Discriminant Analysis Feature Extraction Research Articles

Efficient Iris Image Recognition System Based on Machine Learning Approach

Iris identification is a well‐known technology used to detect striking biometric identification techniques for recognizing human beings based on physical behavior. The texture of the iris is essential and its anatomy varies from individual to individual. Humans have distinctive physical characteristics that never change. This has resulted in a considerable advancement in the field of iris identification, which inherits the random variation of the data and is often a dependable technological area. This research proposes three algorithms to examine the classifications in machine learning approaches using feature extraction for the iris image. The applied recognition system used many methods to enhance the input images for iris recognition using the Multimedia University (MMU) database. Linear Discriminant Analysis (LDA) feature extraction method is applied as an input of three algorithms of machine learning approaches that are OneR, J48, and JRip classifiers. The result indicates that the OneR classifier with LDA achieves the highest performance with 94.387 % accuracy, while J48 and JRip reached to 90.151% and 86.885% respectively. Index Terms— Iris recognition, LDA, OneR, J48, JRip.

Principal component and linear discriminant analyses for the classification of hominoid primate specimens based on bone shape data.

In this study, we tested the hypothesis that machine learning methods can accurately classify extant primates based on triquetrum shape data. We then used this classification tool to observe the affinities between extant primates and fossil hominoids. We assessed the discrimination accuracy for an unsupervised and supervised learning pipeline, i.e. with principal component analysis (PCA) and linear discriminant analysis (LDA) feature extraction, when tasked with the classification of extant primates. The trained algorithm is used to classify a sample of known fossil hominoids. For the visualization, PCA and uniform manifold approximation and projection (UMAP) are used. The results show that the discriminant function correctly classified the extant specimens with an F1-score of 0.90 for both PCA and LDA. In addition, the classification of fossil hominoids reflects taxonomy and locomotor behaviour reported in literature. This classification based on shape data using PCA and LDA is a powerful tool that can discriminate between the triquetrum shape of extant primates with high accuracy and quantitatively compare fossil and extant morphology. It can be used to support taxonomic differentiation and aid the further interpretation of fossil remains. Further testing is necessary by including other bones and more species and specimens per species extinct primates.

A robust geometric mean-based subspace discriminant analysis feature extraction approach for image set classification

Discriminant analysis technique is an important research topic in image set classification because it can extract discriminative features. However, most existing discriminant analysis methods almost fail to work for feature extraction of data because there is only a small amount of valid discriminant information. The main weakness of most existing discriminant analysis models is that the class mean vector is constructed by using class sample average. That is not sufficient to provide an accurate estimation of the class mean. In this paper, we proposed a robust geometric mean-based subspace discriminant analysis feature extraction method for image set classification. This method is a combination of geometric mean vector, subspace and covariance matrix to jointly represent an image set because they are contain different discriminative information. The above three representation ways lie on different spaces. To reduce the dissimilarity between the heterogeneous spaces, a robust geometric mean-based subspace discriminant analysis learning (GMSDA) framework is developed, which includes three steps operation. At first, uses a new geometric mean vector to construct geometric between-class scatter matrix (Sgb) and geometric within-class scatter matrix (Sgw) instead of traditional mean vector of many methods. Secondly, maximizes the geometric between-class scatter matrix to increase the difference between extracted features. Finally, maximizes the subspace between-class scatter (SgbS) and minimizes the subspace within-class scatter (SgwS) simultaneously in the subspace of original space. Experiments on five datasets illustrate that our proposed GMSDA method works the best in small sample size situation by using maximum likelihood classification (MLC).

