Seafloor Classification of Multibeam Sonar Data Using Neural Network Approach

In this study, the self-organizing map (SOM), which is an unsupervised clustering algorithm, and a supervised proportional learning vector quantization (PLVQ), are employed to develop a combined method of seafloor classification using multibeam sonar backscatter data. The PLVQ is a generalized learning vector quantization based on the proportional learning law (PLL). The proposed method was evaluated in an area where there are four types of sediments. The results show that the performance of the proposed method is better than the SOM and a statistical classification method.     

协同主动学习和半监督方法的海冰图像分类

海冰遥感光谱影像分类中标签样本难以获取,导致海冰分类精度难以提高,但是大量包含丰富信息的未标签样本却没有得到充分利用,针对这种情况,提出一种协同主动学习和半监督学习方法用于海冰遥感图像分类。在主动学习部分,结合最优标号和次优标号、自组织映射神经网络以及增强的聚类多样性算法来选择兼具不确定性和差异性的样本参与训练;在半监督学习部分,利用直推式支持向量机,并且融合主动学习思想从大量未标签样本中选取相对可靠且包含一定信息量的样本进行迭代训练;然后协同主动学习分类结果和半监督分类结果,通过一致性验证保证所加入伪标签样本的正确性。为了验证方法的有效性,分别采用巴芬湾地区30 m分辨率的Hyperion高光谱数据(验证数据为15 m分辨率的Landsat-8数据)和辽东湾地区15 m分辨率的Landsat-8数据(验证数据为4.77 m分辨率的Google Earth数据)进行海冰分类实验。实验结果表明,相对其他传统方法,该协同分类方法可以在只有少量标签样本的情况下,充分利用大量未标签样本中包含的信息,实现快速收敛,并获得较高的分类精度(两个实验的总体精度分别为90.003%和93.288%),适用于海冰遥感图像分类。     

SOM拓扑保序机理的模拟弹性力分析

针对SOM内在原理和基本特性中一些未解决的重要问题,定义了输出阵列上神经元i承受的、正比于模式向量及权向量之间距离的2种模拟弹性力：输入模式向量对胜者i的拉力,以及阵列中胜者j对邻近i的牵动力;分析了弹性力的9个性质,对SOM学习过程中神经元权向量的演变过程及拓扑映射保序的机理给出了易于理解的、形象的解释。     

Sea-floor classification using multibeam echo-sounding angular backscatter data: a real-time approach employing hybrid neural network architecture

The presently studied numerical model, e.g., composite roughness, is successful for the purpose of seafloor classification employing processed multibeam angular backscatter data from manganese-nodule-bearing locations of the Central Indian Ocean Basin. Hybrid artificial neural network (ANN) architecture, comprised of the self-organizing feature map and learning vector quantization (LVQ), has been implemented as an alternative technique for sea-floor roughness classification, giving comparative results with the aforesaid numerical model for processed multibeam angular backscatter data. However, the composite-roughness model approach is protracted due to the inherent need for processed data including system-gain corrections. In order to establish that tedious processing of raw backscatter values is unessential for efficient classification, hybrid ANN architecture has been attempted here due to its nonparametric approach. In this technical communication, successful employment of LVQ algorithm for unprocessed (raw) multibeam backscatter data indicates true real-time classification application.     

基于模糊C均值聚类-支持向量机的海底沉积物分类识别

在总结了目前海底底质分类研究的基础之上,率先提出利用计算机数值模拟技术对海底底质进行分类识别研究。相较于目前海底底质分类研究中所使用的水槽实验法,提出采用计算机数值正演技术模拟实际地震勘探中数据采集过程。在分类识别算法上,分别采用支持向量机(SVM)和模糊C均值聚类(FCM)算法对采集的数据进行分类,为使支持向量机分类识别率达到最大,引入差分进化算法对支持向量机中关键参数进行最优化搜索,并研究了向原始地震记录中加入10%,30%,50%的高斯白噪音时算法的稳定性。在分析了这两种算法分类识别的正确率及其各自的优缺点后,提出了海底底质分类识别的两步法,即(1)先利用模糊C均值聚类进行一粗糙的预测分类,在每一类中挑选聚类性较好的数据作为支持向量机的训练样本;(2)将上一步中筛选的样本作为支持向量机的训练样本,并用差分进化算法优化支持向量机分类参数,再利用训练好的支持向量机对其余数据做预测分类。鉴于计算机数值模拟的可重复性、高效快速性及本文提出的模糊C均值聚类-支持向量机方法的鲁棒性,为便于开展进一步研究,归纳总结了一套行之有效的采用计算机数值模拟技术开展海底底质分类识别研究的一般化流程。     

