支持向量机的建筑物激光脚点提取方法  

Support vector machine classification for detecting laser points of building

作　　者：罗伊萍[1,2] 姜挺[1] 王鑫[1] 陈文锋[3] 张锐[1] 

机构地区：[1]信息工程大学测绘学院 [2]61081部队 [3]73603部队

出　　处：《测绘科学》2011年第4期173-175,共3页Science of Surveying and Mapping

摘　　要：本文提出了一种基于全色波段航空影像和激光雷达数据的建筑物检测方法。如何从激光点云数据中提取出建筑物激光脚点,是建筑物三维重建和轮廓提取的难点问题之一。植被密集区域以及与建筑物紧密相邻的树木的激光点很难与建筑物激光点区分开。本文利用支持向量机对单个激光点的特征进行两分类,特征向量包括激光点的高程、高程变化信息以及与激光点配准的影像光谱信息。实验表明,基于支持向量机的点态分类算法能够有效提取建筑物激光脚点,影像光谱信息能明显提高分类精度。A building detection method based on aerial image with panchromatic bands and airborne laser scanning data was pres- ented in the paper. Detecting building footprints from laser points cloud data is one of the most difficult problems in building modeling and edge detection. Tree data points in mountainous area or connecting to building points are usually too confused to know from building points. A binary classification was done to distinguish building points from tree points. It used the support vector machine to the features of single laser point, including height, height variation information and spectrum information from LiDAR point cloud and registered aerial image. The experiments showed that the pointwise classification based on support vector machine could efficiently detect the building points from LiDAR data, and the spectrum information would improve the accuracy of classification.

关 键 词：激光雷达 支持向量机 建筑物检测 点态分类 影像特征 

分 类 号：TP751.1[自动化与计算机技术—检测技术与自动化装置]