A modified stochastic neighbor embedding for multi-feature dimension reduction of remote sensing images

In automated remote sensing based image analysis, it is important to consider the multiple features of a certain pixel, such as the spectral signature, morphological property, and shape feature, in both the spatial and spectral domains, to improve the classification accuracy. Therefore, it is essential to consider the complementary properties of the different features and combine them in order to obtain an accurate classification rate. In this paper, we introduce a modified stochastic neighbor embedding (MSNE) algorithm for multiple features dimension reduction (DR) under a probability preserving projection framework. For each feature, a probability distribution is constructed based on t-distributed stochastic neighbor embedding (t-SNE), and we then alternately solve t-SNE and learn the optimal combination coefficients for different features in the proposed multiple features DR optimization. Compared with conventional remote sensing image DR strategies, the suggested algorithm utilizes both the spatial and spectral features of a pixel to achieve a physically meaningful low-dimensional feature representation for the subsequent classification, by automatically learning a combination coefficient for each feature. The classification results using hyperspectral remote sensing images (HSI) show that MSNE can effectively improve RS image classification performance.     

基于Parsivel天气现象仪资料的郑州“7·20”罕见特大暴雨微物理特征分析

基于2016-2018年ECMWF模式温度预报和浙江省72个国家基本站观测资料,根据温度日变化特征,采用K-近邻(KNN)回归算法进行误差订正,改进浙江省172 h精细化温度预报。在KNN回归算法中,将模式起报时刻的温度视作“背景”,由模式预报减去起报时刻温度消除“背景”影响,得到温度日变化曲线,通过温度日变化曲线构建差异指标,选取历史相似个例。根据历史相似个例的误差特征,对温度预报进行订正,得到改进的温度预报。检验结果表明,KNN方案的温度预报平均绝对误差较ECMWF和30 d滑动平均误差订正方案(OCF)的分别减小26.2%和5.2%;日最高和最低温度预报误差绝对值小于2℃,准确率较ECMWF的分别提高14.8%和4.3%,较OCF的分别提高3.0%和1.3%。KNN方案对地形复杂地区的温度预报改进效果更为明显,对冷空气活动和夏季高温等天气过程预报改善效果也较稳定。     

面向对象的高分辨率影像特征选择研究

本文主要研究面向对象的高分辨影像信息提取中的特征选择问题.文中分别选择光谱、纹理、形状等特征57个和28个进行特征优化,得到两组分别由46个和4个特征组成的不同的最优特征集,并利用这些特征集采用K近邻、模糊与K近邻级联两种不同的面向对象分类策略进行分类研究.最后从合理性、效率和精度三方面进行了对比分析.实验结果表明,对...     

多层次格网模型的近邻指数聚类生态区划算法与实验——以新疆北部地区区划为例

在综合考虑气候、植被、地貌等凶素的基础上,提出一种基于多层次格网模型的最近邻指数-模糊聚类生态区域划分算法(Nearest Neighbor Index Fuzzy clustering,NNI-FIC).该算法采用"自下而上"的方式,首先,利用离散格网单元之间的严格相似性形成区划的核心分区;然后,通过最近邻指数统计分...     

A continental analysis of correlations between tree patterns in African savannas and human and environmental variables

This study analyses possible relationships between natural processes taking place in savannas and the tree patterns found in savannas. This can lead to new hypotheses about which processes are driving savanna physiognomy. To do so tree patterns were quantified for African savannas from historical aerial photographs applying frequently used landscape metrics. Also, additional data for these areas were collected to quantify the processes taking place at these locations. Correlations between tree pattern indices and explaining factors were analysed. We found a negative trend between tree cover and density of sheep and goats, but no relationship between tree cover and density of cattle, suggesting that small livestock have an effect on tree cover, but that larger livestock (or obligate grazers) do not. Also, a positive correlation between human population density and tree cover was found. Possible explanations for the found relations are discussed. Subsequent ways to analyse the latter correlation are discussed, and the potential of the presented historical database of aerial photographs is highlighted.     

基于最近邻支撑向量特征线融合算法的核爆地震识别

为增强核爆地震模式分类器的泛化能力以提高对核爆炸事件的准确识别能力,论文提出了一种选择支撑向量样本集来表征训练样本集的最近邻支撑向量特征线分类算法,用以训练时扩展核爆地震的训练样本库,提高分类器的泛化能力.该算法用于核爆炸和地震的识别结果发现,和最近邻特征线分类器相比,提出的算法降低了计算复杂度,但识别能力却有些许降低.对新算法的分析发现,纯粹的支撑向量集不能完全代表原始样本空间集,支撑向量比例在其中有重要作用,为发挥支撑向量比例的作用以提高核爆分类器的识别能力,提出了最近邻支撑向量特征线融合算法.最后以核爆地震数据库对上述算法进行了检验和分析,理论分析和识别结果证实,在相同的训练样本选择条件下,最近邻支撑向量特征线融合算法对于核爆炸的识别来说具有较好的泛化能力,正确识别率达到90.3%,且优于支持向量机算法和最近邻特征线算法.     

The spatial ecology of deer–vehicle collisions

Deer–vehicle collisions (DVCs) are a constant problem in North America that cost millions of dollars of damage and loss of human life. This study investigated the spatial relationship of DVCs for a county in western Indiana using nearest neighbor analysis, chi-square test of independence, and a landscape metric: the percentage of like adjacencies. The study was able to show that DVCs are not randomly located, and that habitat type and structure may play an important role in the location of DVCs.     

Delineation of geological facies from poorly differentiated data

The ability to delineate geologic facies and to estimate their properties from sparse data is essential for modeling physical and biochemical processes occurring in the subsurface. If such data are poorly differentiated, this challenging task is complicated further by the absence of a clear distinction between different hydrofacies at locations where data are   available. We consider three alternative approaches for analysis of poorly differentiated data: a k$k$-means clustering algorithm, an expectation–maximization algorithm, and a minimum-variance algorithm. Two distinct synthetically generated geological settings are used to analyze the ability of these algorithms to assign accurately the membership of such data in a given geologic facies. On average, the minimum-variance algorithm provides a more robust performance than its two counterparts, and when combined with a nearest neighbor algorithm, it also yields the most accurate reconstruction of the boundaries between the facies.     

非参数统计方法在水文水资源中的应用与展望

非参数统计方法在水文水资源系统中的应用和研究已取得了一些进展,依研究内容不同可分三类:(1)在水文频率分析中的应用和研究;(2)在水文水资源预报中的应用和研究;(3)在水文水资源模拟中的应用和研究.对非参数统计方法在水文水资源系统中的应用和研究现状作了较全面的介绍,并展望了今后的研究趋势和发展方向.