Recently, convolutional neural network based methods have been studied for ship detection in optical remote sensing images. However, it is challenging to apply them microwave synthetic aperture radar (SAR) First, most of the regions inshore scene include scattered spots and noises, which dramatically interfere with detection. Besides, SAR images contain targets different sizes, especially small ships dense distribution. Unfortunately, fewer distinguishing features making difficult be...

In remote sensing scene classification (RSSC), features can be extracted with different spatial frequencies where high-frequency usually represent detailed information and low-frequency global structures. However, it is challenging to extract meaningful semantic for RSSC tasks by just utilizing high- or features. The composition of images (RSIs) more complex than that natural images, the scales objects vary significantly. this article, a multiscale feature fusion covariance network...

Most SAR ship detectors based on convolutional neural networks (CNNs) needed preset anchor boxes to object classification and bounding box coordinate regression. However, the sparsity unbalanced distribution of ships in images mean that most are redundant. Thus, settings directly affect performance generalization ability detector. In addition, a variety scales substantial interference inshore backgrounds bring significant challenges detector's improvement. this letter, novel anchor-free...

To infer unknown remote sensing scenarios, for scene classification (RSSC) most existing deep neural networks (DNNs) are trained on closed datasets. When the acquisition speed and quantity of images increases rapidly, these models cannot be used to classify new scenes. Currently, incremental learning as an effective solution solving <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">catastrophic forgetting</i> issue, but ignoring...

We propose herein to make use of rotating electric fields for achieving flexible control on the hydrodynamic behavior two miscible co-flowing water solutions in straight microchannels, context a new manipulation tool stratified liquid contents microfluidic systems. Our theoretical analysis indicates that, while fluids distinct electrical conductivities and identical permittivity are parallel pumped into mainchannel, circularly traveling field, as emitted from four-phase electrode array...

Since the dimension of combined feature set for partial discharge (PD) pattern recognition is higher, corresponding sample size increases, as does required amount storage space and calculation, there are features with less category-related characteristics in parameters, which may contain redundant information between them. To solve problem higher complicated classification model identification insulation defect type this paper. Random forest sequential forward selection method based on...

Projecting the point cloud on 2D spherical range image transforms LiDAR semantic segmentation to a task image. However, is still naturally different from regular RGB image; for example, each position encodes unique geometry information. In this paper, we propose new projection-based pipeline that consists of novel network structure and an efficient post-processing step. our structure, design FID (fully interpolation decoding) module directly upsamples multi-resolution feature maps using...

Landslides are one of the most widespread disasters and threaten people’s lives properties in many areas worldwide. Landslide susceptibility mapping (LSM) plays a crucial role evaluation extenuation risk. To date, large number machine learning approaches have been applied to LSM. Of late, high-level convolutional neural network (CNN) has with intention raising forecast precision The primary contribution research was present model which based on CNN for LSM methodically compare its capability...

Deep learning is being increasingly employed for hyperspectral classification, although such use often predicated on the availability of a sufficiently large set labeled samples training. To improve classification performance under limited training-set size, semi-supervised network with end-to-end local–global active (AL) based graph convolutional networks (GCNs) proposed. The proposed AL extracts both global as well local graph-based features to gauge discriminative information in unlabeled...

We describe the synthesis of hexagonal NaNbO3 by mixing reactants at high temperature under hydrothermal conditions. The morphological and structural evolution crystalline was studied in detail using X-ray diffraction, scanning electron microscopy, transmission selected area high-resolution Raman techniques. A variety products form irregular Na8Nb6O19·nH2O bars, microporous monoclinic Na2Nb2O6·nH2O fibers, self-assembled dandelion-like structures, octahedral-shaped have been prepared....

Remote sensing image (RSI) scene classification plays an active role in many application areas. Due to the excellent performance of convolutional neural networks (CNNs), which have widely applied RSI recent years. However, most existing methods improve accuracy by improving model parameters or fusing features CNNs. This will make whole very complicated and unable extract multiscale at a more granular level. letter proposes novel lightweight depthwise network (MSDWNet) with efficient spatial...

Convolutional networks have been widely used for the classification of hyperspectral images; however, such are notorious their large number trainable parameters and high computational complexity. Additionally, traditional convolution-based methods typically implemented as a simple cascade convolutions using single-scale convolution kernel. In contrast, lightweight multiscale convolutional network is proposed, capitalizing on feature extraction at multiple scales in parallel branches followed...

The rapid development of additive manufacturing technology has offered a new avenue for designing and fabricating high wave-absorbing meta structures. In this study, the mechanical properties broadband absorption performance Poly-Ether-Ether-Ketone (PEEK)–based electromagnetic wave–absorbing composite materials was investigated. high-performance polymer PEEK used as matrix, with wave loss, such reduced graphene oxide, Carbonyl Iron (CI), Flake CI (FCI), were absorbers. Based on theory...

With the development of deep learning, remote sensing image semantic segmentation has produced significant advances. The majority existing methods use fully convolutional network (FCN) that lacks fine-grained multi-scale representation and fails to extract global context information. Thus, we improve FCN by adding two modules—multi-scale attention (MSA) non-local filter (NLF). MSA module enhances network's capability allows modeling inter-dependencies feature maps among different channels....

Deep learning (DL) plays an increasingly important role in earth observation by multi-source remote sensing. However, the current DL-based methods do not make fully use of complementary information among sensing data, such as hyperspectral image (HSI) and light detection ranging (LiDAR) lack consideration multi-scale, directional fine-grained features. To address these issues, a multi-scale multi-direction feature extraction network is proposed this article. Specifically, spatial (MSSpaF)...

In order to meet the highway guidance demand, this work studies short-term traffic flow prediction method of highway. The Yu-Wu which is main road in Chongqing, China, time series taken as study object. It uses phase space reconstruction theory and Lyapunov exponent analyze nonlinear character flow. A new Volterra based on model reduction via quadratic-linear systems (QLMOR) applied predict Compared with Taylor-expansion-based methods, these QLMOR-reduced models retain more information...

Shandong peninsula, the largest peninsula of China, is prone to severe land subsidence hazards along coastline. In this paper, we provide, for first time, multi-scale and multi-dimensional time series deformation measurements entire with advanced Interferometric Synthetic Aperture Radar (InSAR) techniques. We derive spatiotemporal evolutions by integrating multi-track Sentinel-1A/B RADARSAT-2 satellite images. InSAR are cross validated independent rate results generated from different SAR...

This paper is concerned with the finite-time projective synchronization problem of fractional-order memristive neural networks (FMNNs) mixed time-varying delays. Firstly, under frame differential inclusion and set-valued map, several criteria are derived to ensure FMNNs. Meanwhile, three properties established deal different forms fractional inequation, which greatly extend some results on estimation settling time In addition traditional Lyapunov function 1-norm form in Theorem 1, a more...

Abstract This article considers the problem of exponential synchronization inertial neural networks (INNs) with mixed delays via a novel hybrid control scheme consisting pinning and periodic intermittent control. Both discrete delay distributed are taken into account in network model. Through proper variable substitution, original system is transferred simply formed differential equation. Meantime, some new sufficient conditions derived by Lyapunov stability theory for INNs. Eventually,...

In this paper, the adaptive synchronization of fractional-order complex-valued neural networks with time-varying delays (FOCVNNTDs) is investigated. First, two novel differential inequalities time are established, which can be seen as an extension Halanay inequality. Besides, complete and quasi-projective FOCVNNTDs investigated based on using a controller. addition, instead separating into real-valued networks, non-decomposition method adopted to study FOCVNNTDs, avoids difficulty complexity...