共查询到20条相似文献，搜索用时 125 毫秒
1.
学习向量量化(LVQ)在地震和爆破识别中的应用 总被引：2，自引：1，他引：2
介绍了竞争神经网络和学习向量量化（LVQ）算法。此算法应用于对北京及周围地区地震和爆破的识别中，在对38个事件的应用中，得到的结果是，误识为3个，结果较好，说明在识别中是有效的。 相似文献
2.
3.
遗传优化神经网络方法在桥梁震害预测中的应用 总被引：5，自引：1，他引：4
本文将遗传算法与神经网络相结合,从而建立了一种高效的、实用的桥梁震害预测方法。根据遗传算法具有局部寻优的特点,为避免BP神经网络陷入局部极小值,本文将二者结合起来形成GA-BP混合算法,以GA优化神经网络的初始权值和阈值,对网络进行训练。在大量收集梁式桥震害资料的基础上,将此算法引入桥梁的震害预测中,并与传统的单独BP神经网络相比较,结果表明该方法能够有效、准确地对桥梁结构进行震害预测。 相似文献
4.
目的：利用深度学习技术，全自动标注病变的计算机断层扫描（CT）数据，开发准确快速区分新型冠状病毒感染（COVID-19）和其他社区获得性肺炎的人工智能模型。方法：回顾性分析248例COVID-19患者及347例其他肺炎患者的资料，进行COVID-19与其他肺炎分类；在人工智能肺分割提取后将异常的CT图像特征降维，输入几种经典强化机器学习模型、三维卷积神经网络（3D CNN）和注意力多示例学习（Attention-MIL）深层神经网络架构中，模型诊断性能利用受试者工作特性（ROC）曲线、精确召回率（PR）曲线、曲线下面积（AUC）、敏感性、特异性、准确性指标进行评价。结果：在经典机器学习模型中K邻近算法（KNN）具有较好的效果，在外部测试集上的AUC值和平均精度（AP）值分别为0.79和0.89，平衡F分数（F1）值为0.76，准确率为0.75，敏感性为0.76，精确率为0.77；经典的3D CNN在外部测试集上效果良好，AUC值和AP值分别为0.64和0.82,F1值为0.71，准确率为0.78，敏感性为0.66，精确率为0.62;Attention-MIL模型在外部测试集上表现出更好... 相似文献
5.
将遗传算法(GA)和反向传播算法(BP算法)相结合成为GA-BP算法，以此建立了遗传BP神经网络．并将以BP算法为基础的BP神经网络及以GA-BP算法为基础的遗传BP神经网络用于对地震和爆破的识别中．得到的结果表明:遗传BP网络比BP网络对事件的识别能力略好些． 相似文献
6.
致密砂岩流体识别难度大,智能算法能够较好地建立其流体识别模型.相较于单一智能算法,分类委员会机器通过联合多个专家(智能算法)有助于提升智能模型整体性能.而针对分类委员会机器中单个专家性能难以提升的问题,添加门网络构建动态分类委员会机器是一种更有效的模块化学习方式.本研究首先采用门网络将输入数据划分为多个子数据集,然后联合决策树、概率神经网络、贝叶斯分类、BP神经网络、最近邻算法分别训练子数据集得到多个子模型,最后利用组合器最优化子模型组合得到最佳的流体识别模型.针对塔里木盆地库车坳陷大北、克深、博孜地区致密砂岩地层测井数据和测试数据,采用平均影响值法优选敏感测井系列作为输入,构建了动态的测井流体识别模型,其训练、验证准确率分别为96.29%和91.39%.利用此模型以BZ9井为例进行流体类型判别,预测结果与测试结果一致.该方法将无监督与有监督学习相结合,引入门网络提高了数据集利用效率,避免了数据集分布不均衡对模型构建的影响;采用投票机制集成多种专家,建立了子模型与专家的适应关系,流体识别模型预测精度和泛化能力大大提高. 相似文献
7.
