Improving automatic detection of driver fatigue and distraction using machine learning
URL: http://arxiv.org/abs/2309.16742
archive: archived pipeline: cataloged verified
Abstract
Changes and advances in information technology have played an important role in the development of intelligent vehicle systems in recent years. Driver fatigue and distracted driving are important factors in traffic accidents. Thus, onboard monitoring of driving behavior has become a crucial component of advanced driver assistance systems (ADAS) for intelligent vehicles. In this article, we present techniques for simultaneously detecting fatigue and distracted driving behaviors using vision-based and machine learning-based approaches. In driving fatigue detection, we use facial alignment network (FAN) to identify the 68 facial feature points in the image. Calculating facial feature points' distance to detect the opening/closing of the eyes and mouth. In distraction detection, we use a convolutional neural network (CNN) based on the MobileNet architecture to identify various distracted driving behaviors. Experiments are performed on the PC with a webcam and using public datasets and our own datasets for training and testing.
Summary
MSc thesis (University of Birmingham, Dubai) developing a vision-based pipeline that simultaneously detects driver fatigue and distraction. A Facial Alignment Network (FAN) localizes 68 facial landmarks to compute eye- and mouth-aperture distances (PERCLOS-style fatigue indicators), while a MobileNet-based CNN classifies distraction behaviors. Training and testing used both public datasets and authors' own webcam-captured datasets on a PC; reported accuracy and computation time exceeded prior approaches.
Key finding
Combining FAN-derived facial landmarks for PERCLOS-style fatigue estimation with a MobileNet CNN for distraction classification yields better accuracy and inference speed than prior single-purpose detectors when run on a webcam-equipped PC.
Methodology
System development and benchmarking. Vision pipeline with two heads (FAN-based fatigue indicators from eye/mouth landmark distances; MobileNet CNN for distraction-behavior classification). Models were trained and evaluated on public datasets plus author-collected datasets, with comparisons of accuracy and computation time against baseline approaches.
Quality score: 5 / 5