Multi-Scale Texture Modeling

Patterns of crystallographic preferred orientation are referred to as texture. The specific subject of texture analysis is the experimental determination and interpretation of the statistical distribution of orientations of crystals within a specimen of polycrystalline material, which could be metals or rocks. The objective is to relate an observed pattern of preferred orientation to its generating processes and vice versa. In geosciences, texture of minerals in rocks is used to infer constraints on their tectono-metamorphic history. Since most physical properties of crystals, such as elastic moduli, the coefficients of thermal expansion, or chemical resistance to etching depends on crystal symmetry and orientation, the presence of texture imparts directional properties to the polycrystalline material. A major issue of mathematical texture analysis is the resolution of the inverse problem to determine a reasonable orientation density function on SO(3) from measured pole intensities on , which relates to the inverse of the totally geodesic Radon transform. This communication introduces a wavelet approach into mathematical texture analysis. Wavelets on the two-dimensional sphere and on the rotational group SO(3) are discussed, and an algorithms for a wavelet decomposition on both domains following the ideas of Ta-Hsin Li is given. The relationship of these wavelets on both domains with respect to the totally geodesic Radon transform is investigated. In particular, it is shown that the Radon transform of these wavelets on SO(3) are again wavelets on . A novel algorithm for the inversion of experimental pole intensities to an orientation density function based on this relationship is developed.     

Automatic classification of volcanic earthquakes by using Multi-Layered neural networks

The application of neural networks as classifiers of seismic events is described with the aim of developing an automatic system for the classification of explosion quakes at the Stromboli volcano. The architecture of the network that we trained to identify four different classes of shocks was a Multi-Layer Perceptron, using the Back Error Propagation algorithm. Five different approaches for representing the information embedded in the seismograms, both in the time and in the frequency domain, were considered, and the results compared. The direct use of the time series of the shocks was not satisfactory. The auto-correlation function worked well, but in some cases it was misleading. A better performance was obtained with a frequency domain representation. Finally, the use of the envelope function did not work well. Combining parameters such as the auto-correlation and envelope functions can improve one source of error, but it may introduce new ones. The performance obtained highlights the importance of the data attributes used for the training of the network. Topologies with eight neurons in a single hidden layer gave, on average, the best results among the considered neural network structures. The overall results provide a large number of events (89% with the best performance) correctly classified, indicating that this automatic technique is reliable, and encouraging further applications in the field of volcanic seismology.     

地震综合效应场函数及其在地震预报中的应用──（一）地震综合效应场函数及其计算方法

在系统地分析了目前各种测震学地震预报方法科学思路的基础上,认为测震学地震预报方法基本上可以分为两大类。一类是以已经发生的一些地震作为未来可能发生的地震的“因”,即由于已经发生的地震对区域应力场的影响,导致未来发生较强地震。这一类包括的预报方法较多,如空区、条带、ｂ值、地震迁移、相关地震等等及其由此衍生出来的各种方法。另一类是把已经发生的一些地震作为区域应力场增强的“果”,即已经发生的地震是区域应力场增强过程中的一种反映,而未来地震不一定是已经发生的地震所导致的结果。这一类包括“地震窗口”、小震群活动等方法。针对第一类方法,各种预报方法都是力图从地震三要素中提取未来地震的信息,而具体作法又都是利用地震三要素这个多维空间的某个剖面。为了从地震活动诸要素的多维空间提取综合信息,我们对每个地震加入了破裂面方位,构成了地震第四要素,并依据地震４要素建立了地震综合效应场函数。地震综合效应场函数概括了多种测震学地震预报方法的科学思路和预报经验,从而可以形成测震学的综合预报方法。     

Estimation of covariance matrix from geochemical data with observations below detection limits

