Abstract
Traffic flow information electromechanical detected system is an important part intelligent transportation system. The collected traffic flow information is the guarantee of intelligent transportation guidance system, providing important basis. However, the traffic information is usually collected by the sensors buried at the node of traffic grid, however, defective information is always occurred due to the sensor failure of constructing, transmitting or processing and lack of sensor. Therefore, in order to provide complete information to display and explore the law of traffic distribution deeply, the defective information need to be mended.
According to the analysis, the defective information can be divided into incomplete information and non-information, which can be repaired by SARBF neural network fitting method and Lagrange interpolation method respectively. The first step of SARBF neural network fitting method to repair the incomplete is selecting the data-complete intersections by autocorrelation analysis in spatial database, then obtain the RBF neural network training sample data of participated intersections to mend the defective data. When the non-information is repaired by Lagrange interpolation method, the interpolation nodes are selected by spatial encoding in GIS.