An Onboard Multimodal Sparse Perception Framework for Railway Intrusion Detection
IEEE Sensors Journal · 2026-08 · 26 (16) · 24478-24490
DOI: 10.1109/jsen.2026.3708154 ↗Kurz gefasst
Fachartikel in der Zeitschrift IEEE Sensors Journal, 2026-08. Verfasst von Xiying Song und 4 weiteren Autor:innen. Die Arbeit nennt 62 Literaturangaben. Gefördert von: National Natural Science Foundation of China.
Abstract (Originalsprache)
Railway foreign object intrusion (FOI) detection is a safety-critical onboard monitoring task that requires long-range multimodal sensing and reliable real-time response. However, this task is hindered by the mismatch between continuous dense sensing and sparse intrusion events, as well as by the scarcity and diversity of foreign objects. To address these challenges, this article proposes a multimodal sparse perception framework for a railway onboard sensing system. The framework first extracts intrusion-aware responses from camera and light detection and ranging (LiDAR) measurements and then forwards candidate anomalous regions to downstream recognition. Specifically, PIOADNet serves as an anomaly-sensitive detection module, while the sparse representation-guided prototype alignment (SRGPA) module supports fine-grained and cross-sample-consistent foreign object recognition. Experiments on a self-collected railway dataset show an image AUROC of 93.11% and a competitive point cloud AUROC of 94.59%. The results also demonstrate more than 78% reductions in inference time and GPU memory consumption at a 10% anomaly ratio.
Förderung
- National Natural Science Foundation of China
Zitieren
APA
Song, X., Song, H., Liu, S., Zhang, Y., & Dong, H. (2026). An Onboard Multimodal Sparse Perception Framework for Railway Intrusion Detection. IEEE Sensors Journal, 26(16), 24478-24490. https://doi.org/10.1109/jsen.2026.3708154
Vancouver
Song X, Song H, Liu S, Zhang Y, Dong H. An Onboard Multimodal Sparse Perception Framework for Railway Intrusion Detection. IEEE Sensors Journal. 2026;26(16):24478-24490. doi:10.1109/jsen.2026.3708154
BibTeX
@article{Song2026,
author = {Song, Xiying and Song, Haifeng and Liu, Shaoqing and Zhang, Yu and Dong, Hairong},
title = {An Onboard Multimodal Sparse Perception Framework for Railway Intrusion Detection},
journal = {IEEE Sensors Journal},
year = {2026},
volume = {26},
number = {16},
pages = {24478--24490},
doi = {10.1109/jsen.2026.3708154}
}
EUVIMED