EUVIMEDEUVIMED entdecken
Crossref

YOLO26-Mamba3: Selective State Space With Complex State Space for Dense Railway Pedestrian Detection

Longxia Liao, Long Chen

IEEE Signal Processing Letters · 2026 · 33 · 3876-3880

DOI: 10.1109/lsp.2026.3734849 ↗

Kurz gefasst

Fachartikel in der Zeitschrift IEEE Signal Processing Letters, 2026. Verfasst von Longxia Liao und 1 weiteren Autor:innen. Die Arbeit wurde bisher 1-mal zitiert (Crossref) und nennt 29 Literaturangaben.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Dense pedestrian detection in railway scenes remains challenging because of frequent crowd gatherings and severe occlusion. Existing detectors in the You Only Look Once (YOLO) family often suffer from missed detections and false positives when pedestrians are tightly clustered. To address this issue, we propose YOLO26-Mamba3, a hybrid architecture that integrates YOLO26 with the Mamba3 Selective State Space Model (SSM) and its complex state representation. The method retains the efficient CNN backbone and NMS-free end-to-end inference pipeline of YOLO26 while replacing selected C2f modules in the neck network with sequence-aware Mamba3 modules. Specifically, the Mamba3 module first serializes the two-dimensional feature map into a one-dimensional sequence, uses the Selective SSM to focus on more informative pedestrian features under occlusion, and further employs the complex state space to encode relative spatial phase differences between adjacent individuals. This design helps distinguish closely adjacent pedestrians that are easily merged by conventional CNN-based detection heads. Experimental results on the self-built Dense Railway Pedestrian Detection Dataset show that YOLO26-Mamba3 improves mAP and recall under highly occluded conditions over the baseline YOLO26.

Zitieren

APA

Liao, L., & Chen, L. (2026). YOLO26-Mamba3: Selective State Space With Complex State Space for Dense Railway Pedestrian Detection. IEEE Signal Processing Letters, 33, 3876-3880. https://doi.org/10.1109/lsp.2026.3734849

Vancouver

Liao L, Chen L. YOLO26-Mamba3: Selective State Space With Complex State Space for Dense Railway Pedestrian Detection. IEEE Signal Processing Letters. 2026;33:3876-3880. doi:10.1109/lsp.2026.3734849

BibTeX
@article{Liao2026,
  author = {Liao, Longxia and Chen, Long},
  title = {YOLO26-Mamba3: Selective State Space With Complex State Space for Dense Railway Pedestrian Detection},
  journal = {IEEE Signal Processing Letters},
  year = {2026},
  volume = {33},
  pages = {3876--3880},
  doi = {10.1109/lsp.2026.3734849}
}
Kostenlos starten →