Local Binary Patterns in Focal-Plane Processing. Analysis and Applications

Feature extraction is the part of pattern recognition, where the sensor data is transformed into a more suitable form for the machine to interpret. The purpose of this step is also to reduce the amount of information passed to the next stages of the system, and to preserve the essential information in the view of discriminating the data into different classes. For instance, in the case of image analysis the actual image intensities are vulnerable to various environmental effects, such as lighting changes and the feature extraction can be used as means for detecting features, which are invariant to certain types of illumination changes. Finally, classification tries to make decisions based on the previously transformed data. The main focus of this thesis is on developing new methods for the embedded feature extraction based on local non-parametric image descriptors. Also, feature analysis is carried out for the selected image features. Low-level Local Binary Pattern (LBP) based features are in a main role in the analysis. In the embedded domain, the pattern recognition system must usually meet strict performance constraints, such as high speed, compact size and low power consumption. The characteristics of the final system can be seen as a trade-off between these metrics, which is largely affected by the decisions made during the implementation phase. The implementation alternatives of the LBP based feature extraction are explored in the embedded domain in the context of focal-plane vision processors. In particular, the thesis demonstrates the LBP extraction with MIPA4k massively parallel focal-plane processor IC. Also higher level processing is incorporated to this framework, by means of a framework for implementing a single chip face recognition system. Furthermore, a new method for determining optical flow based on LBPs, designed in particular to the embedded domain is presented. Inspired by some of the principles observed through the feature analysis of the Local Binary Patterns, an extension to the well known non-parametric rank transform is proposed, and its performance is evaluated in face recognition experiments with a standard dataset. Finally, an a priori model where the LBPs are seen as combinations of n-tuples is also presented

Learning from Geometry

Subspaces and manifolds are two powerful models for high dimensional signals. Subspaces model linear correlation and are a good fit to signals generated by physical systems, such as frontal images of human faces and multiple sources impinging at an antenna array. Manifolds model sources that are not linearly correlated, but where signals are determined by a small number of parameters. Examples are images of human faces under different poses or expressions, and handwritten digits with varying styles. However, there will always be some degree of model mismatch between the subspace or manifold model and the true statistics of the source. This dissertation exploits subspace and manifold models as prior information in various signal processing and machine learning tasks.

A near-low-rank Gaussian mixture model measures proximity to a union of linear or affine subspaces. This simple model can effectively capture the signal distribution when each class is near a subspace. This dissertation studies how the pairwise geometry between these subspaces affects classification performance. When model mismatch is vanishingly small, the probability of misclassification is determined by the product of the sines of the principal angles between subspaces. When the model mismatch is more significant, the probability of misclassification is determined by the sum of the squares of the sines of the principal angles. Reliability of classification is derived in terms of the distribution of signal energy across principal vectors. Larger principal angles lead to smaller classification error, motivating a linear transform that optimizes principal angles. This linear transformation, termed TRAIT, also preserves some specific features in each class, being complementary to a recently developed Low Rank Transform (LRT). Moreover, when the model mismatch is more significant, TRAIT shows superior performance compared to LRT.

The manifold model enforces a constraint on the freedom of data variation. Learning features that are robust to data variation is very important, especially when the size of the training set is small. A learning machine with large numbers of parameters, e.g., deep neural network, can well describe a very complicated data distribution. However, it is also more likely to be sensitive to small perturbations of the data, and to suffer from suffer from degraded performance when generalizing to unseen (test) data.

From the perspective of complexity of function classes, such a learning machine has a huge capacity (complexity), which tends to overfit. The manifold model provides us with a way of regularizing the learning machine, so as to reduce the generalization error, therefore mitigate overfiting. Two different overfiting-preventing approaches are proposed, one from the perspective of data variation, the other from capacity/complexity control. In the first approach, the learning machine is encouraged to make decisions that vary smoothly for data points in local neighborhoods on the manifold. In the second approach, a graph adjacency matrix is derived for the manifold, and the learned features are encouraged to be aligned with the principal components of this adjacency matrix. Experimental results on benchmark datasets are demonstrated, showing an obvious advantage of the proposed approaches when the training set is small.

