Abstract. This paper addresses classification of 3D point cloud data from natural environments based on voxels. The proposed model uses multi-layer perceptrons to classify voxels based on a statistic geometric analysis of the spatial distribution of inner points. Geometric features such as tubular structures or flat surfaces are identified regardless of their orientation, which is useful for unstructured or natural environments. Furthermore, the combination of voxels and neural networks pursues faster computation than alternative strategies. The model has been successfully tested with 3D laser scans from natural environments.Keywords: Multi-layer perceptron · 3D classification · Mobile robot · Voxel map
IntroductionKnowledge of geometric features in three-dimensional (3D) scenes is useful for object recognition in challenging applications in natural and unstructured environments, such as robotics for search and rescue (SAR) and planetary exploration [1,2]. In these applications, scenes are usually obtained through laser scanners [3] and stereo vision [4] as large and complex point clouds. Object recognition from point clouds usually involves three main steps: segmentation, feature extraction, and classification. Three main approaches have been adopted for object recognition in point clouds. The first approach avoids point-based computations by reducing the scene to a 2D representation which can be processed with standard artificial vision algorithms. The objects may be classified based on local and global statistics features of each object from a range image [5] or from a 2D deviation map when classification is based on texture analysis [6]. The classification can be also performed with image with depth data (RGB-D) by fusing results from separate 2D and 3D segmentation and feature extraction processes [7].