Research On Data Fusion Algorithm In Transportation System

Traffic flow is the basic core issue of intelligent traffic system detection.Obtaining accurate traffic flow,vehicle speed,and other traffic parameters through vehicle detection is the basis for implementing effective road guidance,traffic control,and other traffic management.Traditional vehicle detection methods generally use single-type sensors.Because the sensors are limited by themselves and the surrounding environment,the test results cannot fully reflect the actual situation.This research topic integrates vehicle information obtained by magnetic sensors and video sensors to solve the problem of inaccurate information of vehicles under different conditions and improve the accuracy of vehicle detection.At present,video sensor detection has been widely used in traffic roads,but video sensors are sensitive to light changes and are easily affected by weather,smoke,light and other factors,and magnetic sensors are not affected by the above factors,but will be subject to the vehicle's own magnetic field Impacts,detection data also have some deviations,and often buried in the lane,inconvenience maintenance and construction.By analyzing and comparing the detection data obtained by these two types of sensors,a fusion framework and a fusion center structure based on video detection and magneto-intensity detection are designed and implemented on the fusion server development platform.The fusion framework is divided into two stages of detection and fusion.With the advantages of coupling and extensibility,the fusion center structure adopts a modular design,and feature extraction,data correlation and fusion calculation are performed hierarchically.The fusion algorithm is based on the D-S(Dempster-Shafer)evidence fusion theory of data source reliability.It fuses the data collected by the video sensor and the magnetic sensor,and compensates the inconsistency of the data in the classical theory through reliability evaluation and probability redistribution.Problems increase the accuracy of the test.The real-time video data is collected by the video data acquisition client,and the magnetic sensor data acquisition client collects the real-time magnetic data,and the data is uploaded to the server in real time for integration and judgment.The experimental results show that under different traffic conditions,compared with the single sensor detection results in daytime and nighttime conditions,the accuracy rate of the traffic flow obtained by the data fusion method can be increased by 2.2% to 53.4%.Can significantly improve the accuracy of nighttime harsh environmental conditions,and at the same time in the case of adequate light during the day the test results have improved.