Anti-cross validation technique for constructing and boosting random subspace neural network ensembles for hyperspectral image classification

Capturing images in thousands of contiguous spectral bands has been made simpler with the emergence of technology in the field of hyperspectral remote sensing. Despite of these huge data available for analysis, Hyperspectral images (HSI) face many challenges due to high dimensionality, noise, spectral mixing and computational complexity. Several preprocessing methods can be used to overcome the above mentioned issues. In this paper, an enhancement technique using 2D-Empirical Wavelet Transform (EWT) is used as a preprocessing step for the HSI reconstruction prior to sparsity based classification (Subspace Pursuit and Orthogonal Matching Pursuit). The effectiveness of the proposed method is proved by comparing the classification results obtained with and without applying preprocessing. Experimental analysis shows a significant improvement in the classification accuracies i.e., for 40 $$\%$$ of training samples, OMP shows an improvement in overall classification accuracy from 66.12 $$\%$$ to 93.20 $$\%$$ and SP shows an improvement from 66.36 $$\%$$ to 92.74 $$\%$$ .

In land-use classification of hyperspectral remote sensing (RS) images, traditional classification methods often experience large amount of datasets and low efficiency. To solve these problems, a fast machine-learning method, the extreme learning machine (ELM) algorithm, was introduced. However, basic use of the ELM usually encounters problems of unstable classification results and low classification accuracy. Hence, in this paper, optimization methods for ELM-based RS image classification were mainly discussed and applied to solve the bottleneck problems. From the three perspectives of ensemble learning, making full use of image texture features, and deep learning, three classification optimization methods have been designed and implemented. The results show that: 1) To some extent, all the three methods can achieve a balance between classification accuracy and efficiency, i.e., they can maintain the advantage of ELM algorithm in classification efficiency and speed while have better classification accuracy; 2) The image texture feature optimization method (LBP-KELM) solves the problem of unsatisfactory classification results experienced by the ensemble learning optimization method (Ensemble-ELM) and further improves classification accuracy. However, the classification results are sensitive to the type of dataset; and 3) Fortunately, the optimization method combined with deep learning (CNN-ELM) can meet the application needs of multiple datasets. Furthermore, it can also further improve classification accuracy.

Hyperspectral images contain information from a wider range of the electromagnetic spectrum than natural images which gives them potential for better classification ability. However, hyperspectral datasets are typically small due to the expensive equipment needed to obtain the images, which can limit classification performance. One solution to this problem is transfer learning, in which a model trained on one dataset is reused for a separate dataset. Research has shown that transfer learning between hyperspectral datasets can give improved performance over models without transfer learning when training data are limited. Since extra hyperspectral data are not always available, the solution proposed here is to instead use networks pretrained on natural image (i.e., red, blue, green, or RGB) datasets for transfer learning. By using various feature selection and feature extraction methods, extracted hyperspectral samples are transformed into a three-channel format to imitate an RGB image and are used for fine tuning the well-known ResNet, DenseNet, and VGG networks. Feature extraction methods include techniques like principal component analysis, which create lower dimensional features from high dimensional spectral data. Alternatively, feature selection methods aim to find the best set of existing channels to use for classification. Experimental results are obtained using two well-known hyperspectral datasets, showing 73.6% accuracy on Pavia University and 82.8% accuracy on Salinas with 25 training samples per class. Additional ensemble methods are implemented that utilize multiple networks and show an increase in accuracy of 4.4% and 3% for Pavia University and Salinas, respectively. These results demonstrate that networks pretrained on RGB datasets are suitable for transfer learning with hyperspectral image datasets and can achieve desirable performance given the proper preprocessing technique.

