共查询到17条相似文献,搜索用时 93 毫秒
1.
2.
3.
利用最小二乘向量机(LS-SVM)算法构造海底趋势面的过程中,由于算法解缺乏稀疏性,使得异常测深训练样本对最终构造的函数模型也产生影响。为了解决该问题,在对留一样本交叉检核法研究的基础上提出了LS-SVM稀疏算法,由于留一样本交叉检核法求解的残差序列可以有效地表示函数预测值偏离实际水深的程度,因此利用该原则重新修剪后的样本数据不仅使算法具有稀疏特性,而且构造的函数模型更合理。为了检验算法的有效性,选取实测的多波束测深数据进行验证,计算结果表明留一样本交叉检核法能够合理地筛选出对函数模型构造贡献程度大的测深训练样本,使得构造的函数模型更合理。 相似文献
4.
5.
在验证趋势面滤波是最小二乘支持向量机算法取特定参数解的基础上,利用 LS-SVM 所构造的海底趋势面对测深异常值进行剔除。 为了克服 LS-SVM 解非稀疏性的缺点,同时抑制偏差较大的训练样本对海底趋势面构造的影响,提出并实现了一种基于局部样本中心距离的训练样本优化方法。 为了检验该算法的有效性,选取实测的多波束测深数据进行验证,结果表明在训练样本优化的基础上,通过调整 LS-SVM 的参数可以得到更为合理的海底趋势面,测深异常值地剔除也更为有效。 相似文献
6.
溢油事件的发生会给海洋环境的保护和经济发展带来巨大的影响。运用现代化的监测手段和技术进行监测,及时发现溢油现象和违规行为,保护海洋环境是非常重要的。合成孔径雷达(SAR)技术是溢油检测的有效工具,在SAR图像中溢油表现为黑色的区域,但是黑色区域也可能会由其他的因素引起。本文提出了一种基于二维经验模态分解(BEMD)的方法来识别溢油和疑似溢油。首先通过BEMD方法将感兴趣的区域分解为局部窄带的各分量—内蕴模函数(BIMF)之和,并对分解后得到的各分量IMF进行Hilbert变换,通过Hibert谱分析得到64维的特征空间,然后使用Relief方法得到5个特征向量,最后利用马氏距离分类器进行分类。通过实验结果表明,该方法能够有效、准确地检测出溢油,准确率超过90%。 相似文献
7.
为提高非线性和非平稳海水温度时间序列的预测能力,提出了一种基于经验模态分解(Empirical Mode Decomposition.简称EMD)的BP神经网络预测方法.该方法首先对原始序列进行经验模态分解,将其分解为多个平稳性得到很大改善的本征模态函数(Intrinsic Mode Function,简称IMF)之和,然后时每个本征模态函数进行预测,最后再根据EMD方法的完备性把预测结果相加得出原始序列的预测结果.预测试验结果表明.基于EMD的BP神经网络预测的精度比单纯用BP神经网络预测有很大提高. 相似文献
8.
异常事件对EMD方法的影响及其解决方法研究 总被引:4,自引:1,他引:4
赵进平 《中国海洋大学学报(自然科学版)》2001,31(6):805-814
作者指出异常事件在数据中形成局部的高频信号 ,运用经验模态分解 (EMD)方法分析这种存在异常事件干扰的数据 ,就会产生本征模函数 (IMF)的频率混叠现象 ,而造成物理过程的重叠 ,使得难以用时间过程曲线表现特定的物理过程。这一问题是 EMD方法中尚未妥善解决的问题。为解决这一问题 ,作者利用干扰信号极值及其两边的极大与极小值位置与原始数据有明显对应关系的特征 ,将相关 IMF中的异常信息直接滤除 ,再用 Spline插值方法弥补滤除时段的数据 ,得到重新拟合的该 IMF数据。采用这种方法可以提取出异常信号 ,提取的精度与异常信号的时段长度有关。而且 ,拟合结果消除了异常干扰 ,可以将该 IMF与其余 IMF一起叠加成没有异常干扰的数据。将滤除了异常干扰的数据再次进行 EMD分解 ,可以得到新的 IMF系列 ,而它与不加校正的分解结果有相当大的差别 ,可靠地反映了真实物理过程。结果表明 ,只有在有效滤除异常干扰的情况下才能获得可靠的 IMF系列 ,并准确地描述各种尺度的现象 ;消除了异常干扰的 IMF可以任意单独或组合使用 ,表现各种时间尺度的变化与过程 ;所讨论的方法只适合异常时段较小的情形。对于异常时段接近或大于正常变化周期的干扰还需要探讨其他方法 相似文献
9.
10.
《海洋技术学报》2021,40(4)
电阻率测量是海底沉积物工程地质勘察的主要原位观测方法之一,作为一种间接测量方法,需要建立沉积物物性参数与沉积物电阻率的回归模型。为提高建模精度,本文提出了一种基于鲸鱼算法优化的最小二乘支持向量机(Whale Optimization Algorithm-Least Squares Support Vector Machine,WOA-LSSVM)的海底沉积物物性参数与电阻率回归建模方法。该方法建立了海底沉积物电阻率与沉积物4种基本物性参数(含水率、密度、孔隙比、塑性指数)的单输入、单输出最小二乘支持向量机(Least Squares Support Vector Machine,LSSVM)回归模型,利用WOA算法对LSSVM参数进行寻优取值。对比研究了WOA算法、遗传算法(Genetic Algorithm,GA)、粒子群(Particle Swarm Optimization,PSO)算法优化的LSSVM建模结果,结果表明,基于WOA-LSSVM建立的海底沉积物物性参数与电阻率的回归模型具有更好的预测效果,均方根误差降低1.1%~14.9%,平均绝对百分比误差降低0.4%~19.9%。 相似文献
11.
