LiDAR Point Cloud Recognition of Overhead Catenary System with Deep Learning.

Area of Science:

• Engineering
• Computer Science
• Transportation Technology

Background:

• High-speed railways rely on overhead catenary systems (OCS) for power.
• Manual OCS inspection is inefficient and costly, necessitating automated solutions.
• 3D point cloud data offers valuable geometric information for OCS inspection.

Purpose of the Study:

• To develop a deep learning-based method for recognizing OCS components from point cloud data.
• To automate the geometric parameter measurement for OCS inspection.
• To improve the efficiency and accuracy of high-speed railway inspection.

Main Methods:

• Utilized a convolutional neural network (CNN) to identify context in single-frame point clouds.
• Combined single-frame data based on classification results.
• Employed a segmentation network for OCS component identification.
• Created a dedicated point cloud dataset of OCS components with eight categories.

Main Results:

• The proposed deep learning method accurately detects OCS components from point cloud data.
• Experimental results demonstrate high accuracy in component recognition.
• The method successfully identifies eight distinct categories of OCS components.

Conclusions:

• The developed method offers a significant advancement for automated OCS inspection.
• This approach has practical applications in real-world OCS component detection.
• The research contributes to the automation of high-speed railway maintenance and safety.