Abstract:Fatigue driving is one of the main factors which affect the safety of drivers and passengers in mountain freeway. To improve the driving safety, the application of fatigue driving detection system is a crucial measure. Accuracy, speed and robustness are key performances of fatigue detection system. However, most researches pay attention to one of them, instead of taking care of them all. It has limitation in practical application. This paper proposes a novel three-layered framework, named Real-time and Robust Detection System. Specifically, the framework includes three modules, called facial feature extraction, eyes regions extraction and fatigue detection. In the facial feature extraction module, the paper designs a deep cascaded convolutional neural network to detect the face and locate eye key points. Then, a face tracking sub-module is constructed to increase the speed of the algorithm, and a face validation submodule is applied to improve the stability of detection. Furthermore, to ensure the orderly operation of each sub-module, we designed a recognition loop based on the finite state machine. It can extract facial feature of the driver. In the second module, eyes regions of the driver were captured according to the geometric feature of face and eyes. In the fatigue detection module, the ellipse fitting method is applied to obtain the shape of driver's pupils. According to the relationship between the long and short axes of the ellipse, eyes state (opening or closed) can be decided. Lastly, the PERCLOS, which is defined by calculating the number of closed eyes in a period, is used to determine whether fatigue driving or not. The experimental results show that the comprehensive accuracy of fatigue detection is 95.87%. The average algorithm rate is 32.29 ms/f in an image of 640×480 pixels. The research results can serve the design of a new generation of driver fatigue detection system to mountain freeway.

Illumination Intelligent Adaptation and Analysis Framework: A comprehensive solution for enhancing nighttime driving fatigue monitoring

Advancing driver fatigue detection in diverse lighting conditions for assisted driving vehicles with enhanced facial recognition technologies

Efficient detection of driver fatigue state based on all-weather illumination scenarios

Nighttime Driving Safety Improvement Via Image Enhancement for Driver Face Detection

Highway Safety with an Intelligent Headlight System for Improved Nighttime Driving

A Multi-Feature Fusion and Situation Awareness-Based Method for Fatigue Driving Level Determination

A Driving Risk Assessment Framework Considering Driver’s Fatigue State and Distraction Behavior

Improving automatic detection of driver fatigue and distraction using machine learning

Driver fatigue detection method based on multi-feature empirical fusion model

Bio-Inspired Dark Adaptive Nighttime Object Detection

Remote Photoplethysmography and Motion Tracking Convolutional Neural Network with Bidirectional Long Short-Term Memory: Non-Invasive Fatigue Detection Method Based on Multi-Modal Fusion

R2DS: A novel hierarchical framework for driver fatigue detection in mountain freeway

Research Progress on the Development of Fatigue Driving Detection Based on Deep Learning

Fatigue Driving Detection Based on Deep Learning and Multi-Index Fusion

Real-Time System for Driver Fatigue Detection by RGB-D Camera

Multimodal Driver Condition Monitoring System Operating in the Far-Infrared Spectrum

Research on a Real-Time Driver Fatigue Detection Algorithm Based on Facial Video Sequences

An fNIRS labeling image feature-based customized driving fatigue detection method

Advancements in the Intelligent Detection of Driver Fatigue and Distraction: A Comprehensive Review

Design and Implementation of a Drowsiness-Fatigue-Detection System Based on Wearable Smart Glasses to Increase Road Safety

An Improved Fatigue Detection System Based on Behavioral Characteristics of Driver