Reply on RC2

Deep learning technologies have been successfully applied to hyperspectral (HS) image classification with remarkable performance. However, compared with traditional machine learning methods, neural networks usually need more data. In remote sensing (RS) research, obtaining a large number of labeled HS data is very difficult and expensive work. Simultaneously, the distribution of feature information is bound to be unbalanced, and tends to conform to the long tail. At present, the neighborhood information of unlabeled samples is usually ignored in HS image classification tasks based on semi-supervised learning. In this letter, we propose a new semi-supervised long-tail learning framework based on spatial neighborhood information (SLN-SNI), which can complete the HS image classification task under unbalanced small sample data. Specifically, a new semi-supervised learning strategy is proposed. On this basis, a new method to determine the label of unlabeled samples based on spatial neighborhood information (SNI) is proposed. The coarse classification results divided into three situations are judged again, and the accuracy of pseudo labels is improved. The performance of the proposed method is tested on three public HS image datasets. Compared with the current advanced methods have achieved a certain improvement.

Traditional remote sensing image surveys rely on field observations, which have shortcomings such as low efficiency and limited coverage, making it difficult to meet the needs of urban planning and management. In response to this, this article conducts in-depth research on remote sensing image classification methods based on graph neural networks (GNN). This study adopts a joint framework of GAT (Graph Attention Network)+SLC (Simple Linear Iterative Clustering)+Hybrid Pooling. Use SLIC algorithm to perform superpixel segmentation on hyperspectral data and construct a graph structure that integrates spectral spatial features. Utilizing GAT's attention mechanism to enhance feature interaction among neighboring nodes, combined with a hybrid pooling strategy to balance feature dimensionality reduction and key information preservation. The experiment was conducted on the public datasets WHU-RS19 and UC Merced, using traditional CNN, random forest, and minimum distance methods as baseline models. The significance of the differences was verified through independent sample t-test. By expanding the sample coverage, optimizing the model regularization strategy, and continuously iterating training, the generalization performance of the model has been significantly improved. Therefore, from the comparison of classification performance, graph NN classification is superior to minimum distance classification. This method utilizes perspective remote sensing data to achieve accurate ground target classification and dynamic change information extraction, providing reliable data support for urban land survey and management.

Hyperspectral image classification based on combining texture and multiscale guided filter feature of principal component

Hyperspectral remote sensing image classification based on residual generative Adversarial Neural Networks

Lung transplantation is an established therapy in selected patients with advanced cystic fibrosis lung disease. Resumption of employment after lung transplantation is generally supported. In Switzerland, there are no data on long-term employment in people with cystic fibrosis undergoing lung transplantation. In a single-centre, cross-sectional study at a Swiss university hospital, clinical data from lung transplant recipients with cystic fibrosis, covering the transplantation period from January 1996 to December 2016, were analysed retrospectively. The potential influence of pre-lung transplantation factors (age, sex, lung function, body mass index, six-minute walk test distance, lung transplantation wait list time, paid employment on the wait list, education, relationship status, housing situation) and post-lung transplantation factors (chronic allograft dysfunction [CLAD], dialysis, cancer diagnosis [except skin cancer]) on paid employment and work percentage after lung transplantation were investigated using mixed logistic and linear regression models. Descriptive analyses of paid employment were performed for various periods after lung transplantation (<1, 1–3, 3–5, 5–10, >10 years). Data are reported as odds ratios (ORs) or coefficients (β) with their 95% confidence intervals (CIs). Eighty-four subjects (46.4% female) with a mean ± SD age of 29.9 ± 8.4 years were included in the study. Mean wait time for lung transplantation was 42.7 ± 40.2 weeks. The number (percentage) of subjects employed <1 year, 1–3 years, 3–5 years, 5–10 years and >10 years after lung transplantation was n = 23 (28%), n = 51 (65%), n = 44 (75%), n = 30 (68%) and n = 21 (75%), respectively. In mixed logistic regression models, pre-lung transplantation paid employment (OR 24.03, 95% CI 6.08 to 164.39, p <0.0001), academic education (OR 7.81, 95% CI 1.66 to 48.66, p = 0.01) and time post lung transplantation (on log scale, OR 5.81, 95% CI 3.15 to 12.78, p <0.0001) were the main factors influencing post-lung transplantation paid employment status. In mixed linear regression models, pre-lung transplantation paid employment (β = 21.40, 95% CI 10.98 to 31.81, p = 0.00014), academic education (β = 12.54, 95% CI 0.48 to 24.55, p = 0.05) and time post lung transplantation (on log scale, β = 8.96, 95% CI 6.17 to 11.82, p <0.0001) were the main factors influencing work percentage post lung transplantation. No evidence for an influence of clinical factors such as CLAD, cancer or dialysis on post-lung transplantation employment and work percentage was found. Pre-transplant employment is the dominant factor influencing lung transplantation employment in people with cystic fibrosis. People with cystic fibrosis undergoing lung transplantation should be encouraged to work for as long as their health status permits. Professional reintegration after successful lung transplantation should be supported by a multi-disciplinary lung transplant team.

