基于KPCA-RBF模型的风暴潮灾害经济损失预测

作为破坏性最强的海洋灾害,风暴潮灾害每年都给我国沿海地区造成了巨大的经济损失,运用科学的方法模型合理预测风暴潮灾害经济损失对指导沿海地区的防灾减灾工作意义深远。本文基于风暴潮灾害的成灾特点建立了风暴潮灾害直接经济损失预评估指标体系,由于评估指标数据高度非线性,采用核主成分分析（KPCA）对高维非线性数据进行降维优化,并利用径向基函数（RBF）神经网络对降维后的数据进行训练,从而实现对风暴潮灾害直接经济损失的预测。选取广东省1996—2018年的32个风暴潮灾害损失样本对模型进行仿真测试,结果表明,KPCA-RBF预测模型集成了核主成分分析和径向基函数神经网络的优势,预测结果精度高,学习收敛速度快,对风暴潮灾害数据序列有较好的非线性拟合能力。     

基于深度学习和时空特征融合的海洋渔船密度预测方法

为了从海量渔船轨迹数据中挖掘隐含的信息和知识,进而为渔业行政主管部门的决策提供科学依据,本研究以AIS渔船轨迹数据为研究对象,提出了一种基于深度学习和面向时空特征融合的海洋渔船密度预测方法：首先,利用渔船轨迹数据集对渔船行驶区域进行网格划分;其次,筛选出渔船高密度区域进行研究,避免数据稀疏性问题;再次,根据渔船轨迹数据的时空分析,构建三维时空融合矩阵;最后,通过卷积循环神经网络模型捕获渔船分布的时间和空间特征,并利用卷积神经网络的堆叠加强对空间特征的学习。实验通过东海海域渔船真实轨迹数据进行具体测试,结果表明渔船密度预测值与真实值非常接近,平均绝对误差为4×10-4,模型较好地拟合了渔船密度分布特征,有效地提高了渔船捕捞热点预测的准确性和鲁棒性。     

地理信息系统支持下的山区遥感影像决策树分类

山区遥感影像分类是遥感研究的一大难题。本文利用一种决策树生成算法（C 4.5算法）自动提取知识，基于知识建立决策树用于山区影像分类，并结合研究区土地利用类型与DEM空间统计关系的先验知识，在GIS空间分析的基础上进行影像分类的后处理。与传统的最大似然法分类结果相比，该方法极大地改善了山区地表覆被分类的精度，得到试验区较为可靠的遥感分类图像。     

基于粗糙集与人工神经网络的变压器故障诊断

根据电力变压器故障诊断问题,提出了基于粗糙集与人工神经网络的变压器故障诊断模型,分析了该模型的实现步骤.采用Kohonen网络对连续属性值进行离散化,应用粗糙集理论对特征参数进行属性约简,并把约简结果生成规则作为BP网络的输入.仿真结果表明,把经过粗糙集理论预处理过的数据送入BP网络训练,提高了学习速度和故障诊断正确率,减少了训练时间.     

结合SOM神经网络和混合像元分解的高光谱影像分类方法研究

本文对SOM神经网络算法进行改进,在标类的过程中采用3个策略加以控制,对初始产生的自组织映射图进行调整。通过改进,那些映射到可靠神经元的像素得到了很好的分类,而那些映射到不可靠神经元的像素都被作为不可分像元而提取出来。继而,从混合像元分解的角度来对这些不可分像元进行处理,按类型分解的思想确定混合像元的类别,实现对不可分像元的分类。将SOM神经网络和混合像元分解相结合的分类方法应用于高光谱图像的分类中,通过实验表明了该方法能较好地改善分类效果,提高分类精度。     

Floc size variability under strong turbulence: Observations and artificial neural network modeling

The flocculation of cohesive sediment in the presence of waves is investigated using high-resolution field observations and a newly-developed flocculation model based on artificial neural networks. Vertical profiles of suspended sediment concentration and turbulent intensity are estimated using measurements of current profile and acoustic backscatter. The vertical distribution of floc size is estimated using an artificial neural network (ANN) that is trained and validated using floc size measurements at one vertical level. Data analysis suggests a linear correlation between suspended sediment concentration and turbulence intensity. Observations and numerical simulations show that floc size is inversely related to sediment concentration, turbulence intensity and water temperature. The numerical results indicate that floc growth is supported by low concentration and low turbulence. In the vertical direction, mean size of flocs decreases toward the bottom, suggesting floc breakage due to increasing turbulence intensity toward the bed. A significant decrease in turbulent shear could occur within the bottom few-cm, related to increased damping of turbulence by sediment induced density stratification. The results of the numerical simulations presented here are consistent with the concept of a cohesive sediment particle undergoing aggregation-fragmentation processes, and suggest that the ANN can be a precise tool to study flocculation processes.     

