Urban road status perception information fusion using Support Vector Regression

This paper analyzes the complementarity in perception parameter, data accuracy, coverage range, acquisition costs between fixed detectors and mobile detectors in road traffic field and presents the actual and technical demands for multi-source traffic sensor information fusion. The paper focuses on traffic status perception data of floating cars and magnetic loops as two types of typical scenes, and proposes a traffic status data fusion model based on Support Vector Regression (SVR) algorithm. The fusion model can be built in self-adaptive optimization mode by a perception dataset with SVR machine learning methods, which can effectively solve small samples, nonlinear and uncertainty problems for multi-source data fusion modeling. We acquire simulation data of different sampling interval and road types on GPS floating cars and magnetic loops in PTV microscopic traffic platform environment and experimentally verify feasibility of the fusion model for multi-source traffic status perception data.