融合影像信息的LiDAR点云多特征分类方法  

LiDAR Point Cloud Multi-feature Classification Method Based on the Fusion Image Information

作　　者：叶刚[1] 

机构地区：[1]郑州市规划勘测设计研究院,河南郑州450000

出　　处：《地理空间信息》2018年第6期70-75,共6页Geospatial Information

基　　金：国家科技支撑计划(2014BAL05B00)

摘　　要：针对单一Li DAR点云数据分类精度不高的问题,提出一种融合影像信息的激光点云多特征分类方法。该方法根据应用目的以及地物分类的需求对航空影像所提供的光谱、形状等特征和Li DAR数据提供的几何特征进行研究分析,确定参与分类器中的特征空间,并作为设定相应分类规则的先验知识,然后根据特征描述子之间的空间距离进行空间聚类,最终成功将点云分类为建筑物、树木、草地、道路以及不确定地物等5类,分类精度达到95.3%,kappa系数达0.935。此外,还分别引入基于影像的SVM分类和基于terrasolid软件的点云分类方法,以验证本文算法的有效性。Because that the classification accuracy of single Li DAR point cloud data was not high due to the complexity and category diversity of the object, we put forward a Li DAR point cloud multi-feature classification method based on the fusion image information. We accomplished this method as the three steps. Firstly, we determined the feature space according to the requirements of application purpose and feature classification by analyzing the spectral and shape features of the image, and the geometrical features of Li DAR data, which were as a prior knowledge of corresponding classification rule set. And then, according to the distance between feature descriptors, we determined the space spatial clustering. Finally, we obtained the point cloud classification of five categories(e.g. buildings, trees, grass, road and uncertain features). The classification accuracy is 95.3%, the kappa coefficient is 0.935. In addition, we introduced the SVM method based on image classification, and point cloud classification method based on Terrasolid software respectively to verify the effectiveness of the algorithm in this paper.

关 键 词：LIDAR点云 航空影像 多特征 融合数据 分类 

分 类 号：P228[天文地球—大地测量学与测量工程]