JOURNAL OF TRANSPORTATION SYSTEMS ENGINEERING AND INFORMATION TECHNOLOGY Volume 13, Issue 6, December 2013 Online English edition of the Chinese language journal Cite this article as: J Transpn Sys Eng & IT, 2013, 13(6), 52í59. Received date: May 2, 2013; Revised date: Jun 5, 2013; Accepted date: Jun 13, 2013 *Corresponding author. E-mail: joujou1980@163.com Copyright © 2012, China Association for Science and Technology. Electronic version published by Elsevier Limited. All rights reserved. DOI: 10.1016/S1570-6672(13)60128-2 RESEARCH PAPER Traffic Incident Prediction on Intersections Based on HMM ZHOU Jun* 1 , CHENG Lin 2 , ZHOU Lingyun 1 , CHU Zhaoming 2 1. Faculty of Transportation Engineering, Huaiyin Institute of Technology, Huai’an 223003, Jiangsu, China; 2. School of Transportation, Southeast University, Nanjing 210096, China Abstract: The intersection is an area where many accidents occur. Reasons for accidents are due to complicated intersection designs and the congested travel conditions. For these reasons, traffic incident detection is more complicated. This paper uses the intersections of Huaihai South Road and Jiefang Road as an example. According to the vehicle operation and phase timing, the situations of two vehicle's relative movement on four phase intersections are summarized. The motion vectors of the conflicting vehicle are quantized on the basis of vehicle tracking. Then, the HMM is used to classify the traffic conflicts of the intersection. Finally, numerical experiments verify that the algorithm is able to classify the conflict when the traffic is normal. Furthermore, the algorithm can forecast the traffic accidents (such as bumping, tandem, and stop) which occurred in the intersection. Key Words: intelligent transportation; incident detection; hidden Markov model; intersection; traffic conflict 1 Introduction In recent years, the rate of urban road traffic accidents remains high, especially in urban intersections. Except for the complicated intersection designs, the other factor is that the travel condition is often congested. For example, vehicles always travel freely to different directions at multilane intersections, yet the driver may park their cars illegally in the middle of the road when an accident occurs. The car will need to be parked in a safe position in order to allow traffic to return to its normal status. Moreover, because of this, the ambulance and police have to work for a long time to deal with the accidents. These obstructions makes it difficult for the traffic incident detection of the intersection [1] . Until now, several research institutions abroad have investigated the simulation experiment of collision detection at intersections. They used data mining technology to find a traffic flow mode suitable for various intersections. They also contrasted the traffic flow conditions of the intersection to predict the incidents [2] . Cuchira et al. [3] utilized the traffic rules’ inference to reflect a simple traffic condition on a one-way street. Jung et al. [4] adopted computer technology to track vehicles from a traffic image and to enhance the accuracy of traffic information detection. Oikawa et al. [5] identified traffic congestions based on the video sequence. In China, current research on intersection traffic conditions is in the design stage of traffic controllers. Moreover, building the detecting coil and predicting incidents via the direct usage of the highway traffic flow mode must still be designed. In view of the lack of intersection traffic incidents detection in China, this study designs a set of traffic conflict detection systems suitable for urban intersections. 2 Analysis of the relative motion between two vehicles with conflict at the intersection Fig. 1 displays the intersection plan for Huaihai South Road and Jiefang Road in Huai’An city. The signal is controlled by four phases. According to the vehicle running and phase timing, this paper summarizes the relative motion condition between two conflicting vehicles at the intersection controlled by four phases, as shown in Fig. 2. Fig. 2(a) indicates north-south and east-west through vehicles; Fig. 2(b) indicates north-south and east-west left-turn vehicles; Fig. 2(c) indicates the traffic conflict between two adjacent direction’s vehicles; one is the through vehicle, and the other is the right-turn vehicle; Fig. 2(d) indicates the traffic conflict between two vehicles of north-south or east-west direction; one is the left-turn vehicle, and the other is the right-turn vehicle, as shown in Fig.3.