北极冰下水声通信中改进的最小均方/四次方直接自适应均衡器

本文提出了一种适用于北极冰下水声通信的最小均方/四次方直接自适应均衡器（LMS/F-DAE）。它能处理基带复信号，与LMS相比，具有更好的均衡效果。考虑到均衡器的稀疏特性，在其代价函数中加入自适应范数（AN）作为约束。它能根据均衡器系数的大小自适应变化：对于小系数，此约束项存在以加快收敛速度；对于大系数，此约束项不存在以减小均衡误差。利用第九次中国北极科学考察得到的实验数据验证AN-LMS/F-DAE的性能。结果表明，与传统的LMS/F-DAE相比，AN-LMS/F-DAE能提升均衡器的稀疏性且均衡性能更优。     

Phase-coherent digital communications for underwater acousticchannels

High-speed phase coherent communications in the ocean channel are made difficult by the combined effects of large Doppler fluctuations and extended, time-varying multipath. In order to account for these effects, we consider a receiver which performs optimal phase synchronization and channel equalization jointly. Since the intersymbol interference in some underwater acoustic channels spans several tens of symbol intervals, making the optimal maximum-likelihood receiver unacceptably complex, we use a suboptimal, but low complexity, decision feedback equalizer. The mean squared error multiparameter optimization results in an adaptive algorithm which is a combination of recursive least squares and second-order digital phase and delay-locked loops. The use of a fractionally spaced equalizer eliminates the need for explicit symbol delay tracking. The proposed algorithm is applied to experimental data from three types of underwater acoustic channels: long-range deep water, long-range shallow water, and short-range shallow water channels. The modulation techniques used are 4- and 8-PSK. The results indicate the feasibility of achieving power-efficient communications in these channels and demonstrate the ability to coherently combine multiple arrivals, thus exploiting the diversity inherent in multipath propagation     

ROC evaluation of adaptive beamforming in a simulated shallow waterenvironment

An optimal evaluation of adaptive beamforming techniques in a reverberation-limited shallow water environment is presented. A comprehensive simulation, using the sonar simulation toolset (SST) software in conjunction with the generic sonar model (GSRT) software, is used to create realistic beam data complete with target, noise, and reverberation. Adaptive beamforming techniques from the recursive least squares (RLS) family are applied to enhance detection performance via interference rejection. Two techniques are considered: linearly constrained beamforming using the minimum variance distortionless response (MVOR) beamformer and constrained adaptive noise cancelling (ANC) using the joint process least squares lattice (JPLSL) algorithm. Target detection trials, summarized in the form of receiver operator characteristics (ROC), are used to evaluate performance of the two adaptive beamformers. Results demonstrate mixed performance in reverberation-limited shallow water environments     

光谱预处理方法对硝酸盐测量模型影响实验研究及软件设计

基于Savitzky-Golay(SG)卷积平滑、正态变量变换(SNV)、一阶导数法、附加散射校正(MSC)和小波去噪(WDS)等信号处理方法,对水中硝酸盐偏最小二乘(PLS)测量模型的影响进行实验研究,采用评估均方差(RMSEE)、预测标准差(RMSEP)、相关系数(R)、预测值与样本浓度值回归关系显著性F检验对预处理效果进行考察。实验结果表明,经SG卷积平滑预处理的PLS模型预测准确性优于其他处理方法。同时编写了该5种光谱预处理方法软件,实现了光谱数据采集与预处理、谱图绘制和光谱保存等功能。     

典型高光谱图像端元提取算法在黄河口湿地应用评价研究

端元提取是混合像元分解的基础,也是高光谱遥感的研究热点。对于特定区域的高光谱图像应该使用哪种端元提取算法,需要对各种端元提取算法进行客观地评价。作者针对黄河口湿地CHRIS高光谱图像,使用了重建图像与原图像的均方根误差、有效端元数量两个指数对PPI、N-FINDR、VCA、OSP、IEA和SISAL六种典型的端元提取算法进行了评价。结果表明,SISAL算法重建误差最小,仅有其他算法误差的10%~28%;OSP算法识别了具有物理意义的6种有效端元,多于其他算法识别的地物类型,而SISAL算法识别的端元缺乏物理意义。     

