Scene understanding and pedestrian/vehicle tracking from traffic cameras

The increase of vehicles in urban environments has raised the importance of decreasing car accidents, managing traffic congestion and building automated transportation systems. In this repository, an automatic traffic surveillance system is proposed that uses traffic camera videos of different resolutions. The objective is implemented in three core steps, the first step is the object detection on the scene, the second objective is multiple object tracking and thirdly is the depth estimation which translates coordinates on the image to coordinates on the road plane. The automatic traffic surveillance system provides, obtained trajectories of vehicles and pedestrians, estimated instantaneous and average speeds and translation of tracked trajectories to OpenScenario format.

Author: Spyridon Couvaras

Imperial College London