摘要
基于激光雷达传感器,提出了一种道路路锥识别方法。首先,在传统DBSCAN聚类算法的基础上改进算法搜寻核心点的方式,对雷达所采集的数据进行快速地分割、聚类。接着,分析类簇,对每帧数据的类簇进行特征采样并赋予标签值。最后,通过支持向量机(SVM)训练样本数据,利用网格化搜索与交叉验证法优化SVM参数,得到类簇分类模型,用于识别路锥。实验结果表明,改进后的DBSCAN算法计算效率有了显著提升,并且对点云的聚类更具有针对性。经过多次随机数据集检测,分类模型的准确率保持在93%以上,实现了对路锥的有效识别。
In this paper,a traffic cone recognition method based on radar sensors is proposed.Firstly,based on the traditional DBSCAN clustering algorithm,the method of searching core points is improved,and the data collected by the radar is quickly segmented and clustered.Then,the cluster is analyzed,and the clusters of each frame of data are sampled for features and assigned label values.Finally,SVM(support vector machine)is used to train the sample data,and the SVM parameters are optimized by grid search and cross-validation to obtain cluster-like classifiers for identifying road cones.The experimental results show that the efficiency of the improved DBSCAN algorithm is significantly improved,and the clustering of point cloud is more targeted.After several random data sets detection,the accuracy of the classifier mode is maintained at over 93%,which realized the effective recognition of road cones.
作者
王兆权
陈天炎
王水发
吴宁钰
吴海彬
WANG Zhao-quan;CHEN Tian-yan;WANG Shui-fa;WU Ning-yu;WU Hai-bin(Management Center of Experiment and Training,Minjiang University,Fuzhou 350108,China;Fujian Chuanzheng Communications College,Fuzhou 350007,China;College of Mechanical Engineering and Automation,Fuzhou University,Fuzhou 350108,China)
出处
《激光与红外》
CAS
CSCD
北大核心
2022年第12期1796-1803,共8页
Laser & Infrared
基金
福建省交通运输科技项目(No.2020030)
福建省教育厅项目(No.JAT200431)资助。
