Search

With the rapid development of converters in a variety of industrial fields, the fault diagnosis of power switching devices has become an important factor in ensuring the safe and reliable operation of related systems. In recent years, machine learning has performed well in many fault diagnosis tasks. The success of these advanced methods depends on sufficient marked samples for each fault type. However, in most industrial applications, it is expensive and difficult to collect fault samples, and the fault diagnosis model trained under the limited samples cannot meet the requirements of fault diagnosis accuracy. In order to solve this problem, this study proposes a few-shot learning method based on fault sample generation to realize the open-circuit fault diagnosis of IGBT in a three-phase PWM converter. This method is the deformation of the auto-encoder called the disturbance auto-encoder generation model. By designing the model structure and training algorithm constraints, the encoder learns the nonlinear transferable disturbance from the normal operating sample pairs. Then, the disturbance is applied to the decoder to synthesize new fault samples to realize the training of the fault diagnosis model with limited samples. The biggest advantage of this method is that it can achieve 95.90% fault diagnosis accuracy by only collecting the samples in the normal operating conditions of the target system. Finally, the feasibility and advantages of the proposed method are verified by comparative experiments.

Hyperspectral anomaly detection (HAD) is an important hyperspectral image application. HAD can find pixels with anomalous spectral signatures compared with their neighbor background without any prior information. While most of the existed researches are related to statistic-based and distance-based techniques, by summarizing the background samples with certain models, and then, finding the very few outliers by various distance metrics, this review focuses on the HAD based on machine learning methods, which have witnessed remarkable progress in the recent years. In particular, these studies can generally be grouped into the traditional machine learning and deep-learning-based methods. Several representative HAD methods, including both traditional machine and deep-learning-based methods, are then conducted on four real HSIs in the experiments. Finally, conclusions regarding HAD are summarized, and prospects and future development direction are discussed.

As the COVID-19 epidemic began to worsen in the first months of 2020, stringent lockdown policies were implemented in numerous cities throughout the world to control human transmission and mitigate its spread. Although traffic density reduction inside the city was felt subjectively, there has thus far been no objective and quantitative study of its variation to reflect the intracity population flows and their corresponding relationship with lockdown policy stringency from the view of remote sensing (RS) images with the high resolution under 1 m. Accordingly, we here provide a quantitative investigation of the traffic density reduction before and after lockdown was implemented in six cities (Wuhan, Milan, Madrid, Paris, New York, and London) around the world heavily affected by the COVID-19 epidemic, which is accomplished by extracting vehicles from the multitemporal high-resolution RS images. A novel vehicle detection model combining unsupervised vehicle candidate extraction and deep learning identification was specifically proposed for the images with the resolution of 0.5 m. Our results indicate that traffic densities were reduced by an average of approximately 50% (and as much as 75.96%) in these six cities following lockdown. The influences on traffic density reduction rates are also highly correlated with policy stringency, with an ${{\boldsymbol{R}}^2}$ value exceeding 0.83. Even within a specific city, the traffic density changes differed and tended to be distributed in accordance with the city's land-use patterns. Considering that public transport was mostly reduced or even forbidden, our results indicate that city lockdown policies are effective at limiting human transmission within cities.

This study presents a novel algorithm which combines active learning (AL) and transfer learning for medical data classification. The main idea of the proposed algorithm is iteratively querying a small number of informative unlabelled target samples, and, at the same time, removing the source samples which do not fit with the posterior probability distributions in the target domain, so as to combine the basic idea of AL with transfer learning. The experimental results obtained in the classification of the datasets from the University of California Irvine (UCI) Machine Learning Repository and The Cancer Imaging Archive (TCIA) confirm the effectiveness of the proposed algorithm.