针对传统趋势面滤波方法中多项式拟合曲面系数向量的求取和作为阈值的均方根误差的求取都受到异常数据的影响,使该方法在异常测深数据较多的情况下滤波效果不佳的问题,提出了一种中值滤波加权修正的改进方法。在构造趋势面之前,对水深数据进行加权修正,以前后两次修正后数据的拟合优度的变化量作为是否进行下一步水深修正的依据,利用最终修正后的水深数据求取多项式拟合曲面系数向量和均方根误差,大幅降低了异常数据的影响,具有很强的抗差性。经仿真模拟数据和多波束实测数据滤波试验,该方法在异常数据较多的情况下依然良好,能够保持良好的滤波效果,明显优于传统趋势面滤波;同时,该方法能够保持较高的运算效率,适用于海量多波束测深数据的自动滤波。 相似文献
12.
13.
Some errors and noises are often present in multibeam swath bathymetric data. Echo detection error (EDE) is one of the main errors. It causes the depth error to become bigger in outer beams and looks like sound refraction. But depth errors due to EDEs have a trumpet-shaped appearance, instead of a curved appearance that is caused by the sound refraction errors. EDEs, including systematic acoustic signal detection errors and internal noises, cannot be removed during the correction of sound refraction. It causes depth inconsistencies between adjacent swaths and degrades precision of outer beams. Sometimes, the bathymetric errors caused by EDEs do not even meet the requirements of IHO (International Hydrographic Organization). Therefore, a post-processing method is presented to minimize the EDEs by filtering outliers and compressing outer beams of multibeam bathymetric data. The outliers caused by internal noises are removed by an automatic filter algorithm first. Then the outer beams are compressed to reduce systematic acoustic signal detection errors according to their depths, the calculated depth line and standard deviations (SDs). The automatic filter process is important for calculating the depth line. The selection of inner beams to calculate the average SD of beam depths is crucial to achieving compressing goals. The quality of final bathymetric data in outer beams can be improved by these steps. The method is verified by a field test. 相似文献
14.
The frequency attenuation gradient method can provide important information for hydrocarbon detection. In this paper, a method using Complete Ensemble Empirical Mode Decomposition (CEEMD), Hilbert transform and the least-squares curve-fitting is proposed for seismic attenuation estimation as an effective frequency attenuation gradient estimation approach. We first use CEEMD to obtain the different Intrinsic Mode Functions (IMFs), which have a narrow band and can enhance the physical meaning of instantaneous attributes trace by trace. The time-frequency spectrum, which is computed using a Hilbert transform of each IMF, is represented as a spectrum with a single-peak that has narrow side lobes, which is conducive to frequency attenuation gradient estimation. Second, for each time sample, the frequency-amplitude spectrum of each IMF trace is extracted from the time-frequency spectrum to conduct the attenuation gradient computation. Then, the logarithm operation is performed for each IMF trace. Due to the very narrow bands of some IMFs in some seismic traces, a variable frequency window is adopted along the IMF trace according to the local data characteristics. Finally, the attenuation gradient for each IMF in a seismic trace can be computed using least-squares fitting. A different IMF reflects a seismic trace with a different spatiotemporal scale and can highlight different geologic and stratigraphic information. The correlation weighted average operation is used to highlight some useful details in seismic trace and obtains the attenuation gradient for each seismic trace. Field data examples demonstrate our method and its effectiveness. The proposed method can stably estimate the frequency attenuation gradient. 相似文献
15.
Hsien-Chueh Peter YANG Tsung-Hao CHEN Cheng-Wu CHEN Chen-Yuan CHEN Chun-Te LIU 《中国海洋工程》2008,22(3):421-429
Logit regression analysis is widely applied in scientific studies and laboratory experiments, where skewed observations on a data set are often encountered. A number of problems with this method, for example, oudiers and influential observations, can cause overdispersion when a model is fitted. In this study a systematic statistical approach, including the plotting of several indices is used to diagnose the lack-of-fit of a logistic regression model. The outliers and influential observations on data from laboratory experiments are then detected. Specifically we take account of the interaction of an internal sohtary wave (ISW) with an obstacle, i.e., an underwater ridge, and also analyze the effects of the ridge height, the lower layer water depth, and the potential energy on the amplitude-based transmission rate of the ISW. As concluded, the goodness-of-fit of the revised logit regression model is better than that of the model without this approach. 相似文献
16.
17.
Multibeam echosounders have commonly been employed for a wide range of applications including offshore survey, navigation, hydrogeology, and oceanography. Because the tremendous volume of the bathymetric data is demanding for some purposes and requires significant storage space, the data reduction plays a prominent role in practice. Additionally, the multibeam soundings are inevitably contaminated with sporadic outliers, and as such, the data cleaning can be challenging especially in shallow waters. We present a speedily robust method for reliably reducing the volume of the bathymetric data within grid cells. In this respect, robust M-estimators are recursively applied to the data in a patch-wise manner to alleviate the undesirable effects of the outlying observations. Accordingly, the reduced bathymetry is automatically made unaffected by the possible outliers once their equivalent weights have been downweighted. The performance of the presented method has been demonstrated by synthetic datasets and an experimental dataset collected by an ATLAS FS 20/100 kHz shallow-water multibeam echosounder in the offshore waters of Kish wharf. The reliability, efficiency, and capability of the proposed method have been verified, which makes it quite possible to meet the IHO requirements for special-order seafloor mapping. 相似文献