Stereo Vision 3D Tracking of Multiple Free-Swimming Fish for Low Frame Rate Video

Palconit, Maria Gemel B. and Concepcion II, Ronnie S. and Alejandrino, Jonnel D. and Pareja, Michael E. and Almero, Vincent Jan D. and Bandala, Argel A. and Vicerra, Ryan Rhay P. and Sybingco, Edwin and Dadios, Elmer P. and Naguib, Raouf N.G. (2021) Stereo Vision 3D Tracking of Multiple Free-Swimming Fish for Low Frame Rate Video. Journal of Advanced Computational Intelligent and Intelligent Informatics. ISSN 1343-0130 (Accepted for Publication)

[thumbnail of FINAL - Palconit_JACIII_Fish Tracking_Final Revision.pdf] Text
FINAL - Palconit_JACIII_Fish Tracking_Final Revision.pdf - Accepted Version
Restricted to Repository staff only until 31 December 2021.
Available under License Creative Commons Attribution No Derivatives.

Download (635kB) | Request a copy

Abstract

3D multiple fish tracking has gained a significant growing research interest to quantify fish behavior. However, most tracking techniques have used a high frame rate that is currently not viable for real-time tracking applications. This study discusses multiple fish tracking techniques using low frame rate sampling of stereo video clips. The fish are tagged and tracked based on the absolute error of predicted indices using past and present fish centroid locations and a deterministic frame index. In the predictor sub-system, the linear regression and machine learning algorithms intended for nonlinear systems, such as Adaptive Neuro-Fuzzy Inference System (ANFIS), symbolic regression, and Gaussian Process Regression (GPR), were investigated. Results have shown that in the context of tagging and tracking accuracy, the symbolic regression attained the best performance, followed by the GPR, i.e., 74% to 100% and 81% to 91%, respectively. Considering the computation time, symbolic regression resulted in the highest computing lag of approximately 946 ms per iteration, whereas GPR achieved the lowest computing time of 39 ms.

Item Type: Article
Keywords: Multiple Object Tracking, Fish Tagging and Tracking, Multigene Genetic Programming, ANFIS, Gaussian Process Regression, Stereovision
Faculty / Department: Faculty of Science > Mathematics and Computer Science
Depositing User: Raouf Naguib
Date Deposited: 16 Sep 2021 09:27
Last Modified: 16 Sep 2021 09:27
URI: https://hira.hope.ac.uk/id/eprint/3356

Actions (login required)

View Item View Item