Internet-of-Things Based Drowsiness and Distraction Detection of Drivers Using Computer Vision

Adil, Muhammad

DSpace Home
→
E-Theses
→
SINES
→
Computation Science & Engineering
→
MS
→
View Item

dc.contributor.author	Adil, Muhammad
dc.date.accessioned	2021-11-29T11:27:06Z
dc.date.available	2021-11-29T11:27:06Z
dc.date.issued	2020-10-01
dc.identifier.other	RCMS003235
dc.identifier.uri	http://10.250.8.41:8080/xmlui/handle/123456789/27760
dc.description.abstract	Drowsiness and distraction are one of the biggest reasons for road traffic casualties. According to the World Health Organization (WHO), 1.35 million people die annually due to traffic accidents. The situation of traffic accidents is even worse in developing and underdeveloped countries. In some of the developing countries, traffic accidents cost above 3 percent of their Gross Domestic Product (GDP). Many research efforts have been done in the recent past to solve this issue through driver monitoring systems. Most of the scientific techniques developed in this regard are customized for developed countries. A research gap exists in some aspects of existing techniques e.g. efficiency, high cost of development and deployment, dependency on standard road and vehicle conditions, reproducibility of lab testing accuracy in field testing. In this work, an Internet-of-Things (IoT) based technique is proposed to detect drowsiness and distraction of drivers in real-time using a computer vision approach. A decision tree algorithm is implemented due to its compatibility with applications built on decision control logic and suitability for edge devices. A detailed performance comparison is also made by implementing the application on devices like Raspberry pi-3, Nvidia Jetson Nano and 7th generation core i7 system. The edge device is also integrated with the cloud end for remote visualization of real-time data. The results of proposed technique are also tested on two different datasets of inattentive drivers. The first dataset used for this purpose is DROZY Dataset which is captured through an Infrared camera, hosted by Open Repository and Bibliography (ORBi). The second dataset used for measuring accuracy is UTA Real-time Drowsiness Dataset. This dataset contains videos captured through an RGB camera and hosted by the University of Texas, Austin.	en_US
dc.description.sponsorship	Dr. Muhammad Tariq Saeed	en_US
dc.language.iso	en_US	en_US
dc.publisher	RCMS NUST	en_US
dc.subject	Drowsiness Detection, Distraction Detection, Driver Monitoring System, Computer Vision, IoT, Decision Tree.	en_US
dc.title	Internet-of-Things Based Drowsiness and Distraction Detection of Drivers Using Computer Vision	en_US
dc.type	Thesis	en_US