Dimensionality reduction based on binary encoding for hyperspectral data

ABSTRACTBinary encoding is an approach that aims at summarizing the information contained in various spectral bands into a single image that stores the meaningful information of the bands. In this paper, it is introduced a feature extraction approach to reduce the dimensionality of hyperspectral data with binary encoding for classification purposes. Different options to reduce the radiometric information of the pixels are introduced, such as using a single threshold or multiple thresholds. After the dimensionality reduction, the separation of the spectral classes was analysed and the thematic classification of the reduced data was performed. In order to evaluate the performance of the proposed approach, experiments on AVIRIS (Airborne Visible/Infrared Imaging Spectrometer) image, ROSIS (Reflection Optics System Imaging Spectrometer) hyperspectral image and HYDICE (Hyperspectral Digital Imagery Collection Experiment) hyperspectral image are presented. In the experiments, neighbouring spectral bands are grouped and coded and the results of the classification are compared. The results show that the use of binary encoding based on three thresholds by spectral region is more efficient than with the use of one threshold. The thematic mapping of the hyperspectral data with reduced dimension confirms the competitiveness of the binary encoding method compared with other dimension reduction methods, such as the Principal Component Analysis (PCA), the Principal Component Analysis – Fisher’s Linear Discriminant Analysis (PCA-LDA), the Discriminant Analysis Feature Extraction (DAFE) and the Non-parametric Weighted Feature Extraction (NWFE). In this context, the present methodology shows to be promising, because it reduces the computational complexity and improves performance.

An unsupervised feature extraction method based on band correlation clustering for hyperspectral image classification using limited training samples

ABSTRACTDue to the advancement of sensor technology, hyperspectral sensors are now able to record reflectance of the earth’s surface in hundreds of bands. This makes hyperspectral images a rich source for a diverse range of applications, including classification tasks. Based on Hughes phenomenon, classification accuracy decreases as dimensionality increases while a limited number of training samples are within reach. In a hyperspectral image adjacent and non-adjacent bands may have a high correlation; therefore removing this redundancy reduces the amount of data for further analysis including classification. Extraction of a few features which lead to high classification accuracy is a need. In this article, we propose an unsupervised feature extraction method called band correlation clustering (BCC). The proposed method includes three main steps: 1) Calculation of the bands’ correlation coefficient and generating correlation coefficient matrix, 2) Grouping the bands based on correlation coefficient matrix using the k-means clustering algorithm, and 3) Mean calculation of each cluster as a newly extracted feature. The proposed method is evaluated by two different criteria: classification accuracy and time consumption. For performance evaluation, the extracted features are fed to two supervised classifiers of non-parametric SVM and parametric ML. The results are compared with the ones obtained by unsupervised extracted features using principle component analysis (PCA), independent component analysis (ICA), and minimum noise fraction (MNF) and supervised ones using generalized discriminant analysis (GDA) and clustering-based feature extraction (CBFE). Comparison of the results shows promising performance for proposed BCC both in terms of computational costs and classification accuracy.

Classification of Hyperspectral Images with Robust Regularized Block Low-Rank Discriminant Analysis

Classification of Hyperspectral Images (HSIs) has gained attention for the past few decades. In remote sensing image classification, the labeled samples are insufficient or hard to obtain; however, the unlabeled ones are frequently rich and of a vast number. When there are no sufficient labeled samples, overfitting may occur. To resolve the overfitting issue, in this present work, we proposed a novel approach for HSI feature extraction, called robust regularized Block Low-Rank Discriminant Analysis (BLRDA), which is a robust and efficient feature extraction method to improve the HSIs’ classification accuracy with few labeled samples. To reduce the exponentially growing computational complexity of the low-rank method, we divide the entire image into blocks and implement the low-rank representation for each block respectively. Due to the symmetric matrix requirements for the regularized graph of discriminant analysis, the k-nearest neighbor is applied to handle the whole low-rank graph integrally. The low-rank representation and the kNN can maximally capture and preserve the global and local geometry of the data, respectively, and the performance of regularized discriminant analysis feature extraction can be apparently improved. Extensive experiments on multi-class hyperspectral images show that the proposed BLRDA is a very robust and efficient feature extraction method. Even with simple supervised and semi-supervised classifiers (nearest neighbor and SVM) and randomly given parameters, the feature extraction method achieves significant results with few labeled samples, which shows better performance than similar feature extraction methods.

