There are a large number of ancient trees and urban trees in China,so how to build a reasonable digital model of trees is an important prerequisite for evaluating the wind resistance and safety of trees.In view of the variability of wind direction,it is necessary to obtain the different data of tree crown from any angles based on this model.Due to the high price and strong professionalism of Lidar,it is not convenient for its widely using,and a cheap and convenient mobile phone to capture tree images was used in this paper.In view of the fact that the traditional three-dimensional image reconstruction technology was not suitable for trees with complex structure,high reflectivity and porous objects,the researchers used nerf(neural radiation field)technology,which uses voxels to render the model volume and carry out three-dimensional digital reconstruction of trees,which can lay the foundation for obtaining tree images and canopy parameters from any angle.At the same time,the paper used instant ngp acceleration algorithm to analyze the influence of background,lighting conditions,image number and other factors on the quality of modeling.The results show that the tree digital reconstruction method based on nerf can effectively avoid the problems of cavity and deformation in traditional three-dimensional reconstruction,which is a feasible method for building three-dimensional digital model of trees.
NETL
NSTL
万方数据
