基于一维激光雷达数据的交通车辆分类研究

Car Classification Based on One-Dimensional Lidar Data

为满足特殊场景中交通管理系统的高时效性、高准确率、低成本需求,本文以一维激光雷达距离数据为基础,将RGB图像、SAR图像等视觉图像处理方法中使用的角点特征的概念用于一维离散数据,从而获取一维离散数据的轮廓特征。本文提出了利用均值差变和离差获得离散角点数据的方法,然后通过对数据样本的长、宽、高、离散角点数据等信息进行分析,获得每类目标的统计特征,进一步调整基于决策树的分类系统参数,提高目标分类的准确率。实验结果表明,该方法对目标分类的正确率在91%以上,能够满足特定环境场景的需求。

To meet the high timeliness,high accuracy and low cost requirements of the traffic management system in special scenarios.This paper uses the one-dimensional lidar distance data as the basis,and applies the corner features from the visual image processing methods such as RGB images and SAR images to one-dimensional discrete data,thereby obtaining the contour features of the one-dimensional discrete data.In this paper,a method of obtaining discrete corner data using mean difference variation and dispersion is proposed.Then,by analyzing the length,width,height and discrete corner data,etc,the statistical characteristics of each type of target are obtained.It is used to adjust the parameters of the classification system that based on decision tree to improve the accuracy of the target classification.The experimental results show that the correct rate of the target classification is over 91%,which can meet the needs of specific environmental scenarios.