Deep learning-based methods have demonstrated significant breakthroughs in the application of hyperspectral image (HSI) classification. However, some challenging issues still exist, such as the overfitting problem caused by the limitation of training size with high-dimensional feature and the efficiency of spectral–spatial (SS) exploitation. Therefore, to efficiently model the relative position of samples within the generative adversarial network (GAN) setting, we proposed a dual-channel SS fusion capsule generative adversarial network (DcCapsGAN) for HSI classification. Dual channels (1-D-CapsGAN and 2-D-CapsGAN) are constructed by integrating the capsule network (CapsNet) with GAN for eliminating the mode collapse and gradient disappearance problem caused by traditional GAN. Meanwhile, octave convolution and multiscale convolution are integrated into the proposed model for further reducing the parameters of the CapsNet and extracting multiscale features. To further boost the classification performance, the SS channel fusion model is constructed to composite and switch the feature information of different channels, thereby facilitating the accuracy and robustness of the whole classification performance. Three commonly used HSI data sets are utilized to investigate the performance of the proposed DcCapsGAN model, and the performance of the experiment demonstrates that the proposed model can efficiently improve the classification accuracy and performance.

The COVID-19 pandemic has necessitated the development of accurate and efficient classification models for diagnosis and prognosis. While deep learning has shown promising results in various medical applications, its combination with ensemble methods, which amalgamate the predictions of multiple classifiers, can further enhance the model's accuracy. This research paper introduces a novel approach called Deep Neural Ensemble Classification (DNEC) to tackle the challenge of Developing an enhanced ensemble model using deep learning algorithms and comparing its performance with existing ensemble methods. The research problem stems from the literature gap, where existing studies primarily focus on single-model approaches, lacking in-depth exploration of ensemble methods for COVID-19 classification. Motivated by the potential improvement in classification accuracy through ensemble methods, this study aims to create a deep neural ensemble classification model to improve the classification accuracy tailored for COVID-19 data. A set of diverse classifiers, including k-nearest neighbour (IBK), decision tree (J48), naïve bayes (NB), support vector machine (SVM), and sequential minimal optimization (SMO), are utilized in the proposed ensemble method. The accuracy improvement of the ensemble classifiers is evaluated using various metrics such as precision, recall, F1 score, confusion matrix, and processing time. The proposed method demonstrates that IBK+SVM+NB emerges as the top-performing deep neural ensemble classifier with an accuracy score of 99.29% and a total run time of 27.61 seconds. The innovative ensemble techniques introduced in this research contribute to the existing body of knowledge by filling the identified literature gap and offering a novel and highly accurate approach for COVID-19 classification.

