Creating Three-Dimensional Models using Lidar Data Classification

This article discusses the issue of control, protection and conservation of forest resources, which is very relevant in the modern world. One of the technologies that is widely used in the field of forest tracking is LiDAR, which is capable of creating point clouds based on surrounding objects. However, in order to fully work with the data, point clouds need to be segmented, after which 3D models can be created based on the resulting clusters, which can be used in the future. However, point cloud segmentation does not always provide a good output result, as a result, it is necessary to manually select positive results after segmentation. To solve this problem, was hypothesize that the use of neural network based on PointNet architecture can speed up the process of creating 3D models and also increase their quality.