Related Experiment Video
Updated: Jul 15, 2026
03:31
End-To-End Deep Neural Network for Salient Object Detection in Complex Environments
Published on: December 15, 2023
Spatial-aware energy learning for out-of-distribution segmentation in railway intrusion detection
Runliang Tian1, Ruifeng Ding1, Dongyu Fan2
1School of Mechanical Engineering, Shijiazhuang Tiedao University, Shijiazhuang, 050043, Hebei, China.
Scientific Reports
|July 13, 2026
Summary
This study introduces a novel spatial-aware energy learning (SAEL) method to improve railway intrusion detection by better identifying unknown objects. The new approach enhances safety and reliability in railway operations.
Area of Science:
- Computer Vision
- Artificial Intelligence
- Railway Engineering
Background:
- Railway intrusion detection is vital for operational safety.
- Current deep learning methods struggle with detecting unknown foreign objects.
- Existing out-of-distribution (OoD) detection lacks spatial context for railway environments.
Purpose of the Study:
- To develop an uncertainty-driven, spatial-aware method for robust railway intrusion detection.
- To address the limitations of existing methods in handling unknown objects and spatial awareness.
Main Methods:
- Proposed a spatial-aware energy learning (SAEL) method incorporating distance transform and Gaussian smoothing.
- Introduced a texture representation module to capture anomalous region details.
- Evaluated the method on public and self-constructed datasets.
Main Results:
- SAEL significantly outperformed state-of-the-art baseline methods on the OoD-RID test set.
- Achieved a 10.68% reduction in FPR95, a 5.40% improvement in AuPRC, and a 1.73% increase in AuROC.
- Demonstrated superior detection performance and computational efficiency.
Conclusions:
- The SAEL method offers a robust solution for detecting unseen intrusions in complex railway scenes.
- This approach balances detection performance with computational efficiency for practical applications.
- Provides a practical out-of-distribution segmentation solution for railway safety.
