PointTrackNet: An End-to-End Network for 3-D Object Detection and Tracking from Point Clouds

Sukai Wang; Yuxiang Sun; Chengju Liu; Ming Liu

doi:10.1109/LRA.2020.2974392

PointTrackNet: An End-to-End Network for 3-D Object Detection and Tracking from Point Clouds

Sukai Wang
, Yuxiang Sun
, Chengju Liu
, Ming Liu^*

^*Corresponding author for this work

Research output: Journal Publications and Reviews › RGC 21 - Publication in refereed journal › peer-review

67 Citations (Scopus)

Abstract

Recent machine learning-based multi-object tracking (MOT) frameworks are becoming popular for 3-D point clouds. Most traditional tracking approaches use filters (e.g., Kalman filter or particle filter) to predict object locations in a time sequence, however, they are vulnerable to extreme motion conditions, such as sudden braking and turning. In this letter, we propose PointTrackNet, an end-to-end 3-D object detection and tracking network, to generate foreground masks, 3-D bounding boxes, and point-wise tracking association displacements for each detected object. The network merely takes as input two adjacent point-cloud frames. Experimental results on the KITTI tracking dataset show competitive results over the state-of-the-arts, especially in the irregularly and rapidly changing scenarios. © 2020 IEEE.

Original language	English
Article number	9000527
Pages (from-to)	3206-3212
Journal	IEEE Robotics and Automation Letters
Volume	5
Issue number	2
Online published	17 Feb 2020
DOIs	https://doi.org/10.1109/LRA.2020.2974392
Publication status	Published - Apr 2020
Externally published	Yes

Funding

This work was supported in part by the National Natural Science Foundation of China under Grants U1713211 and 61673300, in part by the Basic Research Project of Shanghai Science and Technology Commission under Grant 18DZ1200804, and in part by HKUST ECE Start-up Grant from HKUST for Heterogeneous Navigation System.

Research Keywords

autonomous vehicles
end-to-end
multiple-object tracking
Point cloud

Access Output

10.1109/LRA.2020.2974392

Other Access Options

Check CityUHK Library

Permanent Link

Right click to copy link

Cite this

@article{1242fdd46ad04c0b85d51e7946add0d9,

title = "PointTrackNet: An End-to-End Network for 3-D Object Detection and Tracking from Point Clouds",

abstract = "Recent machine learning-based multi-object tracking (MOT) frameworks are becoming popular for 3-D point clouds. Most traditional tracking approaches use filters (e.g., Kalman filter or particle filter) to predict object locations in a time sequence, however, they are vulnerable to extreme motion conditions, such as sudden braking and turning. In this letter, we propose PointTrackNet, an end-to-end 3-D object detection and tracking network, to generate foreground masks, 3-D bounding boxes, and point-wise tracking association displacements for each detected object. The network merely takes as input two adjacent point-cloud frames. Experimental results on the KITTI tracking dataset show competitive results over the state-of-the-arts, especially in the irregularly and rapidly changing scenarios. {\textcopyright} 2020 IEEE.",

keywords = "autonomous vehicles, end-to-end, multiple-object tracking, Point cloud",

author = "Sukai Wang and Yuxiang Sun and Chengju Liu and Ming Liu",

year = "2020",

month = apr,

doi = "10.1109/LRA.2020.2974392",

language = "English",

volume = "5",

pages = "3206--3212",

journal = "IEEE Robotics and Automation Letters",

issn = "2377-3766",

publisher = "IEEE",

number = "2",

}

TY - JOUR

T1 - PointTrackNet

T2 - An End-to-End Network for 3-D Object Detection and Tracking from Point Clouds

AU - Wang, Sukai

AU - Sun, Yuxiang

AU - Liu, Chengju

AU - Liu, Ming

PY - 2020/4

Y1 - 2020/4

N2 - Recent machine learning-based multi-object tracking (MOT) frameworks are becoming popular for 3-D point clouds. Most traditional tracking approaches use filters (e.g., Kalman filter or particle filter) to predict object locations in a time sequence, however, they are vulnerable to extreme motion conditions, such as sudden braking and turning. In this letter, we propose PointTrackNet, an end-to-end 3-D object detection and tracking network, to generate foreground masks, 3-D bounding boxes, and point-wise tracking association displacements for each detected object. The network merely takes as input two adjacent point-cloud frames. Experimental results on the KITTI tracking dataset show competitive results over the state-of-the-arts, especially in the irregularly and rapidly changing scenarios. © 2020 IEEE.

AB - Recent machine learning-based multi-object tracking (MOT) frameworks are becoming popular for 3-D point clouds. Most traditional tracking approaches use filters (e.g., Kalman filter or particle filter) to predict object locations in a time sequence, however, they are vulnerable to extreme motion conditions, such as sudden braking and turning. In this letter, we propose PointTrackNet, an end-to-end 3-D object detection and tracking network, to generate foreground masks, 3-D bounding boxes, and point-wise tracking association displacements for each detected object. The network merely takes as input two adjacent point-cloud frames. Experimental results on the KITTI tracking dataset show competitive results over the state-of-the-arts, especially in the irregularly and rapidly changing scenarios. © 2020 IEEE.

KW - autonomous vehicles

KW - end-to-end

KW - multiple-object tracking

KW - Point cloud

UR - https://www.scopus.com/pages/publications/85081680591

UR - https://www.scopus.com/record/pubmetrics.uri?eid=2-s2.0-85081680591&origin=recordpage

U2 - 10.1109/LRA.2020.2974392

DO - 10.1109/LRA.2020.2974392

M3 - RGC 21 - Publication in refereed journal

SN - 2377-3766

VL - 5

SP - 3206

EP - 3212

JO - IEEE Robotics and Automation Letters

JF - IEEE Robotics and Automation Letters

IS - 2

M1 - 9000527

ER -

PointTrackNet: An End-to-End Network for 3-D Object Detection and Tracking from Point Clouds

Abstract

Funding

Research Keywords

Access Output

Other Access Options

Permanent Link

Other Files and Links

Fingerprint

Cite this