Employing unlabeled data to improve the classification performance of SVM, and its application in audio event classification

Regularization has appeared explicitly in hyperspectral image (HSI) classification community, which serves as a promising paradigm for leveraging labeled and unlabeled information, computer's automation and user's interaction, spectral and spatial information, and so on. Graph-based regularization is capable of modeling the nonlinear structures embedded in high-dimensional space, with the great potential for HSI classification. However, traditional methods exhibit low capacity when facing noisy and large-scale data, thus posing a big challenge for their successful use in this community. In this paper, we present two novel sparse graph regularization methods, SGR and SGR with total variation (TV-SGR). In SGR, the labels of large unknown data are propagated based on the fraction matrix and the prediction function, where the fraction matrix is obtained using an effective sparse representation (SR) algorithm with respect to the dictionary, and the prediction function is estimated by optimizing a typical graph-based regularization problem. In contrast, TV-SGR is an extension of SGR by considering spatial information modeled by total variation in SR. Propagating the prediction function from dictionary to large unknown data using the fraction matrix is the essence of the paradigm. SGR and TV-SGR can be equipped with semisupervised learning, active learning, and spectral-spatial classification with large flexibility. The experimental results with two popular hyperspectral data sets indicate that the proposed methods outperform some state-of-the-art approaches in terms of computational efficacy, classification accuracy, and robustness to noise.

In this paper, we propose a novel semi-supervised learning classification framework using box-based smooth ordering and multiple 1D-embedding-based interpolation (M1DEI) in [J. Wang, Semi-supervised learning using multiple one-dimensional embedding-based adaptive interpolation, Int. J. Wavelets Multiresolut. Inf. Process. 14(2) (2016) 11 pp.] for hyperspectral images. Due to the lack of labeled samples, conventional supervised approaches cannot generally perform efficient enough. On the other hand, obtaining labeled samples for hyperspectral image classification is difficult, expensive and time-consuming, while unlabeled samples are easily available. The proposed method can effectively overcome the lack of labeled samples by introducing new labeled samples from unlabeled samples in a label boosting framework. Furthermore, the proposed method uses spatial information from the pixels in the neighborhood of the current pixel to better catch the features of hyperspectral image. The proposed idea is that, first, we extract the box (cube data) of each pixel from its neighborhood, then apply multiple 1D interpolation to construct the classifier. Experimental results on three hyperspectral data sets demonstrate that the proposed method is efficient, and outperforms recent popular semi-supervised methods in terms of accuracies.

A semi-supervised deep learning approach for cropped image detection

In this paper discrete wavelet transform (DWT) and empirical mode decomposition (EMD) are employed as a preprocessing stage in a multiclassifier and decision fusion system. The proposed method consists of three steps. In the first step, 2D-EMD is performed on each hyperspectral image band in order to obtain useful spatial information. Then, useful spectral information is obtained by applying the 1D-DWT to each signature of 2D-EMD performed bands. A novel feature set is generated using both spectral and spatial information. In the second step, each feature is independently classified by support vector machines (SVM), creating a multiclassifier system. In the last step, classification results are fused using a decision fusion criterion to produce one final classification. The proposed method improves overall classification accuracy over independent classifiers when reduced number of features are employed.

