A new algorithm of incident detection on freeways

With the high development speed of the Chinese economy, lots of new freeways have been constructed. Hence, incident management becomes an important issue in the freeway traffic management system. Because incident detection is a kind of pattern recognition problem, in this paper we employed the probabilistic neural network (PNN) to solve it. Applying traffic simulation software FRESIM, a wide range of incidents that include different patterns under a variety of flow conditions and traffic periods were generated to train and evaluate the performance and the transferability of the proposed PNN-based algorithm. It was proved that the models of our proposed algorithm built on one segment can be used for other segments, and all three performance measures indicated the potential of practical use of them