Search

Subspace detection from high dimensional hyperspectral image (HSI) data cube has become an important area of research for efficient identification of ground objects. Standard feature extraction met...

The hyperspectral remote sensing images (HSIs) are acquired to encompass the essential information of land objects through contiguous narrow spectral wavelength bands. The classification accuracy i...

A hyperspectral image (HSI), which contains a number of contiguous and narrow spectral wavelength bands, is a valuable source of data for ground cover examinations. Classification using the entire original HSI suffers from the “curse of dimensionality” problem because (i) the image bands are highly correlated both spectrally and spatially, (ii) not every band can carry equal information, (iii) there is a lack of enough training samples for some classes, and (iv) the overall computational cost is high. Therefore, effective feature (band) reduction is necessary through feature extraction (FE) and/or feature selection (FS) for improving the classification in a cost-effective manner. Principal component analysis (PCA) is a frequently adopted unsupervised FE method in HSI classification. Nevertheless, its performance worsens when the dataset is noisy, and the computational cost becomes high. Consequently, this study first proposed an efficient FE approach using a normalized mutual information (NMI)-based band grouping strategy, where the classical PCA was applied to each band subgroup for intrinsic FE. Finally, the subspace of the most effective features was generated by the NMI-based minimum redundancy and maximum relevance (mRMR) FS criteria. The subspace of features was then classified using the kernel support vector machine. Two real HSIs collected by the AVIRIS and HYDICE sensors were used in an experiment. The experimental results demonstrated that the proposed feature reduction approach significantly improved the classification performance. It achieved the highest overall classification accuracy of 94.93% for the AVIRIS dataset and 99.026% for the HYDICE dataset. Moreover, the proposed approach reduced the computational cost compared with the studied methods.

The remote sensing hyperspectral images (HSIs) usually comprise many important information of the land covers capturing through a set of hundreds of narrow and contiguous spectral wavelength bands....

Hyperspectral image classification has become a very important area of research in recent years. However the classif cation become very challenging as the input dimension is very high. This is because the cost increases with the number of input features used to describe the pixel vectors. This problem can be addressed by reducing the irrelevant features for the task of classification. Moreover all the features are no equally important for classification. Principal Component Analysis (PCA) extracts feature however it depends on global variance. To overcome these limitations, a class-oriented subspace detection method is proposed which measures the relevancy of the selected subspace using Normalized Mutual Information (NMI) and Cross Cumulative Residual Entropy (CCRE). The application of NMI and CCRE over PCA images maximize the relevance as well as provided an uncorrelated subspace. Experimental analysis was performed to assess the effectiveness of the proposed method and the selected subspace was evaluated using the Support Vector Machine (SVM). Our proposed method achieves 87.61% and 88.57% classification accuracy respectively.

Hyperspectral Image (HSI) is a rich source of information for the analysis of the earth's surface. HSI produces a rich set of both spectral and spatial information for possible recognition of earth materials, minerals and vegetation categories. Since HSI has high dimensional spectral information so that, feature extraction methods has been used to reduce the dimensions. The most widely used feature extraction method Principal Component Analysis (PCA) is applied in HSI for dimension reduction. The aim of this paper is to analyze PCA and its different variants Segmented-PCA (SPCA), Folded-PCA (FPCA), and its nonlinear approach Kernel-PCA (KPCA) for effective feature extraction and classification of HSI. Moreover, the noise adjusted methods Minimum Noise fraction (MNF) and its variants segmented MNF is also studied for comparing the feature extraction methods. For comparing the robustness of the studied methods, two real HSI is used for the experiments. The experiments show that the classification accuracy of the MNF method are 95.94% and 97.61% for AVIRIS and HYDICE datasets respectively which outperforms that other PCA based methods.

This paper is concerned about the detection procedure of drowsiness that causes fatal road accidents leading to death. Increasing lack of awareness of traffic rules is doubling the number of accidents daily. However, detection and indication of driver fatigue is an active area of research. In this concussion, both inside and outside individuals of the car are victimized. However, prevention of drowsiness requires a technique to detect the deterioration legitimately along with a warning mechanism of the vehicle operator. Although the existing solutions are created using some distinctive methods, there are some problems such as costly sensors and handling of information. The objective of this research is to create an improved, innovative, cost efficient and real time strategy for solving this problem of blinking, yawn, and head bending. A pre-trained model by a histogram-oriented gradient (HOG) and a linear vector support machine (SVM) extracts the eye, nose and mouth position and assesses the eye aspect ratio (EAR), moutt opening ratio (MOR) and nose length ratio (NLR). These pieces of information are then compared with the value threshold adapted from the sleeping or drowsy face models aspect ratio data set.

Finding an informative subspace of the original in hyperspectral images has become very essential due to its comprehensive applications in ground object recognition. Information extraction from hyperspectral images is a challenging work on account of its high correlation among the image bands in both the spatial and spectral redundancy. A feature reduction approach combining both the feature extraction and feature selection is proposed in this paper. A combination of Minimum Noise Fraction (MNF) and Mutual Information (MI) is proposed to select the subspace of the original datacube with regard to achieve improved classification accuracy. In the proposed method, feature ranking is improved by scaling the mutual information to a specific range in order to avoid redundant features. The proposed technique (MNF-nMI) is tested on two hyperspectral images captured by NASA AVIRIS sensor and HYDICE sensor. The experimental results typically indicate the noticeable improvement pertaining to classification accuracy. The proposed technique shows the classification accuracy of 96.8% and 99.3% on AVIRIS and HYDICE hyperspectral data respectively which is greater than the conventional methods studied.

Subspace detection is an essential step which is used as a preprocessing for the task of hyperspectral image classification, and ground surface identification. An informative subspace can be obtained through feature extraction/feature selection or using both. This paper proposed an efficient subspace detection technique using a both segmented principal component analysis (SPCA) and normalized mutual information (NMI) measure. At first, the original dataset is partitioned into several groups using NMI measure and then perform the principal component transform (PCT) on each group. Finally, the NMI is utilized to select the most informative images to obtain a resultant subspace and this method is named as (SPCA-nMI). The proposed method is tested on two real hyperspectral images, the experimental results shows the superiority of the proposed approach and obtain 95.47% classification accuracy on dataset 1 and (99.026%) on dataset 2 which is best among the methods studied.

Hyperspectral images are extensively used, now-a-days, for governing several geo-spatial fields. Due to the high number of dimensions or spectral bands the classification accuracy demotes which is known as the “Hughes Phenomenon” or the “Curse of Dimensionality”. To overcome this obstacle dimensionality reduction approaches need to be performed or simply the number of the spectral bands needs to be reduced. As a preprocessing step feature extraction or feature selection can be performed that reduces the computational complexity of the hyperspectral data classification. Highly correlated features are omitted and only the informative ones are considered for the classification. In this paper, we considered the feature extraction approaches, namely, the Principal Component Analysis, the Linear Discriminant Analysis and both of them combined. We have applied these feature extraction methods on the dataset individually and then classified the dataset using Support Vector Machine classifier. The experimental results show that the LDA approach provides the best accuracy (86.53%) among the three dimension reduction techniques applied. All the files and codes used in our work can be found at https://github.com/joybiS31/multiclassSVM classification u sing PCA LDA/

In many computer vision based problems, multiview 3D object retrieval is very useful with many application possibilities. Actually multiview 3D object is represented by a set of different views of 2D images. There are many hand-crafted features extraction techniques. Rather than using them, deep embedding network and autoencoder are used to extract features and calculate Euclidean distance to measure the similarity. This paper emphasizes on the process of retrieving 3D object from multi-view 2D images. Two established deep learning based solutions are used to retrieve images of multi-view 3D object. Finally different evaluation metrics are used to compare image retrieval performance accuracy & compare computation time and space complexity for both autoencoder and deep embedding network techniques. Here, dimension reduction algorithms PCA & t-SNE are also used to interpret the retrieval results. The experimented results show that deep embedding network gained 98% accuracy & autoencoder gained 97% accuracy to retrieve multi-view 2D images using RGB-D dataset.

Information extraction from spatial big data faces challenges in data relevancy analysis and heterogeneous data modeling. When the interested targets are more than one, the relevant analysis is often compromised. In this paper, a one-class oriented approach for effective feature selection and classification of remote sensing images is proposed. Mutual information (MI) is used as the feature selection criterion to cope with a wide range of data types. Then a cluster space (CS) representation is applied to model multimodal data and classifies each target class in turn. Hyperspectral and LiDAR data sets were used in the experiments. The test results demonstrate the advantage in terms of classification accuracies by focusing on one class at a time as compared to considering all classes simultaneously in classification.

