Forest Parameter Extraction Using LiDAR Data And Digital Camera Image

LiDAR (light detection and ranging) is an active remote sensing technique, LiDAR instruments measure the roundtrip time for a pulse of laser energy to travel between the sensor and a target, provide a distance or range from the instrument to the object. The first LiDAR sensor of the world was made in the late of sities of the 21th century, and since that, it has attracted more and more attention from the researchers. Since the mid of 1980s, the use of LiDAR for forestry applications has advanced with technology. It can acquire the 3D geographical coordinates for estimating forestry characteristics, especially in tree height and spatial structure et al. Many researches on abroad have proved the importance of LiDAR data in the application of forest surveys and monitoring, such as the estimation of stand heights, base area and timber volume et al. However, the forest application of LiDAR technology is in the first phase in Chinese. In this dissertation, the major research contents are as follows.1) Current research and development on home and abroad are summarized seriously by reading much of reference, and the main algorithms of forest paremters inversion are analyzied deeply. And so, some important technique to resolve are presented, and the prospect of the application of LiDAR technology in forestry have been reviewed for the future.2) LiDAR point clouds are used to orthorectify digital camera imagery. Two methods are applied in the orthorectification of camera imagery. They are different in generating digital elevation models (DEM). In contrast to traditional photogrammetry, LiDAR data can generate DEM simply, quickly and accurately.3) According to the terrain, vegetation and other things of the surface, the LiDAR point clouds are filtered and classified. The main algirithums include Tin Filter, Polynomial Filter, Planner Filter, Z Clip Filter, Z Proximity Filter and et al. The classified vegetation points are overlapped on the digital camera imagery to check up the classification result visually.4) Percentile value in different heights are considered as the variable, and the correlation with ground-truth stand height are analysed. The result indicates that 75 percentile value are the best correlated with the field value. The liner regression equation is established, and the average evaluation precision is up to 90.59 %.5) Object-oriented methods are used to identify the sigle tree in digital camera imagery.Multi-scale and canopy mode are applied in the segementation, and the nearest neigours and member function methods are the main classification algorithm. Edge optimization of canopy polygons can improve the precision of sigle tree identification.6) LiDAR data are used to build Digital Surface Modal (DSM) and Digital Elevation Modal (DEM), and subtration algorithm is applied to generate Canopy Height Modal (CHM). CHM and canopy polygons are overlayed, and the max height in the polygon is as the tree height. The regression equation is established between LiDAR-derived tree height and field-derived height, and the average evaluation precision is 74.89%.