4D- Distracted Driver Detection Device

Overview

Every year there are 6 million car accidents in just the United States alone. The leading cause of these accidents is distracted driving, usually due to cell phone usage or drowsiness. Roughly 9 people are killed and 1,000 people are injured daily in accidents where a driver is distracted. A study by the AAA Foundation for Traffic Safety concluded that drowsiness was the main cause of 9.5% of accidents in general, and 10.8% of fatal accidents. The main reason for this gross underestimation is because it is hard to determine if the driver was asleep or distracted behind the wheel. The purpose of this project is to use readily available technology to create a device that detects if the driver is either distracted or falling asleep behind the wheel, and alerts them. A Raspberry PI3 minicomputer, camera module, buzzer, LED lights, and other electrical components were used to build a prototype device. The design goals were to detect the driver’s face in the camera, determine if the driver’s eyes are opened or closed and if the driver’s mouth is open in a yawn, determine if the driver’s eyes are on the road or if they are distracted, and give out a visual and a loud audible alert repeatedly until the device detects that the driver has their eyes on the road. The Raspberry PI and the components were configured to detect driver’s face in the camera’s video stream and use the OpenCV, Dlib and facial landmarks packages on OpenCV along with Python code to implement facial recognition. Since the goal was to process a live video, it was determined that OpenCV’s Haar cascade face detector handled video feed much better than dlib HOG + Linear SVM face detector. Using Haar Cascade results in a slight loss in accuracy compared to HOG, but that is made up for by the increased sampling on a video feed enabled by faster processing. The prototype was tested extensively to determine the optimal threshold for drowsiness and distraction. Since it detects eye movements, it was tested on people with and without glasses, of different genders and skin tones, as well as in 4 different light conditions; very bright light, regular daylight, dim light, and darkness. After analyzing this data and optimizing the algorithm thresholds, the prototype was found to successfully work more than 90% of the time. Installing this type of device in all existing cars is relatively inexpensive. The cost of materials of this prototype was about $55, and it would significantly reduce the number of deaths due to vehicle crashes caused by drowsy and distracted driving.

Competition history

• AJAS 2020 Category not listed

Related projects

Closest projects by meaning, across every fair and year in the corpus.

Browse more like this

• AJAS None projects
• Every AJAS 2020 project