Get started
Work

SSLD-DET: improving symbol detection in station signalling layout drawings with multi-scale feature fusion and advanced attention mechanisms

Year: 2026
Type: article
Abstract: Accurate and efficient detection of signalling symbols in station signalling layout drawings (SSLDs) is essential for intelligent railway design, engineering review, and digital management. However, existing detection methods often struggle with densely distributed graphical elements, substantial scale variations, and high visual similarity among different signalling symbols, resulting in insufficient feature discrimination and limited robustness in complex drawing environments. To address these challenges, we propose SSLD-DET, an efficient detection framework for station signalling layout drawings. First, the Efficient Channel Interaction C2f (ECI-C2f) module enhances channel interaction and improves the discrimination of signalling symbols while suppressing redundant structural interference. Second, the Global Context Enhancement Module (GCEM) introduces joint channel-spatial attention to strengthen global contextual perception, enabling more accurate representation of visually similar graphical elements. Finally, an enhanced Bidirectional Feature Pyramid Network (BiFPN) is incorporated into the neck to improve multi-scale feature fusion through adaptive bidirectional information propagation. Experimental results on a self-constructed SSLD symbol dataset demonstrate that SSLD-DET achieves a precision of 90.6% and a mAP of 93.4%, outperforming the YOLOv26n baseline by 3.9% and 4.1%, respectively. These results indicate that SSLD-DET provides an accurate and efficient solution for automatic signalling symbol detection in complex railway engineering drawings.
DOI: 10.1038/s41598-026-79115-8
Language: English
Open Access status: gold
Estimated APC paid: $2,690
OpenAlex ID: W7225617487
FWCI:
Cites:
Cited by:
Related to: