Dual Codebook- and Scene Change-Based Loop Closure Detection Using SVM Outlier Removal in Verification
IEEE/CAA Journal of Automatica Sinica · 2026-10 · 13 (10) · 2312-2326
DOI: 10.1109/jas.2026.125762 ↗Kurz gefasst
Fachartikel in der Zeitschrift IEEE/CAA Journal of Automatica Sinica, 2026-10. Verfasst von Kuo-Liang Chung und 2 weiteren Autor:innen. Die Arbeit nennt 43 Literaturangaben. Gefördert von: Ministry of Science and Technology, Taiwan.
Abstract (Originalsprache)
Loop closure detection (LCD) is a key component of simultaneous localization and mapping (SLAM) systems, essential for mitigating accumulated errors and ensuring map consistency. This paper presents a novel dual codebook- and scene change-based LCD algorithm, enhanced by a fast support vector machine (SVM)-based outlier removal method during the verification process. Different from using only one static codebook in the existing methods, our first novelty is the proposal of a dual codebook in which the super-features that often appear are saved in the primary codebook, but some rarely used super-features are saved in the secondary codebook for increasing the LCD accuracy. Once the frequency of one codeword in the secondary code-book is larger than the frequency of one codeword in the primary codebook, a codeword-swapping operation is performed to exchange the two concerned codewords between the two code-books. The second novelty introduces a scene change- and code-word histogram-based method to efficiently retrieve the most relevant candidate images for each query image, significantly reducing computation time. The third novelty presents an SVM-based outlier removal method that accelerates the LCD verification step while preserving high detection accuracy. Based on nine large-scale video sequences, comprehensive experimental results demonstrate that our algorithm outperforms state-of-the-art methods in both LCD accuracy and execution efficiency. Furthermore, an ablation study validates the effectiveness of the three proposed contributions. The source code of our algorithm is available at https://github.com/hankjian7/DC-SC-SVM-LCD.git.
Förderung
- Ministry of Science and Technology, Taiwan
Zitieren
APA
Chung, K., Jian, Q., & Zhang, Y. (2026). Dual Codebook- and Scene Change-Based Loop Closure Detection Using SVM Outlier Removal in Verification. IEEE/CAA Journal of Automatica Sinica, 13(10), 2312-2326. https://doi.org/10.1109/jas.2026.125762
Vancouver
Chung K, Jian Q, Zhang Y. Dual Codebook- and Scene Change-Based Loop Closure Detection Using SVM Outlier Removal in Verification. IEEE/CAA Journal of Automatica Sinica. 2026;13(10):2312-2326. doi:10.1109/jas.2026.125762
BibTeX
@article{Chung2026,
author = {Chung, Kuo-Liang and Jian, Qi-Han and Zhang, Yu-Qi},
title = {Dual Codebook- and Scene Change-Based Loop Closure Detection Using SVM Outlier Removal in Verification},
journal = {IEEE/CAA Journal of Automatica Sinica},
year = {2026},
volume = {13},
number = {10},
pages = {2312--2326},
doi = {10.1109/jas.2026.125762}
}
EUVIMED