Breast cancer remains one of the most common cancers amongst women globally. Early detection is crucial for improving survival rates. While mammography is widely used and an effective imaging technique, it can sometimes yield false positive or false negatives. Mammogram interpretation is highly operator-dependent, introducing variability and the potential for diagnostic errors. Additionally, mammographic images have limitations, such as low contrast in breast tissue and overlapping structures that can obscure lesions or mimic abnormalities. These limitations can lead to unnecessary biopsies or delayed diagnosis. These challenges highlight the needs for advanced and data driven diagnostic tools to support and enhance current screening practices.</p> <p>In recent years, machine learning has emerged as a transformative force in the health care industry offering innovative solution to complex medical challenges. One of the most promising applications of this technology is in the early and accurate detection of breast cancer particularly through the mammography image analysis. Machine learning algorithms, especially deep learning models, have shown significant potential in identifying patterns in mammography that may indicate abnormal cases. This study utilizes the power of deep learning particularly Convolutional Neural Networks (CNNs) to enhance the classification of breast ultrasound images.</p> <p>This research proposes a novel two-level classification approach. At the first level classification, the system distinguishes normal from abnormal cases. Next, the second level classification, it further classifies abnormal cases into benign or malignant tumors. To address challenges such as limited and imbalanced datasets, the study incorporates comprehensive preprocessing techniques including contrast enhancement via Contrast Limited Adaptive Histogram Equalization (CLAHE) and region of interest (ROI). The preprocessing step were evaluated separately and in combination to assess their individual contributions. Additionally, data augmentation is utilized to improve model generalization. To further enhance performance and robustness, multiple CNN architectures are combined using an ensemble mechanism. Soft voting and hard voting leverage the strengths of diverse CNNs reducing individual model biases and improving overall accuracy.</p> <p>The main contributions of this research are: (1) the design of a two-level classification system (2) the utilization of multiple preprocessing techniques to improve image quality, (3) the application of voting ensemble learning to improve predictive stability and accuracy, and (4) an extensive evaluation on the BUSI dataset. This breaks down the complex diagnostic task into smaller, manageable steps to allow for more accurate comparisons, analysis, and improvements at each stage.</p> <p>The first preprocessing step involved the application of CLAHE. In the first level classification (normal vs abnormal), performance of various models has been evaluated. The models achieved an 82.05% accuracy for AlexNet, 84.62% for DenseNet, 85.26% for Inception, 82.69% for MobileNet, 82.05% for ResNet and 82.69% for VGG16. The accuracy of soft voting is 96.20% and accuracy of hard voting 92.09%. In the second level classification (benign vs malignant), the models achieved an 75.00% accuracy for AlexNet, 80.47% for DenseNet, 35.94% for Inception, 41.41% for MobileNet, 35.94% for ResNet and 82.81% for VGG16. The accuracy of soft voting is 90.84% and hard voting is 85.76%.</p> <p>The second preprocessing step involved the ROI was implemented. This step helped isolate the regions containing potential abnormalities, such as tumors, while filtering out irrelevant areas. After the application of ROI (normal vs abnormal), the system evaluated the performance of various models. In the first level classification (normal vs abnormal), the models achieved an 82.69% accuracy for AlexNet, 82.69% for DenseNet, 83.97% for Inception, 87.18% for MobileNet, 82.69% for ResNet and 82.05% for VGG16. The accuracy of soft voting is 100.00% and hard voting is 98.08%. In the second level classification (benign vs malignant), the system evaluated the performance of various models. The models achieved an 75.00% accuracy for AlexNet, 80.47% for DenseNet, 35.94% for Inception, 41.41% for MobileNet, 35.94% for ResNet and 82.81% for VGG16. Soft voting achieved 89.15% and hard voting achieved 80.95%.</p> <p>The final step in the preprocessing stage combines both CLAHE and ROI. The combination led to an improvement in classification performance. In the first level classification (normal vs abnormal), the models achieved an 91.77% accuracy for AlexNet, 92.72% for DenseNet, 95.57% for Inception, 96.52% for MobileNet, 76.90% for ResNet and 91.14% for VGG16. The accuracy of soft voting is 90.88% and hard voting is 93.99%. In the second level classification (benign vs malignant), the system evaluated the performance of various models. The models achieved an 87.69% accuracy for AlexNet, 99.23% for DenseNet, 96.92% for Inception, 93.08% for MobileNet, 94.62% for ResNet and 97.69% for VGG16. Soft voting achieved 96.92% and hard voting achieved 97.69%.</p> <p>These results indicate that every preprocessing step contributed to an improvement in classification accuracy across all models. This study demonstrates that a two-level, ensemble-based deep learning system can significantly improve the diagnostic accuracy of breast ultrasound images, offering a promising tool to assist radiologists and support early detection efforts, particularly in resource-constrained healthcare environments.

There is a relative paucity of studies that seek to nonparametrically describe or categorize hyperspectral image (HSI) data in the most natural context within which those data are best represented: n-dimensional (n-D) hyperspace, where n is the number of bands in an HSI data cube. Statistical techniques-and specifically those based on second-order statistics-are predominant and generally do not provide (nor are they utilized to provide or 'leave behind') an understanding of the distribution of spectra in hyperspace for the purpose of discovering patterns, comparing data sets, developing a deeper understanding of HSI data, etc. With the terabytes of HSI data in archives today representing a huge diversity of geographical settings, it seems natural to enquire as to whether there are universal patterns among HSI data sets. If such patterns exist, new techniques may be developed to exploit them, e.g., to enhance information extraction speed and accuracy. Since an understanding of the nature of HSI data has rarely gone beyond cursory, visual assessments of geographical setting; i.e., assessments of whether the data are of desert, forest, urban, littoral, etc., environments, it is perhaps worthwhile to develop methods to probe n-D hyperspace with the intent to identify and describe universal data commonalities, differences, classifications, and patterns. Building upon earlier work (Resmini, 2003 [1], 2006 [2]), an algorithm was created to describe the volume (and label the individual volume elements) of hyperspace occupied by an HSI data set. The algorithm essentially discretizes hyperspace assigning addresses to boxes in which data points (i.e., spectra) reside. The algorithm has been applied to visible/near-infrared to shortwave infrared (VNIR/SWIR) and longwave infrared (LWIR) HSI data cubes. Implemented in C code, the algorithm may also be used to calculate the fractal dimension of any array of points in any n-D hyperspace because it is essentially an implementation of box counting. The algorithm is described as are results of its application to actual HSI data sets. The fractal dimension of HSI data derived from the algorithm are also presented and compared to principal componentsbased estimates of data dimensionality. Application of the algorithm for developing a deeper understanding of the nature of HSI data as viewed as points in hyperspace is discussed as are practical applications for 'traditional' HSI exploitation activities such as anomaly and target detection.

