Reaserch On Ocean-Land Waveform Classification Of Airborne Oceanic LiDAR

Airborne oceanic Li DAR(AOL)enables high-resolution,efficient and flexible measurements in shallow waters such as coastal zones by actively transmitting laser pulses and receiving return signals.According to the number of laser wavelengths emitted,AOL systems are divided into single-frequency AOL systems emitting green laser and dual-frequency AOL systems emitting infrared(IR)and green laser.AOL system can obtain the ocean and land return waveform by receiving the laser pulse signal reflected from ocean and land,and realizes the integrated measurements of the ocean and land.Due to the different processing purposes,techniques,and methods of ocean and land waveforms,it is necessary to classify the original waveforms collected by AOL system.Ocean–land waveform classification(OLWC)is a key basic work in single-/dual-frequency AOL data processing.The realization of OLWC is of great significance for AOL waveform data processing and its application in hydrographic surveying and charting.Due to the different optical properties of IR and green lasers,the OLWC of single-/dual-frequency AOL has the following problems,respectively.(1)Amplitude feature of IR laser waveform of dual-frequency AOL systems is a common waveform feature to classify ocean and land,but the OLWC performance of other waveform features(such as full width at half maxima,area and width)has not been analyzed and evaluated experimentally;meanwhile,due to the complex ocean–land environmental factors,the use of IR laser in coastal areas for OLWC often results in misclassification caused by special waveforms.(2)The use of more economical and flexible single-frequency AOL system to detect the surface and seafloor is a future development trend.Although the OLWC based on IR laser has achieved satisfactory accuracy,the accuracy of OLWC based on green laser can not meet the actual requirements due to the waveform merging effect in very shallow water and the special terrain of land.For this purpose,high-accuracy OLWC of single-frequency and dual-frequency AOL studies are carried out in this paper,respectively.The main work and innovations of this paper are as follows.(1)A high-accuracy OLWC method based on dual-clustering for dual-frequency AOL is proposedThe amplitude feature of IR laser waveform of dual-frequency AOL system is a common feature to classify ocean and land,but the OLWC performance of other waveform features has not been analyzed and evaluated experimentally.Thus,the IR laser waveform feature selection experiment is conducted,and the shows that amplitude is the optimal waveform feature for OLWC using IR laser.Due to the complex ocean–land environmental factors,the use of IR laser in coastal areas for OLWC often results in misclassification caused by special waveforms.This paper proposes a high-accuracy OLWC method based on dual-clustering for IR laser on the basis of feature selection.The amplitude features of IR laser waveforms are first clustered to achieve rough OLWC,and then density-based clustering algorithms are used to identify and correct misclassified ocean and land waveforms based on the center position of the laser points,respectively.The experimental results show that the dual-clustering method can reach an overall accuracy of 99.73%,and the number of misclassified waveforms is reduced by 48%compared with the traditional K-means clustering,which effectively improves the accuracy of OLWC for IR laser,and can be applied to the high-accuracy OLWC for dual-frequency AOL in near-shore coastal waters.(2)A high-accuracy OLWC method based on multichannel voting for single-frequency AOL is proposedDue to the waveform merging effect in very shallow water and the special terrain of land,the accuracy of green laser OLWC is low and cannot meet the practical needs.This paper proposes an OLWC method based on multichannel voting,i.e.,multichannel voting convolutional neural network(MVCNN).First,the multichannel green laser waveforms collected in the deep and shallow channels are input into the corresponding one-dimensional convolutional neural network(1D CNN)module through a multichannel input module.Second,each 1D CNN module processes each channel waveform separately to obtain the predicted category of each channel waveform.Finally,a multichannel voting module is used to vote on the predicted category of each 1D CNN module to determine the final category of the waveform.The experimental results show that the overall accuracy,kappa coefficient,and standard deviation of the overall accuracy for the OLWC using MVCNN can reach 99.41%,0.9800,and 0.03%,respectively,which are better than the existing OLWC methods for green laser,and provide a novel means to realize OLWC with high accuracy for single-frequency AOL.