Stochastic optimization makes it possible to track a slowly varying subspace underlying streaming data. By approximating local neighborhoods using affine subspaces, a slowly varying manifold can be efficiently tracked as well, even with corrupted and noisy data. The more the local neighborhoods, the better the approximation, but the higher the computational complexity. A multiscale approximation scheme is proposed, where the local approximating subspaces are organized in a tree structure. Splitting and merging of the tree nodes then allows efficient control of the number of neighbourhoods. Deviation (of each datum) from the learned model is estimated, yielding a series of statistics for anomaly detection. This framework extends the classical {\em changepoint detection} technique, which only works for one dimensional signals. Simulations and experiments highlight the robustness and efficacy of the proposed approach in detecting an abrupt change in an otherwise slowly varying low-dimensional manifold.

Varianzanalysen - Prüfen der Voraussetzungen und nichtparametrische Methoden sowie praktische Anwendungen mit R und SPSS. Version 1.1 (2.3.2015)

Die Voraussetzunen der parametrischen 1- und mehrfaktoriellen Varianzanalyse, mit und ohne Messwiederholungen, werden besprochen. Ferner werden eine Reihe von alternativen Verfahren vorgestellt, insbesondere einige nichtparametrische, darunter RT (rank transform), INT (inverse normal transform), ART (aligned rank transform), Puri & Sen (L statistic), van der Waerden und Akritas & Brunner (ATS anova type statistic), die sich auf die parametrische Varianzanalyse zurückführen lassen, sowie dichotome und ordinale logistische Regression. Hierzu werden Lösungen mit R und SPSS ausführlich gezeigt.

Analysis Of The Dna Fourier Transform-infrared Microspectroscopic Signature Using An All-reflecting Objective.

The Fourier transform-infrared (FT-IR) signature of dry samples of DNA and DNA-polypeptide complexes, as studied by IR microspectroscopy using a diamond attenuated total reflection (ATR) objective, has revealed important discriminatory characteristics relative to the PO2(-) vibrational stretchings. However, DNA IR marks that provide information on the sample's richness in hydrogen bonds have not been resolved in the spectral profiles obtained with this objective. Here we investigated the performance of an all reflecting objective (ARO) for analysis of the FT-IR signal of hydrogen bonds in DNA samples differing in base richness types (salmon testis vs calf thymus). The results obtained using the ARO indicate prominent band peaks at the spectral region representative of the vibration of nitrogenous base hydrogen bonds and of NH and NH2 groups. The band areas at this spectral region differ in agreement with the DNA base richness type when using the ARO. A peak assigned to adenine was more evident in the AT-rich salmon DNA using either the ARO or the ATR objective. It is concluded that, for the discrimination of DNA IR hydrogen bond vibrations associated with varying base type proportions, the use of an ARO is recommended.

POSITIVITY PROPERTIES OF THE FOURIER TRANSFORM AND THE STABILITY OF PERIODIC TRAVELLING-WAVE SOLUTIONS

In this paper we establish a method to obtain the stability of periodic travelling-wave solutions for equations of Korteweg-de Vries-type u(t) + u(p)u(x) - Mu(x) = 0, with M being a general pseudodifferential operator and where p >= 1 is an integer. Our approach uses the theory of totally positive operators, the Poisson summation theorem, and the theory of Jacobi elliptic functions. In particular we obtain the stability of a family of periodic travelling waves solutions for the Benjamin Ono equation. The present technique gives a new way to obtain the existence and stability of cnoidal and dnoidal waves solutions associated with the Korteweg-de Vries and modified Korteweg-de Vries equations, respectively. The theory has prospects for the study of periodic travelling-wave solutions of other partial differential equations.

LATENT RANK CHANGE DETECTION FOR ANALYSIS OF SPLICE-JUNCTION MICROARRAYS WITH NONLINEAR EFFECTS

Alternative splicing of gene transcripts greatly expands the functional capacity of the genome, and certain splice isoforms may indicate specific disease states such as cancer. Splice junction microarrays interrogate thousands of splice junctions, but data analysis is difficult and error prone because of the increased complexity compared to differential gene expression analysis. We present Rank Change Detection (RCD) as a method to identify differential splicing events based upon a straightforward probabilistic model comparing the over-or underrepresentation of two or more competing isoforms. RCD has advantages over commonly used methods because it is robust to false positive errors due to nonlinear trends in microarray measurements. Further, RCD does not depend on prior knowledge of splice isoforms, yet it takes advantage of the inherent structure of mutually exclusive junctions, and it is conceptually generalizable to other types of splicing arrays or RNA-Seq. RCD specifically identifies the biologically important cases when a splice junction becomes more or less prevalent compared to other mutually exclusive junctions. The example data is from different cell lines of glioblastoma tumors assayed with Agilent microarrays.

