Discriminant Function Classification Research Articles

Brightness Values-Based Discriminant Functions for Classification of Degrees of Organic Matter Decomposition in Soil Thin Sections

The decomposition of organic matter represents a fundamental pedogenetic process, since it impacts the carbon cycle and the release of nutrients to the soil. However, quantitative research aimed at micro-scale in situ analysis is scarce, despite its relevance in the decomposition process. Therefore, the objectives of this research were to generate discriminating functions of the degrees of organic matter decomposition, based on the brightness values associated with each morphological stage, and from this step, to generate thematic maps. Soil thin sections of forest and compost soils were selected, and petrographic microscope images with three light sources were taken: plane polarized light (PPL), crossed-polarized light (XPL), and crossed polarizers and a retardation plate (gypsum compensator) inserted (XPLλ). Subsequently, the RGB (red, green, blue) image was broken down into three bands, resulting in nine bands for each image. Two thousand sampling points were generated for each band, obtaining brightness values for each decomposed organic matter stage. The points were classified into four categories based on their degree of decomposition: no (A), light (B), moderate (C), and strong (D), in addition to porosity (P). Linear discriminant analysis was performed to obtain classification models for each level of decomposition. The results show that each degree of organic matter decomposition can be highlighted through specific light sources and a set of bands, with an overall accuracy of >94% and kappa coefficients of >0.75 for all classes. In addition, the resulting functions were validated in training images and high-resolution mosaics to create final thematic maps. The use of linear models automated the production and quality of thematic maps at the microscopic level, which can be useful in monitoring the organic matter decomposition process.

Robust Classification through a Nonparametric Kernel Discriminant Analysis

The problem of classification in situations where the assumption of normality in the data is violated, and there are non-linear clustered structures in the dataset is addressed. A robust nonparametric kernel discriminant classification function, which is able to address this challenge, has been developed and the misclassification rates computed for various bandwidth matrices. A comparison with existing parametric classification functions such as the linear discriminant and quadratic discriminant is conducted to evaluate the performance of this classification function using simulated datasets. The results presented in this paper show good performance in terms of misclassification rates for the kernel discriminant classifier when the correct bandwidth is selected as compared to other identified existing classifiers. In this regard, the study recommends the use of the proposed kernel discriminant classification rule when one wishes to classify units into one of several categories or population groups where parametric classifiers might not be applicable.

Classification method based on the deep structure and least squares support vector machine

Support vector machines (SVMs) are one of the most representative shallow network models and have good generalisation abilities in small data sets. In this Letter, a new classification method based on the deep structure and least squares SVM (LSSVM) is proposed. For large-scale data sets, the method builds the structures of a multi-layer SVM. Using edge detection and the K-means algorithm, the sample set is compressed into a smaller sample set, which is used to train the LSSVM model of each layer and the discriminant classification function is obtained. Finally, this method is applied to UCI data sets and compared with several density-dependent quantised LSSVM methods and other methods. The experimental results show that the method has good performance in solving the large-scale data set classification problem.

A New Rotation Machinery Fault Diagnosis Method Based on Deep Structure and Sparse Least Squares Support Vector Machine

In this paper, a fault diagnosis method that is based on the deep structure and the sparse least squares support vector machine (SLSSVM) is proposed. This method constructs the structure of a multi-layer support vector machine (SVM). First, the SVM on the first layer is trained by using the training samples, and it learns the shallow features of the data. Then, the “feature extraction formula” is used to generate a new expression of the sample, which is used as input of the next layer. The new layer of the SVM trains on the new sample, and it extracts and learns the deep features of the signal layer by layer; eventually, after multiple feature mapping, it outputs the diagnostic results on the last layer. Because of the deep structure, the algorithm complexity and operation time increase. Therefore, in this paper, the least squares support vector machine (LSSVM) is combined with the sparse theory. By constructing the approximate maximal linearly independent vector set in the feature space, we conduct the sparse expression of samples and obtain the discriminant function for classification, which effectively solves the problem of sparsity deficiency for the LSSVM. Last, the method is used to diagnose centrifugal pump faults and rolling bearing faults and compares with the several methods of the SVM, the SLSSVM, deep SVM, and convolutional neural networks. The diagnostic results indicate that the method in this paper has good performance.

The Method for Comprehensive Quality Evaluation of Tests. Part 2

In the article, the description of the complex evaluation method is given, as well as the classical method of Data Mining and Item Response Theory (IRT). In the general method there are six steps. This article describes steps 4-6.
 The fourth step of the method is to evaluate the reliability of the test. A universal two-step procedure is proposed – the assessment of the reliability of individual test tasks based on the coefficient of internal coherence of Kjuder – Richardson and the evaluation of the reliability of the test as a whole by the coefficient of generalization. The first of the coefficients is considered acceptable at the level of 0.7 and above, the second – at the level of 0.8 and above. Two-factor ANOVA variance analysis without repeated measurements in SPSS was used to calculate the second coefficient.
 At the fifth stage of the methodology, the quality of students' differentiation is assessed by a test that is being studied. The tool for this is selected hierarchical cluster procedures, classification trees and classification discriminant functions. The calculations were performed by means of Statistica and SPSS. Three clusters of students with high, medium and low academic performance were identified. It is shown that the test under study allows the differentiation of students.
 At the last, sixth stage, a study of the quality of the test is described based on the one-parameter model of Rash. The levels of the difficulty of the test assignment and the mastering of the student's study material are measured in logics. The analytical task of the characteristic individual curve of the test assignment and the characteristic individual curve of the student, as well as the auxiliary formulas for their calculations, are given. The description is illustrated by a specific example. It is noted that the characteristic curves of students based on the Rash model by means of MathCAD, can clearly divide the latter into two groups – strong (have positive logic) and weak (have negative logic). Recommendations on the interpretation of the obtained results for certain test tasks are formulated. In particular, in case of overlap of the characteristic curves of various test tasks, they must be deleted (normative-oriented test) or reconstructed (criterion-oriented test). This paper does not consider how to determine which test question is to be deleted or corrected, but it is indicated that this can be established with the help of a two-parameter Birnbaum model. If the density of the characteristic curves of the test tasks is not the same; It is recommended to add a test task (in the case of a normative-oriented test) or thus change the duplicate test questions (in the case of a normative-oriented test) to fill the gaps of the abscissa, where there are no characteristic curves.
 By the practical implementation of this technique, the authors determine the development of a separate plug-in that is compatible with the Moodle distance learning platform.
 The prospect of further research in the theoretical framework is determined by the authors of the study of the boundaries of the use of two-parameter and three-parameter models of Birnbaum to improve the process and test results of students in distance learning systems.

Topography-driven satellite imagery analysis for landslide mapping

ABSTRACTWe describe a semi-automatic procedure for the classification of satellite imagery into landslide or no landslide categories, aimed at preparing event landslide inventory maps. The two-steps procedure requires knowledge of the occurrence of a landslide event, availability of a pre- and post- event pseudo-stereo pair and a digital elevation model. The first step consists in the evaluation of a discriminant function, applied to a combination of well-known change detection indices tuned on landslide spectral response. The second step is devoted to discriminant function classification, aimed at distinguishing the only landslide class, through an improvement of the usual ‘thresholding’ method. We devised a multi-threshold classification, in which thresholding is applied separately in small subsets of the scene. We show that using slope units as topographic-aware subsets produces best classification performance when compared to the ground truth of a landslide inventory prepared by visual interpretation. The method proved to be superior to the use of a single threshold and to any multi-threshold procedure based on topography-blind subdivisions of the scene, especially in the validation stage. We argue that the improved classification performance and limited training requirements represent a step forward towards an automatic, real-time landslide mapping from satellite imagery.

Development and Assessment of Postcranial Sex Estimation Methods for a Guatemalan Population.

