Currently, cross-scene hyperspectral image (HSI) classification has drawn increasing attention. It is necessary to train a model only on source domain (SD) and directly transferring the target (TD), when TD needs be processed in real time cannot reused for training. Based idea of generalization, Single-source Domain Expansion Network (SDEnet) developed ensure reliability effectiveness extension. The method uses generative adversarial learning SD test TD. A generator including semantic...

A super-resolution (SR) method based on compressive sensing (CS), structural self-similarity (SSSIM), and dictionary learning is proposed for reconstructing remote images. This aims to identify a that represents high resolution (HR) image patches in sparse manner. Extra information from similar structures which often exist images can be introduced into the dictionary, thereby enabling an HR reconstructed using CS framework. We use K-Singular Value Decomposition obtain orthogonal matching...

Polyimide (PI) films with extremely high breakdown strength (451 kV/mm), energy density (5.2 J/cm3), and discharge efficiency (86.7%) at the room temperature are fabricated by simple solution casting method. Barium titanate (BaTiO3) nanoparticles introduced into PI matrix, giving rise to enhanced dielectric permittivity (6.8) low loss (0.012). Dielectric storage performances of BTO/PI nanocomposites thoroughly investigated up 200 °C. The strengths both pure decrease dramatically increase in...

Cloud and snow detection has significant remote sensing applications, while they share similar low-level features due to their consistent color distributions local texture patterns. Thus, accurately distinguishing cloud from in pixel level satellite images is always a challenging task with traditional approaches. To solve this shortcoming, letter, we proposed deep learning system classify fully convolutional neural networks level. Specifically, specially designed network was introduced learn...

Ultrasonography is a valuable diagnosis method for thyroid nodules. Automatically discriminating benign and malignant nodules in the ultrasound images can provide aided suggestions, or increase accuracy when lack of experts. The core problem this issue how to capture appropriate features specific task. Here, we propose feature extraction based on convolution neural networks (CNNs), try introduce more meaningful semantic classification. Firstly, CNN model trained with massive natural dataset...

Hyperspectral images (HSIs) have been used in a wide range of fields, such as agriculture, food safety, mineralogy, and environment monitoring, but being corrupted by various kinds noise limits its efficacy. Low-rank representation (LRR) has proved effectiveness the denoising HSIs. However, it just employs local information for denoising, which results ineffectiveness when is heavy. In this paper, we propose an approach group low-rank (GLRR) HSI denoising. our GLRR, divided into overlapping...

Spectral library based spectral super-resolution is an effective but challenging way to obtain high-spatial hyperspectral images from multispectral images. However, the incomplete coverage of response functions makes it impossible comprehensively sense information in imaging model, thus greatly limits performance super-resolution. To deal with this problem, a new method under conditions proposed paper. More specifically, strategy for acquiring typical set spectra proposed, trying provide...

Spectral or spatial dictionary has been widely used in fusing low-spatial-resolution hyperspectral (LH) images and high-spatial-resolution multispectral (HM) images. However, only using spectral is insufficient for preserving information, vice versa. To address this problem, a new LH HM image fusion method termed OTD optimized twin dictionaries proposed paper. The problem of formulated analytically the framework sparse representation, as an optimization spectral-spatial their corresponding...

Previous studies on crop classification methods based deep learning for multitemporal images had already determined the number of inputs multi temporal in network structure design stage. However, reality, due to satellite revisit cycles, weather, and other reasons, stable clear remote sensing (RSIs) cannot be continuously obtained. Once a period image is missing from sequence, entire method used. Although such as interpolation using instead can used address this issue, they greatly reduce...

Hyperspectral image classification is mainly based on the spectral information of land covers, but water vapor or Rayleigh scattering will weaken surface reflectance under effect adjacent pixels, and thus lead to reducing discriminative for subsequent tasks. Atmospheric correction weakened bands one most traditional ways deal with this issue, as a complete atmospheric both them difficult, maybe systematic exclusion severely affected base quantitative evaluation better choice. In paper, an...

Utilizing a spectral dictionary learned from couple of similar-scene multi- and hyperspectral image, it is possible to reconstruct desired image only with one single multispectral image. However, the differences between similar scene make difficult directly apply training domain task domain. To this end, compensation matrix based transfer method for super-resolution proposed in paper, trying more accurate high spatial resolution Specifically, scheme established by using similarity...

Based on the spatial dependence assumption, super-resolution mapping can predict location of land cover classes within mixed pixels. In this paper, we propose a novel method via multi-dictionary based sparse representation, which is robust to noise in both learning and class allocation process. To better distinguish different classes, distribution modes are learned separately. A spectral distortion constraint introduced, combining with reconstruction errors as metrics perform classification....

Fusion of the high-spatial-resolution hyperspectral (HHS) image using low-spatial- resolution (LHS) and multispectral (HMS) is usually formulated as a spatial super-resolution problem LHS with help an HMS image, that may result in loss detailed structural information. Facing above problem, fusion nonlinear spectral mapping from to HHS novel cluster-based method multi-branch BP neural networks (named CF-BPNNs) proposed, ensure more reasonable for each cluster. In training stage, considering...

Purpose Computer‐aided diagnosis (CAD) systems assist in solving subjective problems that typically rely on personal experience. A CAD system has been developed to differentiate malignant thyroid nodules from benign ultrasound images based deep learning methods. The diagnostic performance was compared between the and experienced attending radiologists. Methods image dataset for training included 651 386 while database testing 422 128 nodules. All were confirmed by pathology results. In...

Ultrasonography is a valuable diagnosis method for thyroid nodules. Automatically discriminating benign and malignant nodules in the ultrasound images can provide aided suggestions, or increase accuracy when lack of experts. The core problem this issue how to capture appropriate features specific task. Here, we propose feature extraction based on convolution neural networks (CNNs), try introduce more meaningful classification. A CNN model trained with ImageNet data transferred image domain,...

Endmember extraction and spectral unmixing is a very challenging task in multispectral/hyperspectral image processing due to the incompleteness of information. In this paper, new method for endmember hyperspectral images proposed, which called as minimum distance constrained nonnegative matrix factorization (MDC-NMF). After being compared with newly developed named MVC-NMF, MDC-NMF not only has been demonstrated more reasonable theory but also shows promising results experiments.

High-spatial hyperspectral (HH) image reconstruction using both high-spatial multispectral (HM) and low-spatial (LH) over the same scene is widely used in many real applications. Nevertheless, pair of HM LH hard to obtain. To solve this problem, a new HH method spectral library-based dictionary learning (named as HIRSL) proposed paper, only from one image. The above problem formulated framework sparse representation, an estimation band matching matrix, dictionary, coefficients. More...

To generate a high-spatial-resolution hyperspectral (HHS) image from multispectral (HMS) image, both spatial information and spectral should be considered simultaneously if we want to build more accurate mapping HMS HHS. this end, jointed super-resolution method is proposed in letter using an end-to-end learning strategy for each subspace with the cluster-based multibranch backpropagation neural network (BPNN). More specifically, addition spectra similarity, modified superpixel segmentation...

When the system blurring is unknown, we propose a novel super-resolution approach of hyperspectral images by low-rank and group-sparse modeling. No high spatial resolution auxiliary data or prior information about isna needed. The proposed method imposes model with predefined spectral subspace group sparse on different types frequency components to take advantage shared structure across all bands. desired image kernel are optimized alternatively according cost function. Experimental results...