Combining geometrical and intensity information to recognize vehicles from super-high density UAV-LiDAR point clouds

Area of Science:

• Geospatial technology
• Remote sensing
• Computer vision

Background:

• Traditional airborne LiDAR vehicle recognition relies solely on geometric data, limiting accuracy in complex urban scenes.
• The effectiveness of existing methods on super-high density Unmanned Aerial Vehicle LiDAR (UAV-LiDAR) point clouds remains unclear.

Purpose of the Study:

• To develop and evaluate a novel 3D algorithm for accurate vehicle recognition using super-high density UAV-LiDAR point clouds.
• To integrate both geometric and intensity information for enhanced vehicle detection capabilities.

Main Methods:

• Conversion of raw point clouds into a 3D multi-value image fusing intensity, elevation, and density.
• Extraction of potential vehicle voxels based on consistent intensity, elevation, and density.
• Individual vehicle recognition via spatially connected voxels and vehicle size constraints.

Main Results:

• The proposed algorithm achieves high accuracy in vehicle recognition from UAV-LiDAR data.
• Demonstrated average quality (Kappa coefficient) of 96.58% (96.04%) across varying point cloud densities.
• Performance remains robust despite occlusion and dense vehicle arrangements.

Conclusions:

• The developed 3D algorithm effectively recognizes vehicles in super-high density UAV-LiDAR point clouds.
• Combining geometric and intensity data significantly improves recognition accuracy and robustness.
• The method shows strong potential for applications in urban surveillance and traffic monitoring.