| Tree parameters are an important content in forest resources investigation.With the continuous development of 3D LiDAR technology,terrestrialLiDAR system has been successfully applied to forest resources investigation as a modern data acquisition method.This system not only completely subverts the traditional manual measurement method,but also can realize fast,convenient and efficient data acquisition.However,in the face of forest areas with complex forest environment,terrestrialLiDAR system and manual measurement can not easily go deep into forest areas to collect data.How to accurately and quickly collect data and obtain parameters of DBH and tree height in complex forest areas is an urgent problem to be solved by modern forestry LiDAR technology,the emergence of backpack LiDAR system makes up for this deficiency.Backpack LiDAR systemas a new mobile LiDAR technology can freely go deep into forest areas with complex forest environment for data collection,and has been applied to forest structure parameter collection at present.In order to obtain the parameters of tree diameter at breast height and tree height with high precision,this paper takes Wangyedian forest farm in karakin banner of Inner Mongolia Autonomous Region as the research area,selected three sample plots to collect cloud data of sample sites by backpack LiDAR,and extracted the parameters of tree diameter at breast height and tree height.The main research contents and experimental results are as follows:(1)An accurate extraction algorithm of DBH parameters based on backpack LiDAR point cloud intensity is proposed.The operation principle of backpack LiDAR system relies on the implementation of SLAM technology.Due to the deviation of SLAM system positioning accuracy,there are a large number of discrete low-quality points in LiDAR point cloud data of forest backpack.In the process of extracting DBH parameters,too many low-quality points can affect the accuracy of parameter extraction.In this study,firstly,the collected backpack LiDAR point cloud data is processed based on the point cloud intensity,and the point cloud data is divided into intervals according to the point cloud intensity and the best point cloud intensity interval is determined,so as to reduce the influence of low-quality points on the extraction accuracy of DBH.The results show that the best point cloud intensity interval is[5,10].Compared with the complete point cloud extraction results under the same conditions,the point cloud data DBH extraction accuracy in the best point cloud intensity interval is more advantageous,and the operation efficiency is also significantly improved.(2)The accuracy of extracting DBH from backpack LiDAR point cloud data is related to the thickness of DBH slice.The low-quality points account for a small proportion in the data of DBH slice point cloud,but they still affect the extraction of high-precision DBH.In this paper,by analyzing the accuracy of extracting DBH with three slice thicknesses,it is found that the smaller the slice thickness of point cloud is,the greater the influence of low-quality points on the high-quality point cloud when it is projected on the two-dimensional plane,and the worse the extraction result of DBH is.On the contrary,the greater the thickness of point cloud slice,the higher the proportion of high-quality points in DBH slice,which makes the result of DBH closer to the true value,so the higher the slice thickness,the higher the extraction accuracy of DBH.In this study,DBH was extracted under three slice thickness conditions:0.2m,0.4m and 0.6m.When the slice thickness was 0.6m,the extraction accuracy of DBH in three sample plots was the best.(3)Aalgorithm of clustering of layer-by-layer based on KDtree is proposed for individual tree segmentation to extract tree height.According to the natural growth form of tree crown,starting from the DBH coordinates of single tree,each layer of single tree point cloud is obtained by direct filtering,and each layer of point cloud is clustered by KDtree nearest neighbor search method.The initial layer center is updated by calculating the center of gravity of the layer point cloud,and the offset angle of the layer center is determined.The layer center curve of single wood is obtained layer by layer,so as to complete the single wood segmentation.The tree height parameters can be extracted by the segmented single tree point cloud by using the outer box algorithm.The extraction accuracy of tree height in three plots is 0.91,0.91 and 0.82,and RMSE is 0.63m,0.56m and 0.72m,respectively. |