基于双极化SAR图像的U-Net海冰多分类模型

北极多年冰在近几十年有明显的减少趋势,与北极海冰的厚度、体积和夏季最小海冰范围的减少密切相关。合成孔径雷达(synthetic aperture radar, SAR)具有全天时、全天候成像能力,基于SAR卫星影像的海冰分类对监测北极多年冰具有重要意义。基于深度学习U-Net模型,以SAR图像的双极化信息为模型输入,构建了像素级的海水、一年冰和多年冰多分类模型。与已有SAR图像海冰分类方法(支持向量机、随机森林和卷积神经网络)进行对比,基于双极化SAR图像的U-Net海冰分类模型的准确率、平均重叠度和Kappa系数,分别达到了90.73%、0.831和0.849,优于其他对比模型,分别提升了4.08%～19.04%, 0.063～0.321和0.111～0.335。此外,针对SAR图像水平-垂直极化(horizontal-vertical polarization, HV)有明显的条状热噪声和水平-水平极化(horizontal-horizontal polarization, HH)受入射角效应而亮度不均匀的特点,设计敏感性实验,研究HV噪声、入射角和灰度共生矩阵(gray leve...     

结合纹理特征的SVM海冰分类方法研究

海冰分类是遥感监测领域中的重要应用之一,海冰分类的准确性对于评估海冰冰情、保证航海安全和开辟北极航道具有重要的意义。针对海冰分类问题,本文选用Sentinel-1遥感数据,结合纹理特征分析,提出了一种改进的SAR海冰分类方法。该方法选用灰度共生矩阵提取特征值,通过实验得到适宜用于海冰分类的多特征组合,在此基础上利用支持向量机开展SAR海冰类型的分类研究。实验结果表明,该方法可以实现对海冰SAR图像中一年冰、多年冰和海水3种类型识别,与传统的海冰分类方法神经网络和最大似然法相比较,使用SVM分类方法,结合纹理特征开展海冰类型监测是可行的,同时也表明多特征组合有利于提高SAR图像的分类精度,从而验证了本方法的有效性,为海冰分类提供了一种新思路。     

Rapid environmental assessment in the South China Sea: Improved inversion of sound speed profile using remote sensing data

Complex perturbations in the profile and the sparsity of samples often limit the validity of rapid environmental assessment (REA) in the South China Sea (SCS). In this paper, the remote sensing data were used to estimate sound speed profile (SSP) with the self-organizing map (SOM) method in the SCS. First, the consistency of the empirical orthogonal functions was examined by using k-means clustering. The clustering results indicated that SSPs in the SCS have a similar perturbation nature, which means the inverted grid could be expanded to the entire SCS to deal with the problem of sparsity of the samples without statistical improbability. Second, a machine learning method was proposed that took advantage of the topological structure of SOM to significantly improve their accuracy. Validation revealed promising results, with a mean reconstruction error of 1.26 m/s, which is 1.16 m/s smaller than the traditional single empirical orthogonal function regression (sEOF-r) method. By violating the constraints of linear inversion, the topological structure of the SOM method showed a smaller error and better robustness in the SSP estimation. The improvements to enhance the accuracy and robustness of REA in the SCS were offered. These results suggested a potential utilization of REA in the SCS based on satellite data and provided a new approach for SSP estimation derived from sea surface data.     

Mapping of rock types using a joint approach by combining the multivariate statistics,self-organizing map and Bayesian neural networks: an example from IODP 323 site

