Research On Multi-sensor Data Fusion Model Based On Road Traffic

Traffic data fusion is the basic and core issue of intelligent transportation. Byaccess to the read-time traffic data such as speed, flow and be integrated to implementeffective route guidance, traffic control is the core of intelligent traffic management.More and more companies and organizations began in the intelligent transportationfield proposed respective solutions.The traditional way to obtain information generally using a single type ofsensor, but due to their own constrains: a single sensor cannot get enoughcomprehensive information and fusion model is relatively simple, mostly predictresults by extracting original information and comparing historical data. While thesetechnologies have been used in different traffic scenarios, it still has many problems.The multi-sensors can get more accurate road traffic information including speed,traffic, environment, which is the first choice in the fusion solution.(1)In this thesis, the multi-sensor information fusion algorithm and existingmodels are analyzed and compared in depth, respectively analyzes the usagescenarios and deficiency of existing technical scheme. And on this basis, we mainlyfocus on Kalman’s filtering principle of in-depth study and analysis theparameters traffic flow information and relationships, then estimate thesystem structure and information fusion optimal criterion and combining with theactual situation of the topic.(2) Proposed the Separated Twice Federal Kalman filtering algorithm byanalyzing the structure of fusion model and optimal rule for fusion with this researchtopic. The error model is designed and the final fusion model of each work unit isapplied for the improved algorithm, which verify the feasibility of the model intheory. The design and implementation of the simulation system is developed usingC++on OpenCV.(3)In order to evaluate the performance of the new fusion model, in this study,two model cases were developed applying the improved algorithm and the originalalgorithm respectively. At the same time when collecting traffic data, we usetwo case processing, consumption of precision and performance comparison of twocases of fusion result. By performing experiments and the results were analyzed and compared, using the improved algorithm for both accuracy and performance arebetter than the original algorithm scheme. From then proves feasible and rationalityof the new scheme.This research won the Hubei Natural Science Fund:wireless sensor networkalgorithm of optimal path based on the selection of Worker(serial number:2012FFB05006) support.