Achieving high classification accuracy in hyperspectral image classification is a challenging task. This problem can be addressed by reducing the irrelevant features for the task of classification. Principal Component Analysis (PCA) is a popular feature extraction technique but it depends solely on global variance which makes it limited for some application. To address this, a target class oriented feature reduction method is proposed which incorporates the normalized Mutual Information (NMI) over PCA images to maximize the relevance of the selected subspace. Experimental analysis is performed to assess the effectiveness of the proposed method and the selected subspace is evaluated using kernel Support Vector Machine (KSVM) classifier. The proposed approach can achieve 96.57%classification accuracy on real hyperspectral data which is better than the standard approaches studied.

Subspace detection for hyperspectral images is getting more interest now days because of the challenges of dealing with high dimensional feature space for reliable classification. The objective of supervised dimension reduction technique is to find a subspace of reliable features that preserves maximal information about the target objects. Principal Component Analysis and Mutual Information are two well-known feature extraction and feature selection method respectively, however, a combination of both could be a better approach which can significantly improve the feature reduction performances. In this paper, a hybrid approach is proposed which combines both the Principal Component Analysis (PCA) and Quadratic Mutual Information (QMI). The proposed method named (PCA-QMI) is tested on two real hyperspectral datasets and finally the classification accuracy is measured using kernel support vector machine classifier. The proposed method can detect subspace effectively and therefore the classification accuracy achieved is more than 99% which is better than the standard benchmark techniques.

Remote sensing hyperspectral image (HSI) contains important information of ground surface as a set of hundreds of narrow and contiguous spectral bands. For effective classification of hyperspectral images, feature reduction techniques through feature extraction and feature selection approaches are applied to improve the classification performance. Principal Component Analysis (PCA) is the widely used feature extraction method for dimensionality reduction. In this paper, PCA and its linear variants such as segmented-PCA (SPCA) and folded-PCA (FPCA) together with nonlinear variants kernel-PCA (KPCA) and Kernel Entropy Component Analysis (KECA) have been studied to effectively extract the features for classification task. The feature selection over the new transformed features was carried out using cumulative-variance accumulation based approach except for KECA that employs Renyi entropy based feature selection. The studied methods are compared using real hyperspectral image. The experimental result shows that the classification accuracy of KPCA (95.9245%) and KECA (95.6262%) outperforms FPCA (95.1292%). However, the FPCA provides the less space complexity.

Remote sensing hyperspectral image (HSI) retains significant information of ground surface which is actually acquired as a set of hundreds narrow and contiguous spectral bands. Though it is quite difficult to extract features from these bands, dimensionality reduction techniques through feature extraction and feature selection are used to improve the classification performance of the HSI. Principal Component Analysis (PCA) is the commonly adopted feature extraction technique for dimensionality reduction of HSI. However, PCA can be failure to extract local characteristics of the HSI due to considering global variance. Thus, segmented-PCA (SPCA) and folded-PCA (FPCA) are introduced to effectively extract the local structures in different ways. In this paper, feature extraction using FPCA, termed as segmented FPCA (SFPCA), has been improved through applying it on the highly correlated bands' segments of the real HSI rather than not applying on the whole dataset directly. The feature selection over the new transformed features was carried out using cumulative-variance accumulation based approach. The experimental result shows that the classification accuracy of SFPCA (95.6262%) outperforms conventional FPCA (95.1292%), SPCA (93.837%) and PCA (93.7376%). Moreover, it provides the least space complexity.

Remote sensing hyperspectral image (HSI) contains significant information of ground surface which is actually acquired as a set of immense narrow and contiguous spectral bands. Proper classification approach can only give us the required knowledge from the hundreds of bands of HSI. Though it is quite difficult to extract features from these bands, dimensionality reduction techniques through feature extraction and feature selection are used to improve the classification performance of the HSI. Principal Component Analysis (PCA) is usually adopted as unsupervised linear feature extraction method for feature reduction. However, PCA can be failure to extract local characteristics of the HSI due to considering global variance. Thus, segmented-PCA (SPCA) and folded-PCA (FPCA) are used to efficiently extract the local structures in different ways. In this paper, feature extraction using FPCA, termed as segmented FPCA (SFPCA), has been improved through applying it on the highly correlated bands' segments of the real HSI rather than not applying on the whole dataset directly. The effectiveness of SFPCA is additionally compared with the unsupervised nonlinear feature extraction methods kernel-PCA (KPCA) and Kernel Entropy Component Analysis (KECA). The experimental result shows that the classification accuracy of SFPCA (95.6262%) outperforms conventional FPCA (95.1292%), SPCA (93.837%) and PCA (93.7376%) providing the least space complexity. Moreover, it attempts to fix the nonlinearity like very similar to KPCA (95.9245%) and KECA (95.6262%).

