Driver Safety Monitoring System

Miigwech!

This is my story!

I live in Jackhead First Nation, on the western shore of Lake Winnipeg. Driving isn't just for fun it's how we get everything. Medical appointments, groceries, school, work. That means hours on the highway. Getting tired isn't a maybe. It's a when.

I've seen it happen. People come back talking about how they almost drifted off. That scared me.

What pushed me to do this project: vehicle crashes are a leading cause of death for Indigenous youth under 25. Those aren't just statistics they're our brothers and sisters.

I wanted to build something that helps. Something that works with what people already have.

So I asked: can a regular webcam detect when a driver is about to fall asleep?

I built a system using Python and free software that tracks eye closure using two key measurements: EAR (Eye Aspect Ratio) to detect when eyes are closing, and MAR (Mouth Aspect Ratio) to detect yawning. If your eyes close for 5 seconds, it beeps and flashes yellow. At 10 seconds, a red alarm.

This project is for Jackhead. For every driver who leaves home with a family waiting. It won't fix everything, but if it saves one person, it's worth it.

How did you perform your experiment or develop your solution?

First, I did background research. I looked up drowsy driving stats on trusted websites like Transport Canada and the Canadian government. I also read studies about how scientists measure tiredness using eye closure that's where I learned about Eye Aspect Ratio and PERCLOS, which is the gold standard for drowsiness detection.

Then I started building. I used a laptop webcam, Python, and Google's MediaPipe Face Mesh a free tool that maps 468 points on your face in real time. I wrote code to calculate how open or closed someone's eyes are.

To test my system, I ran it on myself and family members in different conditions: normal room light, bright sunlight, and dim lighting. I also tested with and without glasses. We simulated tired driving by purposely closing eyes for different lengths of time 2 seconds, 5 seconds, 10 seconds and recorded how the system reacted.

I collected data on:

How long eyes stayed closed

How many yawns were detected

How far you'd travel blind at highway speed

False alarms (times it beeped when eyes were open)

I controlled variables by keeping the webcam in the same position for each test and using the same lighting setup for each round.

Materials used:

Laptop with webcam

Python (free software)

USB Arduino with LED lights and buzzer

Tell us your results! What did I find out?

My system works. A regular laptop webcam really can detect when a driver is getting too tired behind the wheel.

How the prototype works:

The system uses free face-mapping software (MediaPipe) that puts 468 tiny dots on your face. My code watches the dots around your eyes and calculates something called Eye Aspect Ratio basically a number that tells me how open or closed your eyes are.

If EAR drops below 0.21 (eyes closed), a timer starts. At 5 seconds closed: yellow warning light + one beep. At 10 seconds closed: red light + continuous buzzer. The system also counts yawns and tracks if your head turns away for too long.

My main results

I tested the system on myself and family members in different conditions. Here's what I found:

Eye closure detection: The system catches when eyes close with about 92-95% accuracy in normal lighting.

Warning timing: At 5 seconds of eye closure, the alert triggers reliably. At highway speed (110 km/h), that's 153 metres traveled blind longer than a football field.

Different lighting: Works best in normal room light (90-95%). Dim lighting drops to about 70-80% accuracy.

Glasses: Non-reflective glasses are fine (85-90%). Reflective glasses are harder (70-80%).

What the blind distance numbers mean

I calculated how far you'd travel with your eyes closed at highway speed:

3 seconds = 92 metres (1 football field)

5 seconds = 153 metres (1.5 football fields)

10 seconds = 306 metres (3 football fields)

This matters because even a 3-second microsleep at highway speed means you've traveled the length of a football field without seeing the road.

What I learned from my results

First, affordable drowsiness detection is possible. You don't need a $50,000 luxury car. A laptop, webcam, and free software can work.

Second, the 5-second warning is the most important feature. It gives drivers enough time to recognize they're tired and pull over before something bad happens.

My system won't stop all drowsy driving crashes. But it can give drivers a wake-up call before it's too late. And for remote communities like Jackhead, that could save lives.

My results show that a regular webcam really can detect when a driver is getting too tired.

The system catches eye closure at 5 seconds and gives a warning yellow light and a beep. At 10 seconds, red alarm.

At highway speed, 5 seconds with eyes closed means you've traveled 153 metres blind. That's longer than a football field.

What I learned is that the warning itself is what matters most. Drivers don't need the system to drive for them. They just need a wake-up call before it's too late. A 5-second warning gives someone time to pull over and rest.

But before any technology kicks in, there is traditional knowledge. In our communities, smudging with sage or sweetgrass cleanses the eyes, ears, and mouth before a journey. It clears the mind and offers protection for the road ahead. My system honors that it watches over your eyes so you can see clearly, listens for yawns so you know when you're tired, and speaks a warning to wake you up. Technology and tradition working together.

My conclusion is simple: affordable drowsiness detection is possible. You don't need a luxury car. A laptop, a webcam, and free software can save lives—especially in remote communities like Jackhead where people drive hours just to get basic necessities.

This won't stop all crashes. But if it saves one person brings one driver home that's a win. Smudge. Drive. Get home safe.

Future Extensions

Next, I would replace the laptop with a Raspberry Pi—a small, cheap computer that fits on any dashboard. I would add an infrared camera so the system works at night. The code could also send a text alert to a family member if an emergency happens.

I would test the system with real drivers on reserve roads, not just in my living room. And I would add a voice warning: "Wake up. Pull over."

Every driver who leaves home deserves to come back. This system is just the beginning.

Tell us about the great people who helped with your project!

First, Thank you to my teacher who helped me learn to code. I wrote most the system myself, but when I got stuck like when eye detection wasn't working my teacher showed me how to find and fix my mistakes. That help made my project work.

Thank you to my family for supporting me for letting me test the system on you and believing in this project.

Finally, thank you to the Canada-Wide Science Fair for creating a space where young people can share ideas that matter to their communities. :)

This project is for Jackhead. Every driver who leaves home deserves to make it back.

Canadian Council of Motor Transport Administrators. (n.d.). Drowsy driving. Retrieved from https://ccmta.ca

Google. (n.d.). MediaPipe Face Mesh. Retrieved from https://developers.google.com/mediapipe/solutions/vision/face_landmarker

OpenAI. (2025). ChatGPT. Used for refining ideas, and generating image prompts. https://chat.openai.com

Leonardo.ai. (2025). AI image generation platform. Used for creating symbolic illustrations and project visuals. https://leonardo.ai

Grammarly. (2025). Grammarly writing assistant. Used for grammar checking and polishing written content. https://grammarly.com

Sawant, A. G., et al. (2025). A real-time driver drowsiness detection system using

MediaPipe and eye aspect ratio. arXiv. https://arxiv.org/abs/2511.13618

Transport Canada. (2023, November 16). *Commercial vehicle collisions in Canada, 2012-2021*. Retrieved from https://tc.canada.ca/en/road-transportation/statistics-data/commercial-vehicle-collisions/2021/commercial-vehicle-collisions-canada-2012-2021