Multivariate statistical analyses have been extensively applied to geochemical measurements to analyze and aid interpretation of the data. Estimation of the covariance matrix of multivariate observations is the first task in multivariate analysis. However, geochemical data for the rare elements, especially Ag, Au, and platinum-group elements, usually contain observations the below detection limits. In particular, Instrumental Neutron Activation Analysis (INAA) for the rare elements produces multilevel and possibly extremely high detection limits depending on the sample weight. Traditionally, in applying multivariate analysis to such incomplete data, the observations below detection limits are first substituted, for example, each observation below the detection limit is replaced by a certain percentage of that limit, and then the standard statistical computer packages or techniques are used to obtain the analysis of the data. If a number of samples with observations below detection limits is small, or the detection limits are relatively near zero, the results may be reasonable and most geological interpretations or conclusions are probably valid. In this paper, a new method is proposed to estimate the covariance matrix from a dataset containing observations below multilevel detection limits by using the marginal maximum likelihood estimation (MMLE) method. For each pair of variables, sayY andZ whose observations containing below detection limits, the proposed method consists of three steps: (i) for each variable separately obtaining the marginal MLE for the means and the variances, , , , and forY andZ: (ii) defining new variables by and and lettingA=C+D andB=C–D, and obtaining MLE for variances, and forA andB; (iii) estimating the correlation coefficient _YZ by and the covariance _YZ by . The procedure is illustrated by using a precious metal geochemical data set from the Fox River Sill, Manitoba, Canada.     

Estimation for partially observed Markov processes

Many stochastic process models for environmental data sets assume a process of relatively simple structure which is in some sense partially observed. That is, there is an underlying process (X_n, n 0) or (X_t, t 0) for which the parameters are of interest and physically meaningful, and an observable process (Y_n, n 0) or (Y_t, t 0) which depends on the X process but not otherwise on those parameters. Examples are wide ranging: the Y process may be the X process with missing observations; the Y process may be the X process observed with a noise component; the X process might constitute a random environment for the Y process, as with hidden Markov models; the Y process might be a lower dimensional function or reduction of the X process. In principle, maximum likelihood estimation for the X process parameters can be carried out by some form of the EM algorithm applied to the Y process data. In the paper we review some current methods for exact and approximate maximum likelihood estimation. We illustrate some of the issues by considering how to estimate the parameters of a stochastic Nash cascade model for runoff. In the case of k reservoirs, the outputs of these reservoirs form a k dimensional vector Markov process, of which only the kth coordinate process is observed, usually at a discrete sample of time points.     

Application of X-ray computed tomography for analyzing cleat spacing and cleat aperture in coal samples

New techniques to determine distributions of cleat aperture, cleat orientation and cleat spacing from CT scans have been developed. For cleat orientation and spacing distributions, two different coal blocks were scanned. The CT scans have been analyzed for the three orthogonal directions. Histograms of the cleat orientations are bimodal, expressing the typical cleat texture of face and butt cleats and bedding perpendicular relaxation fractures. Deviations up to 20° from the peak values in the cleat orientation distributions were used as input for automated image analysis of cleat spacing. Distributions of the cleat spacing measurements are related to the face and butt cleat directions. The term “relevant cleat length” is introduced as a measure to extract the amount of cleat length involved with the cleat spacing measurements. The ratio ranges from 0.03 to 0.38 and expresses the difference in cleat texture in both samples. Cleat spacing versus relevant cleat length shows sample specific patterns for face cleat, butt cleat and bedding. To describe cleat aperture quantitatively, peak height and missing attenuation have been used. The image of a cleat was seen as a convolution of a rectangular fracture profile with a Gaussian point spread function.     

Modeling of the classification behavior of the diaphragms used in multi-chamber cement mills

A new approach has been developed for describing the classification that arises at the diaphragm in multi-chamber ball mills. Fourteen industrial scale cement mills have been sampled in this study and after mass balancing, model fits of the equipment were done. Successive mill product predictions for the sampled conditions were achieved, and requirements to build up an optimization capable model are discussed.     

Modèles géostatistiques de concentrations ou de débits le long des cours d'eau

Estimating concentrations or flow rates along a stream network requires specific models. Two classes of models, recently proposed in the literature, are generalized, to the intrinsic case in particular. We present a global construction by ‘streams’, i.e. on the whole set of paths between sources and outlet. Combining stationary or intrinsic one-dimensional random functions leads to stationary or intrinsic models on segments, with discontinuities at the forks. A construction from outlet to sources, leads to stationary or intrinsic models on each stream, without any discontinuity at the forks. The linear variogram is found as a particular case. The extension to the linear model of coregionalization is immediate, allowing a multivariate modelling of concentrations. To cite this article: C. de Fouquet, C. Bernard-Michel, C. R. Geoscience 338 (2006).