Citation
Pecharromán Galindo, Antonio and Sánchez Almodóvar, Nuria and Torres Arjona, Juan and Menendez Garcia, Jose Manuel
(2011).
Real-time incidents detection in the highways of the future.
In: "15th Portuguese Conference on Artificial Intelligence (EPIA 2011)", 10/10/2011 - 13/10/2011, Lisbon, Portugal. pp. 108-121.
Abstract
Due to ever increasing transportation of people and goods, automatic traffic surveillance is becoming a key issue for both providing safety to road users and improving traffic control in an efficient way. In this paper, we propose a new system that, exploiting the capabilities that both computer vision and machine learning offer, is able to detect and track different types of real incidents on a highway. Specifically, it is able to accurately detect not only stopped vehicles, but also drivers and passengers leaving the stopped vehicle, and other pedestrians present in the roadway. Additionally, a theoretical approach for detecting vehicles which may leave the road in an unexpected way is also presented. The system works in real-time and it has been optimized for working outdoor, being thus appropriate for its deployment in a real-world environment like a highway. First experimental results on a dataset created with videos provided by two Spanish highway operators demonstrate the effectiveness of the proposed system and its robustness against noise and low-quality videos.