Enhanced Edge Feature Fusion and Visual State Space for Precise Detection of Switch Tip Close Fitting
Switches are critical components in railway transportation, and the detection of close fitting between switch tips is essential for ensuring operational safety. Current methods primarily rely on manual on-site measurements, which are time-consuming and prone to errors. To address this, we propose DU...
Saved in:
| Main Authors: | , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
IEEE
2025-01-01
|
| Series: | IEEE Access |
| Subjects: | |
| Online Access: | https://ieeexplore.ieee.org/document/10982269/ |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | Switches are critical components in railway transportation, and the detection of close fitting between switch tips is essential for ensuring operational safety. Current methods primarily rely on manual on-site measurements, which are time-consuming and prone to errors. To address this, we propose DU-Net, an improved image segmentation and detection algorithm based on the U-Net model. DU-Net integrates edge feature enhancement and visual state space to accurately segment tracks in rail transit switch images and detect the close-fitting degree between the basic track and switch tip. An Edge Feature Enhancement Module uses dilated convolutions to enhance local feature extraction, while a Visual State Space Module in the skip connections captures contextual spatial information. A Dynamic Coefficient Fusion Edge Line Detection Module is also incorporated to improve the precision of switch point close contact measurement. Experimental results demonstrate that DU-Net achieves high segmentation and measurement accuracy, with an average Intersection over Union of 97.58% for rail recognition, a Dice similarity coefficient of 98.80%, and an average Hausdorff Distance of 0.426 mm. Furthermore, comparison with manual measurements reveals that measurement errors are within 1 mm, meeting practical application requirements. The proposed method represents a significant step forward in automating and improving the accuracy of switch tip close-fitting detection. |
|---|---|
| ISSN: | 2169-3536 |