In the previous years, vision transformer has demonstrated a global information extraction capability in field of computer that convolutional neural network (CNN) lacks. Due to lack inductive bias transformer, it requires large amount data support its training. remote sensing, costs lot obtain significant number high-resolution sensing images. Most existing change detection networks based on deep learning rely heavily CNN, which cannot effectively utilize long-distance dependence between...

Traditional building and water segmentation methods are vulnerable to noise interference, hence they could not avoid missed false detections in the detection process. Excessive deep learning downsampling would lead significant loss of feature map information, image location information offset, overall effect falling apart. To address these issues, a Multi-Scale Location Attention Network (MSLA) is proposed. Location-spatial channel particularly important for edge detail cover. The network...

The segmentation algorithm for buildings and waters is extremely important the efficient planning utilization of land resources. temporal space range remote sensing pictures growing. Due to generic convolutional neural network’s (CNN) insensitivity spatial position information in images, certain location edge details can be lost, leading a low level accuracy. This research suggests double-branch parallel interactive network address these issues, fully using interactivity global Swin...

Remote sensing image change detection plays an important role in urban planning and environmental monitoring. However, the existing algorithms have limited ability feature extraction, relationship understanding, capture of small target features edge detail features, which leads to loss some information features. To this end, a new dual attention-guided multiscale aggregation network is proposed. In encoding stage, fully convolutional dual-branch structure used extract semantic different...

Change detection (CD) aims to explore surface changes in co-aligned image pairs. However, many existing networks primarily focus on learning deep features, without considering the impact of attention and fusion strategies performance. Therefore, a new Frequency-Temporal-Aware Network (FTA-Net) is proposed, it recognizes by means frequency-domain temporal module supervised multilevel time-difference while reducing model size. Frequency Temporal Fusion Module designed introduce frequency...

The biometric information contained in a face image is affected by many factors such as living environment, racial differences, and genetic diversity, this complexity leads to the nonstationary of age estimation. In order reduce overlap features between adjacent ages improve accuracy prediction, multi-stage feature constraints learning method proposed for gradually refines through three constraint stages. each stage, algorithm continuously updates center its corresponding range, minimizes...

Remote sensing image change detection is an essential aspect of remote technology application. Existing algorithms based on deep learning do not distinguish between changed and unchanged areas explicitly, resulting in serious loss edge detail information during detection. Therefore, a new attentional network Siamese U-shaped structure (SUACDNet) proposed this paper. In the feature encoding stage, three branches are introduced to focus global information, difference similarity bitemporal...

Changes on lakes and rivers are of great significance for the study global climate change. Accurate segmentation is critical to their changes. However, traditional water area methods almost all share following deficiencies: high computational requirements, poor generalization performance, low extraction accuracy. In recent years, semantic algorithms based deep learning have been emerging. Addressing problems associated a very large number parameters, accuracy, network degradation during...

Detailed information regarding land utilization/cover is a valuable resource in various fields. In recent years, remote sensing images, especially aerial have become higher resolution and larger span time space, the phenomenon that objects an identical category may yield different spectrum would lead to fact relying on spectral features only often insufficient accurately segment target objects. convolutional neural networks, down-sampling operations are usually used extract abstract semantic...

With the continuous improvement of segmentation effect for natural datasets, some studies have gradually been applied to high-resolution remote sensing images (HRRSIs). Due a large amount ground object information contained, even objects same type present diversity and complexity features in different periods or locations. The existing algorithms semantic are limited by short-range context, details, especially edges, couldnot be fully recovered. Aiming at problem, multi-level aggregation...

Water information extraction is always an important aspect of remote sensing image analysis. However, in the actual water images sensing, backgrounds areas are mostly complex buildings and vegetation, which interferes with detection. In addition, traditional detection methods were not able to accurately identify small tributaries edge information. order improve accuracy segmentation, a dense skip connections network multi-scale features fusion attention mechanism proposed for segmentation....

Water area segmentation is an important branch of remote sensing image segmentation, but in reality, most water images have complex and diverse backgrounds. Traditional detection methods cannot accurately identify small tributaries due to incomplete mining insufficient utilization semantic information, the edge information rough. To solve above problems, we propose a multi-scale feature aggregation network. In order improve ability network process boundary design deep extraction module using...

The use of remote sensing images for land cover analysis has broad prospects. At present, the resolution aerial is getting higher and higher, span time space larger larger, therefore segmenting target objects enconter great difficulties. Convolutional neural networks are widely used in many image semantic segmentation tasks, but existing models often simple accumulation various convolutional layers or direct stacking interfeature reuse up- downsampling, network very heavy. To improve...

Cloud and cloud shadow segmentation is one of the most important issues in remote sensing image processing. Most images are very complicated. In this work, a dual-branch model composed Transformer convolution network proposed to extract semantic spatial detail information respectively solve problems false detection missed detection. To improve model’s feature extraction, Mutual Guidance Module introduced so that Branch Convolution can guide each other for mining. Finally, view problem rough...

Extracting buildings and roads from remote sensing images is very important in the area of land cover monitoring, which great help to urban planning. Currently, a deep learning method used by majority building road extraction algorithms. However, for existing semantic segmentation, it has limitation on receptive field high-resolution images, means that can not show long-distance scene well during pixel classification, image features compressed down-sampling, meaning detailed information...

Understanding surface changes requires the ability to identify in high resolution remote sensing images. Because current deep learning-based change detection algorithms are not able accurately discriminate between altered and unmodified areas, which leads problem of edge uncertainty small target missing process. To images, this research proposes an unique Attention-Guided Siamese Network (SAGNet). In network, bitemporal images' highly representative semantic features retrieved using a fully...

Because clouds and snow block the underlying surface interfere with information extracted from an image, accurate segmentation of cloud/snow regions is essential for imagery preprocessing remote sensing. Nearly all sensing images have a high resolution contain complex diverse content, which makes task more difficult. A multi-branch convolutional attention network (MCANet) suggested in this study. double-branch structure adopted, spatial semantic image are extracted. In way, model’s feature...

In recent years, with the rapid development of Internet technology, number credit card users has increased significantly. Subsequently, fraud caused a large amount economic losses to individual and related financial enterprises. At present, traditional machine learning methods (such as SVM, random forest, Markov model, etc.) have been widely studied in detection, but these are often difficulty demonstrating their effectiveness when faced unknown attack patterns. this paper, new Unsupervised...