Modeling and classification of the subsurface lithology is very important to understand the evolution of the earth system. However, precise classification and mapping of lithology using a single framework are difficult due to the complexity and the nonlinearity of the problem driven by limited core sample information. Here, we implement a joint approach by combining the unsupervised and the supervised methods in a single framework for better classification and mapping of rock types. In the unsupervised method, we use the principal component analysis (PCA), K-means cluster analysis (K-means), dendrogram analysis, Fuzzy C-means (FCM) cluster analysis and self-organizing map (SOM). In the supervised method, we use the Bayesian neural networks (BNN) optimized by the Hybrid Monte Carlo (HMC) (BNN-HMC) and the scaled conjugate gradient (SCG) (BNN-SCG) techniques. We use P-wave velocity, density, neutron porosity, resistivity and gamma ray logs of the well U1343E of the Integrated Ocean Drilling Program (IODP) Expedition 323 in the Bering Sea slope region. While the SOM algorithm allows us to visualize the clustering results in spatial domain, the combined classification schemes (supervised and unsupervised) uncover the different patterns of lithology such of as clayey-silt, diatom-silt and silty-clay from an un-cored section of the drilled hole. In addition, the BNN approach is capable of estimating uncertainty in the predictive modeling of three types of rocks over the entire lithology section at site U1343. Alternate succession of clayey-silt, diatom-silt and silty-clay may be representative of crustal inhomogeneity in general and thus could be a basis for detail study related to the productivity of methane gas in the oceans worldwide. Moreover, at the 530 m depth down below seafloor (DSF), the transition from Pliocene to Pleistocene could be linked to lithological alternation between the clayey-silt and the diatom-silt. The present results could provide the basis for the detailed study to get deeper insight into the Bering Sea’ sediment deposition and sequence.     

基于支持向量机预测安达曼海南部内孤立波传播特性

内孤立波是发生在密度稳定层化海水中的一种特殊的海洋内波.预测内孤立波传播难度较大.本文提出了一种方法,利用美国麻省理工学院大气环流模型(MITgcm)的内孤立波模型计算了大量模拟数据,建立数据库.采用机器学习的方法,建立一个基于支持向量机(support vector machine,SVM)的安达曼海南部内孤立波传播...     

基于最优特征集的HY-1C卫星海岸带成像仪影像海冰分类方法研究

基于海洋一号C（HY-1C）卫星海岸带成像仪（CZI）遥感影像,提出了一种基于最优特征集的支持向量机海冰分类方法。分别提取CZI影像的光谱特征和纹理特征,采用基于距离可分性的判据进行特征选择,得到最优特征集,以最优特征集作为支持向量机分类器输入,分别对3期辽东湾海域CZI影像开展海冰分类实验和结果分析。结果表明:本文方法得到的海冰分类结果精度优于仅利用光谱特征或纹理特征的海冰分类精度;基于本文方法的3期影像的海冰分类精度均较高,2020年12月19日、2021年1月10日与2021年1月16日的海冰分类总体精度分别为93.67%、91.75%、84.89%,均在80%以上;利用海冰分类结果图估算海冰面积,发现3期辽东湾海冰面积依次增大,最大约为11 998.98 km2。     

小样本机器学习算法的特性分析与应用

基于经典统计学的机器学习算法,在解决小样本学习问题时表现得不能令人满意。在总结分析小样本机器学习算法特点的基础上,以支持向量机(SVM)学习算法为例,定量分析了影响其泛化性能、学习性能的几个因素,实验结果与理论分析结论取得了良好的一致性;SVM用于解决KTH-TIPS纹理图像分类问题,取得了很好的实验结果。     

基于深度置信网络(DBN)的赤潮高光谱遥感提取研究

赤潮是严重的海洋灾害,有效监测赤潮对于保护海洋生态环境具有重要意义。高光谱遥感具有光谱分辨率高、图谱合一等优势,适合于海洋赤潮监测。深度学习是机器学习领域的前沿,为高光谱遥感分类提供了新的思路。深度置信网络(Deep Belief Network,DBN)兼具监督分类与非监督分类的特点,通过构建DBN模型,将DBN应用于赤潮灾害遥感监测中,应用渤海机载高光谱遥感数据开展赤潮分类,以达到提取高光谱图像中赤潮水体范围的目的。通过设置对照实验,对比经典的SVM监督分类方法与ISODATA非监督分类方法,发现DBN模型在相同实验条件下具有更高的分类精度,赤潮遥感提取精度提高了3%～11%。     

一种联合光谱和纹理特征的滩涂提取技术

提出一种联合光谱和纹理特征的支持向量机分类算法,先通过计算灰度共生矩阵得到纹理影像,然后将纹理波段与光谱波段进行叠加形成一幅多波段影像,再使用支持向量机分类算法对该影像进行分类,从而得到最终的滩涂提取结果。实验结果表明,该方法对于滩涂及周边地物具有较好的分类效果。     

