Physical Structure Expression for Dense Point Clouds of Magnetic Levitation Image Data.

Area of Science:

• Intelligent Transportation Systems (ITS)
• Computer Vision
• 3D Reconstruction

Background:

• Intelligent magnetic levitation (ML) transportation systems are a key area within ITS.
• ML systems require precise 3D structural data for applications like digital twins.
• Existing methods may lack the accuracy or robustness needed for complex ML track geometries.

Purpose of the Study:

• To develop and validate a robust 3D reconstruction system for magnetic levitation tracks.
• To leverage unmanned aerial vehicle (UAV) oblique photography and advanced computer vision techniques.
• To generate high-accuracy dense point clouds representing detailed track structures.

Main Methods:

• Acquisition of ML track imagery using UAV oblique photography.
• Image preprocessing and feature extraction.
• Incremental Structure from Motion (SFM) for sparse point cloud generation and camera pose recovery.
• Multi-view Stereo (MVS) for dense point cloud, depth, and normal map estimation.
• Bundle adjustment optimization for enhanced accuracy.

Main Results:

• Successful generation of 3D sparse and dense point clouds of magnetic levitation tracks.
• Accurate representation of physical track structures including turnouts, turns, and linear sections.
• Demonstrated robustness and high accuracy through comparison with traditional Building Information Models (BIM).

Conclusions:

• The developed 3D reconstruction system, based on incremental SFM and MVS, is robust and accurate.
• The system effectively captures complex physical structures of magnetic levitation tracks.
• This technology provides crucial data support for intelligent transportation systems and digital twin development.