Faint Echo Extraction from ALB Waveforms Using a Point Cloud Semantic Segmentation Model

As an active remote sensing technology, airborne LIDAR can work at all times while emitting specific wavelengths of laser light that penetrate seawater. Airborne bathymetry (ALB) records object’s full return waveform, including the water surface, column, seafloor, and objects on it. Due to seawater’s absorption scattering seafloor’s reflectivity effect, amplitude seafloor echoes varies greatly. Seafloor with low signal-to-noise ratios are not easily detected using waveform processing methods, which lead insufficient topography depth incomplete coverage. To extract faint echoes, we proposed a extraction method based PointConv deep learning model, called FWConv. The assumed spatially adjacent were correlated. We converted multi-frame waveforms into point cloud. Each represented bin value in points’ properties contained spatial coordinates waveform. In semantic segmentation these clouds models, considered only each centroid’s amplitude, but also its neighboring distance amplitude. This enriched centroids’ features allowed model better discriminate between background noise echoes. results showed FWConv could experimental area was affected by noise, correctness reached 99.82%. number increased 158%, elevation accuracy 0.20 m concerning multibeam echo sounder data.