人工神经网络反演珠江口海域叶绿素浓度

利用2003年1月、2004年1月在珠江口海域的叶绿素浓度和辐射同步实测资料,建立了反演珠江口海域叶绿素浓度的人工神经网络模型。应用该模型由SeaWiFS资料获取珠江口海域叶绿素浓度分布图,并与SeaBAM推荐的OC2和OC4这2种统计算法的反演结果进行比较,结果表明人工神经网络模型的反演效果明显优于统计算法。人工神经网络模型的均方根差是0．2899,可决系数是0．8848;而统计算法的均方根差大于0．5,可决系数小于0．6。     

人工神经网络在南海近海面气温反演中的应用研究

基于人工神经网络方法,利用海面水温、海面风速以及海面气压反演南海近海面气温,采用的基础数据集是国际综合海洋-大气数据集(International Comprehensive Ocean-Atmosphere Data Set,2.4 Release,ICOADS2.4)1981—2008年的观测资料,其中1981—2000年的观测资料用来建立模型,2001—2008年的观测资料用来进行模型检验。采用的人工神经网络方法是引入动量因子并采用批处理梯度下降法的BP(Back propagation)算法。试验结果表明,基于人工神经网络建立的近海面气温反演方法明显优于多元线性回归方法,尤其是在春季和冬季,海面水温、海面风速以及海面气压与近海面气温之间存在较强的非线性关系,人工神经网络的优势更加明显。总体而言,人工神经网络在各月的反演效果较均衡,均方根误差介于1.5—1.8℃之间,平均绝对误差为1.1—1.3℃。     

Measured and remotely sensed estimates of primary production in the Atlantic Ocean from 1998 to 2005

Primary production (PP) was determined using ¹⁴C uptake at 117 stations in the Atlantic Ocean to validate three PP satellite algorithms of varying complexity. An empirical satellite algorithm based on log chlorophyll-a had the highest bias and root-mean square error compared with measured ¹⁴C PP and tended to under-estimate PP. The vertical generalised production model improved PP estimates and was the most accurate algorithm in the Eastern Tropical Atlantic (ETRA) and Western Tropical Atlantic (WTRA), but tended to over-estimate PP in eutrophic provinces. A photosynthesis-light wavelength-resolved model was the most accurate over the Atlantic basin, having the lowest mean log-difference error, root-mean square error and bias, and exhibited a superior performance in six out of the nine ecological provinces surveyed. Using this algorithm and mean monthly SeaWiFS fields, a PP time series was generated for the Atlantic Ocean from 1998 to 2005 which was compared with Advanced Very High Resolution Radiometer (AVHRR) sea-surface temperature (SST) data. There was a significant negative correlation between SST and PP in the North Atlantic Subtropical Gyre Province (NAST), North Atlantic Tropical Gyre (NATR), and WTRA suggesting that recent warming trends in these provinces are coupled with a decrease in phytoplankton production.     

Influence of the type of sea waves on the backscattered radar cross section at medium incidence angles

We consider the influence of the sea surface state on the backscattered radar cross section and the accuracy of the wind speed retrieval from the scatterometer data. We used a joint set of radars and buoys to determine the type of sea waves. Three types of sea waves were distinguished: developing wind waves, fully developed wind waves, and mixed sea. It is shown that the retrieval error of the near surface wind speed using a one-parameter algorithm is minimal in the case of fully developed wind waves. We compared these data with the results of radio-altimeter data analysis and showed that in both cases underestimation of the retrieval wind speed exists for developing wind waves and overestimation occurs for mixed sea. A variety of swell parameters (length of the dominating wave, swell height, swell age) significantly influence the backscattered radar cross section, leading to a growth in the mean square error of the retrieved wind speed during vertical sounding (radio-altimeter data), and only slightly influence the mean square error of the scatterometer data (medium incidence angles). It is necessary to include the information about the parameters of sea waves in the algorithms and take into account the regional wave properties to increase the accuracy of wind speed retrieval.     

