Comparison of 3D interest point detectors and descriptors for point cloud fusion

Abstract

ABSTRACT:The extraction and description of keypoints as salient image parts has a long tradition within processing and analysis of 2D images. Nowadays, 3D data gains more and more importance. This paper discusses the benefits and limitations of keypoints for the task of fusing multiple 3D point clouds. For this goal, several combinations of 3D keypoint detectors and descriptors are tested. The experiments are based on 3D scenes with varying properties, including 3D scanner data as well as Kinect point clouds. The obtained results indicate that the specific method to extract and describe keypoints in 3D data has to be carefully chosen. In many cases the accuracy suffers from a too strong reduction of the available points to keypoints.

Development of a Change Detection Method with Low-Performance Point Cloud Data for Updating Three-Dimensional Road Maps

“…In addition to these, methods that use not only geometrical features but also color information have been developed, such as Color-SHOT [15] and PFHRGB [16]. Prior research has shown that using feature points reduces calculation costs, but does not always improve registration results [17]. Moreover, it is reported that the performance of descriptors depends on the shape of the objects [18].…”

Registration of RGB-D images: comparison of pairwise registration variants

“…Zainteresowanego czytelnika odsyłamy do artykułów przeglądowych i porównujących cechy punktowe, np. do klasycznego porównania deskryptorów fotometrycznych [21], deskryptorów binarnych [7] lub do artykułu przeglądowego omawiającego deskryptory geometryczne [9].…”

Revue des descripteurs tridimensionnels (3D) pour la catégorisation des nuages de points acquis avec un système LiDAR de télémétrie mobile

“…Les articles passant en revue les performances d'un nombre important de descripteurs 3D s'inscrivent le plus souvent dans le cadre de travaux de recherche visant le guidage de robots autonomes (Golovinskiy et al 2009;Himmelsbach et al 2009) ou bien la reconnaissance d'objets individuels ou domestiques (Chen et al 2016). Ces travaux impliquent en général des données issues de système de télémétrie statique ou de caméra de profondeur (c.-à-d., données rouge-vert-bleu profondeur, RGB-D) (Hänsch et al 2014;Chen et al 2016). Ces contextes ne présentent pas les mêmes difficultés que celles inhérentes aux scènes urbaines (cf.…”