Segmentação de Objetos utilizando LiDAR e os Métodos de Clusterização DBSCAN e HDBSCAN

Henrique Nunes Poleselo; Tiago T. Ribeiro; Andre G. S. Conceicao

doi:10.20906/SBAI-SBSE-2023/3945

Henrique Nunes Poleselo Programa de Pós-Graduação em Engenharia Elétrica, Universidade Federal da Bahia, Salvador-BA
Tiago T. Ribeiro Departamento de Engenharia Elétrica e de Computação, Universidade Federal da Bahia, Salvador-BA
Andre G. S. Conceicao Departamento de Engenharia Elétrica e de Computação, Universidade Federal da Bahia, Salvador-BA

DOI: https://doi.org/10.20906/SBAI-SBSE-2023/3945

Keywords: Unsupervised Clustering, Segmentation, DBSCAN, HDBSCAN, Autonomous navigation

Abstract

In the field of robotics, object segmentation is one of the main activities for the robot to create a sense of its environment. This task is a precursor to other activities such as autonomous navigation in an environment. Through sensors such as LiDAR, it is possible to generate high-resolution three-dimensional maps of the environment in which the robot is inserted, thus enabling their interpretation so that tasks such as object segmentation can be performed. In this article, the DBSCAN and HDBSCAN unsupervised clustering methods are explored. Results in a simulated environment in framework Gazebo together with the ROS for capturing sensory data from LiDAR Livox Mid-70 coupled to a mobile robot show the performance of such techniques through comparisons.