交通运输系统工程与信息

交通运输系统工程与信息 ›› 2006, Vol. 6 ›› Issue (4): 70-74 .

• 系统工程理论与方法 • 上一篇    下一篇

事件检测概率神经网络模型的建立与验证

覃频频1，2   

  1. 1西南交通大学交通运输学院，成都610031； 2.广西大学机械工程学院，南宁530004
  • 收稿日期:2006-02-28 修回日期:1900-01-01 出版日期:2006-08-20 发布日期:2006-08-20

Establishment and Verification of PNN Model for Incident Detection

QIN Pin-pin1，2   

  1. 1.College of Transportation, Southwest Jiaotong University, Chengdu 610031, China; 2.College of Mechanical Engineering, Guangxi University, Nanning 530004, China
  • Received:2006-02-28 Revised:1900-01-01 Online:2006-08-20 Published:2006-08-20

PDF

摘要：

在对概率神经网络(PNN)的分类机理、输入向量选取和网络设置分析的基础上，建立了用于识别两类事件模式(无事件模式和有事件模式)的事件检测PNN模型。采用高速公路路段Ⅰ-880实地线圈数据集和事件数据集验证模型，通过比较PNN模型与多层前向神经网络(MLF)模型的结果，发现无论对于向北、向南或混合方向的高速公路事件检测,PNN模型的检测率(DR)比MLF模型高；平均检测时间(MTD)比MLF模型短；但误报率(FAR)较高。概率神经网络是高速公路事件检测的一种有效算法，其在理论基础、算法和学习速度等方面比多层前向神经网络具有优势。

关键词: 事件检测, 概率神经网络, 多层前向神经网络

Abstract: This paper discusses classification, input variables and setting of a probabilistic neural network and develops a PNN incident detection model. A comparative evaluation between PNN and MLF model on I-880 site coil database and incident detection model id presented. The results show that Dr and MTTD is achieved by PNN model are better than MLF model; FAR is inferior than MLF model whether in northward, southward and the two directions. PNN is an effective algorithm in incident detection and is superior to MLF in theory, algorithm and learning speed.

Key words: incident detection, PNN(probabilistic neural network), MLF(multi-layer feed-forward neural network)