3D object detection and tracking

Research output: Contribution to book/anthology/report/proceedingBook chapterResearchpeer-review

1 Citation (Scopus)

Abstract

3D object detection and tracking are important tasks in robotics and autonomous systems because the application of 2D object detection methods cannot provide enough understanding of the 3D world that a robot (or agent) operate in. In this chapter, we describe 3D object detection and tracking methods, grouping them based on the sensors they use for collecting information about the environment. We describe in detail the most popular Lidar-based methods, the most affordable camera-based methods with monocular and binocular images, and data fusion methods that exploit information collected by multiple sensors at once. We also describe the most popular public datasets and metrics used for performance evaluation of detection and tracking methods.

Original languageEnglish
Title of host publicationDeep Learning for Robot Perception and Cognition
EditorsAlexandros Iosifidis, Anastasios Tefa
Number of pages28
PublisherElsevier
Publication date2022
Pages313-340
ISBN (Print)9780323885720
ISBN (Electronic)9780323857871
DOIs
Publication statusPublished - 2022

Keywords

  • 3D object detection
  • 3D object tracking
  • Depth estimation
  • Lidar
  • Pillar
  • Point cloud
  • Sensor fusion
  • Voxel

Fingerprint

Dive into the research topics of '3D object detection and tracking'. Together they form a unique fingerprint.

Cite this