重、磁勘探具有效率高、成本低、工作范围广等优点,已在地球物理勘探中得到了广泛应用.前人大多在不考虑重、磁勘探观测精度的条件下进行了垂向识别能力的研究,但在考虑重、磁观测精度条件下,重力（重力异常、重力张量）与磁力（磁力异常、磁力三分量、磁力张量）对孤立异常的垂向识别能力如何则需要进行深入的理论研究.本文从重、磁场正演理论出发,以球体（点源模型）和无限延伸水平圆柱体（线源模型）为例,考虑给定观测精度条件下,以重力和磁力幅值大小与观测精度的关系来研究垂向识别能力,从而消除了背景场的影响,提高了研究结果的可靠度.通过研究表明,对于孤立异常,重力张量在浅部一定深度内比重力异常的垂向识别能力强,该深度与重力异常和重力张量观测精度的比值成正比;垂直磁化磁力张量在浅部一定深度内比化极磁力异常的垂向识别能力强,该深度与磁力异常与磁力张量观测精度的比值成正比;磁力在浅部一定深度内比重力的垂向识别能力强,该深度与地质体的磁化强度和剩余密度比值、重力观测精度和磁力观测精度比值成正比.通过重力和磁力垂向识别能力的研究将为重、磁勘探的实际应用起到指导作用. 相似文献
8.
王道阔 《CT理论与应用研究》2012,21(1):43-52
本文利用计算机辅助进行在役管线焊故障缝缺陷检测,在缺陷特征提取中提出圆形度、长宽比、填充度、尖部尖锐度、对称度、灰度比以及缺陷的重心坐标相对焊缝中心的位置等7个参数作为缺陷的特征值,可有效地分类识别不同故障缺陷。在缺陷分类的解决方案上,采用具有自组织、自适应的3层前馈式神经网络,运用改进的BP算法,以焊缝缺陷的特征参数作为神经网络的训练样本。本文还通过实验的方法,分析了初始权值、隐含层的神经元数量、动量系数、误差水平及学习速率对网络训练的影响。 相似文献
9.
在地震学研究中地震检测与震相识别是最基础的环节,其拾取速度和精度直接影响其在地震精确定位以及地震层析成像中的应用效率和精度。近年来,机器学习在地震学领域中引起广泛关注。机器学习可以改进传统地震检测和震相识别方法,使它们能达到更加准确,识别率更高的效果。把机器学习方法按照监督学习和无监督学习分类介绍,并对机器学习方法流程进行总结,并对目前在地震检测与震相识别方面应用较为广泛的机器学习方法(卷积神经网络、指纹和相似性阈值、广义相位检测、PhaseNet、模糊聚类)进行综述。结果表明:机器学习在地震事件检测和震相识别将会是主要的手段。数据驱动的机器学习在地震学中的应用和物理模型的联合运用将是未来的发展趋势。 相似文献
10.
低渗透砂岩储层孔隙结构复杂，储层有效性识别及饱和度准确计算难度较大.笔者以东营凹陷南坡沙四段（Es4）低渗透砂岩为研究对象，根据压汞、物性、薄片及核磁等资料，将研究区孔隙结构分为三大类、五小类.在岩样孔隙结构分类基础上，明确了孔隙结构类型与岩电参数之间存在确定的关系，而核磁共振T2谱定量特征参数在一定程度上能够表征孔隙结构类型及其细节信息，通过提取T2谱中T2几何平均值（T2g）、T2均值（T2）、峰度（KG）、可动流体分量（Smf）及区间孔隙分量等孔隙结构参数，建立了基于核磁T2谱特征参数的孔隙结构识别图版，显示核磁T2谱孔隙结构参数对不同类型的储层有较好的识别效果，进而探讨核磁孔隙结构参数和岩电参数之间的关系，结果表明，T2谱峰度值与孔隙胶结指数（m）值相关性较高，进一步确定了岩电参数m的核磁计算公式.最终，将该套方法应用于研究区井筒剖面中，有效地提高了饱和度计算精度，也为东营凹陷南坡低渗透砂岩油藏储量估算与高效开发提供了依据. 相似文献
11.
