基于FART神经网络的高速公路交通事件检测法

Highway traffic incident detection method based on FART neural network

突发性交通事件是造成高速公路交通拥堵的主要原因,为了快速并准确地检测交通事件,有效提高高速公路的利用率和通行能力,将一种模糊论与自适应谐振神经网络相结合的聚类算法—FuzzyART神经网络算法用于检测高速公路交通事件。根据高速公路事件检测相关参数之间的关系,选用了交通流密度残差和平均速度残差的组合向量作为神经网络的输入,给出了事件检测的算法步骤和输出判断方法,并进行了仿真实验。仿真结果表明,该算法不仅能够检测出交通事件的发生及其造成的拥堵程度,还可以快速识别已有的交通事件类型,记忆未知的交通事件类型,实现边工作、边学习,且检测结果具有较高的稳定性和准确性。

Unexpected traffic incident is the main cause of highway traffic congestion, so detection traffic incidents quickly and accurately can effectively improve the utilization and capacity of highway. A clustering algorithm combined adaptive resonance theory with fuzzy theory-Fuzzy ART neural network algorithm was used to detect highway traffic incidents. According to the rela- tions of detection parameters, the vectors of traffic density differences and average speed differences were selected as network＇ s input. At the same time, the detection steps and how to judge the output of the network were described. The results of simula- tions show that, the algorithm can detect occurring and the congestion degree of traffic incidents, and also can quickly identify learned types and store the unlearned types. The neural network（NN） combines work and study,so the resuhs of detection have higer stablility and accuration.