This study tests whether postcranial sex estimation methods generated from Hispanic, and mainly Mexican samples, can be successfully applied to other increasingly common migrant populations from Central America. We use a sample of postcranial data from a modern (1980s) Guatemalan Maya sample (n=219). Results indicate a decrease in classification accuracies for previously established univariate methods when applied to the Guatemalan study sample, specifically for males whose accuracies ranged from 30 to 84%. This bias toward inaccuracies for Guatemalan males is associated with the smaller skeletal sizes for the Guatemalan sample as compared to the samples used in the tested sex estimation methods. In contrast, the tested multivariate discriminant function classification yielded less sex bias and improved classification accuracies ranging from 82 to 89%. Our results highlight which of the tested univariate and multivariate methods reach acceptable levels for accuracy for sex estimation of cases where the region of origin may include Guatemala.

Morphometric Differences Between Sexes In the White-Faced Ibis (Plegadis chihi)

ABSTRACT Sexual dimorphism in birds may express itself as differences in body size, plumage, color and/or behavior. Many species are monomorphic in color, which makes sex determination difficult in the field. In order to develop a tool to distinguish between male and female specimens of the White-faced Ibis, Plegadis chihi, by using external body measurements, the objective of this work was to quantify morphometric differences between sexes in adults of this species. The following variables were measured: culmen length, bill height and width, tarsus length, middle toe (with claw) length, wing chord, tail length and body mass. Males were larger than females in all of these variables, and presented statistically significant differences for six out of eight parameters. Three discriminant classification functions were obtained with an accurate total classification rate in >85% of the cases. The length of the culmen, tarsus, middle toe with claw, and wing chord, were among the most useful variables to discrimi...

New methodology to assess the quantity and quality of humic substances in organic materials and commercial products for agriculture

The traditional method to determine humic content (humic and fulvic acids) in commercial fertilizers, biostimulants, and organic materials is based on the oxidation of the organic carbon contained in the basic-soluble but acid-insoluble fraction (humic acids) and the basic-acid soluble fraction (fulvic acids) of their alkaline water extracts. This methodology, merely operational, makes it impossible to distinguish if the quantified carbon corresponds to substances with “humic” chemical nature or to non-humic organic matter but with similar solubility properties to those of humic matter. The aim of this work is to develop a new methodology that not only quantifies the humic content in commercial products (and raw materials) but also assesses the humic quality of the quantified organic matter. To this end, humic and fulvic (-like) fractions have been isolated/purified from several humic and non-humic materials and characterized by means of elemental analysis and UV-visible, fluorescence, and infrared spectroscopies, and these data have been used to perform a discriminant analysis (DA). The model obtained from the DA is able to discriminate humic and fulvic fractions from apparently humic or fulvic ones and provides discriminant classification functions that have proven to successfully predict the “humic quality” of the fractions isolated from commercial products, after their elemental and spectroscopic characterization. Therefore, the combination of the fractionation, characterization, and evaluation by the DA is proposed as an effective methodology for quantifying and assessing the quality of the humic content claimed in the labels of commercial products.

Diverse Early Life-History Strategies in Migratory Amazonian Catfish: Implications for Conservation and Management.

