Our team was growing concerned with the rising accident toll of bicycle riders. In many city roads, bikes occupy the same lane as cars. However, many drivers are often not paying attention to bikes and this causes collisions. If we could alert the biker of an approaching car before the collision, we could shave crucial seconds off reaction time and potentially save lives.
What it does
It detects cars through haar cascade and a standard android camera.
How we built it
The proof of concept was done using the OpenCV libraries with python. Car detection was done using a haar cascade for cars found online. Detection worked on a sample video, after we restricted the area of detection to the lane directly ahead of the car. Then we ported the program to Android Java, modeling it after a sample face recognition program using a haar cascade. On top of the application, we wanted to add pebble functionality. We added onto the java program to send notifications alerting about oncoming cars as well as writing a pebble app to allow the user to start and end the java program.
Challenges I ran into
Haar cascade had problems with precision and false positives.
Accomplishments that I'm proud of
Using open cv and pebble for the first time ever. This is our first hackathon for all four of us and we are proud that we made something that worked.
What I learned
This was our first hackathon and we definitely took on a large task. We learned how to work with Pebble as well as vision technology.
What's next for 1Lane
Improve cascade file and UI.