利用多通道加权投票的机载绿激光海陆波形分类  

Ocean-Land Waveform Classification Based on Multichannel Weighted Voting of Airborne Green Laser

作　　者：赵兴磊 梁刚 赵建虎[2] 周丰年 Zhao Xinglei;Liang Gang;Zhao Jianhu;Zhou Fengnian(College of Information Science and Engineering,Shandong Agricultural University,Tai’an 271018,Shandong,China;School of Geodesy and Geomatics,Wuhan University,Wuhan 430079,Hubei,China;The Survey Bureau of Hydrology and Water Resources of Yangtze Estuary,Shanghai 200136,China)

机构地区：[1]山东农业大学信息科学与工程学院,山东泰安271018 [2]武汉大学测绘学院,湖北武汉430079 [3]长江水利委员会水文局长江口水文水资源勘测局,上海200136

出　　处：《激光与光电子学进展》2024年第9期183-191,共9页Laser & Optoelectronics Progress

基　　金：国家自然科学基金(41906166)。

摘　　要：为解决复杂海陆环境下机载绿激光海陆波形分类精度低的问题,本文提出了一种利用多通道加权投票的绿激光海陆波形分类方法,即多通道加权投票卷积神经网络(MWV-CNN)。首先,将绿激光深水和浅水通道采集的多通道波形经一个多通道输入模块分别输入到本文搭建的一维卷积神经网络(1D CNN)模块中;然后,各1D CNN模块对每个通道波形分别进行处理,获得各通道波形属于海洋和陆地类别的预测得分;最后,将各通道预测得分视为权值,利用一个多通道融合模块进行加权投票,确定波形最终类别。采用Optech CZMIL对中国连云港市沿海水域的实测数据进行实验验证,结果表明,MWV-CNN的总体分类精度、Kappa系数和总体精度标准差分别为99.45%、0.982和0.02%。与传统绿激光海陆波形分类方法相比,本文方法具有更好的分类精度和鲁棒性,为实现机载绿激光高精度海陆波形分类提供了一种新的有效途径。In order to improve the accuracy of oceanland waveform classifications of airborne green lasers in complex oceanland environments,an oceanland waveform classification method based on multichannel weighted voting[i.e.,multichannel weighted voting convolutional neural network(MWVCNN)]is proposed.First,the multichannel green laser waveforms collected in the deep and shallow channels are input into the proposed onedimensional 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 scores for each channel waveform belonging to the ocean and land categories.Finally,the predicted score of each channel is treated as weight,and a multichannel fusion module is used to determine the final waveform category via weighted voting.The measured data in the coastal waters of Lianyungang,China are verified by experiment using Optech CZMIL.The results indicate that the overall classification accuracy,Kappa coefficient,and overall accuracy standard deviation of MWVCNN are 99.45%,0.982,and 0.02%,respectively,and as compared with traditional oceanland waveform classification methods,the proposed method exhibits better classification accuracy and robustness,thus providing a new effective way for realizing oceanland waveform classification of airborne green laser with high accuracy.

关 键 词：大气光学与海洋光学 机载激光雷达测深 海陆波形分类 绿激光多通道波形 深度学习 加权投票 

分 类 号：P229[天文地球—大地测量学与测量工程]