A Hyperspectral Image (HSI) contains a great number of spectral bands for each pixel; however, the spatial resolution of HSI is low. Hyperspectral image super-resolution is effective to enhance the spatial resolution while preserving the high-spectral-resolution by software techniques. Recently, the existing methods have been presented to fuse HSI and Multispectral Images (MSI) by assuming that the MSI of the same scene is required with the observed HSI, which limits the super-resolution reconstruction quality. In this paper, a new framework based on domain transfer learning for HSI super-resolution is proposed to enhance the spatial resolution of HSI by learning the knowledge from the general purpose optical images (natural scene images) and exploiting the cross-correlation between the observed low-resolution HSI and high-resolution MSI. First, the relationship between low- and high-resolution images is learned by a single convolutional super-resolution network and then is transferred to HSI by the idea of transfer learning. Second, the obtained Pre-high-resolution HSI (pre-HSI), the observed low-resolution HSI, and high-resolution MSI are simultaneously considered to estimate the endmember matrix and the abundance code for learning the spectral characteristic. Experimental results on ground-based and remote sensing datasets demonstrate that the proposed method achieves comparable performance and outperforms the existing HSI super-resolution methods.

A hyperspectral image (HSI) contains a great number of spectral bands for each pixel, which will limit the conventional image classification methods to distinguish land-cover types of each pixel. Dimensionality reduction is an effective way to improve the performance of classification. Linear discriminant analysis (LDA) is a popular dimensionality reduction method for HSI classification, which assumes all the samples obey the same distribution. However, different samples may have different contributions in the computation of scatter matrices. To address the problem of feature redundancy, a new supervised HSI classification method based on locally weighted discriminant analysis (LWDA) is presented. The proposed LWDA method constructs a weighted discriminant scatter matrix model and an optimal projection matrix model for each training sample, which is on the basis of discriminant information and spatial-spectral information. For each test sample, LWDA searches its nearest training sample with spatial information and then uses the corresponding projection matrix to project the test sample and all the training samples into a low-dimensional feature space. LWDA can effectively preserve the spatial-spectral local structures of the original HSI data and improve the discriminating power of the projected data for the final classification. Experimental results on two real-world HSI datasets show the effectiveness of the proposed LWDA method compared with some state-of-the-art algorithms. Especially when the data partition factor is small, i.e., 0.05, the overall accuracy obtained by LWDA increases by about 20 % for Indian Pines and 17 % for Kennedy Space Center (KSC) in comparison with the results obtained when directly using the original high-dimensional data.

Hyperspectral image (HSI) classification is a widely used application to provide important information of land covers. Each pixel of an HSI has hundreds of spectral bands, which are often considered as features. However, some features are highly correlated and nonlinear. To address these problems, we propose a new discrimination analysis framework for HSI classification based on the Two-stage Subspace Projection (TwoSP) in this paper. First, the proposed framework projects the original feature data into a higher-dimensional feature subspace by exploiting the kernel principal component analysis (KPCA). Then, a novel discrimination-information based locality preserving projection (DLPP) method is applied to the preceding KPCA feature data. Finally, an optimal low-dimensional feature space is constructed for the subsequent HSI classification. The main contributions of the proposed TwoSP method are twofold: (1) the discrimination information is utilized to minimize the within-class distance in a small neighborhood, and (2) the subspace found by TwoSP separates the samples more than they would be if DLPP was directly applied to the original HSI data. Experimental results on two real-world HSI datasets demonstrate the effectiveness of the proposed TwoSP method in terms of classification accuracy.

The endmember extraction algorithm, which selects a collection of pure signature spectra for different materials, plays an important role in hyperspectral unmixing. In this paper, the endmember extraction algorithm is described as a combinatorial optimization problem and a novel Mutation Operator Accelerated Quantum-Behaved Particle Swarm Optimization (MOAQPSO) algorithm is proposed. The proposed approach employs Quantum-Behaved Particle Swarm Optimization (QPSO) to find endmembers with good performances. To the best of our knowledge, this is the first time that QPSO has been introduced into hyperspectral endmember extraction. In order to follow the law of particle movement, a high-dimensional particle definition is proposed. In addition, in order to avoid falling into a local optimum, a mutation operation is used to increase the population diversity. The proposed MOAQPSO algorithm was evaluated on both synthetic and real hyperspectral data sets. The experimental results indicated that the proposed method obtained better results than other state-of-the-art algorithms, including Vertex Component Analysis (VCA), N-FINDR, and Discrete Particle Swarm Optimization (D-PSO).