计算机与现代化

• 应用与开发 • 上一篇    下一篇

PSO-SVM理论在路面识别中的应用

  

  1. (云南省交通科学研究院,云南昆明650011)
  • 收稿日期:2016-01-20 出版日期:2016-09-12 发布日期:2016-09-13
  • 基金资助:
    姚庆华(1976-),男,云南昆明人,云南省交通科学研究院高级工程师,硕士,研究方向:智能交通,软件开发; 蒲雯,女,高级工程师,硕士,研究方向:智能交通,交通规划与统计。

Application of PSO-SVM Theory in Pavement Recognition

  1. (Yunnan Transportation Research Institute, Kunming 650011, China)
  • Received:2016-01-20 Online:2016-09-12 Published:2016-09-13

摘要: 道路的起伏度严重制约着汽车的行驶速度,也在一定程度上影响道路的交通安全。在路面的识别检测中,路面数据中的噪声点和野值点是困扰识别检测的一大难题。本文利用支持向量机理论对于路面数据中的噪声点和野值点具有敏感性的特点,提出一种改进的PSO-SVM识别算法,首先利用参数优化超平面方程,然后利用粒子群(PSO)算法优化支持向量机的核函数及其参数,最后进行路面的识别检测。实验结果表明,本文提出的算法对于路面起伏度的检测计算具有速度快,识别准确率高(可达到92%)的特点。

关键词: SVM理论, PSO算法, 路面识别

Abstract: The road has a serious impact on the speed of the car, and affects the road traffic safety to some extent. In the identification and detection of pavement, the noise points and outliers are difficult problems in the identification and detection of pavement data. In this paper, with the support vector machine for road noise in the data points and outliers being of sensitive features, we present an improved PSO-SVM recognition algorithm. The algorithm firstly uses parameter to optimize hyperplane equation, then optimizes SVM kernel functions and their parameters by using Particle Swarm Optimization (PSO) algorithms, and finally recognizes and detects the pavement. The experimental results show that the proposed algorithm has the advantages of faster speed for the detection and calculation of pavement fluctuation, and high recognition accuracy which can reach up to 92%.

Key words: SVM theory, PSO algorithm, pavement recognition

中图分类号: