Urban traffic incident detection with mobile sensors based on SVM

Traffic accident detection is an important component in Intelligent Transportation System (ITS). Compared with freeway, urban accident detection is more complicated. The mean reason is that there are full of flow-disruptive entities (traffic signals, intersections and bus stops) in urban roads, which can disrupt the traffic flow in a similar way that an accident occurs. In this paper, an effective method for urban traffic accident detection has been designed to recognize the abnormal traffic state based on mobile sensors and Support Vector Machine (SVM). We regard the whole dynamic process of the vehicle passing through a road as an instance. Our results show that SVM is a good method to detect the urban traffic accident. In addition, we investigate three traffic variables (speed, acceleration and lane-changing) and their effect for urban traffic accident detection.