Multi-Size Voxel Cube (MSVC) Algorithm – A Novel Method for Dense Point Clouds Terrain Filtering Using Deep Neural Network

Many filters for ground filters have been developed. However, when filtering highly rugged terrain from dense point clouds (particularly in technical applications such as civil engineering), most widely used filtering approaches yield suboptimal results. Here, we proposed and tested a novel ground-filtering algorithm utilizing a deep neural network. It is based on the voxelization of the cloud, classification of individual voxels as ground or non-ground, and gradual reduction of voxel size. Individual voxels are classified using surrounding voxels (a “voxel cube” of 9x9x9 voxels). We have tested this algorithm on two dense point clouds capturing highly rugged areas with dense vegetation cover. The performance of the multi-size voxel cube (MSVC) algorithm was compared with that of the widely used cloth simulation filter (CSF). Manually classified terrain was used as the reference. MSVC consistently outperformed the CSF filter in terms of correctly identified ground points, correctly identified non-ground points, balanced accuracy, and F-score. Another advantage of this filter lies in its easy adaptability to any type of terrain, enabled by the utilization of machine learning. The only disadvantage lies in the necessity to prepare training data containing features that are present in the point cloud to be filtered. On the other hand, we aim to account for this in the future by producing neural networks trained for individual landscape types thus eliminating this phase of the work.