Change detection is a basic task of remote sensing image processing. The research objective to identify the change information interest and filter out irrelevant as interference factors. Recently, rise in deep learning has provided new tools for detection, which have yielded impressive results. However, available methods focus mainly on difference between multitemporal images lack robustness pseudochange information. To overcome resistance current pseudochanges, this article, we propose...

Recently, supervised deep learning has achieved a great success in remote sensing image (RSI) semantic segmentation. However, for segmentation requires large number of labeled samples, which is difficult to obtain the field sensing. A new paradigm, self-supervised (SSL), can be used solve such problems by pretraining general model with unlabeled images and then fine-tuning it on downstream task very few samples. Contrastive typical method SSL that learn invariant features. most existing...

Training a modern deep neural network on massive labeled samples is the main paradigm in solving scene classification problem for remote sensing, but learning from only few data points remains challenge. Existing methods few-shot sensing are performed sample-level manner, resulting easy overfitting of learned features to individual and inadequate generalization category segmentation surfaces. To solve this problem, should be organized at task level rather than sample level. Learning tasks...

Synthetic aperture radar (SAR) has all-day and all-weather characteristics plays an extremely important role in the military field. The breakthroughs deep learning methods represented by convolutional neural network (CNN) models have greatly improved SAR image recognition accuracy. Classification based on CNNs can perform high-precision classification, but there are security problems against adversarial examples (AEs). However, research AEs is mostly limited to natural images, remote sensing...

Convolutional neural networks (CNNs) have recently been widely used in remote-sensing scene classification. Additionally, it is becoming very popular to automatically learn specific CNN architectures for data sets. The rich contextual information high-resolution images (RSIs) critical intelligent understanding tasks. However, architecture learning approaches tend simplify the original (i.e., resizing smaller resolution) efficiency, yet result loss of RSIs. In this article, we proposed a...

Self-supervised learning achieves close to supervised results on remote sensing image (RSI) scene classification. This is due the current popular self-supervised methods that learn representations by applying different augmentations images and completing instance discrimination task which enables convolutional neural networks (CNNs) invariant augmentation. However, RSIs are spatial-temporal heterogeneous, means similar features may exhibit characteristics in scenes. Therefore, performance of...

Tree species classification is an important and challenging task in image recognition the management of forest resources. Moreover, tree based on remote sensing images can significantly improve efficiency survey save costs. In recent years, many large models have achieved high accuracy airborne remote-sensing manner, but due to their fixed geometric structure, traditional convolutional neural networks are inherently limited local receptive field only provide segmental context information....

Remote sensing image (RSI) scene classification is the foundation and important technology of ground object detection, land use management geographic analysis. During recent years, convolutional neural networks (CNNs) have achieved significant success are widely applied in RSI classification. However, crafted images that serve as adversarial examples can potentially fool CNNs with high confidence hard for human eyes to interpret. For increasing security robust requirements classification,...

Remote sensing image scene classification is a fundamental but challenging task in understanding remote images. Recently, deep learning-based methods, especially convolutional neural network-based (CNN-based) methods have shown enormous potential to understand CNN-based meet with success by utilizing features learned from data rather than designed manually. The feature-learning procedure of CNN largely depends on the architecture CNN. However, most architectures used for are still hand which...

Airborne Laser Scanning (ALS) point clouds have complex structures, and their 3D semantic labeling has been a challenging task. It three problems: (1) the difficulty of classifying around boundaries objects from different classes, (2) diversity shapes within same class, (3) scale differences between classes. In this study, we propose novel double self-attention convolutional network called DAPnet. The includes attention module (PAM) group (GAM). For problem (1), PAM can effectively assign...

In view of the growing popularity and increasing presence public bicycle sharing systems fact that bikes are widely used as a convenient means transportation well an equipment for exercise. This proposed system is based on IoT sensor network in which data PM2.5 collected sent to cloud monitoring use. Each bike user with this mobile app contributes collection enjoys benefit staying informed air quality along various routes from big retrieved real time cloud. addition, integrated platform...

The detection of arbitrarily rotated objects in aerial images is challenging due to the highly complex backgrounds and multiple angles objects. Existing detectors are not robust relative varying angle because CNNs do explicitly model orientation’s variation. In this paper, we propose an Orientation Robust Detector (OrtDet) solve problem, which aims learn features that change accordingly with object’s rotation (i.e., rotation-equivariant features). Specifically, introduce a vision transformer...

Detection and pose estimation of texture-less objects still faces several challenges such as foreground occlusions, background clutter, multi-instance objects, large scale changes to name but a few. In this paper, we present an improved method for template-based detection LINEMOD, in order improve the robustness with partial occlusions. For template creation, divided into four equal parts. After process, patches are matched independently each other. And use image pyramid searching fast...

In the present work, a nanocomposite scaffold was prepared on basis of collagen, hyaluronic acid (HA) and nanobioactive glass (NBAG) by freeze drying method for application to tissue engineering materials. The biological property assays, including von Kossa staining, tetracycline hemolysis, platelet adhesion test, pyrogen tests acute toxicity were performed according requirements ISO 10993 standards evaluate its performance as applicable engineering. results indicated that collagen-HA/NBAG...

Analyzing the urban spatial structure of a city is core topic within geographical information science that has ability to assist planning, site selection, location recommendation, etc. Among previous studies, comprehending functionality places central and corresponds understanding how people use places. With help big geospatial data which contain affluent about human mobility activity, we propose novel multiple subspaces-based model interpret functional regions. This based on assumption...

Multimodal Named Entity Recognition (MNER) is a crucial task for information extraction from social media platforms such as Twitter. Most current methods rely on attention weights to extract both text and images but are often unreliable lack interpretability. To address this problem, we propose incorporating uncertainty estimation into the MNER task, producing trustworthy predictions. Our proposed algorithm models distribution of each modality Normal-inverse Gamma distribution, fuses them...