在海洋可控源电磁法勘探中,接收站常置于海底.在进行海洋电磁场模拟时,由于海水和海底介质存在显著电性差异,这给海底接收点处场值的求取带来困难.本文提出一种新的接收点插值算法,该算法考虑到海底电场法向分量不连续性问题,用法向电流分量进行插值以准确求取海底任意接收点处电磁场值.本文利用交错网格有限差分法实现了二维介质中频率域海洋可控源法(CSEM)正演.对构造走向做傅里叶变换,将三维电磁模拟问题转换为波数域2.5维问题,即三维场源激励下针对二维地电模型的电磁模拟问题.使用交错网格有限差分法,基于一次场/二次场分离方法导出波数域二次电场离散形式,并进一步求得波数域电磁场.采用本文提出的改进的插值算法可求得海底任意接收点处波数域电磁场,采用傅里叶逆变换对波数域电磁场进行积分可得到接收点处空间域电磁场.模型算例表明,与常规的线性插值和严格插值算法相比,本文提出的改进的插值算法具有更高的精度. 相似文献
12.
为基于谱比方法研究海底地震动场地效应,选取日本DONET1台网的20个海底台站2014—2021年记录的1634组地震数据,对其进行筛选和处理后,利用水平与竖向谱比（HVSR）方法考虑不同布设对海底5组节点台站（KMA、KMB、KMC、KMD、KME）谱比特征的影响。研究结果表明:KMA与KME节点台站具有相似的场地特征,KMB与KMD节点台站分散布置在2种场地,KMC节点台站场地与其他节点均不相似,这与长期地质调查结果相似;海底台站谱比曲线呈多峰值现象,其中KMB、KMC、KMD分组台站利用HVSR方法识别的主频变异性较高,KMA、KME分组台站主频较稳定;相同地形条件下,布设方式相同的海底台站谱比曲线随频率分布相似,海底复杂场地条件下,采用装沙沉底方式布置的台站识别场地条件时出现偏差;海底复杂因素对掩埋沉箱方式布设的台站谱比曲线的影响主要集中在频率<5 Hz的低频处;海底复杂因素对未埋入海底台站谱比曲线的影响主要集中在频率为5—10 Hz的高频处。研究结果可为海底地震动场地效应研究提供参考。 相似文献
13.
A method that links acoustic mapping data to underwater video observations of seafloor substrate is described for use in defining fish habitat. Three study areas in the Aleutian Islands were acoustically mapped using sidescan and multibeam sonar. The sidescan sonar data were used to compute average reflectivity (hardness) and seafloor complexity. The multibeam depth data were used to determine local slope, rugosity (seafloor roughness) and relative height. Underwater video was collected from three to four transects in each of the three study areas. The underwater video was used to classify the seafloor into nine observed primary and secondary substrate classes. A statistical relationship between the observed (video) and the remotely sensed (acoustic) seafloor characteristics was estimated using a classification tree. The best classification tree utilized rugosity, reflectivity and complexity data and produced misclassification rates of less than 25% overall. Mean grain size of sediment samples was not strongly related to the acoustic data. Error rates were highest for those substrate classes with the smallest number of data points. The results highlight the need for adequate sample sizes and coverage of all potential substrate types when groundtruthing acoustic maps. 相似文献
14.
The construction of S-wave velocity models of marine sediments down to hundreds of meters below the seafloor is important
in a number of disciplines. One of the most significant trends in marine geophysics is to use interface waves to estimate
shallow shear velocities which play an important role in determining the shallow crustal structure. In marine settings, the
waves trapped near the fluid–solid interface are called Scholte waves, and this is the subject of the study. In 1998, there
were experiments on the Ninetyeast Ridge (Central Indian Ocean) to study the shallow seismic structure at the drilled site.
The data were acquired by both ocean bottom seismometer and ocean bottom hydrophone. A new type of seafloor implosion sources
has been used in this experiment, which successfully excited fast and high frequency (>500 Hz) body waves and slow, intermediate
frequency (<20 Hz) Scholte waves. The fundamental and first higher mode Scholte waves have both been excited by the implosion
source. Here, the Scholte waves are investigated with a full waveform modeling and a group velocity inversion approach. Shear
wave velocities for the uppermost layers of the region are inferred and results from the different methods are compared. We
find that the full waveform modeling is important to understand the intrinsic attenuation of the Scholte waves between 1 and
20 Hz. The modeling shows that the S-wave velocity varies from 195 to 350 m/s in the first 16 m of the uppermost layer. Depths
levels of high S-wave impedance contrasts compare well to the layer depth derived from a P-wave analysis as well as from drilling
data. As expected, the P- to S-wave velocity ratio is very high in the uppermost 16 m of the seafloor and the Poisson ratio
is nearly 0.5. Depth levels of high S-wave impedance contrasts are comparable to the layer depth derived from drilling data. 相似文献
15.
