Detecting traffic signal lights (e.g., red) is an important subject of intersection safety since many accidents are the result of road users' non-conforming behavior to traffic signals. This work shows that traffic signal phases can be inferred through traffic cameras in order to detect temporal violations of road users. The idea is to understand the traffic phase by learning the moving features of road users. Moving features are extracted and labeled according to two traffic signal lights (i.e., green, red) and different learning methods such as K Nearest Neighborhood (KNN), Naive Bayes (NB), Neural Network (NN), Deep Neural Network (DNN) and Support Vector Machine (SVM) are applied on training data. The experimental results of two different intersection videos shows that an accuracy of higher than 90% can be achieved by DNN and SVM when the feature size is appropriately selected.
Traffic phase inference using traffic cameras
2017 IEEE Intelligent Vehicles Symposium (IV) ; 1565-1570
2017-06-01
423953 byte
Conference paper
Electronic Resource
English
Traffic Phase Inference Using Traffic Cameras
British Library Conference Proceedings | 2017
|Traffic Control by Digital Imaging Cameras
British Library Conference Proceedings | 2014
|