Hyperspectral image analysis is becoming an important field of research interest because of its wide range of applications in ground surface identification. New technology is being developed to capture hyperspectral images to cover more spectral bands and finer spectral resolution but also increases challenges to process those images for high correlation among data and both the spectral and spatial redundancy. This paper proposed a feature mining approach for the relevant feature selection as well as efficient classification of the hyperspectral dataset. Principal Component analysis and Mutual Information is two widely used feature reduction techniques utilized in conjunction for the feature reduction of the remote sensing data set. The kernel support vector machine classifier is used to assess the effectiveness of the detected subspace for classification. The proposed feature mining approach is able to achieve 99.3% classification accuracy on real hyperspectral data which higher than the standard approaches studied.

Remote sensing hyperspectral images are blessings of technology through which the ground objects can be detected effectively with the cost of computer processing. For classification of hyperspectral images finding an effective subspace is very important to classify them efficiently. In recent years, many researchers have drawn their interest to extract data more effectively from hyperspectral dataset. In this research, an approach has been proposed to find the effective subspace by measuring the relevance of individual features through Cross Cumulative Residual Entropy from the Principal Component images. The Support Vector Machine has been used as the classifier for the assessment of the feature reduction performance. Experiment has been completed on real hyperspectral dataset and achieved 97% of accuracy which is better than the standard approaches studied.

The objective of this study is to develop a hybrid nonlinear subspace detection technique in which Kernel Principal Component Analysis (KPCA) is combined with a Closest Class Pair (CCP) measure for the task of hyperspectral image classification. In the proposed approach, KPCA is applied first to generate the new features from original dataset then the CCP is applied to rank the features that are able to separate the complex or overlapping classes. Finally, the two ranked scores such as KPCA and CCP are combined to select a subset of features which is relevant and able to provide better discrimination among the input classes of interest. Experiments are performed on a real hyperspectral image acquired by the NASA Airborne Visible Infrared Imaging Spectrometer (AVIRIS) sensor and it can be seen that the proposed approach obtained the best classification accuracy 84.58%.

In this study, a subspace detection technique is developed using a hybrid approach which combines both feature extraction and feature selection for the task of hyperspectral image classification. The proposed approach applies Kernel Principal Component Analysis (KPCA) at the first step, then feature selection from the KPCA images is accomplished by combining the KPCA score with a Jeffries-Matusita (JM) distance based ranking score. Experimental analysis is carried out on a hyperspectral image acquired by the AVIRIS sensor and the results show the advantage of the proposed approach in terms of classification accuracy in the tested cases.

Finding the most informative features from high dimensional space for reliable class data modeling is one of the most challenging problems in hyperspectral image classification. The problem can be address using two basic techniques: feature selection and feature extraction. One of the most popular feature extraction methods is Principal Component Analysis (PCA), however its components are not always suitable for classification. In this paper, we present a feature reduction method (MI-PCA) which uses a nonparametric mutual information (MI) measure on the components obtained via PCA. Supervised classification results using a hyperspectral data set confirm that the new MI-PCA technique provides better classification accuracy by selecting more relevant features than when using either PCA or MI on the original data.

Nowadays it is obvious that speakers can be identified from their voices. In this paper detail of speaker identification from the real-time system point of view is described. Firstly, it have been reviewed the well-known techniques used in speaker identification then the details of every step in identification process and explains the ideas, which leaded to these techniques. We start from the basic definitions used in DSP, then we move to the feature extraction step. Being widely used in pattern recognition tasks, neural networks have also been applied in speaker recognition. In this study, we developed a text-independent speaker identification system based on Back-propagation Neural Network (BPN). BPNs supply flexibility and straightforward design which make the system easily operable along with the successful classification results. In order to analyze the system in practice we made appropriate software and using real data we ran several tests. Empirical results show that proposed approach greatly improves identification speed in feature matching step. From the experiment it is found that the system correctly identify 96% of the speakers, using less then one second of test samples from each speaker.