Inspiration
With engines getting quieter and quieter these days it is harder for pedestrians to anticipate oncoming traffic. I am a road biker and a runner, and I often find myself puzzling about how to make roads safer for everyone. Often during unanticipated slowdowns, cars will veer into the shoulder of the road and that can cause bodily injury to pedestrians if any are there. Furthermore, some areas of road don't have shoulders, and a runner must be extra cautious to get around these mismatches in the road. Knowing the status of oncoming vehicles would be a huge god send during these more hazardous pursuits.
What it does
This helps predict high traffic areas and traffic bottlenecks. Hopefully this information could be integrated into apps like Strava which plan running-routes for pedestrians so they can avoid high risk areas. Otherwise, it could be used as a recommender agent to suggest an upcoming change in average speed. This is just a starting point to potentially grow into a process for mapping 'safe' pedestrian roads around a local area.
How we built it
Using the public video footage available on 511.com, I imported the stream of exit 9 on the i87. This is because this exit is more often than not the most congested in upstate New York. While this footage does show traffic ways that pedestrians are not allowed to run on, the hope is to develop a baseline process to extend onto backroad areas more frequented.
Challenges we ran into
The cost of setting up a network of cameras on the specific roadway in question would be greater than any of the awards could reimburse so I had to look for existing camera networks to run the code. The problem with the existing camera networks is that they are not continuous so there can be no geopoint algorithms for continuous maps. Furthermore, the cameras that are situated at exits change perspective, sometimes including different patches of the roadway. In exit 8a for instance, there is a perfect shot of a blind curve with two directions of traffic, this I used, but other days the cameras that were operating including just one way traffic on a straight away, or two-way traffic at a stop light.
Accomplishments that we're proud of
Calculating an impact time to a moving location on the screen.
What we learned
I learned a lot about creating JSON files in order to bridge scripts and applications, especially with using Elastic and its many capabilities.
What's next for Traffic Snake Alert System
In the future once a consistent algorithm can be used to predict oncoming traffic, I'd like to get a grant to provide cameras on back roads that are popular amongst runners and bikers. These cameras could be integrated to a local app for people to be more cautious on the road.
Log in or sign up for Devpost to join the conversation.