Voice Recognition Application by Using Fisher’s Linear Discriminant Analysis (FLDA) Feature Extraction

Voice Recognition Application by Using Fisher’s Linear Discriminant Analysis (FLDA) Feature Extraction

A New Feature Extraction Method Based on the Information Fusion of Entropy Matrix and Covariance Matrix and Its Application in Face Recognition

The classic principal components analysis (PCA), kernel PCA (KPCA) and linear discriminant analysis (LDA) feature extraction methods evaluate the importance of components according to their covariance contribution, not considering the entropy contribution, which is important supplementary information for the covariance. To further improve the covariance-based methods such as PCA (or KPCA), this paper firstly proposed an entropy matrix to load the uncertainty information of random variables similar to the covariance matrix loading the variation information in PCA. Then an entropy-difference matrix was used as a weighting matrix for transforming the original training images. This entropy-difference weighting (EW) matrix not only made good use of the local information of the training samples, contrast to the global method of PCA, but also considered the category information similar to LDA idea. Then the EW method was integrated with PCA (or KPCA), to form new feature extracting method. The new method was used for face recognition with the nearest neighbor classifier. The experimental results based on the ORL and Yale databases showed that the proposed method with proper threshold parameters reached higher recognition rates than the usual PCA (or KPCA) methods.

Classification of hyperspectral data using extended attribute profiles based on supervised and unsupervised feature extraction techniques

The classification of remote sensing data based on the exploitation of spatial features extracted with morphological and attribute profiles has been recently gaining importance. With the development of efficient algorithms to construct the profiles for large datasets, such methods are becoming even more relevant. When dealing with hyperspectral imagery, the profiles are traditionally built on the first few principal components computed from the data. However, it needs to be determined if other feature reduction approaches are better suited to create base images for the profiles. In this article, we explore the use of profiles based on features derived from three supervised feature extraction techniques (i.e. Discriminant Analysis Feature Extraction, Decision Boundary Feature Extraction and Non-parametric Weighted Feature Extraction) and two unsupervised feature-extraction techniques (i.e. Principal Component Analysis (PCA) and Kernel PCA) in classification and compare the classification accuracies obtained by using different techniques for two different classification methods. The obtained results indicate significant improvements in the accuracies using the supervised feature extraction methods. However, the choice of the method affects the quality of the results for different datasets depending on the availability of the training samples.

CLASSIFICATION OF ROOF MATERIALS USING HYPERSPECTRAL DATA

Abstract. Mapping of surface materials in urban areas using aerial imagery is a challenging task. This is because there are numerous materials present in relatively small regions. Hyperspectral data features a fine spectral resolution and thus has a significant capability for automatic identification and mapping of urban surface materials. In this study an approach for identification of roof surface materials using hyperspectral data is presented. The study is based on an urban area in Ludwigsburg, Germany, using a HyMap data set recorded during the HyMap campaign in August, 2010. Automatisierte Liegenschaftskarte (ALK) vector data with a building layer is combined with the HyMap data to limit the analysis to roofs. A spectral library for roofs is compiled based on field and image measurements. In the roof material identification process, supervised classification methods, namely spectral angle mapper and spectral information divergence and the object oriented ECHO (extraction and classification of homogeneous objects) approach are compared. In addition to the overall shape of spectral curves, position and strength of absorptions features are used to enhance material identification. The discriminant analysis feature extraction method is applied to the HyMap data in order to identify features (band combinations) suitable for discriminating between the target classes. The identified optimal features are used to create a new data set which is later classified using the ECHO classifier. The classification results with respect to material types of roofs are presented in this study. The most important results are evaluated using orthophotos, probability maps and field visits.