Estimation of hexenuronic acids and kappa number in kraft pulps of Eucalyptus globulus by Fourier transform near infrared spectroscopy and multivariate analysis

Fourier transform near infrared (FT-NIR) spectroscopy was evaluated as an analytical too[ for monitoring residual Lignin, kappa number and hexenuronic acids (HexA) content in kraft pulps of Eucalyptus globulus. Sets of pulp samples were prepared under different cooking conditions to obtain a wide range of compound concentrations that were characterised by conventional wet chemistry analytical methods. The sample group was also analysed using FT-NIR spectroscopy in order to establish prediction models for the pulp characteristics. Several models were applied to correlate chemical composition in samples with the NIR spectral data by means of PCR or PLS algorithms. Calibration curves were built by using all the spectral data or selected regions. Best calibration models for the quantification of lignin, kappa and HexA were proposed presenting R-2 values of 0.99. Calibration models were used to predict pulp titers of 20 external samples in a validation set. The lignin concentration and kappa number in the range of 1.4-18% and 8-62, respectively, were predicted fairly accurately (standard error of prediction, SEP 1.1% for lignin and 2.9 for kappa). The HexA concentration (range of 5-71 mmol kg(-1) pulp) was more difficult to predict and the SEP was 7.0 mmol kg(-1) pulp in a model of HexA quantified by an ultraviolet (UV) technique and 6.1 mmol kg(-1) pulp in a model of HexA quantified by anion-exchange chromatography (AEC). Even in wet chemical procedures used for HexA determination, there is no good agreement between methods as demonstrated by the UV and AEC methods described in the present work. NIR spectroscopy did provide a rapid estimate of HexA content in kraft pulps prepared in routine cooking experiments.

Power Transformer Differential Protection Based on Clarke`s Transform and Fuzzy Systems

The power transformer is a piece of electrical equipment that needs continuous monitoring and fast protection since it is very expensive and an essential element for a power system to perform effectively. The most common protection technique used is the percentage differential logic, which provides discrimination between an internal fault and different operating conditions. Unfortunately, there are some operating conditions of power transformers that can affect the protection behavior and the power system stability. This paper proposes the development of a new algorithm to improve the differential protection performance by using fuzzy logic and Clarke`s transform. An electrical power system was modeled using Alternative Transients Program (ATP) software to obtain the operational conditions and fault situations needed to test the algorithm developed. The results were compared to a commercial relay for validation, showing the advantages of the new method.

Identification of protein coding regions using the modified Gabor-wavelet transform

An important topic in genomic sequence analysis is the identification of protein coding regions. In this context, several coding DNA model-independent methods based on the occurrence of specific patterns of nucleotides at coding regions have been proposed. Nonetheless, these methods have not been completely suitable due to their dependence on an empirically predefined window length required for a local analysis of a DNA region. We introduce a method based on a modified Gabor-wavelet transform (MGWT) for the identification of protein coding regions. This novel transform is tuned to analyze periodic signal components and presents the advantage of being independent of the window length. We compared the performance of the MGWT with other methods by using eukaryote data sets. The results show that MGWT outperforms all assessed model-independent methods with respect to identification accuracy. These results indicate that the source of at least part of the identification errors produced by the previous methods is the fixed working scale. The new method not only avoids this source of errors but also makes a tool available for detailed exploration of the nucleotide occurrence.

