Multi-strategy arithmetic optimization algorithm for global optimization and uncertain motion tracking

Previous trackers mostly assume that the target has a smooth motion and perform target matching within a local window. However, targets often exhibit uncertain movements in real-world scenarios. Once the tracked target undergoes abrupt motion and moves outside the predefined local window, these trac...

Full description

Saved in:
Bibliographic Details
Published in:Cluster computing Vol. 28; no. 1; p. 14
Main Authors: Gao, Zeng, Zhuang, Yi, Gu, Jingjing
Format: Journal Article
Language:English
Published: New York Springer US 01.02.2025
Springer Nature B.V
Subjects:
Accuracy
Algorithms
Arithmetic
Benchmarks
Computer Communication Networks
Computer Science
Design
Exploitation
Global optimization
Mathematical analysis
Operating Systems
Optical tracking
Optimization algorithms
Processor Architectures
Qualitative analysis
Sampling
Uncertain motion tracking
Arithmetic optimization algorithm
Dynamic stratification strategy
Multi-strategy
ISSN:1386-7857, 1573-7543
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:Previous trackers mostly assume that the target has a smooth motion and perform target matching within a local window. However, targets often exhibit uncertain movements in real-world scenarios. Once the tracked target undergoes abrupt motion and moves outside the predefined local window, these trackers often fail. To address this issue, this paper introduces a multi-strategy arithmetic optimization algorithm (MSAOA) for global optimization and uncertain motion tracking. MSAOA is a high-performance optimizer that effectively solves uncertain motion in visual tracking. For MSAOA, we first design a dynamic stratification strategy to divide the population into three subpopulations. Then the mathematical model of each subpopulation is modified to improve the exploration and exploitation performance. Finally, extensive experiments over 23 benchmark functions and CEC2020 benchmark problems show that MSAOA is better than other algorithms. For the MSAOA tracker (MSAOAT), we utilize the proposed MSAOA as a joint local sampling-global search to generate candidate targets and match the best targets by a fitness function. More importantly, we design a verifier to unite local sampling and global search to form a complete tracking framework, which can effectively address smooth and abrupt motion in visual tracking. The qualitative and quantitative analyses on the general motion group and the abrupt motion group demonstrate that the MSAOAT can outperform other trackers.
Bibliography:ObjectType-Article-1
SourceType-Scholarly Journals-1
ObjectType-Feature-2
content type line 14
ISSN:1386-7857
1573-7543
DOI:10.1007/s10586-024-04730-x