A New Weighted Fuzzy C-Means Clustering Algorithm for Remotely Sensed Image Classification

Fuzzy clustering model is an essential tool to find the proper cluster structure of given data sets in pattern and image classification. In this paper, a new weighted fuzzy C-Means (NW-FCM) algorithm is proposed to improve the performance of both FCM and FWCM models for high-dimensional multiclass pattern recognition problems. The methodology used in NW-FCM is the concept of weighted mean from the nonparametric weighted feature extraction (NWFE) and cluster mean from discriminant analysis feature extraction (DAFE). These two concepts are combined in NW-FCM for unsupervised clustering. The main features of NW-FCM, when compared to FCM, are the inclusion of the weighted mean to increase the accuracy, and, when compared to FWCM, the centroid of each cluster is included to increase the stability. The motivation of this work is to meliorate the well-known fuzzy C-Means algorithm (FCM) and a recently proposed fuzzy weighted C-Means algorithm (FWCM). Our finding is that the proposed algorithm gives greater classification accuracy and stability than that of FCM and FWCM. Experimental results on both synthetic and real data demonstrate that the proposed clustering algorithm will generate better clustering results than those of FCM and FWCM algorithms, in particularly for hyperspectral images.

Object-oriented subspace analysis for airborne hyperspectral remote sensing imagery

An object-oriented mapping approach based on subspace analysis of airborne hyperspectral images was investigated in this paper. Hyperspectral features were extracted based on subspace learning approaches, in order to reduce the redundancy of spectral space and extract the characteristic images for the further object-oriented classification. In this paper, three kinds of spectral feature extraction (FE) methods were utilized to obtain the subspace of airborne hyperspectral data: (1) unsupervised FE, such as PCA (principal component analysis), ICA (independent component analysis) and MNF (maximum noise fraction); (2) supervised FE, e.g. DBFE (decision boundary feature extraction), DAFE (discriminant analysis feature extraction) and NWFE (nonparametric weighted feature extraction); and (3) linear mixture analysis. Afterwards, the extracted subspace features were fed into the object-based classification system. The FNEA (fractal net evolution approach) was utilized to extract objects from the subspace images and SVM (support vector machines) was then used to classify the object-based features. Experiments were conducted on two airborne hyperspectral datasets: (1) the AVIRIS dataset over the northwest Indiana's Pine with 220 spectral bands (agricultural region), and (2) the ROSIS dataset over Pavia University, northern of Italy with 102 spectral bands (urban region). Results revealed that the proposed object-based approach could give significantly higher accuracies than the traditional pixel-based subspace classification.

Classification and feature extraction for remote sensing images from urban areas based on morphological transformations

Classification of panchromatic high-resolution data from urban areas using morphological and neural approaches is investigated. The proposed approach is based on three steps. First, the composition of geodesic opening and closing operations of different sizes is used in order to build a differential morphological profile that records image structural information. Although, the original panchromatic image only has one data channel, the use of the composition operations will give many additional channels, which may contain redundancies. Therefore, feature extraction or feature selection is applied in the second step. Both discriminant analysis feature extraction and decision boundary feature extraction are investigated in the second step along with a simple feature selection based on picking the largest indexes of the differential morphological profiles. Third, a neural network is used to classify the features from the second step. The proposed approach is applied in experiments on high-resolution Indian Remote Sensing 1C (IRS-1C) and IKONOS remote sensing data from urban areas. In experiments, the proposed method performs well in terms of classification accuracies. It is seen that relatively few features are needed to achieve the same classification accuracies as in the original feature space.

