Research on track fastener positioning method based on local unidirectional template matching.

Area of Science:

• Computer Vision
• Image Processing
• Railway Engineering

Background:

• Existing fastener positioning methods, such as pixel statistics (PS) and template matching (TM), suffer from low accuracy due to complex track backgrounds and difficulties in image segmentation.
• The PS method struggles with determining image segmentation thresholds in complex track environments.
• The TM method's global search is susceptible to background interference, reducing fastener locating accuracy.

Purpose of the Study:

• To propose a novel and accurate fastener positioning method for railway tracks.
• To address the limitations of existing pixel statistics and template matching techniques.
• To develop a method effective for both ballastless and ballasted track types.

Main Methods:

• A hybrid approach, local unidirectional template matching (LUTM), is proposed, combining PS and TM methods.
• Rail positioning is achieved using the PS method based on grayscale vertical projection.
• A 1-shaped rail image is generated, and a symmetrical fastener template is pre-processed before a top-to-bottom search along the rail to calculate correlation for positioning.

Main Results:

• The proposed LUTM method effectively overcomes the limitations of traditional PS and TM methods in complex track backgrounds.
• Accurate fastener positioning is achieved through a unidirectional search and correlation calculation.
• The method demonstrates efficacy for both ballastless and ballasted track systems.

Conclusions:

• The local unidirectional template matching (LUTM) method provides an effective solution for accurate fastener positioning in railway track inspection.
• This hybrid approach enhances robustness against complex backgrounds, improving overall locating accuracy.
• The developed method is suitable for diverse railway track infrastructures.