Multi-drone Multi-object Tracking with RGB Cameras Using Spatio-Temporal Cues

Guanyin Chen, Bohui Fang, Wenxing Fu, Tao Yang

Research output: Chapter in Book/Report/Conference proceedingConference contributionpeer-review

Abstract

This paper proposed a method for multi-drone multi-object tracking (MDMOT) with spatio-temporal cues. When multiple cameras mounted on different drones are used to localize and track aerial objects, false associations between objects from different cameras will lead to the problem of false positive objects in the 3D space. Therefore, we first re-project all the triangulation localization points back to the camera pixel plane and calculate the re-projection errors for the construction of the spatial likelihood matrix of association. Then, the association likelihood is adjusted with temporal information by integrating the historical association results. Finally, we use the likelihood values as the similarity scores for data association. Our method relies only on distributed RGB cameras. The effectiveness is proved by quantitative and qualitative experiments with multi-object tracking metrics.

Original languageEnglish
Title of host publicationProceedings of 3rd 2023 International Conference on Autonomous Unmanned Systems (3rd ICAUS 2023) - Volume IV
EditorsYi Qu, Mancang Gu, Yifeng Niu, Wenxing Fu
PublisherSpringer Science and Business Media Deutschland GmbH
Pages412-421
Number of pages10
ISBN (Print)9789819710904
DOIs
StatePublished - 2024
Event3rd International Conference on Autonomous Unmanned Systems, ICAUS 2023 - Nanjing, China
Duration: 9 Sep 202311 Sep 2023

Publication series

NameLecture Notes in Electrical Engineering
Volume1174 LNEE
ISSN (Print)1876-1100
ISSN (Electronic)1876-1119

Conference

Conference3rd International Conference on Autonomous Unmanned Systems, ICAUS 2023
Country/TerritoryChina
CityNanjing
Period9/09/2311/09/23

Keywords

  • Data association
  • MDMOT
  • RGB Cameras
  • Triangulation

Fingerprint

Dive into the research topics of 'Multi-drone Multi-object Tracking with RGB Cameras Using Spatio-Temporal Cues'. Together they form a unique fingerprint.

Cite this