With more and more advancing technology, it is time to use it to make the road a safer place.
Abstract
With more and more advancing technology, it is time to use it to make the road a safer place. A small and cheap computer that is put on every junction and warns drivers or the pedestrians about dangerous situations, may be the next frontier of fighting against the number of tragedies behind the wheel.
Project Goal
This is a proof of concept project that aims to answer the question: “can we track people using a cheap computer with OpenCV and python?”. If the answer is yes, then putting more research and effort into it can dramatically increase the safety of people.
The System description
A camera attached to Odroid computer is mounted on a junction traffic light or any other place that is stable and high enough to get a good view of all the junction area. A video stream is received from the camera, transformed to greyscale, cropped according to a predefined and configure area of interest and then processed by the main algorithm. At the end, we get a set of coordinates that represent the positions of the pedestrians in real-time.
Success Statistics
- The system correctly identifies the pedestrians on 90% of cases and backgrounds when the configuration file is correctly set
- There is a problem of falsely tracking cars when the find contours returns contours that are smaller than the usual car sizes
- Background reset greatly improved the number of cases and sceneries where this algorithm can be used
- There are assumptions regarding the system: The camera should be mounted on a stable place and the configuration file should be properly set
Future
- Making the configuration file automatically set can greatly improve the usability of the system. Such algorithm would require a project of its own
- Maybe mounting two systems, one to track pedestrians and one to track cars might solve the problem of falsely identifying cars
- There is a built in pedestrian tracking algorithm in OpenCV but it works good only when the camera is close to the pedestrians and mounted on the same height as them, maybe one day this algorithm will improve and then a system that will use it will be robust enough to deal with many kinds of situations