An Automatic Incident of Freeway Detection Algorithm Based on Support Vector Machine

Aimed at the research on freeway detection algorithm has great significance for improving efficiency and effectiveness of freeway traffic management, this paper based on the freeway traffic flow´s characteristics, in accordance with the incident detection´s basic principle, researches on freeway incident detection based on Support Vector Machine (SVM). This paper designs four different simulation experiments based on linearly non-separable SVM, Gauss kernel function and hyperbolic tangent function respectively. Experiments above verify the effectiveness and portability of algorithms. This paper adopts parameters optimization module of Libsvm tool box provided by the associate professor Chih-Jen Lin, after optimal parameters achieved, simulates the above experiments and compared with California algorithm, the simulation results show that choosing appropriate SVM model and kernel function, we can achieve better performances than California algorithm according to different experiments.