Have a personal or library account? Click to login

Robust single target tracking using determinantal point process observations

By:
Open Access
|Feb 2020

Figures & Tables

Figure 1:

Frame 85 of the jogging sequence. At each frame, a greedy mode finding step is performed using Algorithm 1. Rectangles represent ground-truth, state estimates and DPP observations.
Frame 85 of the jogging sequence. At each frame, a greedy mode finding step is performed using Algorithm 1. Rectangles represent ground-truth, state estimates and DPP observations.

Figure 2:

Overall precision plots for the visual tracking sequences.
Overall precision plots for the visual tracking sequences.

Figure 3:

Overall success plots for the visual tracking sequences.
Overall success plots for the visual tracking sequences.

Average precision (th = 20)_

SequenceDPPKCFsKCFStruck
Ball0.3090.2890.246 0.372
Bolt 0.083 0.0170.0170.026
Diving0.0730.0820.087 0.091
Gymnastics 0.710 0.4250.4250.435
Jogging 0.707 0.2310.2310.228
Polarbear 0.946 0.8570.9160.844

Particle Bernoulli-DPP filter_

Particle Bernoulli-DPP filter
Number of particles N 100
Uniform birth probability (πb)0.1
Uniform survival probability (πs)0.99
Newborn particles (Nb)0
Standard deviation for observation model (σo)20.4
Covariance matrix for dynamic model (σx × 1)3.0 × 1

Greedy mode finding_

Greedy mode finding
Acceptance ratio ε 0.7

Average success (th = 0_5)_

SequenceDPPKCFsKCFStruck
Ball0.206 0.211 0.2010.128
Bolt 0.031 0.011 0.0110.017
Diving 0.183 0.110 0.1140.151
Gymnastics 0.560 0.4150.4200.425
Jogging0.205 0.225 0.225 0.225
Polarbear0.7490.747 0.760 0.712
Language: English
Page range: 1 - 8
Submitted on: Jun 5, 2019
Published on: Feb 1, 2020
Published by: Professor Subhas Chandra Mukhopadhyay
In partnership with: Paradigm Publishing Services
Publication frequency: 1 issue per year

© 2020 S. Hernández, P. Sallis, published by Professor Subhas Chandra Mukhopadhyay
This work is licensed under the Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 License.