OPTIMIZED DRIVER ACTIVITY DETECTION USING DEEP  LEARNING

SHEWADIRES, MENAN KEMAL

dc.contributor.author	SHEWADIRES, MENAN KEMAL
dc.date.accessioned	2024-03-21T10:53:43Z
dc.date.available	2024-03-21T10:53:43Z
dc.date.issued	2023-06
dc.identifier.uri	http://ir.bdu.edu.et/handle/123456789/15706
dc.description.abstract	Vehicle accidents are becoming a global public health concern, especially in Ethiopia. According to a World Health Organization report in 2015, Ethiopia is one of the 50 countries with the deadliest roads in the world. There are various attribute towards to vehicle crashes i.e. not respecting traffic laws, high speed, poor diving skills, road problem and undisciplined diver behavior. According to roads authority report of 2005 indicated that 1-3% of the accidents are related to the road but around 81 % of the accidents are related to diver problems. As indicated by the statistics fatigue related accidents are the most common type of accident in Ethiopia. Technological advancement of computer vision is gradually finding applications in different problem domains as like vehicle driver monitoring. Therefore, the implementation of vision technology in such area will have a paramount importance to alert driver and all travelers finally it will minimize vehicle accident related with the driver problem. So, the main purpose of this research paper is to develop a computer vision system that identifies the driver facial expressions like fatigue, chewing chat (ጫት), eating food፣ drinking water, etc and alerting to the driver, assistant and the rest of traveler. The previous works focused on identification of sleep for the driver and as per our review no work is conducted that have many dimension facial reading not only for other country but also in our country Ethiopia. So this paper focused on facial expression reading and eyes location for alerting to the drivers using computer vision techniques and machine learning techniques. We prepare our own dataset and we use CNN for future extraction FFNN and SVM for classification. FFNN and SVM classify 74.24% and 69.69% accuracy respectively. Keywords: CNN, local feature descriptor, feature extraction, SVM, FFNN, thresholding.	en_US
dc.language.iso	en_US	en_US
dc.subject	Information Technology	en_US
dc.title	OPTIMIZED DRIVER ACTIVITY DETECTION USING DEEP LEARNING	en_US
dc.type	Thesis	en_US