树轮长年表在全球变化相关领域中的应用研究进展-以14C曲线校正、环境考古和地质灾害事件定年为例

近几十年树轮年代学得到了迅速发展,现已成功建立了大量长度逾千年的树轮长年表。本文评述树轮长年表在全球变化相关研究领域中的应用。树轮长年表可为¹⁴C定年提供校准曲线,从而将样本的¹⁴C年龄转化为日历年龄,提高了¹⁴C测年的精度;也可为环境考古如古木建筑及墓葬等提供定年服务,确定古木构建筑的建造年代;树轮长年表在地质灾害事件定年方面也发挥了重要作用。然而,树轮长年表在上述研究领域的应用也存在不足,最新的国际¹⁴C会议公布的¹⁴C校正曲线中能够准确定年的树轮校正部分仅为过去12 410年,目前尚无法提供更长时间尺度的树轮校正曲线。在对木构建筑进行定年时受到古木样本量不足的限制等。因此,为了更好地发挥树轮长年表的应用潜力,应不断发掘与延长现有树轮宽度长年表,同时建立长年表的C、O同位素序列,以期为长年表的应用不断做出贡献。     

Combining Decision Trees with Angle Indices to Identify Mangrove Forest at Shenzhen Bay,China

Mangroves are woody plant communities in the intertidal zone of tropical and subtropical coasts that play an important role in these zones. The infrared wave band is one of the key bands in the remote sensing identification of mangrove forest, and ALI (advanced land imagery) has a large number of infrared bands. Two angle indices were proposed based on liquid water absorption at band 5p and band 5 of EO-1 ALI, denoted as β_1.25 and β_1.65 respectively. A decision tree method was adopted to identify mangrove forest using remote sensing techniques for β_1.25-β_1.65 and NDVI (normalized difference vegetation index) for EO-1 ALI imagery acquired at Shenzhen Bay. The results showed that the reflectance of mangrove forests at band 5p and band 5 was significantly lower than that of terrestrial vegetation due to the characteristics of coastal wetlands of mangrove forests. This resulted in a greater β_1.25-β_1.65 value for mangrove forest than terrestrial vegetation. The decision tree method using β_1.25-β_1.65 and NDVI effectively identifies mangrove forest from other land cover categories. The misclassification and leakage rates were 4.29% and 5.11% respectively. ALI sensors with many infrared bands could play an important role in discriminating mangrove forest.     

基于遗传算法优化的BP神经网络在密度界面反演中的应用

BP神经网络方法在二维密度界面的反演中取得了较好的效果,但在反演三维界面时,由于模型更复杂、参数更多,BP神经网络的收敛速度和反演精度都有一定程度的下降。为了改善反演效果,本文利用遗传算法对BP神经网络的权值、阈值选择过程进行优化,获得了更好的网络模型;并将此模型应用于密度界面模型的反演中,预测误差从上百米减小到数十米,同时迭代计算步数减少了近2/3,有效减少了计算时间,反演结果更准确。利用基于遗传算法优化的BP神经网络反演了法国某地区莫霍面深度,预测相对误差仅为1.8%,取得了较好的应用效果。基于遗传算法优化的BP神经网络在密度界面的反演中具有良好的应用价值和研究前景。     

GIS and ANN model for landslide susceptibility mapping

1 IntroductionThe population growth and the expansion of settlements and life-lines over hazardous areas exert increasingly great impact of natural disasters both in the developed and developing countries. In many countries, the economic losses and casualties due to landslides are greater than commonly recognized and generate a yearly loss of property larger than that from any other natural disasters, including earthquakes, floods and windstorms. Landslides in mountainous terrain often occur a…