Registration Of Terrestrial Lidar And Hyperspectral Data For The Application Of Single Tree Structure Parameters Extration

With the rapid development of remote sensing, traditional optical remote sensing can no longer meet the spatial and spectral accuracy of forest resources survey. In recent years, as LiDAR(Light Detection and Ranging, LiDAR) can provide spatial information of high precision and hyperspectrum has rich spectral information, LiDAR and hyperspectrum have been widely used in various industries. With the mutual compensation of spatial information of LiDAR and spectral information of hyperspectrum, the synergy of those two kinds of data is beneficial for improving the inversion accuracy of forest parameters. This paper discusses the following work with the spatial information and spectral information obtained from ground-laser scanner Leica Scanstation C10 and imaging spectrometer SOC710:(1) Summarize the research status of LiDAR data and hyperspectral data; introduce the synergy of two kinds of data in the research of forestry and the research status of the registration of LiDAR and optical remote sensing data;(2) Aiming at the massive and scattered LiDAR point clouds data, we have established LiDAR point cloud indexing mechanism;(3) Aiming at the different imaging modes of LiDAR and hyperspectrum, LiDAR point clouds data is scattered 3D information while hyperspectral data is 2D spectral image. Synergizing those data needs to match the spatial dimension. This paper simulates camera imaging to convert LiDAR point cloud data into 2D image;(4) Due to the difference of LiDAR and hyperspectral imaging, this paper studies the registration methods based on CP(control points),feature points and mutual information and proposes a registration method which hybrid CP and feature-mutual information, using coarse registration parameters obtained from basing on CP registration as the initial parameters of Powell algorithmic and adopts feature-mutual information to measure similarities. This method realizes the registration between Li DAR image and hyperspectral image.(5) Aiming at using LiDAR own intensity information cannot segment leaves and stems and the high calculation using LiDAR own intensity to segment leaves and stems, this paper collaborates terrestrial LiDAR and hyperspectral data to divide leaves and branches, and then estimate LAD(Leaves Area Density) based on VCP(Voxel-based Canopy) algorithm using the leaves’ LiDAR cloud data.In summary, with the characteristics of terrestrial LiDAR point cloud and hyperspectral image, this paper fully utilizes the methods of CP-based image registration, feature-based registration, and mutual information-based registration to realize the registration of LiDAR data and hyperspectral image and at last estimates individual tree’s LAD by synergizing ground LiDAR data and hyperspectral images.