A regional scale soil mapping approach using integrated AVHRR and DEM data

There is an increasing need for reasonably accurate small-scale soil databases. The compilation of a continental or global-scale soil database requires a lot of spatially and thematically accurate soil data. The aim of this study was to test a method for small-scale soil mapping in Italy using Advanced Very High Resolution Radiometer (AVHRR) and digital elevation data. This method was employed in an earlier study in Hungary for a much smaller area and a significantly different soil-forming environment. An integrated, 45-layer AVHRR-terrain database was used for the study, including a digital elevation model (DEM), slope, curvature, aspect, potential drainage density, and the five bands of AVHRR data for eight different dates. The data were processed using the Discriminant Analysis Feature Extraction (DAFE) function, which is based on a canonical analysis procedure. Two types of images (basic and transformed) were classified using the maximum likelihood classifier. Two training sets were chosen that have identical geographic coverage, but differ in the level of soil classification. One set was based on the soil units (SU) of the FAO-revised legend, while the other set represented major soil groupings (MSG). The best 10, 15, 20, 25, 30, 35, 40 and 45 layers were selected using the Bhattachryya feature selection method and were then classified. The results of the different sets were compared. The performance of the purely AVHRR and purely terrain-data-based images, respectively, were also interpreted. The results indicate that the terrain descriptors alone are not sufficient for soil classification. However, the feature selection algorithms always selected the DEM and its derivatives among the first ones, highlighting their importance for soil-landscape characterization. When using AVHRR data alone, test class performances of 49.8 percent (MSG) and 48.6 percent (SU) were achieved. Integration of terrain data into the AVHRR database produced relatively small improvements (4.6 and 2.8 percent). The best test class performances were achieved when all available channels were used for the classification, namely 51.4 for the FAO's SUs and 54.4 for the MSGs on the basic image, and 51.7 and 54.4 respectively on the DAFE-transformed images. The most informative AVHRR bands were found to be from the spring period (April-May), while the most abundant bands were the visible-red (band 1) and bands 3 and 4.

Use of combined digital elevation model and satellite radiometric data for regional soil mapping

Previous reports demonstrated that data from air- and spaceborne sensors are appropriate for delineation of soil patterns. Also, many attempts have been made to use digital elevation model (DEM) for deriving soil information. However, little is known about the potential use of low spatial resolution satellite and digital elevation data in small-scale soil mapping. A case study was conducted to assess the use of integrated terrain and Advanced Very High Resolution Radiometer (AVHRR) databases for small-scale soil pattern delineation. The main objective was to test the effect of the addition of terrain descriptor data to the AVHRR data set on the classification results. Two database were used for the study. The first one was purely AVHRR data and contained the five basic AVHRR channels and the normalized difference vegetation index (NDVI) of five different dates, while in the second database the AVHRR data was complemented with a DEM, a curvature, a slope, an aspect and the potential drainage density layers. The performance of these two databases when employed to derive soil information was compared. These databases were then further processed using the Discriminant Analysis Feature Extraction (DAFE) function (which is based on a canonical analysis procedure), and were then classified using the Fisher linear discriminant, and the ECHO spectral–spatial classifiers. Based on the results, it was concluded that the two reflective bands, the middle infrared, the two thermal bands and the NDVI provided a relatively wide range of detectable soil information. The use of single images or small dimensional AVHRR data sets (less then 10 layers) does not result in acceptable performances, while the use of multispectral and multitemporal databases improved the classification performance very significantly. However, the purely AVHRR-based model could not always delineate soil variations related to terrain differences, and resulted in an overall classification performance of 49.1%. Digital elevation and terrain descriptor data were essential in the model for achieving acceptable results. In the second part of the study an integrated AVHRR-terrain database was used, where five terrain layers were added to the 30 AVHRR channels. Two different spatial resolutions were compared, 500 m and 1 km, respectively. The use of elevation, slope, aspect and curvature as differentiating criteria often lead to a satisfactory result in terrain characterization, particularly in large-scale mapping. However, with those variables extracted from DEM of physiographically complex areas, e.g., — where plain areas and mountainous/hilly regions occur together in the same study — often lose their ability to delineate soil variations of the level lands. Beyond these terrain descriptors we implemented a new function, called potential drainage density (PDD) to improve the performance of the model on the plain areas. The classification accuracy of the integrated AVHRR-terrain database was improved significantly over the case when only AVHRR data was in the model. The classification performances of the three different resolution images were 87.3% for the 500-m resolution image and 70.1% for the 1-km resolution image.