Object detection and tracking

Kateryna Chumachenko, Moncef Gabbouj, Alexandros Iosifidis

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

Abstract

The availability of an increasing amount of computational power and large-scale public data sets has driven the field of object detection and tracking with an unprecedented development speed, finding applications in many areas. This chapter surveys the most prominent methods in the field. We first formulate the problems of object detection and single and multiple object tracking, and then present the most relevant methodologies which have successfully been developed to solve these problems. The chapter includes a comprehensive treatment of two-stage, one-stage and anchor-free object detection methods. Both single object and multiple object tracking methods are reviewed in the chapter. The former includes methods based on correlation filters and deep learning, including similarity learning, whereas the latter present online and offline multiple object tracking techniques. Online methods, including deep features driven and detection-based methods, rely on visual representations, while offline methods are mostly based on graph optimization.

Original languageEnglish
Title of host publicationDeep Learning for Robot Perception and Cognition
EditorsAlexandros Iosifidis, Anastasios Tefas
Number of pages36
PublisherElsevier
Publication date2022
Pages243-278
ISBN (Print)9780323885720
ISBN (Electronic)9780323857871
DOIs
Publication statusPublished - 2022

Keywords

  • Multiple object tracking
  • Object detection
  • Object tracking
  • Single object tracking
  • Visual object tracking

Fingerprint

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

Cite this