Island feature classification for single-wavelength airborne lidar bathymetry based on full-waveform parameters

Area of Science:

• Geomatics
• Remote Sensing
• Optical Engineering

Background:

• Single-wavelength airborne lidar bathymetry (ALB) is cost-effective for shallow water mapping.
• Waveform classification is challenging due to signal mixing and water-air interactions.

Purpose of the Study:

• To develop a robust method for classifying airborne lidar waveforms.
• To accurately differentiate between land and water echoes for improved bathymetric measurements.

Main Methods:

• Utilized the generalized Gaussian model (LM-GGM) to extract 38 waveform parameters.
• Employed random forest feature selection (RFFS) to identify 10 dominant features.
• Applied a random forest classification model for waveform categorization.

Main Results:

• Achieved an overall waveform classification accuracy of 97%.
• Successfully distinguished between isolated, supersaturated, land, and water waveforms.
• Demonstrated the effectiveness of morphological features for water-land division.

Conclusions:

• The proposed method accurately classifies airborne lidar waveforms for shallow water applications.
• Morphological feature analysis combined with random forest provides a reliable approach for ALB signal detection.
• High classification accuracy supports the use of ALB in coastal and inland water surveying.