This paper proposes a novel face representation approach, local Gabor binary mapping pattern (LGBMP), for multi-view gender classification. In this approach, a face image is first represented as a series of Gabor magnitude pictures (GMP) by applying multi-scale and multi-orientation Gabor filters. Each GMP is then encoded as a LGBP image where a uniform local binary pattern (LBP) operator is used. After that, each LGBP image is divided into non-overlapping rectangular regions, from which spatial histograms are extracted. Although an LGBP feature vector can be obtained by fitting together the regional histograms, it can not be employed in pattern classification due to its high dimension. We propose that each regional LGBP feature be mapped onto a one-dimensional subspace independently before they are concatenated as a whole feature vector. This is attractive since we reduce the feature dimension and also preserve the spatial information of LGBP image. Two ways have been proposed to map the regional LGBP feature in this paper. One is so-called LGBMP-LDA using linear discriminant analysis (LDA) for dimensionality reduction while the other is to project the regional LGBP feature onto the class center connecting line, namely, LGBMP-CCL. As a result, despite several decades of Gabor filters, the final feature dimension is even less than that of the feature extracted by using LBP directly on gray-scale images. The classification tasks in our work are performed by support vector machines (SVM). The experimental results on the CAS-PEAL face database indicate that the proposed approach achieves higher accuracy than the others such as SVMs+Gray-scale pixel, SVMs+Gabor and SVMs+LBP approach, more particularly, it has the lowest dimension of feature vector.

Hyperspectral image classification using multi-feature fusion

A novel semi-supervised hyperspectral image classification approach based on spatial neighborhood information and classifier combination

The local binary pattern (LBP) and its variants have shown the effectiveness in texture images classification, face recognition, and other applications. However, most of these LBP methods only focus on the histogram of LBP patterns and ignore the spatial contextual information between LBP patterns. In this paper, we propose a 2D-LBP method which uses a sliding window to count the weighted occurrence number of the rotation invariant uniform LBP pattern pairs to obtain the spatial contextual information. The multi-resolution 2D-LBP features can also be obtained when the radius of 2D-LBP is changed. At last, a two-stage classifier which acts as an ensemble learning step is followed to achieve an accurate classification by combining the predictions on each 2D-LBP with single resolution. Theoretical proof shows that the proposed 2D-LBP is a general framework and can be integrated on other LBP variants to derive new feature extraction methods. Experimental results show that, the proposed method achieves 99.71%, 97.09%, 98.48%, and 49.00% classification accuracy on the public “Brodatz,” “CUReT,” “UIUC,” and “FMD” texture image databases, respectively. Compared with the original LBP and its variants, the proposed method obtains higher classification accuracy under different cases, and simultaneously owns shorter time complexity.

Graph neural network has excellent performance in obtaining the similarity relationship of samples, so it has been widely used in computer vision. But hyperspectral remote sensing image (HSI) has some problems such as data redundancy, noise, lack of labeled samples, and insufficient utilization of spatial information. These problems affect the accuracy of HSI classification using graph neural networks. To solve the above problems, this paper proposes graph-based semi-supervised learning with weighted features for HSI classification. The method proposed in this paper first uses the stacked autoencoder network to extract features, which is used to remove the redundancy of HSI data. Then, the similarity attenuation coefficient is introduced to improve the original feature weighting scheme. In this way, the contribution difference of adjacent pixels to the center pixel is reflected. Finally, to obtain more generalized spectral features, a shallow feature extraction mechanism is added to the stacked autoencoder network. And features that have good generalization can solve the problem of the lack of labeled samples. The experiment on three different types of datasets demonstrates that the proposed method in this paper can get better classification performance in the case of the scarcity of labeled samples than other classification methods.