New Analytic Solution of Boltzmann Transform for Horizontal Water Infiltration into Sand

The use of the Boltzmann transform function, lambda(theta), to solve the Richards equation when the diffusivity, D, is a function of only soil water content,., is now commonplace in the literature. Nevertheless, a new analytic solution of the Boltzmann transform lambda(h) as a function of matric potential for horizontal water infiltration into a sand was derived without invoking the concept or use of D(theta). The derivation assumes that a similarity exists between the soil water retention function and the Boltzmann transform lambda(theta). The solution successfully described soil water content profiles experimentally measured for different infiltration times into a homogeneous sand and agrees with those presented by Philip in 1955 and 1957. The applicability of this solution for all soils remains open, but it is anticipated to hold for soils whose air-filled pore-size distribution before wetting is sufficiently narrow to yield a sharp increase of water content at the wetting front during infiltration. It also improves and provides a versatile alternative to the well-known analysis pioneered by Green and Ampt in 1911.

A finite integral transform technique for solving the diffusion-reaction equation with Michaelis-Menten kinetics

An approximate analytical technique employing a finite integral transform is developed to solve the reaction diffusion problem with Michaelis-Menten kinetics in a solid of general shape. A simple infinite series solution for the substrate concentration is obtained as a function of the Thiele modulus, modified Sherwood number, and Michaelis constant. An iteration scheme is developed to bring the approximate solution closer to the exact solution. Comparison with the known exact solutions for slab geometry (quadrature) and numerically exact solutions for spherical geometry (orthogonal collocation) shows excellent agreement for all values of the Thiele modulus and Michaelis constant.

Characterizing QALYs under a general rank dependent utility model

This paper provides a characterization of QALYs, the most important outcome measure in medical decision making, in the context of a general rank dependent utility model. We show that both for chronic and for nonchronic health states the characterization of QALYs depends on intuitive conditions. This facilitates the assessment of the validity of QALYs in rank dependent non-expected utility theories and a comparison with other utility based measures of health.

Cannabidiol decreases bone resorption by inhibiting RANK/RANKL expression and pro-inflammatory cytokines during experimental periodontitis in rats

Cannabidiol (CBD) is a cannabinoid component from Cannabis sativa that does not induce psychotomimetic effects and possess anti-inflammatory properties. In the present study we tested the effects of CBD in a periodontitis experimental model in rats. We also investigated possible mechanisms underlying these effects. Periodontal disease was induced by a ligature placed around the mandible first molars of each animal. Male Wistar rats were divided into 3 groups: control animals; ligature-induced animals treated with vehicle and ligature-induced animals treated with CBD (5 mg/kg, daily). Thirty days after the induction of periodontal disease the animals were sacrificed and mandibles and gingival tissues removed for further analysis. Morphometrical analysis of alveolar bone loss demonstrated that CBD-treated animals presented a decreased alveolar bone loss and a lower expression of the activator of nuclear factor-kappa B ligand RANKL/RANK. Moreover, gingival tissues from the CBD-treated group showed decreased neutrophil migration (MPO assay) associated with lower interleukin (IL)-1 beta and tumor necrosis factor (TNF)-alpha production. These results indicate that CBD may be useful to control bone resorption during progression of experimental periodontitis in rats. Crown Copyright (C) 2008 Published by Elsevier B.V. All rights reserved.

Expression of bone resorption regulators (RANK, RANKL, and OPG) in odontogenic tumors

Objective. To investigate the expression of bone resorption regulators (receptor activator of nuclear factor kappa B [RANK], RANK ligand [RANKL], and osteoprotegerin [OPG]) in calcifying cystic odontogenic tumor (CCOT), adenomatoid odontogenic tumor (AOT), calcifying epithelial odontogenic tumor (CEOT), odontogenic myxoma (OM), and ameloblastic fibroma (AF). Study design. The expression of these mediators was evaluated by means of immunohistochemistry. Results. All specimens demonstrated positive immunoreactivity to RANK, RANKL, and OPG. The quantification of these mediators in epithelium revealed a similar pattern of expression for RANKL and OPG in CCOT, AOT, CEOT, and AF. With regard to stromal/mesenchymal cells, the majority of AOT and CCOT cases showed a higher content of OPG than RANKL, whereas CEOT, OM, and especially AF had a tendency to present a greater content of RANKL than OPG. Conclusion. Our data indicate that the CCOT, AOT, CEOT, OM, and AF cell constituents express key regulators of bone metabolism that might locally modulate tumor-associated bone resorption. (Oral Surg Oral Med Oral Pathol Oral Radiol Endod 2008;106:548-55)