| Road traffic contingency jam detection can provide support and advice for traffic guidance, emergency treatment, and so on, so as to improve the traffic management level and service level. However, road traffic contingency jam are those random events, when and where they will happen are highly uncertainty. Therefore the existing algorithms can’t obtain acceptable results. In recent years, a huge number of bus GPS data began to accumulate in the field of I TS. These GPS data have many advantages, such as wide area covering, high real time, low maintaining cost and high reliability. Bus GPS data can reflect changes in the process of road traffic states, because of the date Contains abundant information of road traffic states. And without any doubt, it can surely provide something help for road traffic contingency jam detection. Here comes the question, how can we use the GPS data to detect traffic contingency jam, especially on urban road traffic? There is little effective study for the question.Therefore, based on the huge number of bus GPS data, this paper is aimed to study urban road traffic contingency jam detection. The paper first analyses the historical rule of road traffic states. Next, it focused on road traffic states’ real time variation tendency. By means of road traffic states’ historical rule and real time variation tendency, it can realize road traffic contingency jam detection. Based on the method, this paper designs and implements the detection systedem.Hence, this paper mainly focused on 3 research word.Firstly, this paper studies the system structure for road traffic contingency jam detection base on bus GPS data, which divides the main work into several parts and make the main work of each part and the relationship between then more clearly. And in the last, the paper designs and implements the urban road contingency jam detection system.Secondly, the paper studies on developing a method which can help to analyze the rule of traffic states based on huge number of GPS historical data. As researches showed that instantaneous velocity in bus GPS data can’t confidently reveal traffic states, the paper defines a variable which named ‘Road Delay Time Index’, shorted for RDTI. AS RDTI is more exact to describe traffic states, is will be the key variable in the follow-up work. Aimed at the relativity and conditionity of traffic states, the paper puts forward K-Means self-adaptive clustering algorithm base on T test, which is used for traffic conditions classification. Using the algorithm, we can divide traffic condition into 8 classes, and therefore determine the characteristic of traffic condition. And in the meantime, introduce ‘interquartile range’, a common statistics concept, to divide normal traffic state and abnormal traffic state clearly, which will play a significant role in training threshold and detection result assessment.The last main research of the paper is road traffic states’ real time variation tendency analyze and road traffic contingency jam detection. This paper firstly draws 4 different variables from bus GPS data, which can describe the real time traffic states more fully and exactly. Besides, this paper introduces Canonical Variate Analysis(CVA) into road traffic contingency jam detection process. The algorithm using Squared Prediction Error(SPE) as parameter to analyzes the road traffic states’ real time variation tendency, and uses historical traffic contingency jam to train the threshold. By comparing the real time variation tendency and threshold in a number of periods, we can determine whether an abnormal traffic state has happened.Based on the detection system, the paper uses real GPS to test the contingency jam detection result. Final result shows that on the condition of False Alarm Rate(FAR) under 35%, Detection Rate(DR0 reach to 90%, and the Mean Time To Detection(MTTD) is less than 3.2 minutes. This result can be helpful for urban traffic management and service, and has practical value. |
PDF Full Text Request