摘要
针对车载激光雷达(LiDAR)数据中杆状地物分类效果不理想的问题,该文对从车载LiDAR数据中提取的杆状地物进行形态分析与分类研究。首先,利用基于体素的方法对杆状地物进行提取。其次,对提取出的杆状地物进行形态分析,使用ESF特征、几何特征及附属物拓扑特征作为杆状地物的特征向量集。最后,利用随机森林分类器对特征向量集进行重要性分析,构建最优特征子集,对杆状地物进行精细分类。该文在3个数据集上进行试验以验证方法的有效性。结果表明,该文方法对杆状地物有较好的分类效果,准确率分别为91.8%、89.23%和88.51%。
Aiming at the challenging issue of the classification of pole-like objects extracted from mobile light detection and ranging(LiDAR)data,this paper studied on the shape analysis and classification of them.First,the pole-like objects were automatically extracted using a voxel-based method.Second,the shape analysis on the extracted pole-like objects was performed using ensemble of shape functions(ESF)feature descriptor,geometric feature and attached parts topological feature as the feature vector set to represent pole-like objects.Finally,The random forest classifier was used to analyze the feature importance for constructing optimal feature subset,and the constructed optimal feature subset served as input for further classification.Experiments were carried out on three datasets to verify the effectiveness of the method.The results indicated that the proposed method was effective and showed a good classification result on pole-like objects,with the overall accuracies of 91.8%,89.23%,and 88.51%respectively.
作者
杨洲
康志忠
杨俊涛
周梦蝶
孔民
YANG Zhou;KANG Zhizhong;YANG Juntao;ZHOU Mengdie;KONG Min(School of Land Science and Technology,China University of Geosciences(Beijing),Beijing 100083,China;Shanxi Key Laboratory of Resources,Environment and Disaster Monitoring,Jinzhong,Shanxi 030600,China)
出处
《测绘科学》
CSCD
北大核心
2020年第1期69-76,共8页
Science of Surveying and Mapping
基金
国家自然科学基金项目(41872207).
