An evaluation of Guided Regularized Random Forest for classification and regression tasks in remote sensing

New Earth observation missions and technologies are delivering large amounts of data. Processing this data requires developing and evaluating novel dimensionality reduction approaches to identify the most informative features for classification and regression tasks. Here we present an exhaustive evaluation of Guided Regularized Random Forest (GRRF), a feature selection method based on Random Forest. GRRF does not require fixing a priori the number of features to be selected or setting a threshold of the feature importance. Moreover, the use of regularization ensures that features selected by GRRF are non-redundant and representative. Our experiments based on various kinds of remote sensing images, show that GRRF selected features provides similar results to those obtained when using all the available features. However, the comparison between GRRF and standard random forest features shows substantial differences: in classification, the mean overall accuracy increases by almost 6% and, in regression, the decrease in RMSE almost reaches 2%. These results demonstrate the potential of GRRF for remote sensing image classification and regression. Especially in the context of increasingly large geodatabases that challenge the application of traditional methods.     

Properties of Galaxies in the Second Byurakan Survey. II. Characteristic Features of a Separate Sample of Galaxies

Some features of a sample of galaxies from the Second Byurakan Survey are discussed. Most of them are small galaxies with star formation. It is shown that different types of galaxies are in the Survey, with high, medium, and low levels of excitation and continua ranging from blue to relatively red. Many of the galaxies have absorption lines in their spectra; their continua are formed by stars in later spectral classes.     

AN INTEGRATED APPROACH TO AIR MASS CLASSIFICATION IN THE NORTH CENTRAL UNITED STATES

Both subjective and objective schemes for air mass classification offer advantages to climatological studies. An integrated approach is developed and applied to 850 mb level data in the north central United States. Each air mass type was initially delimited by a subjective analysis. Final limits were then determined with an objective technique that separates temperature curves into normal components based on interpretation from the subjective method. Continental air was most frequent in all seasons except summer. The methodology also allows estimation of classification error.     

Discrimination of sedimentary facies by association analysis

The ability of association analysis to discriminate sedimentary facies was tested on Purdy's modal analyses of modern sediments of the Great Bahama Bank. Purdy's data set has served in the past as a standard reference for evaluating various multivariate classification algorithms. In order to adapt Purdy's data to association analysis, the percent abundance of the 12 constituents was converted to binary form by dichotomizing each variable on its mean value. The results obtained by association analysis are virtually identical to those obtained by Purdy and other authors. The same four main sedimentary facies were discriminated; 86% of the samples were identically classified (97% when misclassified borderline cases are counted as matches); the total partition variance of the classification is only negligibly greater (4%); and the grouping of the variables yielded the same four groups. The rank order of the three division-attributes responsible for the sample classification is fines, oolites, and corals. Association analysis has been employed by other authors to differentiate meaningful facies groups in studies of ancient reef carbonates, modern reef sediments, and heavy minerals in stream sediments. In all these studies, the results were found to be compatible with those obtained by using the continuous quantitative measurements, indicating that qualitative binary data may often be sufficient for the purpose of facies discrimination in many branches of geology and that association analysis is an effective method for this purpose.     

Descriptive Analysis and Classification of Benthic Communities in Some Mediterranean Coastal Lagoons (Central Italy)

Abstract. Benthic samples were taken seasonally at 32 stations in five coastal lagoons of Central Italy (Fogliano, Monaci, Caprolace, Fondi, and Lungo) during a two-year period (1982–1984). Composition and distribution of benthic populations in each lagoon as well as seasonal trends of species richness, density, diversity, and redundancy are presented and discussed.
The faunal differentiation appeared to be mainly due to the different degree of water exchange, both with marine and continental systems.     

红树林遥感智能分类方法研究

利用ＧＩＳ信息中的红树林地域分布规律作向导，充分利用械树林区的光谱信息进行红树林植被信息提取。在对比研究常规的红树林图象自理方法的基础上，提出了适用于红树林植被信息遥感监测的智能分类方法。     

利用X射线衍射测定单斜辉石中硬玉组分含量

硬玉是单斜辉石中十分重要的端元组分,是单斜辉石分类的主要基础。利用x射线衍射分析确定的硬玉组分含量计算式为Jd(％)=[A-2.235]×25.741,其中Jd为硬玉组分百分含量,A为2—Theta角为30°(d=2.97)时峰值的相对强度Ⅰ与半高宽的乘积,其计算误差仅6.32％。     

用于数值分类的聚类分析

这里所谓聚类分析包括传统的聚类分析和作者提出的聚类统计检验分析。传统的聚类分析是常用的重要数值分类方法。但是，这个方法由于缺乏显著性检验，在国际上已长期公认是一个在多元统计中并不完整的方法。本文把聚类统计检验分析在其具有显著性检验的优越性上，和传统聚类分析作了对比分析。聚类统计检验分析区分有均匀聚类与中心聚类。前者用（１６）作为聚类判据，后者则用（２７），并且各用相应的公式（１７）－（２１）与（２     

简化模糊H-ARTMAP神经网络在高光谱遥感分类中的应用

人工神经网络作为一种不需估计类别分布参数的遥感影像分类方法,能够克服分类中的不确定性,提高分类精度。模糊ARTMAP人工神经网络具有稳定、泛化性能好、支持增量式学习等特点,通过对简化模糊AR-TMAP神经网络和H-ARTMAP神经网络的分析和集成,构造了一种用于高光谱遥感影像分类的简化模糊H-ART-MAP网络。实验证明该方法在分类效率、运算时间和分类精度等方面都优于最大似然分类、BP神经网络、最小距离分类、光谱角制图模型等分类方法。