Résumé: The development of inexpensive 3D data acquisition devices has promisingly facilitated the wide availability and popularity of point clouds, which attracts increasing attention to the effective extraction of 3D point cloud descriptors for accuracy of the efficiency of 3D computer vision tasks in recent years. However, how to develop discriminative and robust feature representations from 3D point clouds remains a challenging task due to their intrinsic characteristics. In this paper, we give a comprehensively insightful investigation of the existing 3D point cloud descriptors. These methods can be principally divided into two categories according to their advancement: hand-crafted and deep learning-based approaches, which will be further discussed from the perspective of elaborate classification, their advantages, and limitations. Finally, we present the future research directions of the extraction of 3D point cloud descriptors.

Langue: Anglais
Thème Informatique

Mots clés:
3D point cloud
Hand-crafted descriptor
Deep learning-based descriptor