Abstract:
Road traffic accident in Malaysia is a heavy concern in these days. Among the top factors of traffic accidents, the fatigue and drowsiness of drivers often times contributed to the increasing number of cases and fatality rate of accidents. This research aims to develop a computer vision system to detect such fatigue and drowsiness of the drivers and wake them up from the split-second nap. The implementation of this research is to develop a drowsiness detection system implemented in a compact development board to assist drivers to awaken from microsleep during driving on fatigue due to long driving hours and various other reasons. This research used a Raspberry Pi 4 along with the official Raspberry Pi camera module V2 and an active buzzer module as waking mechanism for the system. The development used and experimented on the Haar cascade classifier and Histogram of Oriented Gradient + linear Support Vector Machine in the effort of determining the best suitable model to be used for drowsiness detection in terms of speed and accuracy. Both models were run and tested to work properly. The implementation of the Haar cascade classifier produced the best performance in terms of speed and response time to detect drowsiness. On the other hand, the HOG + SVM had better accuracy when compared to the Haar cascade classifier even in low illumination. Having said that, the response time is significantly slower than Haar model which caused a problem regarding the reaction time of drivers to react on time. To conclude, the Haar cascaded classifier is decided as the most appropriate model to be applied for the development of a drowsiness detection system.