Since hyperspectral images (HSI) captured by different sensors often contain different number of bands, but most of the convolutional neural networks (CNN) require a fixed-size input, the generalization capability of deep CNNs to use heterogeneous input to achieve better classification performance has become a research focus. For classification tasks with limited labeled samples, the training strategy of feeding CNNs with sample-pairs instead of single sample has proven to be an efficient approach. Following this strategy, we propose a Siamese CNN with three-dimensional (3D) adaptive spatial-spectral pyramid pooling (ASSP) layer, called ASSP-SCNN, that takes as input 3D sample-pair with varying size and can easily be transferred to another HSI dataset regardless of the number of spectral bands. The 3D ASSP layer can also extract different levels of 3D information to improve the classification performance of the equipped CNN. To evaluate the classification and generalization performance of ASSP-SCNN, our experiments consist of two parts: the experiments of ASSP-SCNN without pre-training and the experiments of ASSP-SCNN-based transfer learning framework. Experimental results on three HSI datasets demonstrate that both ASSP-SCNN without pre-training and transfer learning based on ASSP-SCNN achieve higher classification accuracies than several state-of-the-art CNN-based methods. Moreover, we also compare the performance of ASSP-SCNN on different transfer learning tasks, which further verifies that ASSP-SCNN has a strong generalization capability.

Despite the importance of data reduction as part of the processing of reflection-based classifications, this study represents one of the first in which the effects of both spatial and spectral data reductions on classification accuracies are quantified. Furthermore, the effects of approaches to data reduction were quantified for two separate classification methods, linear discriminant analysis (LDA) and support vector machine (SVM). As the model dataset, reflection data were acquired using a hyperspectral camera in 230 spectral channels from 401 to 879 nm (spectral resolution of 2.1 nm) from field pea (Pisum sativum) samples with and without internal pea weevil (Bruchus pisorum) infestation. We deployed five levels of spatial data reduction (binning) and eight levels of spectral data reduction (40 datasets). Forward stepwise LDA was used to select and include only spectral channels contributing the most to the separation of pixels from non-infested and infested field peas. Classification accuracies obtained with LDA and SVM were based on the classification of independent validation datasets. Overall, SVMs had significantly higher classification accuracies than LDAs (P < 0.01). There was a negative association between pixel resolution and classification accuracy, while spectral binning equivalent to up to 98% data reduction had negligible effect on classification accuracies. This study supports the potential use of reflection-based technologies in the quality control of food products with internal defects, and it highlights that spatial and spectral data reductions can (1) improve classification accuracies, (2) vastly decrease computer constraints, and (3) reduce analytical concerns associated with classifications of large and high-dimensional datasets.

