Classification Algorithm Of Typical Objects Based On LiDAR And Long-wavelength Infrared Images

In recent years, with the development of science and technology, the concept is presented by the "digital" and "digital city". It has been more and more important to get the region terrain data and building three-dimensional data quickly and accurately. LiDAR technology is developed rapidly in recent decades, and it provides a new way to measure the terrain area. The classification of point cloud is mainly used to construction the city and three-dimensional model for city. According to the classification of the features, it can extract the important information. Airborne LiDAR system can get the ground three-dimensional coordinate data directly. However, the remote sensing image has rich spectral information and texture information. So, it is necessary to combination with the two kinds of data that can make up for monophyletic data to the LiDAR point cloud classification insufficient.This paper divide typical ground objects into ground points, roads, vegetation, buildings and other objects five kinds of objects, and it is based on airborne LiDAR point cloud data and long-wavelength infrared remote sensing image as data sources. Firstly, it is separated in non-ground points and ground points based on orthogonal polynomial filtering algorithm for airborne LiDAR point cloud filtering operation; Secondly, based on object-oriented remote sensing image segmentation algorithm, the road network, vegetation outlines and building's outlines are extracted; Thirdly, it makes the point cloud grid combine with LiDAR point cloud data. LiDAR point cloud rasterize resolution should be consistent with the resolution of the remote sensing image, because it is order to make this two data stack together, and based on the principle of cutting, cutting out the object feature image. Finally, three-dimensional point cloud is extracted from second-dimensional image. It is extracted road points from LiDAR point cloud first of all, then, by region growth algorithm to generate the road 3d model, last, extract vegetation outlines and buildings'outlines and others objects.The LiDAR point cloud and long-wavelength infrared data after Haiti earthquake and the LiDAR point cloud of Changchun city of JiLin Province are employed and processed in the experiment based on the three-dimensional graphics library-OpenGL program and Matlab software. The point cloud data filtering and image cutting procedures are designed in this research, it is verified the feasibility of the algorithm, and achieved good results.