面向城市道路的地磁传感器单节点车型分类

发布时间：2018-05-01 18:40

本文选题：道路工程 + 车型分类　；参考：《吉林大学学报(工学版)》2015年01期

【摘要】：提出了一种利用多功能地磁传感器采集道路环境磁场数据,并基于决策树模型实现车型的在线分类方法。文中提取8种与车速无关的车辆波形时域特征作为模型输入,基于最优最小划分样本数的CART算法对决策树模型进行训练。对训练得到的决策树,基于最小误差剪枝原则进行剪枝,得到具有更高样本鲁棒性的最佳剪枝树。通过在北京市某道路上布设地磁传感器获取了两种车型数据,正、反向测试的平均准确率分别为88.9%和94.4%。与现有多个分类方法进行了对比实验,结果表明:本文方法能够进行在线车型分类,并在分类准确率、样本鲁棒性和算法执行时间等方面更具优势,能够应用于实际城市道路现场进行车型分类。
[Abstract]:This paper presents an on-line classification method based on decision tree model to collect road environment magnetic field data by using multifunctional geomagnetic sensor. In this paper, eight time-domain features of vehicle waveform independent of vehicle speed are extracted as the input of the model, and the decision tree model is trained based on the CART algorithm with the optimal minimum number of samples. The best pruning tree with higher sample robustness is obtained by pruning the decision tree based on the minimum error pruning principle. Two types of vehicle data were obtained by using geomagnetic sensors on a certain road in Beijing. The average accuracy of forward and reverse tests was 88.9% and 94.4% respectively. Compared with many existing classification methods, the results show that this method can be used for online vehicle classification, and has more advantages in classification accuracy, sample robustness and algorithm execution time, etc. Can be applied to the actual urban road scene vehicle classification.
【作者单位】：北京交通大学轨道交通控制与安全国家重点实验室;中铁国际多式联运有限公司;西昌卫星发射中心;
【基金】：“863”国家高技术研究发展计划项目(2012AA112401,2011AA110505) 国家自然科学基金项目(61104164)
【分类号】：U491.116

【相似文献】