Journal of Atmospheric and Environmental Optics, Volume. 20, Issue 3, 410(2025)
A sea surface position extraction method for airborne marine lidar based on bidirectional long short-term memory network
Figures & Tables(12)
Fig. 1. Schematic diagram of airborne lidar seawater depth measurement
Fig. 2. Schematic diagram of peak detection (PD) algorithm and half-peak power (HPP) algorithm results
Fig. 3. Autocorrelation function diagram of single lidar data time series
Fig. 5. Changes in loss function of Bi-LSTM model training set and test set
Fig. 6. Distribution of lidar two-channel (1064 nm and 532 nm) data points along the flight distance in the first set of test dataset. (a) Scatter plot of two-channel sea surface position using PD algorithm; (b) statistical chart of two-channel sea surface position difference using PD algorithm; (c) scatter plot of two-channel sea surface position using Bi-LSTM algorithm; (d) statistical chart of two-channel sea surface position difference using Bi-LSTM algorithm; (e) line chart of two-channel sea surface position using Bi-LSTM algorithm; (f) statistical chart of two-channel sea surface position difference using Bi-LSTM algorithm
Fig. 7. Distribution of lidar two-channel (1064 nm and 532 nm) data points along the flight distance in the second set of test dataset. (a) Scatter plot of two-channel sea surface position using PD algorithm; (b) statistical chart of two-channel sea surface position difference using PD algorithm; (c) scatter plot of two-channel sea surface position using Bi-LSTM algorithm; (d) statistical chart of two-channel sea surface position difference using Bi-LSTM algorithm; (e) line chart of two-channel sea surface position using Bi-LSTM algorithm; (f) statistical chart of two-channel sea surface position difference using Bi-LSTM algorithm
Fig. 8. Range deviation of two methods (PD algorithm and Bi-LSTM algorithm) under different attenuation coefficients of lidar
Table 1. Parameters of Mapper5000 airborne lidar system[19]
View tableView in ArticleTable 1. Parameters of Mapper5000 airborne lidar system[19]
参数 详细说明 激光波长/nm 1064, 532 激光峰值功率/MW 1 激光脉冲宽度/ns 1 激光脉冲重复频率/kHz 5 扫描角度范围/(°) ±15 垂直分辨率/m 0.23 水平分辨率/m 0.26
Table 2. Parameters of model training
View tableView in ArticleTable 2. Parameters of model training
参数 详细说明 迭代次数 200 损失函数 均方误差 优化器 自适应矩估计 双向长短期记忆网络
输入层 (100) 全链接层1 (256), 激活函数 = '线性流整函数' 全链接层2 (64)
Table 3. Statistics of sea surface position results in the testing area
View tableView in ArticleTable 3. Statistics of sea surface position results in the testing area
组别 指标 红外通道 PD算法 Bi-LSTM算法 第一组测试集 距离偏差/m 0.0 -0.21 0.03* 标准差/m ― 0.35 0.04* 计算时间/s ― 15.1 1800.6 第二组测试集 距离偏差/m 0.0 -0.23 0.06* 标准差/m ― 0.37 0.05* 计算时间/s ― 15.1 1798.8
Table 4. Statistical comparison of convergence rates of different methods
View tableView in ArticleTable 4. Statistical comparison of convergence rates of different methods
数据集类型 训练集 测试集 数据量 1064 nm/532 nm通道各10万条原始回波数据 532 nm通道10万条原始回波数据 方法 Bi-LSTM算法 Bi-LSTM算法 PD算法 HPP算法 LLE算法 计算时间/s 1800.6 15.9 15.1 16.5 16.7
Tools
Get Citation
Copy Citation Text
Chen TANG, Ping JIANG, Wenjun ZHAN, Ziyu GUO, Miaoqi DING, Xiaoquan SONG. A sea surface position extraction method for airborne marine lidar based on bidirectional long short-term memory network[J]. Journal of Atmospheric and Environmental Optics, 2025, 20(3): 410
Paper Information
Category: "Advanced technology of lidar and its application in atmospheric environment" Albun
Received: Mar. 31, 2025
Accepted: --
Published Online: Jun. 9, 2025
The Author Email: Xiaoquan SONG (songxq@ouc.edu.cn)
© Copyright 2018-2021 | Chinese Laser Press.
All Rights Reserved 沪ICP备15018463号-20