A pregeneration–recognition method of detecting weak seafloor echoes for full-waveform airborne LiDAR bathymetry

Full-waveform airborne LiDAR bathymetry (ALB) technology, in which the analyzed waveforms reflect the temporal positions and attribute information of targets, is effective in shallow water. However, weak seafloor echoes in full-waveform data induced by environmental and device characteristics are confused with noise signals, leading to difficulties in seafloor detection. This paper proposes a pregeneration–recognition method of detecting weak seafloor echoes for ALB. First, a two-stage local maximum algorithm is developed to identify potential seafloor echoes in waveforms and to pregenerate points. Then, an adaptive ellipsoidal neighborhood related to the point density is used to select neighborhood points, and eigenvalue-based spatial features are calculated. Finally, a back propagation neural network (BPNN) model is constructed using the points generated from surface–seafloor shots, and the seafloor points in seafloor-undefined shots are obtained by optimizing the BPNN results. The proposed method is verified on four swaths collected via the Optech Aquarius ALB system near Wuzhizhou Island and Ganquan Island in the South China Sea. The numbers of additional points detected with the proposed method near these two islands increase by 195.9 % and 40.1 % compared with the Aquarius system, which is better than the Richardson–Lucy deconvolution method. The coverages and maximum depth of seafloor points are improved and the accuracy evaluations demonstrate the credibility of the results. Therefore, the proposed pregeneration–recognition method can effectively improve the detection rate for weak seafloor echoes and the depth performance of ALB systems. Future research will focus on mitigating the impact of seafloor topography on the proposed method to expand its application scenarios.