Coarse-to-Fine Localization for Detecting Misalignment State of Angle Cocks.

Area of Science:

• Railway Engineering
• Mechanical Engineering
• Computer Vision

Background:

• Accurate detection of angle cock states is vital for freight train safety.
• Existing methods often fail to distinguish between fully open and misaligned states, potentially overlooking critical faults.
• Misalignment can lead to abnormal braking, posing a significant safety risk.

Purpose of the Study:

• To develop a robust method for detecting the specific misalignment state of freight train angle cocks.
• To differentiate the misalignment state from other non-closed states for enhanced safety.
• To improve the precision of angle cock fault detection beyond simple normal/abnormal classification.

Main Methods:

• A coarse-to-fine localization approach using YOLOv4 for initial angle cock detection.
• Support Vector Machine (SVM) with Histogram of Oriented Gradients (HOG) features for handle localization and non-closed state detection.
• Binarization, window search, and minimum bounding rectangle analysis to calculate handle tilt angle for misalignment detection.

Main Results:

• The proposed method successfully differentiates between fully open and misaligned angle cock states.
• The system achieves a high accuracy of 96.49% in detecting the misalignment state.
• Experimental validation confirms the effectiveness and robustness of the coarse-to-fine localization technique.

Conclusions:

• The developed coarse-to-fine localization method effectively identifies misaligned angle cocks in freight trains.
• This advanced detection capability enhances train safety by preventing potential brake malfunctions.
• The study highlights the importance of granular fault detection in critical railway components.