This work proposes a framework for road traffic surveillance using computer vision techniques. After a foreground estimation, post processing techniques are applied to the detected vehicles in motion to generate blobs. Then, a tracking approach based on Kalman filters is used to extract instantaneous information throughout a video sequence, including speed and trajectory estimation and imprudent driving detection.
The system has been developed in Python and can be launched in real-time using a standard CPU. The code is available at github:
https://github.com/mcv-m6-video/mcv-m6-2018-team3.