Object-Based Point Cloud Analysis of Full-Waveform Airborne Laser Scanning Data for Urban Vegetation Classification

Area of Science:

• Remote Sensing
• Geospatial Analysis
• Forestry and Ecology

Background:

• Airborne laser scanning (ALS) is crucial for 3D vegetation mapping.
• Full-waveform (FWF) ALS data offers rich information (amplitude, width, multiple echoes) for enhanced point cloud classification.
• Current FWF data utilization for classification remains limited.

Purpose of the Study:

• To present an object-based point cloud analysis (OBPA) approach for classifying tall vegetation in urban environments using FWF ALS data.
• To leverage FWF ALS echo characteristics for improved 3D point cloud segmentation and classification.
• To demonstrate the method's effectiveness and accuracy in detecting trees and shrubs.

Main Methods:

• Object-based point cloud analysis (OBPA) combining segmentation and classification.
• Seeded region growing segmentation based on surface roughness and echo width homogeneity.
• Statistical classification tree derived from training data using aggregated echo features (amplitude, surface roughness).

Main Results:

• The OBPA method successfully segmented and classified tall vegetation (trees, shrubs) in urban environments.
• Achieved over 90% completeness and correctness in point-wise error assessment on validation sites.
• The approach directly processes original 3D points, avoiding data loss associated with gridding.

Conclusions:

• The proposed FWF ALS data processing method enhances tall vegetation detection in urban settings.
• Direct 3D point analysis offers superior precision compared to gridded data methods.
• This technique improves the separability of buildings and terrain occluded by vegetation.