary cause of traffic accidents.
• Existing research lacks models predicting speeding based on road environment features.
• Driver's visual perception significantly influences driving behavior.

Purpose of the Study:

• To develop a speeding prediction model using quantified visual road environment data.
• To enhance pre-warning systems for traffic safety.
• To predict driver speeding behavior and provide timely warnings.

Main Methods:

• Quantified visual road geometry and roadside environment using a visual road environment model.
• Collected naturalistic driving data on rural highways.
• Applied Random Forests and logistic regression for model building and analysis.

Main Results:

• The speeding prediction model achieved over 85% accuracy.
• The model predicts speeding behavior at a future position based on current data.
• The prediction interval allows sufficient time for driver response to warnings.

Conclusions:

• Quantifying the visual road environment is effective for predicting speeding behavior.
• The developed model can be integrated into pre-warning systems to reduce speeding.
• This research contributes to mitigating speeding-related traffic accidents.