Animal migrations provide important ecological functions and can allow for increased biodiversity through habitat and niche diversification. However, aquatic migrations in general, and those of the world’s largest fish in particular, are imperiled worldwide and are often poorly understood. Several species of large Amazonian catfish carry out some of the longest freshwater fish migrations in the world, travelling from the Amazon River estuary to the Andes foothills. These species are important apex predators in the main stem rivers of the Amazon Basin and make up the region’s largest fishery. They are also the only species to utilize the entire Amazon Basin to complete their life cycle. Studies indicate both that the fisheries may be declining due to overfishing, and that the proposed and completed dams in their upstream range threaten spawning migrations. Despite this, surprisingly little is known about the details of these species’ migrations, or their life history. Otolith microchemistry has been an effective method for quantifying and reconstructing fish migrations worldwide across multiple spatial scales and may provide a powerful tool to understand the movements of Amazonian migratory catfish. Our objective was to describe the migratory behaviors of the three most populous and commercially important migratory catfish species, Dourada (Brachyplatystoma rousseauxii), Piramutaba (Brachyplatystoma vaillantii), and Piraíba (Brachyplatystoma filamentosum). We collected fish from the mouth of the Amazon River and the Central Amazon and used strontium isotope signatures (87Sr/86Sr) recorded in their otoliths to determine the location of early rearing and subsequent. Fish location was determined through discriminant function classification, using water chemistry data from the literature as a training set. Where water chemistry data was unavailable, we successfully in predicted 87Sr/86Sr isotope values using a regression-based approach that related the geology of the upstream watershed to the Sr isotope ratio. Our results provide the first reported otolith microchemical reconstruction of Brachyplatystoma migratory movements in the Amazon Basin. Our results indicate that juveniles exhibit diverse rearing strategies, rearing in both upstream and estuary environments. This contrasts with the prevailing understanding that juveniles rear in the estuary before migrating upstream; however, it is supported by some fisheries data that has indicated the presence of alternate spawning and rearing life-histories. The presence of alternate juvenile rearing strategies may have important implications for conservation and management of the fisheries in the region.

Exploring More Representative States of Hidden Markov Model in Optical Character Recognition: A Clustering-Based Model Pre-Training Approach

Hidden Markov Model (HMM) is an effective method to describe sequential signals in many applications. As to model estimation issue, common training algorithm only focuses on the optimization of model parameters. However, model structure influences system performance as well. Although some structure optimization methods are proposed, they are usually implemented as an independent module before parameter optimization. In this paper, the clustering feature of states in HMM is discussed through comparing the mechanism of Quadratic Discriminant Function (QDF) classifier and HMM. Then, through the clustering effect of Viterbi training and Baum–Welch training, a novel clustering-based model pre-training approach is proposed. It can optimize model parameters and model structure by turns, until the representative states of all models are explored. Finally, the proposed approach is evaluated on two typical OCR applications, printed and handwritten Arabic text line recognition. And it is compared with some other optimization methods. The improvement of character recognition performance proves the proposed approach can make more precise state allocation. And the representative states are benefit to HMM decoding.

Surface Defect Detection on Optical Devices Based on Microscopic Dark-Field Scattering Imaging

Methods of surface defect detection on optical devices are proposed in this paper. First, a series of microscopic dark-field scattering images were collected with a line-scan camera. Translation transformation between overlaps of adjacent microscopic dark-field scattering images resulted from the line-scan camera’s imaging feature. An image mosaic algorithm based on scale invariance feature transform (SIFT) is proposed to stitch dark-field images collected by the line-scan camera. SIFT feature matching point-pairs were extracted from regions of interest in the adjacent microscopic dark-field scattering images. The best set of SIFT feature matching point-pairs was obtained via a parallel clustering algorithm. The transformation matrix of the two images was calculated by the best matching point-pair set, and then image stitching was completed through transformation matrix. Secondly, a sample threshold segmentation method was used to segment dark-field images that were previously stitched together because the image background was very dark. Finally, four different supervised learning classifiers are used to classify the defect represented by a six-dimensional feature vector by shape (point or line), and the performance of linear discriminant function (LDF) classifier is demonstrated to be the best. The experimental results showed that defects on optical devices could be detected efficiently by the proposed methods.

PREDICTION OF UNIVERSITY ADMISSION ACHIEVEMENT WITH DISCRIMINANT MODEL: CASE STUDY OF PHANATPITTAYAKARN SCHOOL, CHONBURI, THAILAND

This study was conducted to classify two achievement groups of uni- versity admission (failure and success). High school students studied in Phanat- pittayakarn School during the academic year 2006-2011 participated. Gender, high school program, each of eight O-NET scores (Thai, Social, English, Mathematics, Science, Hygiene, Art and Work) as well as GPA of each subject (Social, English, Mathematics, Science, Hygiene, Art, Work) and high school GPA were investigated as predictors of university admission achievement. Discriminant analysis was per- formed to build a predictive model for achievement of university admission. The correct classification rate (CCR) is utilized to evaluate a model performance. The result presents the discriminant classification function obtained rather highly accu- rate of the average CCR (80.79%) for prediction of university admission achievement in the validation data set.

