Real-Time Traffic Activity Detection Using Mobile Devices
Cuong Van PHAM, Nguyễn Thị Thanh Thủy · 2016
We propose a method and system for detection of traffic activities in real-time using mobile devices. We address to detect a set of traffic activities that can be applicable for both developed countries and developing countries like Vietnam where major transportation means are motorbikes. In addition, this work focuses on real-time implementation of the system for convenient use, as it can run as a background service on the mobile platforms. Our method relies on a two-stage of detection: low- and high-frequency motion event detection and activity classification. The proposed method is evaluated over a dataset collected from 12 subjects under realistic settings. The results demonstrate that traffic activities (including unknown activities) can be detected with precision and recall over 83% under leave-one-subject-out evaluation. These results are very potential for the situated applications such as the interventions of improper use of mobile devices for traffic participants on the road, and therefore would be useful for alleviation of traffic accidents.