Design Of Traffic Detection System Based On Bluetooth And Research Of Traffic Incident Automatic Detection Algorithm

With the rapid development of the transportation industry, the number of cars has increased dramatically, which greatly reduces the traffic capacity of the road, causes frequent occurrence of traffic accidents and seriously affects people's lives and property safety. How to timely and accurate get traffic information, real-time and effective detection of the occurrence of traffic incidents has become a hot issue of concern. Traffic information detection technology and traffic incident detection algorithm are designed to solve these problems and have become a hot topic in the field of intelligent transportation.The paper make a deep research on traffic flow detection system and traffic incident detection algorithm based on the research of traffic flow data characteristics and the principle of traffic incident detection. The main contents include the following three points.1)This paper constructs the traffic flow detection system which is composed of Bluetooth, accelerometer and BD. In order to develop the traffic flow detection system of vehicle parts and electronic bus stop board part and realize the multi sensor velocity measurement function of accelerometer and BD combination, this paper chooses Freescale ARM Cortex-M4 core K60 chip as the main controller based on analyzing the system feasibility. This paper uses the CC2540 module of TI Company as the Bluetooth host and slave. According to the study of the Bluetooth 4.0 protocol stack, the detection of traffic flow is realized through developing the wireless connection between host and slave.2) In this paper, the fuzzy C-means(FCM) is improved: a weight of membership is introduced to optimize the number of cluster C so that the impact of the complexity of traditional FCM and the sensitivity of outliers can be removed. An algorithm of traffic incident detection based on improved fuzzy C-means, coupled with the Support Vector Machine(SVM), is presented in this thesis. At first, the traffic flow, average speed, vehicle occupancy are selected as input parameters, and are classified preliminarily using a modified fuzzy C-means method. Then, those selected data is served to build the classification model as training samples, with which a series of discriminant functions are formed to recognize the traffic incident.3) In order to improve the accuracy of event detection, this paper also propose the method of semantic technology, coupled with the Support Vector Machine, applied to event detection: The domain ontology ontology model about the traffic incident detection is established, then heterogeneous data included of the data collected from the algorithm and collecting location is used as an instance of ontology. A set of geographic ontology reasoning formula is created, and Jess inference engine was used to achieve the application model.