Coupled adversarial learning for fusion classification of hyperspectral and LiDAR data

Ensembles of extreme learning machine (ELM) have been widely used for hyperspectral image classification. The previous studies have shown that the goal of ensemble learning is to train accurate but diverse component classifiers to improve the generalization performance. To approach this goal, this article proposes a novel framework to construct an ELM ensemble model. The proposed framework relies on multiview-based random rotation ensemble pruning (MVRR-EP) and has several novel features. First, to ensure that the subsets of spectral bands can sufficiently learn the target concept, the spectral bands are divided into multiviews by using correlation analysis. Second, random rotation, a new approach of space transformation, is introduced to transform each view into multiple coordinate spaces, which makes the component classifiers trained on the transformed spaces have great diversity. Third, an accuracy guided ensemble pruning strategy is designed for pruning the component classifiers with low complementarity, and consequently, the remaining component classifiers with high complementarity are combined to a construct ensemble classifier. These techniques guarantee that the component classifiers used to construct an ensemble classifier are accurate but diverse, which ultimately improves the performance of the ensemble classifier. To demonstrate the effectiveness of the proposed MVRR-EP, extensive experiments were carried out on four hyperspectral data sets. Experimental results verify that compared with other methods, the proposed method provides competitive results.

Remote sensing Hyperspectral Image (HSI) comprises significant information about the earth’s surface which is actually acquired by hundred of narrow and adjacent spectral bands. The intended performance of classification accuracy does not attain due to the volume of the original HSI dataset and the enormous quantity of spectral bands. As such, dimensionality reduction approaches using feature extraction and selection are typically adopted to enhance classification performance. The unsupervised Principal Component Analysis (PCA), as well as the supervised Linear Discriminant Analysis (LDA), are commonly used as linear feature extraction methods for feature subspace detection. However, due to considering the effects of global variation, both PCA and LDA fail to extract local characteristics of HSI. In this paper, we propose a segmented LDA-based (SLDA) feature extraction where we apply the LDA in a segmented way to extract better local characteristics as well as global characteristics from the HSI. Per-pixel classification using a Support Vector Machine (SVM) is applied to our proposed SLDA method, PCA, Segmented-PCA (SPCA), and LDA on the Indian Pines agricultural HSI dataset. The experimental results show that the overall classification performance of SLDA (90.60%) remarkably outperforms all the other investigated methods: PCA (85.55%), SPCA (86.96%), LDA (86.45%), and the complete original dataset without employing any feature reduction method (83.10%). The proposed SLDA also requires the least amount of space complexity in different implementation scenarios.

Predicting stock returns by classifier ensembles

Recently, graph convolutional network (GCN) has drawn increasing attention in hyperspectral image (HSI) classification, as it can process arbitrary non-Euclidean data. However, dynamic GCN that refines the graph heavily relies on the graph embedding in the previous layer, which will result in performance degradation when the embedding contains noise. In this letter, we propose a novel dual residual graph convolutional network (DRGCN) for HSI classification that integrates two adjacency matrices of dual GCN. In detail, one GCN applies a soft adjacency matrix to extract spatial features, the other utilizes the dynamic adjacency matrix to extract global context-aware features. Subsequently, the features extracted by dual GCN are fused to make full use of the complementary and correlated information among two graph representations. Moreover, we introduce residual learning to optimize graph convolutional layers during the training process, to alleviate the over-smoothing problem. The advantage of dual GCN is that it can extract robust and discriminative features from HSI. Extensive experiments on four HSI data sets, including Indian Pines, Pavia University, Salinas, and Houston University, demonstrate the effectiveness and superiority of our proposed DRGCN, even with small-sized training data.

Hyperspectral images belong to high-dimensional data having a lot of redundancy information when they are directly used to classification. Support vector machine (SVM) can be employed to map hyperspectral data to high dimensional space effectively and make them linearly separable. In this paper, spectral and spatial information of hyperspectral images were used to construct SVM kernel function respectively. This paper proposed a hyperspectral image classification method utilization spatial-spectral combined kernel SVM in order to improve classification accuracy. The proposed method was used to classify AVIRIS hyperspectral images. The results demonstrated that the proposed SVM method can achieve 96.13% overall accuracy for the single category classification and 84.81% overall accuracy for multi-class classification only using ten percent of the total samples as the training samples. That is to say, the proposed method can make full use of the spectral information and spatial information of hyperspectral data, and effectively distinguish different categories compared with the traditional SVM for classification.