基于机载LiDAR与长波红外影像的典型地物分类算法

发布时间：2018-06-22 00:51

本文选题：LiDAR点云 + 高分辨率遥感　；参考：《东北大学》2013年硕士论文

【摘要】：近年来,随着科技的发展,“数字地球”、“数字城市”概念的提出,快速而准确的获取地区地形数据和建筑物三维空间信息显得日趋重要。LiDAR技术是近几十年来快速发展起来的一种新型测量技术,为地区测量提供了新方法。点云数据的分类主要应用于城市规划、城市三维模型的构建等,按照分类的特征,提取地表重要信息。机载LiDAR系统可以直接获取地面三维坐标数据,而遥感影像上有着丰富的光谱信息、纹理信息等属性,将两种数据结合,可以弥补基于单源的遥感影像或LiDAR点云数据进行分类的不足。本文以机载LiDAR点云数据和长波红外遥感影像为数据源,将典型地物分为地面点、道路、建筑物、植被以及其它共五类地物。首先,基于正交多项式滤波算法对机载LiDAR点云进行滤波运算,分离地面点与非地面点；然后基于面向对象的遥感影像分割算法,提取道路网、树木轮廓及建筑物特征；再结合LiDAR点云数据,将点云数据栅格化,栅格化的分辨率要与遥感影像的分辨率保持一致,使两种数据叠加在一起,基于裁剪原理,裁剪出地物特征影像；最后,从二维影像提取三维点云,从LiDAR点云数据中首先分类出道路点,通过区域增长算法生成路面三维模型,再依次提取树木轮廓特征、建筑物特征及其它。本文以海地震后获取的机载LiDAR点云数据与长波红外遥感影像和吉林省长春市机载LiDAR数据为实验数据,利用三维图像库OpenGL与Matlab进行了程序的开发,编写了点云数据的滤波处理与影像裁剪程序,验证了本文算法的可行性,取得了较好的实验结果。
[Abstract]:In recent years, with the development of science and technology, the concepts of "digital earth" and "digital city" have been put forward. Rapid and accurate acquisition of regional terrain data and three-dimensional spatial information of buildings is becoming increasingly important. LiDAR technology is a new type of measurement technology developed rapidly in recent decades, which provides a new method for regional survey. The classification of point cloud data is mainly used in urban planning and the construction of urban three-dimensional model. According to the classification features, the important information of the surface is extracted. The airborne LiDAR system can directly obtain 3D coordinate data on the ground, while remote sensing images have rich spectral information, texture information and other attributes, which combine the two kinds of data. It can make up for the lack of classification based on single source remote sensing image or LiDAR point cloud data. Based on airborne LiDAR point cloud data and long-wave infrared remote sensing images, the typical features are divided into five types: ground points, roads, buildings, vegetation and other features. Firstly, the airborne LiDAR point cloud is filtered based on orthogonal polynomial filtering algorithm to separate ground points from non-ground points, and then the road network, tree contour and building features are extracted based on object-oriented remote sensing image segmentation algorithm. Combined with LiDAR point cloud data, the rasterization of point cloud data and the resolution of rasterization should be consistent with the resolution of remote sensing images, so that the two kinds of data are superimposed together, and based on the principle of clipping, the feature images of ground objects are cut out. The 3D point cloud is extracted from the two-dimensional image and the road points are classified from the LiDAR point cloud data. The road surface 3D model is generated by the region growth algorithm, and then the tree contour feature, the building feature and others are extracted in turn. In this paper, the airborne LiDAR point cloud data and long-wave infrared remote sensing image obtained after the sea earthquake and the airborne LiDAR data of Changchun City of Jilin Province are used as experimental data. The program is developed with OpenGL and Matlab. The program of filtering and image clipping for point cloud data is written, which verifies the feasibility of this algorithm and achieves good experimental results.
【学位授予单位】：东北大学
【学位级别】：硕士
【学位授予年份】：2013
【分类号】：P237

【相似文献】