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.     

Consideration of time-correlated errors in a Kalman filter applicable to GNSS

An algorithm for considering time-correlated errors in a Kalman filter is presented. The algorithm differs from previous implementations in that it does not suffer from numerical problems; does not contain inherent time latency or require reinterpretation of Kalman filter parameters, and gives full consideration to additive white noise that is often still present but ignored in previous implementations. Simulation results indicate that the application of the new algorithm yields more realistic and therefore useful state and covariance information than the standard implementation. Results from a field test of the algorithm applied to the problem of kinematic differential GPS demonstrate that the algorithm provides slightly pessimistic covariance estimates whereas the standard Kalman filter provides optimistic covariance estimates.     

Damage identification method using harmonic excitation force considering both modelling and measurement errors

A method of structural damage identification using harmonic excitation force is presented. It considers the effects of both measurement and modelling errors in the baseline finite element model. Damage that accompanies changes in structural parameters can be estimated for a damaged structure from the change between measured vibration responses and ones calculated from the analytical model of the intact structure. In practice, modelling errors exist in the analytical model due to material and geometric uncertainties and a reduction in the degrees of freedom as well as measurement errors, making identification difficult. To surmount these problems, bootstrap hypothesis testing, which enables statistical judgment without information about these errors, was introduced. The method was validated by numerical simulation using a three‐dimensional frame structure and real vibration data for a three‐storey steel frame structure. Copyright © 2005 John Wiley & Sons, Ltd.     

非线性洋面温度反演

本文从辐射方程出发，简述洋面温度与亮度温度之间的非线性关系，进行非线性洋面温度反演的方案设计，用正演算法 计算出非线性洋面温度反演公式的回归系数，对样进行检验。     

神经网络反演散射计风场算法的研究

建立了一个神经网络反演卫星散射计海面风场的B-P算法,给出了一个神经网络反演风场的模型,并利用该反演算法和模型对实际卫星散射计数据进行了海面风场反演试验,对风向的多解性利用圆中数滤波方法进行排除.对神经网络训练和检验数据集分别采用ERS-1/2散射计数据和欧洲中期天气预报(ECMWF)提供的风场作为配准点数据.把反演的风速和风向与CMCD4和ECMWF的风场作了比较,它们吻合得比较好;研究表明神经网络反演海面风场是可行和高效的.     

基于分块分形编码的多姿态、表情人脸图像检索

利用图像分形编码中定义域块和最小均方误差一一对应的特点，提出了1种基于分形编码的多姿态、表情的人脸图像检索方法。该方法将待检索图像分割为相同大小的值域块，然后将每一值域块按给定的定义域块进行分形编码，得到最小均方误差，计算该最小均方误差与图像库中最小均方误差的欧氏距离。将待检索图像所有值域块的欧氏距离求平均，此平均欧氏距离较小的几幅图像即为检索出的图像。实验证明该方法能够准确地检索出图像库中存储的同一人的不同姿态、表情的图像。     

一类典型模糊阵及其在传递闭包化简中的应用

本文研究模糊矩阵传递闭包的化简问题。通过定义 1类典型阵 ,给出传递闭包能够进一步化简的充要条件 ,为网络最大流分析提供了 1种新的模式。     

Optimum Sample Size to Detect Perturbation Effects: The Importance of Statistical Power Analysis – A Critique

Abstract. The current article describes statistical power analysis as an efficient strategy for the estimation of the optimum sample size. The principle aim is constructively to criticise and enrich the results presented by Mouillot et al. (1999) , who estimate the optimum sample size for evaluating possible perturbations. The authors did not make any reference to statistical power analysis, even though their objective clearly went beyond a simple stock evaluation to assess management strategies in a particular marine ecosystem. Surprisingly, they proposed (a priori) an ANOVA design to test a hypothesis considering both space and temporal scales. However, the authors did not cover important topics related with power analysis and the precautionary principle, both used into environment impact assessment programmes for marine ecosystems. Based on their results and on statistical power analysis, it is demonstrated that the variability (dispersion statistics), a key factor they used to estimate the sample size, is less relevant than the magnitude of perturbation (effect size). Therefore, a greater effort must be devoted to estimate the effect size of a particular phenomenon rather than a desired variability.     

Ghost fishing in European waters: Impacts and management responses

In this paper we review levels of net loss, what happens to the gear once it has been lost, and the resulting levels of ‘ghost catches’ made in passive net fisheries in the EU. We also consider ghost catches resulting from lost gear in other types of fisheries, and the extent to which the value of ghost catches has been quantified. We consider why fishing gear is lost, and profile common management responses. We present a cost benefit model to assess the relative cost effectiveness of different management measures, and suggest that gear retrieval programmes may provide less value for money than other management responses.     

一维扩散方程整体观测资料下的初值变分同化反演

本文考虑一维扩散方程的反问题,利用变分同化方法通过观测资料来确定方程中的未知初值,通过分析观测误差对于初值误差的影响,证明变分同化初值收敛于原问题的真实参数,并得到了参数的收敛精度。同时将得到的初值代入预报模式中,得到预报解,并分析了预报解的收敛性和预报误差。