A Bayesian Filter for Multi-View 3D Multi-Object Tracking With Occlusion Handling

This paper proposes an online multi-camera multi-object tracker that only requires monocular detector training, independent of the multi-camera configurations, allowing seamless extension/deletion of cameras without retraining effort. The proposed algorithm has a linear complexity in the total numbe...

Full description

Bibliographic Details
Main Authors: Ong, Jonah, Vo, Ba Tuong, Vo, Ba-Ngu, Kim, Du Yong, Nordholm, Sven
Format: Journal Article
Language:English
Published: IEEE COMPUTER SOC 2022
Subjects:
Online Access:http://purl.org/au-research/grants/arc/DP170104854
http://hdl.handle.net/20.500.11937/90801
_version_ 1848765432122572800
author Ong, Jonah
Vo, Ba Tuong
Vo, Ba-Ngu
Kim, Du Yong
Nordholm, Sven
author_facet Ong, Jonah
Vo, Ba Tuong
Vo, Ba-Ngu
Kim, Du Yong
Nordholm, Sven
author_sort Ong, Jonah
building Curtin Institutional Repository
collection Online Access
description This paper proposes an online multi-camera multi-object tracker that only requires monocular detector training, independent of the multi-camera configurations, allowing seamless extension/deletion of cameras without retraining effort. The proposed algorithm has a linear complexity in the total number of detections across the cameras, and hence scales gracefully with the number of cameras. It operates in the 3D world frame, and provides 3D trajectory estimates of the objects. The key innovation is a high fidelity yet tractable 3D occlusion model, amenable to optimal Bayesian multi-view multi-object filtering, which seamlessly integrates, into a single Bayesian recursion, the sub-tasks of track management, state estimation, clutter rejection, and occlusion/misdetection handling. The proposed algorithm is evaluated on the latest WILDTRACKS dataset, and demonstrated to work in very crowded scenes on a new dataset.
first_indexed 2025-11-14T11:35:09Z
format Journal Article
id curtin-20.500.11937-90801
institution Curtin University Malaysia
institution_category Local University
language English
last_indexed 2025-11-14T11:35:09Z
publishDate 2022
publisher IEEE COMPUTER SOC
recordtype eprints
repository_type Digital Repository
spelling curtin-20.500.11937-908012023-04-20T05:37:01Z A Bayesian Filter for Multi-View 3D Multi-Object Tracking With Occlusion Handling Ong, Jonah Vo, Ba Tuong Vo, Ba-Ngu Kim, Du Yong Nordholm, Sven Science & Technology Technology Computer Science, Artificial Intelligence Engineering, Electrical & Electronic Computer Science Engineering Three-dimensional displays Cameras Trajectory Bayes methods Detectors Training Visualization Multi-view multi-sensor multi-object visual tracking occlusion handling generalized labeled multi-bernoulli PERFORMANCE EVALUATION MULTITARGET TRACKING VISUAL TRACKING CAMERAS This paper proposes an online multi-camera multi-object tracker that only requires monocular detector training, independent of the multi-camera configurations, allowing seamless extension/deletion of cameras without retraining effort. The proposed algorithm has a linear complexity in the total number of detections across the cameras, and hence scales gracefully with the number of cameras. It operates in the 3D world frame, and provides 3D trajectory estimates of the objects. The key innovation is a high fidelity yet tractable 3D occlusion model, amenable to optimal Bayesian multi-view multi-object filtering, which seamlessly integrates, into a single Bayesian recursion, the sub-tasks of track management, state estimation, clutter rejection, and occlusion/misdetection handling. The proposed algorithm is evaluated on the latest WILDTRACKS dataset, and demonstrated to work in very crowded scenes on a new dataset. 2022 Journal Article http://hdl.handle.net/20.500.11937/90801 10.1109/TPAMI.2020.3034435 English http://purl.org/au-research/grants/arc/DP170104854 http://purl.org/au-research/grants/arc/DP160104662 http://creativecommons.org/licenses/by/4.0/ IEEE COMPUTER SOC fulltext
spellingShingle Science & Technology
Technology
Computer Science, Artificial Intelligence
Engineering, Electrical & Electronic
Computer Science
Engineering
Three-dimensional displays
Cameras
Trajectory
Bayes methods
Detectors
Training
Visualization
Multi-view
multi-sensor
multi-object visual tracking
occlusion handling
generalized labeled multi-bernoulli
PERFORMANCE EVALUATION
MULTITARGET TRACKING
VISUAL TRACKING
CAMERAS
Ong, Jonah
Vo, Ba Tuong
Vo, Ba-Ngu
Kim, Du Yong
Nordholm, Sven
A Bayesian Filter for Multi-View 3D Multi-Object Tracking With Occlusion Handling
title A Bayesian Filter for Multi-View 3D Multi-Object Tracking With Occlusion Handling
title_full A Bayesian Filter for Multi-View 3D Multi-Object Tracking With Occlusion Handling
title_fullStr A Bayesian Filter for Multi-View 3D Multi-Object Tracking With Occlusion Handling
title_full_unstemmed A Bayesian Filter for Multi-View 3D Multi-Object Tracking With Occlusion Handling
title_short A Bayesian Filter for Multi-View 3D Multi-Object Tracking With Occlusion Handling
title_sort bayesian filter for multi-view 3d multi-object tracking with occlusion handling
topic Science & Technology
Technology
Computer Science, Artificial Intelligence
Engineering, Electrical & Electronic
Computer Science
Engineering
Three-dimensional displays
Cameras
Trajectory
Bayes methods
Detectors
Training
Visualization
Multi-view
multi-sensor
multi-object visual tracking
occlusion handling
generalized labeled multi-bernoulli
PERFORMANCE EVALUATION
MULTITARGET TRACKING
VISUAL TRACKING
CAMERAS
url http://purl.org/au-research/grants/arc/DP170104854
http://purl.org/au-research/grants/arc/DP170104854
http://hdl.handle.net/20.500.11937/90801