Morphometric analysis of modern faeces as a tool to identify artiodactyls' coprolites

The identification of the zoological origin of coprolites is important to infer paleoecological characteristics of ancient wildlife. The aim of this study was to design a reliable tool to distinguish coprolites of artiodactyls collected from archaeological sites of Patagonia, Argentina. A reference collection of modern droppings of large size artiodactyls of Patagonia, Lama guanicoe (guanaco) and Hippocamelus bisulcus (huemul), was also studied. Generalized discriminant analysis using shape and size attributes was applied to 538 pellets (nguanaco = 240, nhuemul = 298). Faecal material was photographed and multiple predictor variables were calculated by the Image J (Java 1.6.0_12) Program. Coprolites attributable to camelids by paleoparasitological analyses were also classified by the discriminant classification function obtained from the reference collection. Results indicated that pellet size and shape were able to discriminate droppings of guanaco from those of huemul (Pillai's trace = 0.93, p = 0.0001). The accuracy of the classification function was 100%, with all the modern faeces of guanaco and huemul correctly classified. The accuracy for the identification of camelid coprolites was 100%. This is the first study that used a discriminant classification function of modern dung to identify coprolites. This research also provides information for its potential application to other archaeological deposits of dung in order to confirm the zoological origin of samples.

Improvement of CALIOP Cloud Masking Algorithms for Better Estimation of Dust Extinction Profiles

Mineral dust suspended in the atmosphere affects the Earth’s radiation budget. To accurately predict the effect of dust on the climate system, information regarding its extinction profiles is needed. The Cloud-Aerosol Lidar with Orthogonal Polarization (CALIOP) instrument on the Cloud-Aerosol Lidar and Infrared Pathfinder Satellite Observation (CALIPSO) satellite has enabled the global observation of the vertical distributions of aerosols and clouds since June 2006. To correctly retrieve extinction coefficients from CALIOP signals, the lidar-observed layers must be classified into aerosols or clouds. The cloud masking algorithms of CALIOP should be improved since the cloud mask products occasionally misclassify dense dust as clouds. This study attempts to discriminate misclassified clouds from the CALIOP cloud mask with a discriminant analysis. The training data are collected by tests with the CloudSat cloud mask, the Moderate Resolution Imaging Spectroradiometer (MODIS) cloud mask, and relative humidity. Discrimination of dust from clouds is successful in cases over land and water surfaces during the daytime and nighttime. In contrast, the discrimination model of previous studies was inadequate during the nighttime since training data were not collected during the nighttime. The accuracy rate of the linear discriminant function classification is 91.7 % for misclassified clouds. The cloud mask is most frequently misclassified in the Taklimakan Desert. The proportion of misclassified clouds to the observed dust is ∼34.6 % (below 2 km) in the desert. Comparison of our results with CALIOP level 3 products indicates that the extinction profile using the improved cloud mask is at most twice larger than that of CALIOP level 3 products. This study suggests that the smaller extinction coefficients of CALIOP level 3 products are mainly caused by misclassification of dust as clouds in the vertical feature mask.

Taxonomic Classification of Phytoplankton with Multivariate Optical Computing, Part I: Design and Theoretical Performance of Multivariate Optical Elements

Phytoplankton are single-celled, photosynthetic algae and cyanobacteria found in all aquatic environments. Differential pigmentation between phytoplankton taxa allows use of fluorescence excitation spectroscopy for discrimination and classification. For this work, we applied multivariate optical computing (MOC) to emulate linear discriminant vectors of phytoplankton fluorescence excitation spectra by using a simple filter-fluorometer arrangement. We grew nutrient-replete cultures of three differently pigmented species: the coccolithophore Emiliania huxleyi, the diatom Thalassiosira pseudonana, and the cyanobacterium Synechococcus sp. Linear discriminant analysis (LDA) was used to determine a suitable set of linear discriminant functions for classification of these species over an optimal wavelength range. Multivariate optical elements (MOEs) were then designed to predict the linear discriminant scores for the same calibration spectra. The theoretical performance specifications of these MOEs are described.