目前大地电磁三维正演模拟的主要问题是计算效率偏低.Pankratov等提出了一种精确的、稳定的和宽频的三维电磁场正演计算方法,并成功应用于大地电磁场正演模拟中.该方法使用体积积分方程法,利用改进的Neumann序列(MNS)技术来求解Maxwell方程,成功地避免了解大型的线性方程组.在本文中针对这一主要问题尝试引入了广义双共轭梯度法来迭代求改进的Neumann序列中的解,与传统的迭代方法相比可以提高迭代的效率.同时使用了将格林函数分解为两部分在波数域求解,这样比常规的利用快速汉克尔变换求解效率更高.最后试验了两个模型,并与三维交错网格有限差分法计算结果相比较,证明该方法的正确与有效,并且通过具体计算表明该方法在精度保证的条件下计算速度上具有明显的优势. 相似文献
16.
使用最小二乘支持向量机分类方法建立了两个砂土液化预测模型,预测结果与野外实际情况全部相符,表明该分类方法用于预测砂土液化是可行的,且预测准确率高。 相似文献
17.
Swarm intelligence for classification of remote sensing data 总被引：2，自引：0，他引：2
This paper proposes a new method to classify remote sensing data by using Particle Swarm Optimization (PSO). This method is to generate classification rules through simulating the behaviors of bird flocking. Optimized intervals of each band are found by particles in multi-dimension space, linked with land use types for forming classification rules. Compared with other rule induction techniques (e.g. See5.0), PSO can efficiently find optimized cut points of each band, and have good convergence in the search process. This method has been applied to the classification of remote sensing data in Panyu district of Guangzhou with satisfactory results. It can produce higher accuracy in the classification than the See5.0 decision tree model. 相似文献
18.
19.
We describe an algorithm for inverting magnetotelluric (MT) data in the presence of strong bathymetry or topography. Instead of correcting distortions due to bathymetry or topography we incorporate them directly into the inversion. To achieve a high accuracy in computing MT responses we use finite‐difference approximations that permit sloping discrete boundaries inside model elements. The same approach is applicable to any seafloor electromagnetic calculation and can also be used to incorporate steep topography on land. We test our approach on various topographic features and compare our results to that of a finite element approach. Finally, we present inversion examples that illustrate the effectiveness of our inversion algorithm in recovering true subsurface structures in the presence of strong bathymetry and topography. 相似文献
20.
Unsupervised fuzzy classification and object-based image analysis of multibeam data to map deep water substrates, Cook Strait, New Zealand 总被引：1，自引：0，他引：1
A comprehensive 32 kHz multibeam bathymetry and backscatter survey of Cook Strait, New Zealand (∼8500 km2), is used to generate a regional substrate classification map over a wide range of water depths, seafloor substrates and geological landforms using an automated mapping method based on the textural image analysis of backscatter data. Full processing of the backscatter is required in order to obtain an image with a strongly attenuated specular reflection. Image segmentation of the merged backscatter and bathymetry layers is constrained using shape, compactness, and texture measures. The number of classes and their spatial distribution are statistically identified by employing an unsupervised fuzzy-c-means (FCM) clustering algorithm to sediment samples, independent of the backscatter data. Classification is achieved from the overlay of the FCM result onto a segmented image and attributing segments with the FCM class. Four classes are identified and uncertainty in class attribution is quantified by a confusion index layer. Validation of the classification map is done by comparing the results with the sediment and structural maps. Backscatter (BS) strength angular profiles are used to show acoustic class separation. The method takes us one step further in combining multibeam data with physical seabed data in a complementary analysis to seek correlations between datasets using object-based image analysis and unsupervised classification. Texture within these identified classes is then examined for correlation with typical backscatter angular responses for mud, sand and gravel. The results show a first order correlation between each of the classes and both the sedimentary properties and the geomorphological map. 相似文献