总体最小二乘法在浮游植物色素吸收系数模型估算中的应用

采用常规最小二乘法(LS)和总体最小二乘法(TLS)对浮游植物色素吸收系数与叶绿素a浓度之间的关系进行了研究。利用2003-2012年6个航次243组实测数据,建立了LS和TLS模型,应用模型估算典型波段的色素吸收系数值,并对两种模型进行了验证。采用平均绝对百分误差和均方根误差两种方法进行模型效果评估,结果显示,两种评估方法均能很好地说明,在400～700 nm光谱范围内,TLS拟合结果优于LS拟合结果。总体最小二乘法在综合考虑了参数误差和测量误差的情况下,拟合结果精度更高,更符合实际情况,估算结果的准确性要优于常规最小二乘法。     

一种基于伽利略搜救系统的联合定位算法

分析了伽利略搜救系统的地面站定位模型。提出了一种适用于该系统的，在存在多余观测的情况下，基于信号的时间延迟和多普勒频率测量值加权后，共同参与平差计算的联合定位算法模型及其误差分布模型。仿真结果表明，与单纯使用一种观测值的定位算法相比，该算法可以提高定位精度，是一种适合伽利略搜救地面站使用的定位算法。     

Spatially and temporally varying adaptive covariance inflation for ensemble filters

Ensemble filters are used in many data assimilation applications in geophysics. Basic implementations of ensemble filters are trivial but are susceptible to errors from many sources. Model error, sampling error and fundamental inconsistencies between the filter assumptions and reality combine to produce assimilations that are suboptimal or suffer from filter divergence. Several auxiliary algorithms have been developed to help filters tolerate these errors. For instance, covariance inflation combats the tendency of ensembles to have insufficient variance by increasing the variance during the assimilation. The amount of inflation is usually determined by trial and error. It is possible, however, to design Bayesian algorithms that determine the inflation adaptively. A spatially and temporally varying adaptive inflation algorithm is described. A normally distributed inflation random variable is associated with each element of the model state vector. Adaptive inflation is demonstrated in two low-order model experiments. In the first, the dominant error source is small ensemble sampling error. In the second, the model error is dominant. The adaptive inflation assimilations have better mean and variance estimates than other inflation methods.     

The comparison of altimeter retrieval algorithms of the wind speed and the wave period

With the launch of altimeter,much effort has been made to develop algorithms on the wind speed and the wave period.By using a large data set of collocated altimeter and buoy measurements,the typical wind speed and wave period algorithms are validated.Based on theoretical argument and the concept of wave age,a semi-empirical algorithm for the wave period is also proposed,which has the wave-period dimension,and explicitly demonstrates the relationships between the wave period and the other variables.It is found that Ku and C band data should be applied simultaneously in order to improve either wind speed or wave period algorithms.The dual-band algorithms proposed by Chen et al.(2002) for the wind speed and Quilfen et al.(2004) for the wave period perform best in terms of a root mean square error in the practical applications.     

Improvements in acoustic sediment concentration profiling using anLMS compensation algorithm

In sediment concentration profiling, sound attenuation becomes important for high-frequency acoustic backscattering systems (ABSS) when sediment concentration is high and/or the range of remote sensing is long. In the present study, a test of a least mean square (LMS) compensation algorithm was performed in order to determine the advantages over the more frequently used iteration algorithm. Numerical simulation and laboratory experiments in a mixing tank as well as an open channel suggest that the new algorithm is much more robust to white Gaussian noise and fixed-echo noise, and errors do not accumulate along the profiling path. Good linearity of this instrument in concentrations up to 20 kg/m^-3 is found by comparison with a suction-sampler. In the mixing tank, backscattering intensity profiles in concentrations up to 120 kg/m^-3 vary little along a propagation path of 40 cm and fit well to the expected homogeneous concentration. Furthermore, by using the oscillation errors in the LMS solution, a procedure has been developed to compensate for the errors in the instrument constants evaluated from the attenuation between the transducer and the first measuring volume. The feasibility of the proposed instrument-constant compensation is verified by experiments. This allows instantaneous correction of the influence of sediment deposition on the transducer surface which is important when the system is applied to coastal investigation     

XGBoost算法在多光谱遥感浅海水深反演中的应用

在多光谱遥感浅海水深反演过程中,考虑到水体和底质影响,水深值和海水表面辐射亮度之间的线性关系不成立。本文以甘泉岛南部0～25m范围的沙质区域为研究区域,利用GeoEye-1多光谱遥感影像和多波束实测水深数据构建XGBoost非线性水深反演模型,研究了XGBoost算法用于水深反演的性能。以决定系数(R~2),均方误差(MSE)和平均绝对误差(MAE)作为评价指标,并与3种传统线性回归模型进行了对比分析。结果表明, XGBoost非线性水深反演模型的R~2、MSE和MAE分别为0.991、0.33m和0.44m,拟合程度最好,精度优于线性回归模型。为进一步探究各模型在不同水深的反演精度,将水深范围分成3段(0～8 m, 8～15 m, 15～25 m)分别进行精度验证和误差分析。结果表明, XGBoost模型在各分段的精度均优于线性回归模型, MSE依次为0.56 m, 0.14 m和0.43 m。可见,在单一底质区域下XGBoost模型的水深反演精度更高,且反演效果更稳定。     

