LAUSR

Ocean–land waveform classification (OLWC) is crucial in airborne LiDAR bathymetry (ALB) data processing and can be used for ocean–land discrimination and waterline extraction. However, the accuracy of OLWC for single-wavelength ALB systems is low given the nature of the green laser waveform in complex environments. Thus, in this article, a deep learning-based OLWC method called the multichannel voting convolutional neural network (MVCNN) is proposed based on the comprehensive utilization of multichannel green laser waveforms. First, multiple green laser waveforms collected in deep and shallow channels are input into a multichannel input module. Second, a one-dimensional (1-D) convolutional neural network (CNN) structure is proposed to handle each green channel waveform. Finally, a multichannel voting module is introduced to perform majority voting on the predicted categories derived by each 1-D CNN model and output the final waveform category (i.e., ocean or land waveforms). The proposed MVCNN is evaluated using the raw green laser waveforms collected by Optech coastal zone mapping and imaging LiDAR (CZMIL). Results show that the overall accuracy, kappa coefficient, and standard deviation of the overall accuracy for the OLWC utilizing green laser waveforms based on MVCNN can reach 99.41%, 0.9800, and 0.03%, respectively. Results further show that the classification accuracy of the MVCNN is improved gradually with the increase in the number of laser channels. The multichannel voting module can select the correct waveform category from the deep and shallow channels. The proposed MVCNN is highly accurate and robust, and it is slightly affected by aquaculture rafts and the merging effect of green laser waveform in very shallow waters. Thus, the use of MVCNN in OLWC for single-wavelength ALB systems is recommended. In addition, this article explores the relationships between green deep and shallow channel waveforms based on the analysis of CZMIL waveform data.