Traffic Danger Recognition With Surveillance Cameras Without Training Data

Lijun Yu, Dawei Zhang, Xiaojun Chang, Xiangqun Chen, Alexander G. Hauptmann

Published in ICCV Demo, 2019

Demo Paper

We propose a traffic danger recognition model that works with arbitrary traffic surveillance cameras to identify and predict car crashes. There are too many cameras in the world to monitor manually. Therefore, we developed a model to predict and identify car crashes from surveillance cameras based on a 3D reconstruction of the road plane and prediction of trajectories. For regular traffic, it supports real-time proactive safety checks of speeds and distances between vehicles to provide insights about possible high-risk areas. We achieve accurate prediction and recognition of car crashes without using any labeled training data of crashes. Quantitative experiments on the BrnoCompSpeed dataset show that our model can accurately monitor the road, with mean errors of 1.80% for distance measurement, 2.77 km/h for speed measurement, 0.24 m for car position prediction, and 2.53 km/h for speed prediction. Later experiments on the MEVA dataset also demonstrates a considerable capability for this model to detect interactions between vehicles and pedestrians. It establishes an excellent foundation for the future development of automatic surveillance systems.