基于对向传播网络的交通事件检测算法

针对高速公路交通这个复杂的非线性系统，提出一种基于对向传播网络的交通事件检
测算法。阐述了对向传播网络的结构和学习算法，给出了事件检测的依据，并合理地选择了网络的输入参数，用高速公路管理部门提供的样本数据进行了仿真研究。仿真结果表明，该事件检测算法具有学习速度快、泛化能力好、检测准确率高、平均检测时间短等优点，具有良好的应用前景。
关键词：高速公路；事件检测；对向传播网络；算法
Abstract: Aiming at the complex and nonlinear system of freeway traffic, an incident detection
algorithm based on counter propagation network (CPN) is proposed. The structure and learning algorithm of CPN are formulated. Then the basis of incident detection is given, and the CPN input variables are selected reasonably. Simulation research is carried out with the sample data provided by the freeway administrative office. Simulation results show that this incident detection algorithm has such advantages as fast learning speed, good generalization ability, high detection rate and short mean detection time. It is found to be potentially applicable in practice.
Key words: freeway; incident detection; counter propagation network; algorithm