International Journal of Applied Earth Observations and Geoinformation (Nov 2023)
Waveform classification and index sharing based-effective intensity data extraction of airborne laser bathymetry
Abstract
Alongside widely accepted geometric information, airborne laser bathymetry (ALB) typically captures the temporal profile sample (waveforms) of both the emitted laser pulses and their echoes. These waveforms provide radiometric properties (backscattering intensity) of sensed targets and assist with precise strip registration, classification of fine ground cover (sediment), and advanced geometric modelling. However, effective intensity data extraction is essential because the intensity provided by the LAS file is mixed with invalid signals, multiple echo intensity, and intermittent intensity variations caused by improper interference of automatic gain control, making it unusable for direct analysis. To address this issue, a flexible index sharing mechanism between waveform, coordinate, and intensity data is constructed for interference intensity tracking. A novel waveform classification-based approach is proposed to efficiently extract ALB intensity by dividing waveform data into six categories using morphological differences and topographic data of waveforms. To ensure accurate analysis, duplicate and invalid waveforms are eliminated, leaving only genuine intensity readings. Additionally, a triple spline fit is employed to restore oversaturated waveform segments that were previously suppressed due to exceeding the device's maximum measurable limit. To address the problem of mixing multiple return intensities, waveforms are decomposed using various decomposition models. This approach ensures that different waveform categories retain their respective varying return intensities. The approach is then tested on an ALB dataset collected using Optech Aquarius around Yuanzhi Island in the South China Sea. The results demonstrate a considerable advancement in data quality when compared to LAS file intensity products with a reduction in maximum deviation of 883 [digital number, DN], standard deviation of 350 [DN], and mean absolute error of 310 [DN].
