33,000 Americans die in motor vehicle accidents every year. These accidents are the leading cause of death for people under 25.
What it does
Bumperz aims to make cutting edge accident and prevention and collision avoidance technology available to all drivers starting with those who spend the most time on the road. It warns drivers when they're approaching an area with a historically higher than average accident rate. It also warns fatigued drivers when changing lanes and will employ proximity detection in a future version to warn drivers when they're too close to a vehicle or obstacle in front of them.
How we built it
Historical accident data was downloaded using the NYC open data api and put into a PostgreSQL database. A rails restful API sits on top of that database and waits for requests from the app. This API will return various warning levels based on density of historical accidents in a given radius.
The image processing for lane changing detection was implemented in Python. Our algorithm is loosely based on what was described in a "RESEARCH ON LANE DETECTION TECHNOLOGY BASED ON OPENCV" by Xu Yang Zhang Ling. Our drifting detection happens in 3 phases, filtering, which includes grayscaling, bilateral filtering, and truncating. Then we do feature detection using SURF, and using the movement of those features we determine lane drifting.
Challenges we ran into
Finding and refining a good lane detection algorithm was very difficult.
Accomplishments that we're proud of
We were able to implement this algorithm and test it on real world data that we collected during the hackathon.
What we learned
Lane change detection in real world conditions is very tricky. We created a novel and effective method to detect lane changes.
What's next for Bumperz