基于OLS算法的RBF神经网络高速公路事件探测

高速公路事件是指破坏正常交通流并造成交通阻塞的非重现随机发生的事件。事件发生后对其进行快速可靠的探测对减少交通延误、保障道路安全、减少环境污染具有十分重要的意义。文中提出了一种基于模糊聚类技术和RBF神经网络的混合智能高速公路事件自动探测算法，同时改进了用于RBF神经网络训练的OLS(正交最小二乘)选择算法。仿真实验证明，改进的OLS选择算法大大提高了RBF神经网络的训练速度，同时具有无须事先确定RBF中心的优点，将之运用于公路事件探测可以获得满意的性能。

Freeway Incident Detection Based on OLS and RBF Neural Networks

Freeway incidents are non-recurrent and pseudorandom events that disrupt the normal flow of traffic and create a bottleneck in the road network. Quick and reliable incidents detection is essential to reduce traffic delays, ensure road safety and protect environment. This paper presents a new hybrid intelligence algorithm for automatically detecting freeway incidents, which employs fuzzy clustering and RBF neural computing technique. An improved OLS(Orthogonal Least Squares) selection algorithm for training RBF neural networks is also proposed. The simulation results illustrate that the improved OLS selection algorithm accelerates the training of the RBF neural networks substantially and there is no need to decide the number of RBF centers in advance. The satisfactory performance could be achieved by using this algorithm in freeway incidents detection.