Driver Fatigue and Distraction Analysis Using Machine Learning Algorithms
Fatigue and distraction are the two critical causes which account for the majority of the road accidents leading to innumerable fatalities. There is a need for real-time fatigue and activity detection while the vehicle is in motion. This paper presents an
- PDF / 372,344 Bytes
- 10 Pages / 439.37 x 666.142 pts Page_size
- 101 Downloads / 256 Views
Abstract Fatigue and distraction are the two critical causes which account for the majority of the road accidents leading to innumerable fatalities. There is a need for real-time fatigue and activity detection while the vehicle is in motion. This paper presents an overview of fatigue and distraction analysis using two machine learning algorithms, KNN and CNN. Fatigue symptoms like eyelid closure and yawning are computed by eye and mouth aspect ratios through facial landmarks mapping on a live video to obtain high throughput with alarm generation for different states. Fatigue states are classified using KNN which unlike other algorithms provides a very simplistic approach with adequate accuracy. Distraction types are classified using CNN implemented on a data set containing different driving activities comprising of specific actions performed by a driver inside the vehicle. Images are trained and classified followed by test prediction. VGG16 transferred learning has been applied to CNN to increase the accuracy. Thus, the system is able to efficiently predict and classify driver states based on input feeds through a camera and also drive activities from a model leading to a robust driving-based analysis. Keywords Fatigue · Activity · KNN · CNN · VGG16
R. Rathi (B) · A. Sawant · L. Jain · S. Kulkarni Sardar Patel Institute of Technology, Bharatiya Vidya Bhavan’s Campus, Munshi Nagar, Andheri West, Mumbai 400058, India e-mail: [email protected] A. Sawant e-mail: [email protected] L. Jain e-mail: [email protected] S. Kulkarni e-mail: [email protected] © Springer Nature Singapore Pte Ltd. 2021 D. Gupta et al. (eds.), International Conference on Innovative Computing and Communications, Advances in Intelligent Systems and Computing 1165, https://doi.org/10.1007/978-981-15-5113-0_88
1037
1038
R. Rathi et al.
1 Introduction Denser road networks have led to a higher number of road accidents. As per reports of 2017, 16 citizens were killed and 53 injured every hour on Indian roads with many accidents unreported [1]. Of the many causes of deaths occurring in road accidents, driver drowsiness has caused majority of the accidents. The methods for fatigue detection of a driver are classified as behavioural-based, vehicle-based and physiological based [2]. In physiological-based system, sensors are employed to record driver data [3]. These sensors monitor the EEG, ECG and EOG signals of the driver and determine the fatigue level of the driver which makes it an intrusive method [2, 4]. The connection of such electrode sensors tends to be quite excruciating for the driver and expensive though highly efficient. In this research, using a behaviouralbased approach, a simple camera is used for driver fatigue detection. By tracking of facial visual movements in real time, we can understand the level of alertness of the driver and alarm the driver if found in a tired or sleepy state, thus making it a nonintrusive technique. The eye and mouth aspect ratios are the parameters considered in analyzing states of
Data Loading...
