<abstract xmlns="http://www.w3.org/1999/xhtml">

<p>The efficiency and robustness of modern visual tracking systems are largely dependent on the object detection system at hand. Bernoulli and Multi-Bernoulli filters have been proposed for visual tracking without explicit detections (image observations). However, these previous approaches do not fully exploit discriminative features for tracking. In this paper, we propose a novel Bernoulli filter with determinantal point processes observations. The proposed observation model can select groups of detections with high detection scores and low correlation among the observed features; thus achieving a robust filter.</p>
</abstract>

The efficiency and robustness of modern visual tracking systems are largely dependent on the object detection system at hand. Bernoulli and Multi-Bernoulli filters have been proposed for visual tracking without explicit detections (image observations). However, these previous approaches do not fully exploit discriminative features for tracking. In this paper, we propose a novel Bernoulli filter with determinantal point processes observations. The proposed observation model can select groups of detections with high detection scores and low correlation among the observed features; thus achieving a robust filter.

Robust single target tracking using determinantal point process observations

International Journal on Smart Sensing and Intelligent Systems

This work is licensed under the Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License.

running-head

The efficiency and robustness of modern visual tracking systems are largely dependent on the object detection system at hand. Bernoulli and Multi-Bernoulli...

Sequence	DPP	KCF	sKCF	Struck
Ball	0.309	0.289	0.246	0.372
Bolt	0.083	0.017	0.017	0.026
Diving	0.073	0.082	0.087	0.091
Gymnastics	0.710	0.425	0.425	0.435
Jogging	0.707	0.231	0.231	0.228
Polarbear	0.946	0.857	0.916	0.844

Particle Bernoulli-DPP filter
Number of particles N	100
Uniform birth probability (π_b)	0.1
Uniform survival probability (π_s)	0.99
Newborn particles (N_b)	0
Standard deviation for observation model (σ_o)	20.4
Covariance matrix for dynamic model (σ_x × 1)	3.0 × 1

Sequence	DPP	KCF	sKCF	Struck
Ball	0.206	0.211	0.201	0.128
Bolt	0.031	0.011	0.011	0.017
Diving	0.183	0.110	0.114	0.151
Gymnastics	0.560	0.415	0.420	0.425
Jogging	0.205	0.225	0.225	0.225
Polarbear	0.749	0.747	0.760	0.712

Robust single target tracking using determinantal point process observations

Article Category: Research-Article

Publicado en línea: 01 feb 2020

Páginas: 1 - 8

Recibido: 05 jun 2019

DOI: https://doi.org/10.21307/ijssis-2020-001

Palabras clave
Visual tracking, Bernoulli filter.

© 2020 S. Hernández et al., published by Sciendo

This work is licensed under the Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License.

Figure 1:

Figure 2:

Figure 3:

Average precision (th = 20).

Particle Bernoulli-DPP filter.

Greedy mode finding.

Average success (th = 0.5).

Greedy mode finding
Acceptance ratio ε	0.7

Robust single target tracking using determinantal point process observations

Article Category: Research-Article

Publicado en línea: 01 feb 2020

Páginas: 1 - 8

Recibido: 05 jun 2019

DOI: https://doi.org/10.21307/ijssis-2020-001

Palabras claveVisual tracking, Bernoulli filter.

© 2020 S. Hernández et al., published by Sciendo

This work is licensed under the Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License.

Figure 1:

Figure 2:

Figure 3:

Average precision (th = 20).

Particle Bernoulli-DPP filter.

Greedy mode finding.

Average success (th = 0.5).

Palabras clave
Visual tracking, Bernoulli filter.