文献类型：会议论文

英文题名：Characteristics of Full-Waveform Lidar Data from typical objects and its potential in Point Cloud Classification

作者：Xu, Guangcai[1,2] Pang, Yong[1] Yuan, Lingling[3] Li, Mingyang[2] Fu, Tian[1]

第一作者：Xu, Guangcai

通信作者：Xu, GC[1]

机构：[1]Chinese Acad Forestry, Inst Forest Resource & Informat Technol, Beijing 100091, Peoples R China;[2]Nanjing Forestry Univ, Coll Forest Resources & Environm, Nanjing 210037, Jiangsu, Peoples R China;[3]Jiangsu Prov Jinwei Remote Sensing Data Eng Co Lt, Nanjing 210000, Jiangsu, Peoples R China

会议论文集：6th International Symposium on Digital Earth - Models, Algorithms, and Virtual Reality

会议日期：SEP 09-12, 2009

会议地点：Beijing, PEOPLES R CHINA

语种：英文

外文关键词：full-waveform lidar; Gaussian decomposition; statistic parameters; forestry

年份：2010

摘要：Full-Waveform Lidar systems have already been proved to have large potentialities in forest related applications. Entire backscatter signals of each emitted pulse, which give end users more controls in raw data management and interpretation process compared to traditional discrete return lidar data, are digitized and recorded by the system. Especially in the forest area, more detail information is provided using waveform data and new opportunities are inspired for point cloud classification from waveform characteristics. In this study, full-waveform data were collected by Riegl LMS Q560 system with a point density of 1.4 points/m(2) in Dayekou Watershed (DYK), Gansu province, China. These small footprint airborne full-waveform lidar data were used to extract statistic information (i.e. echo half-width, amplitude and intensity) of different targets such as grass, shrub, forest and bared area, and try to classify the typical targets in test field. Non-linear least square method was adopted to fit a series Gaussian pulse to decompose the raw waveform data. Then the attributes including peak location, half-width, amplitude, intensity of each pulse were calculated. Generally, different objects response to the emitted pulse diversely, which is incarnated in the three attributes described above. The decomposed waveform data were transformed to 3D points with several related attributes. And the field survey information and the same period of the high-resolution multi-spectral images were used to determine the specific location and extent of different features areas (forest, bare land, grassland, construction), then get the statistic value of three attributes for the corresponding regions in the decomposed waveform data. The results showed that three statistical characteristics of different targets are different in some extent, which demonstrated their potential in point cloud classification.