Abstract:
"
In past years, drowsiness of drivers has been one of the major reasons for road accidents.
These accidents can lead to deaths, major physical injuries and considerable financial
losses.People are becoming more and more busy nowadays. So most of the time they
cannot take enough rest before starting the driving. For example most of the time people
drive after work without enough rest. So it is hard to completely prevent these kinds of
situations. Best practical solution we can give is to build an efficient cost effective
system which can identify the drowsiness of the driver as soon as possible.
In this research I have implemented a system which uses only visual inputs to identify
drowsiness situations. Because one purpose of this research is to give a cost effective
solution. Using expensive human based sensors or vehicle based sensors will make that
purpose invalid.
In this research I have used multiple convolution neural networks to identify drowsiness
factors like blinking, yawning and fatigue expressions. This research system will
continuously process video input and decide whether this person is yawning, blinking
or having fatigue expressions. After detecting those factors the system will give a score
for each factor. Based on these scores, the system will decide whether a driver should
be alerted or not"
