The extraction of features from the fully connected layer a convolutional neural network (CNN) model is widely used for image representation. However, obtained by layers are seldom investigated due to their high dimensionality and lack global In this study, we explore uses local description feature encoding deeply features. Given an input image, pyramid constructed, different pretrained CNNs applied each scale extract Deeply descriptors can be concatenating in spatial position. Hellinger...

The classification performance of aerial scenes relies heavily on the discriminative power feature representation from high-spatial resolution remotely sensed imagery. convolutional neural networks (CNNs) have recently been applied to adaptively learn image features at different levels abstraction rather than requiring handcrafted and achieved state-of-the-art performance. However, most these focus multi-stage global learning yet neglect local information, which plays an important role in...

Convolutional neural networks (CNNs) have facilitated impressive improvements in the semantic segmentation of very high-resolution (VHR) remote sensing images. The success depends on an effective receptive field (RF) large enough to cover entire object. Popular methods enlarge RF include dilated filters, subsampling operations, and stacking layers. Unfortunately, are inefficient or able cause grid artifacts. Moreover, although object sizes vary greatly images, size cannot reach a compromise...

Extracting change regions from bitemporal images is crucial to urban planning, land, and resources survey. In the literature, many methods obtaining difference between remote sensing have been proposed. However, there are still some problems due complexity of conditions. order solve above-mentioned problems, we propose a novel network called PPCNET, combining patch-level pixel-level detection for images. This divided into three branches: dual structure used extract features images,...

As one of the best image clustering methods, fuzzy local information C-means is often used for segmentation. The effects noise are avoided by utilizing spatial relationship among pixels, but it generates boundary zones mix pixels around edges. This letter presents an method, called with edge and (FELICM), which reduces degradation introducing weights within neighbor windows. edges extracted at first Canny detection. During detection, two adaptive thresholds obtained multi-Otsu method used....

Change detection (CD) is an essential remote sensing application for Earth observations widely used several monitoring, management, and surveillance-related purposes. Generally, pixel-to-pixel prediction roles are susceptible to position details, as high-resolution bitemporal images contain abundant ground so it needs more precautions extract features. Recently, various efforts have been made deep semantic change features, but the importance of grained features has always ignored, leading...

The Hough forest method is an effective for object detection in ground-shot images that has received increasing research attention. However, this lacks the ability to detect objects with arbitrary orientations. This largely constrains from being used detecting geospatial remotely sensed (RS) since can have many different In order achieve rotation invariance and compensate associated loss of discriminative power, paper presents a novel color-enhanced rotation-invariant (CRIHF) RS images. our...

Infrared (IR) and visible images are heterogeneous data, their fusion is one of the important research contents in remote sensing field. In last decade, deep networks have been widely used image due to ability preserve high-level semantic information. However, lower resolution IR images, learning-based methods may not be able retain salient features images. this article, a novel based on Features & Multiscale Dense Network (IR-MSDNet) proposed content key target from both fused image. It...

Convolutional neural networks (CNNs) have attracted great attention in the semantic segmentation of very-high-resolution (VHR) images urban areas. However, large-scale variation objects areas often makes it difficult to achieve good accuracy. Atrous convolution and atrous spatial pyramid pooling composed can alleviate this problem by exploring multiscale contextual information. Unfortunately, causes gridding artifacts, where actual receptive fields are separated unit sets fail cover all...

Image segmentation is crucial to object-oriented remote sensing imagery analysis. In this paper, a novel texture-preceded algorithm proposed for high-resolution imagery, in which texture clustering first carried out as loose constraint later segmentation. The based on the graph models of region adjacency and nearest neighbor graph, can achieve fast node merging, depending global optimum. Here, combined distance, composed texture, spectral, shape features, established measure similarity...

Linearly nonseparability and class imbalance of very high resolution (VHR) imagery make feature selection for object-oriented classification quite challenging, while such characteristics, especially imbalance, have usually been ignored in open literature. To cope with the challenges, this paper proposes a new graph-based method named locally weighted discriminating projection (LWDP). First, popular criteria are reformulated to present linear or nonlinear mapping space. Second, weight...