基于WOA-BP算法的海底管道腐蚀速率预测

管道腐蚀速率的影响因素众多且各因素协同作用,构成异常复杂的腐蚀体系,很难对其进行准确预测。针对单一BP模型由于初始权值和阈值的选取不当容易陷入局部最优等问题,引入WOA算法优化BP神经网络对海底管道腐蚀速率进行预测,并与GA和PSO算法优化BP预测模型进行对比,验证WOA-BP模型的预测效果和可行性。结果表明：WOA-BP模型的平均绝对百分误差和均方根误差分别为3.689%和0.1537,远低于单一BP、PSO-BP、GA-BP模型,具有较高的预测精度和稳定性,可以为海底管道内腐蚀防护和油气管道流动保障提供决策支持。     

一种多层神经网络的学习算法

提出一种基于序贯预测误差方法（ＳＰＥ）的多层神经网络（ＭＮＮ）的学习算法。经模拟计算，它比传统的基于最陡下降方法的误差反传（ＳＤＢＥＰ）算法具有更好的收敛性能。并对这两种算法进行了模拟计算的比较．     

Retrieving the antarctic sea-ice concentration based on AMSR-E 89 GHz data

Sea-ice concentration is a key item in global climate change research.Recent progress in remotely sensed sea-ice concentration product has been stimulated by the use of a new sensor,advanced microwave scanning radiometer for EOS(AMSR-E),which offers a spatial resolution of 6 km×4 km at 89GHz.A new inversion algorithm named LASI(linear ASI) using AMSR-E 89GHz data was proposed and applied in the antarctic sea areas.And then comparisons between the LASI ice concentration products and those retrieved by the other two standard algorithms,ASI(arctic radiation and turbulence interaction study sea-ice algorithm) and bootstrap,were made.Both the spatial and temporal variability patterns of ice concentration differences,LASI minus ASI and LASI minus bootstrap,were investigated.Comparative data suggest a high result consistency,especially between LASI and ASI.On the other hand,in order to estimate the LASI ice concentration errors introduced by the tie-points uncertainties,a sensitivity analysis was carried out.Additionally an LASI algorithm error estimation based on the field measurements was also completed.The errors suggest that the moderate to high ice concentration areas(>70%) are less affected(never exceeding 10%) than those in the low ice concentration.LASI and ASI consume 75 and 112 s respectively when processing the same AMSR-E time series thourghout the year 2010.To conclude,by using the LASI algorithm,not only the seaice concentration can be retrieved with at least an equal quality as that of the two extensively demonstrated operational algorithms,ASI and bootstrap,but also in a more efficient way than ASI.     

Adaptive beamforming: spatial filter designed blocking matrix

Controlling the resolution in adaptive beamformers is often crucial. A simple method that works for both narrow-band and broad-band arrays is presented. This method is based on the normalized leaky LMS algorithm in conjunction with a generalized sidelobe canceller (GSC) structure, where the GSC is designed using a spatial filtering approach. In essence, the suppression of the spatial filters and the implicit noise of the leaky LMS algorithm together determine the adaptive beamformer. Analytical expressions are given for the Wiener filters and the output spectrum versus frequency and point source location. These expressions are employed in the design specification of the spatial filters and to obtain conditions for a controlled quiescent beamformer response. Simulation results are presented to illustrate the behavior of the array     

Blind equalization based on RLS algorithm using adaptive forgetting factor for underwater acoustic channel

Blind equalization based on adaptive forgetting factor, recursive least squares （RLS） with constant modulus algorithm （CMA）, is investigated. The cost function of CMA is simplified to meet the second norm form to ensure the stability of RLS-CMA, and thus an improved RLS-CMA （RLS-SCMA） is established. To further improve its performance, a new adaptive forgetting factor RLS-SCMA （ARLS-SCMA） is proposed. In ARLS-SCMA, the forgetting factor varies with the output error of the blind equalizer during the iterative process, which leads to a faster convergence rate and a smaller steady-state error. The simulation results prove the effectiveness under the condition of the underwater acoustic channel.