A Novel Fast Mode Decision Algorithm for H.264/AVC Using Particle Swarm Optimization

Video coding plays an important role in human life especially in communications. H.264/AVC is a prominent video coding standard that has been used in a variety of applications due to its high efficiency comes from several new coding techniques. However, the extremely high encoding complexity hinders itself from real-time applications. This paper presents a new encoding algorithm that makes use of particle swarm optimization (PSO) to train discriminant functions for classification based fast mode decision. Experimental results show that the proposed algorithm can successfully reduce encoding time at the expense of negligible quality degradation and bitrate increases.

Object-Based Image Analysis of High-Resolution Satellite Images Using Modified Cloud Basis Function Neural Network and Probabilistic Relaxation Labeling Process

Object-based image analysis is quickly gaining acceptance among remote sensing community, and object-based image classification methods are increasingly being used for classification of land use/cover units from high-resolution satellite images with results closer to human interpretation compared to per-pixel classifiers. The problem of nonlinear separability of classes in a feature space consisting of spectral/spatial/textural features is addressed by kernel-based nonlinear mapping of the feature vectors. This facilitates use of linear discriminant functions for classification as used in artificial neural networks (ANNs). In this paper, performance of a recently introduced kernel called cloud basis function (CBF) is investigated with some modification for classification. The CBF has demonstrated superior performance to the tune of about 4% higher classification accuracy compared to conventional radial basis function used in ANN. The results are further improved by using probabilistic relaxation labeling as a postprocessing step. This paper has potential applications in urban planning and urban studies.

Technical note: Estimating sex using cervical canine odontometrics: A test using a known sex sample

The size of the permanent human canine tooth is one of the few sexually dimorphic features to be present in childhood and as such offers the opportunity to assist in the identification of sex in remains where no other appropriate criteria exist, such as in subadults. However, canine odontometrics are often associated with high levels of interobserver error and can be difficult to access if dentition is in situ. Additionally, appropriate points of measurement can be difficult to identify if the tooth is worn. Alternate measurements of the cervical canine diameters have been proposed as solutions to these issues, but the utility of these measurements in estimating sex has not been conclusively demonstrated. This study uses the buccolingual and mesiodistal cervical diameter of the canines from a known-sex sample from St. Bride's Church, London and a partially known-sex sample from the Old Church, Chelsea, London to classify individuals as male or female. A discriminant function classification using these diameters successfully identifies sex in 93.8% of the known-sex assemblage and 95% of the partially osteologically estimated sex assemblage. It is suggested that cervical canine diameters are highly repeatable measurements with low interobserver error, can be obtained on worn and in situ teeth, and provide as good or better guidance on estimating sex in human remains as standard maximal diameters.

Fault Diagnosis of Rotating Machinery Based on Empirical Mode Decomposition and Fractal Feature Parameter Classification

Rolling bearings are vital elements in rotating machinery and vibration signal is a kind of effective mean to characterize the status of rolling bearing fault. This paper presents a novel intelligent method for fault diagnosis based on empirical mode decomposition, fractal feature parameter extracting and orthogonal quadratic discriminant function classifier. The new method consists of three steps. Firstly, with investigating the feature of impact fault in vibration signals, the raw vibration signals are decomposed into intrinsic mode functions by empirical mode decomposition. Secondly, using the method of time sequences fractal dimension calculating, fractal feature parameters are extracted from intrinsic mode functions. Then, each raw signal sample has a feature set. Finally, training set and testing set are inputted into the orthogonal quadratic discriminant function model in the classification phase to identify different abnormal cases. The proposed method is applied to the fault diagnosis of rolling element bearing, and the test results indicate that the novel intelligent diagnosis method is sensitive to fault severity and capable of fault detection and fault diagnosis.