Dynamic Traffic Information Collection And Data Fusion Techniques

In the advanced Traffic Management System, traffic information is the core content. All the functions of the traffic management system depend on the application of the traffic information. It is information collection,transmission,storage, processing and applications that make the traffic management changing from the simple static management to dynamic intelligent management. The maximally sharing and using of this information by travelers,drivers,traffic managers,traffic researchers and government agencies enables the dynamic optimization operation of the Intelligent Transportation System,which effectively meet people's demands of the development of the transport system.The collection of real-time traffic data and Data Fusion System aims to complete the collection,analysis and management of the main dynamic traffic information, such as traffic load,average travel speed,and lane occupancy. It allows the transportation department timely and accurately accessing to information about traffic conditions and handling it,so as to give better play to the accuracy and control of the traffic management system in the traffic surveillance,traffic control,access control,rescue management. Because of the limitation on the output information of a single traffic detector in ITS,it can not fully reflect the entire road network traffic status information,and the sharing of information on the various subsystems has difficulty,what's more,the traffic management efficiency is very low. Therefore,this system uses a Multi-Sensor Acquisition method and integrated the overall data, which overcomes the uncertainties and limitations by using a single sensor and obtains the consistency of the measured object in description and interpretation. So as to realize the corresponding decision-making and estimate,make the system enjoying more adequate information than its components and enhance the effective performance of the sensor system,to describe the measured object comprehensively and accurately.The main content of thesis is as follows:1) First,the author analyzes the access problem on the video-based vehicle detection and traffic parameters. In this part,against to the problem that the model is vulnerable under the impact of environmental changes,improving vehicle detection method is proposed based on the higher accuracy of classified pixel values in the context of reconstruction algorithm, at the same time using divisional management of background update strategy that will improve the system's real-time reliability,and by using of space-based HIS shadow detection algorithm to eliminate the shadow interference, finally,by setting up multiple virtual coils in the video image to complete traffic detection parameters.2) Based on the research of Vehicle Detection, the author gives a more detailed investigation on multi-sensor data fusion issues and proposes RBF (Radial Basis Function) Network Integration Methods based on improved immune algorithm. First, the author studied the two methods of RBF network, namely fuzzy K-means algorithm and improved immune algorithm,and compared the advantages and disadvantages of the two methods through experiments. In the experiments, the two algorithms can quickly generate their respective cluster center or memory group,but improved immune algorithm iteration process is significantly faster than the K-means clustering algorithm. At the same time, due to the presence of memory,immune algorithm has demonstrated its superiority. It showed a stronger affinity to similar data, making the speed and accuracy of data integration are obviously better than the K-means algorithm.The final chapter summarizes the research work done in this paper,and forecasts the direction of future research on this topic.