The Driver's Attention Level

Abstract

Road accidents are directly proportional to the number of cars on the market. Without car safety systems, this number will keep rising. The main factor for the accidents are drowsiness and fatigue. These can be detected by analysing images with the driver so, an example of a driver monitoring system may include a monitoring camera, mounted in front of the driver. A method based on machine learning and computer vision can be a solution to solve the problem of driver safety. The objectives of our work includes an analysis of different approaches of driver monitoring systems and the implementation of a system based on convolutional neural networks which analyze the images coming from a monochrome infrared monitoring camera placed in front of the driver seat. The goal of this work is to decide if the driver is attentive or not (attentive) on the road. Our research was done by implementing a classifier based on AlexNet architecture and return one of the 6 attention classed. To improve the system accuracy, the face was detected using DNN Face Detector (using OpenCV approach). The final system is able to detect when the driver is not paying attention to the road, based on existing test data.