Multimodal fusion is an essential research area in computer vision application. However, there are still many obstacles the image domain that cause loss key content, due to method limitation or its least efficiency. To solve these problems, a novel pyramid feature attention strategy (PFAF-Net) based on multiscale features with core idea of different level proposed. First, high-level receptive fields extracted by extraction module. Second, and low-level fused techniques, i.e., attention-based...

With the advancement of machine learning, classification methods have been increasingly used in change (or transition) detection. The texton forest (TF)-based method has received increasing research attention because its speed, good generalization characteristics, stability, and especially ability to capture spatial contextual information. In this paper, we propose a TF-based for transition detection remotely sensed imagery. We investigate maximal joint-information gain criterion random...

Generally, some object-based features are more relevant to a thematic class than other features. These strongly features, termed as class-specific would significantly contribute information extraction for very high resolution (VHR) images. However, many existing feature selection methods have been designed select good subset all classes, rather an independent the class. The latter might better meet requirement of former. In addition, lack quantitative evaluation contribution selected classes...

Objective This study aims to analyze the efficacy and safety of different electrical stimulation treatments for post-stroke motor dysfunction, quantitatively advantages between them their possible benefits patients. Methods We will systematically search seven databases. All be retrieved from inception 15, April 2024. Two reviewers evaluation risk bias in all included studies with version 2 Cochrane risk-of-bias assessment tool. Data synthesis performed using a random-effects model network...

A novel target detection method based on affine invariant interest point detection, feature encoding, and large-margin dimensionality reduction (LDR) is proposed for optical remote sensing images. First, four types of detectors are introduced, their performance in extracting low-level descriptors using shape estimation compared. Such a description can deal with significant transformations, including viewpoints. Second, which extends bag-of-words (BOW) by encoding high-order statistics,...

Remote sensing image scene classification has drawn significant attention for its potential applications in the economy and livelihoods. Unlike traditional handcrafted features, convolutional neural networks provide an excellent avenue obtaining powerful discriminative features. Although tremendous efforts have been made so far this domain, there are still many open challenges due to complexity with higher within-class diversity between-class similarity. To solve above-mentioned problems,...

The extraction of activation vectors (or deep features) from the fully connected layers a convolutional neural network (CNN) model is widely used for remote sensing image (RSI) representation. In this study, we propose to learn discriminative convolution filter (DCF) based on class-specific separability criteria linear transformation features. particular, two types pretrained CNN called CaffeNet and VGG-VD16 are introduced illustrate generality proposed DCF. extracted rearranged into form an...

As a sophisticated computing unit, the pyramidal neuron requires sufficient metabolic energy to fuel its powerful computational capabilities. However, majority of previous works focus on nonlinear integration and consumption in individual neurons but seldom effects synaptic transmission dendritic integration. Here, we developed biologically plausible models simulate neurons, exploring relations between energy. We find that not only drives vesicle cycle, also participates regulation this...

Automatic airport detection has received great attention due to the importance of airports in both military and civilian uses. This paper focuses on automatic large-size remote sensing images under a two-step object framework. In first step, geometrical saliency local entropy are improved find more accurate ROIs for detecting images. The is based line features airports, segment detector (LSD) used detect segments. Then, group weighted map generated after connection, created by further...

Manifold learning is one of the representative nonlinear dimensionality reduction techniques and has had many successful applications in fields information processing, especially pattern classification, computer vision. However, when it used for supervised particular hierarchical result still unsatisfactory. To address this issue, a novel approach, namely manifold (HML) proposed. HML takes into account both between-class label within-class local structural training sets simultaneously to...

Semantic segmentation labels each pixel in high-resolution remote sensing (HRRS) images with a category. To tackle the large size and complexity of HRRS images, this letter presents novel multiscale feature aggregation lightweight network (MFALNet) for semantic segmentation. Unlike standard convolution, asymmetric depth-wise separable convolution residual (ADCR) unit is used to reduce parameter makes optimized structure deeper but less complex. The proposed an encoder–decoder structure,...

A semisupervised feature selection method, named asymmetrically local discriminant (ALDS), is proposed to evaluate the class separability of unbalanced sample sets from very high resolution (VHR) imagery in an object-oriented classification. In order cope with imbalance, ALDS incorporates asymmetric misclassification costs classes into weight matrices. Furthermore, this method locally exploits multiple kinds relationships between pairs more accurately assess ability features preserving...