基于DeepLabv3架构的高分辨率遥感图像分类

针对目前使用机器学习解决高分辨率遥感图像分类主要存在下采样导致的细节信息丢失问题,提出了一种基于DeepLabv3架构的小波域DeepLabv3-MRF(Markov random field,MRF)算法。选择当前较为普遍的DeepLabv3架构分类算法,能够获得更为精确的分类结果;采用小波域DeepLabv3-MRF算法,还能够获得更为清晰的边缘细节信息。选取南方某地区高分辨率无人机遥感图像进行分类实验,通过小波变换的方向性、非冗余性以及MRF变换像素空间的交互性这三个方面,将分类结果与原始DeepLabv3架构分类结果对比分析。结果表明,所提出的分类方法精度明显高于原始DeepLabv3架构分类算法的精度,总体精度可提升3%左右,并且可以充分表达高分辨率遥感图像细节信息。     

基于SVM的甲壳动物线粒体基因分析方法

甲壳动物线粒体基因组蕴涵了物种进化历程中重要的遗传信息,如何有效地利用这些保留在基因组中的基因序列和基因顺序信息,是甲壳动物线粒体基因组研究的一个重点方向。为了进一步探讨甲壳动物稳定、可靠的系统发育关系,本文利用支持向量机的分类功能实现了甲壳动物线粒体基因组基因区与基因间区、编码区与非编码区的准确分类和预测,同时为了提高分类学习机的泛化能力,使用了交叉验证方法和粒子群算法优化选取支持向量机相关训练参数。通过MATLAB仿真分析的方法,对10种甲壳动物线粒体基因组序列的基因区和基因间区进行分类,以及对5种甲壳动物进行线粒体基因组序列中编码区和非编码区的分类,获得了较好的分类准确率。仿真结果表明本文方法是可行的和有效的,能够出色地应用于甲壳动物线粒体基因组序列的研究分析。     

融合地理空间认知的珊瑚礁地貌单元高分遥感分类方法

近年来,在人为活动和自然因素的影响下,全球珊瑚礁面临着大规模退化问题,开展珊瑚礁监测研究对珊瑚礁生态系统评估、修复和保护工作具有重要作用。本文以西沙群岛北礁和华光礁为研究区,应用2015年高分二号（GF-2）和WorldView-2高空间分辨率卫星影像和现场调查数据,基于不同珊瑚礁地貌单元的空间位置特征,提出了融合地理空间认知（Geo-Spatial Cognition,GSC）的珊瑚礁地貌单元高分遥感分类方法。研究结果表明：针对因空间位置不同和底质组成高度近似导致珊瑚礁地貌单元漏分和错分的问题,本文提出的方法更能有效获取精准的珊瑚礁地貌单元信息。其中,融合地理空间认知的随机森林（Integrating Geo-Spatial Cognition-Random Forest,GSC-RF）方法展现出了最优的分类表现,在北礁和华光礁珊瑚礁地貌单元分类中总体精度分别为98.06%和91.93%,Kappa系数分别为0.98和0.91。相比于仅使用光谱信息的随机森林（Random Forest,RF）、多元逻辑回归（Multinomial Logistic Regression,MLR）和支...     

Second-order moment advection scheme applied to Arctic Ocean simulation

We apply the second-order moment (SOM) advection scheme of (Prather, M.J. 1986. Numerical advection by conservation of second-order moments. J. Geophys. Res. 91, 6671–6681.) to the simulation of the large-scale circulation of the Arctic Ocean with a coupled ocean–sea-ice model. Compared to three other advection schemes commonly employed in ocean simulations (centred differences, flux corrected transport, and multidimensional positive definite advection transport), the SOM method helps preserve the vertical structure of Arctic water masses. The depth, thickness and hydrographic properties of the Arctic Surface Water and the Arctic Atlantic Layer are better represented with SOM than with any of the other three advection algorithms. We also present a convenient method for calculating the implicit numerical diffusivity of upstream based schemes, such as the SOM method, and discuss three approaches for improving the monotonicity properties of the SOM algorithm.     

海底声学底质分类的ELM-AdaBoost方法

基于自适应增强算法（AdaBoost）结合极限学习机（ELM）,通过迭代、调整、优化ELM分类器之间的权值,从而构建了具有强鲁棒性、高精度的ELM-AdaBoost强分类器,增强了现有的ELM分类器的稳定性。以珠江口海区侧扫声呐图像为实验数据,对礁石、砂、泥3类典型底质进行分类识别,该方法的平均分类精度超过90%,优于单一ELM分类器的平均分类精度85.95%,也优于LVQ、BP等传统分类器,且在分类所耗时间上也远少于传统分类器。实验结果表明,本文构建的ELM-AdaBoost方法可有效应用于海底声学底质分类,可满足实时底质分类的需求。