DENSE MULTIPERSON TRACKING WITH ROBUST HIERARCHICAL LINEAR ASSIGNMENT

Abstract

We introduce a novel dual-stage algorithm for online multi-target tracking in realistic conditions. In the first stage, the problem of data association between tracklets and detections, given partial occlusion, is addressed using a novel occlusion robust appearance similarity method. This is used to robustly link tracklets with detections without requiring explicit knowledge of the occluded regions. In the second stage, tracklets are linked using a novel method of constraining the linking process that removes the need for ad-hoc tracklet linking rules. In this method, links between tracklets are permitted based on their agreement with optical flow evidence. Tests of this new tracking system have been performed using several public datasets.

Flow chart of the stages in our multi-target tracker

 

Results

Comparison of tracker performance using different tracklet linking methods: optical flow based tracklet linking - OF Linking. Tracklet linking with no optical flow linking - No OF Linking. No tracklet linking - 1st stage only.

 

Citation

Dense Multiperson Tracking with Robust Hierarchical Linear

N McLaughlinJ Martinez Del RinconP Miller

 IEEE Transactions on Cybernetics, Vol 45(7)

  

bibtex
@ARTICLE{McLaughlin2015,
author={McLaughlin, N. and Martinez del Rincon, J. and Miller, P.},
journal={Cybernetics, IEEE Transactions on},
title={Dense Multiperson Tracking with Robust Hierarchical Linear
Assignment},
year={2015},
volume={45},
number={7},
pages={1276-1288},
doi={10.1109/TCYB.2014.2348314},
ISSN={2168-2267},
month={July},}

 

Videos

Comparison of our multi-target tracker with tracklet linking enabled / disabled. Note that when tracklet linking is enabled the tracker is more consistent and makes fewer mistakes.

This tracker video shows the feature extraction regions used to build colour based appearance models of each person's head, torso, and upper legs