Change detection was one of the earliest and is also most important applications remote sensing technology. For multispectral images, an effective solution for change problem to exploit all available spectral bands detect changes. However, in practice, temporal variance makes it difficult separate changes nonchanges. In this paper, we propose a novel slow feature analysis (SFA) algorithm detection. Compared with changed pixels, unchanged ones should be spectrally invariant varying slowly...

Change detection has been a hotspot in remote sensing technology for long time. With the increasing availability of multi-temporal images, numerous change algorithms have proposed. Among these methods, image transformation methods with feature extraction and mapping could effectively highlight changed information thus better performance. However, changes images are usually complex, existing not effective enough. In recent years, deep network shown its brilliant performance many fields...

With the rapid development of Earth observation technology, very-high-resolution (VHR) images from various satellite sensors are more available, which greatly enrich data source change detection (CD). Multisource multitemporal can provide abundant information on observed landscapes with physical and material views, it is exigent to develop efficient techniques utilize these multisource for CD. In this article, we propose a novel general deep siamese convolutional multiple-layers recurrent...

With the development of Earth observation technology, a very-high-resolution (VHR) image has become an important data source change detection (CD). These days, deep learning (DL) methods have achieved conspicuous performance in CD VHR images. Nonetheless, most existing models based on DL require annotated training samples. In this article, novel unsupervised model, called kernel principal component analysis (KPCA) convolution, is proposed for extracting representative features from...

Deep learning for change detection is one of the current hot topics in field remote sensing. However, most end-to-end networks are proposed supervised detection, and unsupervised models depend on traditional pre-detection methods. Therefore, we a fully convolutional framework with generative adversarial network, to unify unsupervised, weakly supervised, regional tasks into framework. A basic Unet segmentor used obtain map, an image-to-image generator implemented model spectral spatial...

Benefiting from the developments in deep learning technology, learning-based algorithms employing automatic feature extraction have achieved remarkable performance on change detection (CD) task. However, of existing CD methods is hindered by imbalance between changed and unchanged pixels. To tackle this problem, a progressive foreground-balanced sampling (PFBS) strategy basis not adding information proposed to help model accurately learn features pixels during early training process thereby...

Characterized by tremendous spectral information, hyperspectral image is able to detect subtle changes and discriminate various change classes for detection. The recent research works dominated binary detection, however, cannot provide fine information. And most methods incorporating unmixing multiclass detection (HMCD), yet suffer from the neglection of temporal correlation error accumulation. In this study, we proposed an unsupervised Binary Change Guided Network (BCG-Net) HMCD, which aims...

The rapid advancement of automated artificial intelligence algorithms and remote sensing instruments has benefited change detection (CD) tasks. However, there is still a lot space to study for precise detection, especially the edge integrity internal holes phenomenon features. In order solve these problems, we design Change Guiding Network (CGNet), tackle insufficient expression problem features in conventional U-Net structure adopted previous methods, which causes inaccurate holes. maps...

Scene change detection between multitemporal image scenes can be used to interpret the variation of regional land use, and has significant potential in application urban development monitoring at semantic level. The traditional methods directly comparing independent classes neglect temporal correlation, thus suffer from accumulated classification errors. In this paper, we propose a novel scene method via kernel slow feature analysis (KSFA) postclassification fusion, which integrates with...

Remote sensing change detection has played an important role in many applications. Most traditional methods deal with single-band or multispectral remote images. Hyperspectral images offer more detailed information on spectral changes so as to present promising performance. The challenge is how take advantage of the at such a high dimension. In this paper, we propose subspace-based (SCD) method for hyperspectral Instead dealing band-wise changes, proposed measures changes. SCD regards...

Object tracking is a popular topic in the field of computer vision. The detailed spatial information provided by very high resolution remote sensing sensor makes it possible to track targets interest satellite videos. In recent years, correlation filters have yielded promising results. However, terms dealing with object videos, kernel filter (KCF) tracker achieves poor results due fact that size each target too small compared entire image, and background are similar. Therefore, this letter,...

Fusing multiple change detection results has great potentials in dealing with the spectral variability multitemporal very high-resolution (VHR) remote sensing images. However, it is difficult to solve problem of uncertainty, which mainly includes inaccuracy each candidate map and conflicts between different results. Dempster–Shafer theory (D–S) an effective method model uncertainties combine evidences. Therefore, this paper, we proposed urban for VHR images by fusing methods D–S evidence...

Because of the uncertainty and randomness wind speed, power has characteristics such as nonlinearity multiple frequencies. Accurate prediction is one effective means improving integration. traditional single model cannot fully characterize fluctuating power, scholars have attempted to build other models based on empirical mode decomposition (EMD) or ensemble (EEMD) tackle this problem. However, accuracy these affected by modal aliasing illusive components. Aimed at defects, paper proposes a...

Satellite video target tracking is a new topic in the remote sensing field, which refers to moving objects of interest from satellite real time. The usually occupies only few pixels image, even when train long. Thus, still faces challenges compared with traditional visual tracking, including detection low-resolution targets, features less representation, and targets an extremely similar background. Little research has been done on little known about whether or not existing algorithms can...

Object tracking is a hot topic in computer vision. Thanks to the booming of very high resolution (VHR) remote sensing techniques, it now possible track targets interests satellite videos. However, since videos are usually too small comparison with entire image, and similar background, most state-of-the-art algorithms failed target satisfactory accuracy. Due fact that optical flow shows great potential detect even slight movement targets, we proposed multiframe tracker for object The...

Classifying multi-temporal scene land-use categories and detecting their semantic scene-level changes for imagery covering urban regions could straightly reflect the transitions. Existing methods change detection rarely focus on temporal correlation of bi-temporal features, are mainly evaluated small scale datasets. In this work, we proposed a CorrFusion module that fuses highly correlated components in feature embeddings. We firstly extracts deep representations inputs with convolutional...

Tracking moving objects from space-borne satellite videos is a new and challenging task. The main difficulty stems the extremely small size of target interest. First, because usually occupies only few pixels, it hard to obtain discriminative appearance features. Second, object can easily suffer occlusion illumination variation, making features less distinguishable in surrounding regions. Current state-of-the-art tracking approaches mainly consider high-level deep single frame with low...

Deep learning-based hyperspectral image (HSI) classification methods have recently attracted significant attention. However, features captured by convolutional neural network (CNN) are always partial due to the restrictions of respective fields and loss multiscale information, which lead being discontinuous when extracted. In a departure from existing approaches, in this article, we propose novel Enhanced Multiscale Feature Fusion Network (EMFFN). As deeper wider network, EMFFN can extract...

Unsupervised multimodal change detection is a practical and challenging topic that can play an important role in time-sensitive emergency applications. To address the challenge remote sensing images cannot be directly compared due to their modal heterogeneity, we take advantage of two types modality-independent structural relationships images. In particular, present relationship graph representation learning framework for measuring similarity relationships. Firstly, graphs are generated from...

The fast development of self-supervised learning lowers the bar feature representation from massive unlabeled data and has triggered a series researches on change detection remote sensing images. Challenges in adapting natural images classification to arise difference between two tasks. learned patch-level representations are not satisfying for pixel-level precise detection. In this paper, we proposed novel hyperspectral spatial-spectral understanding network (HyperNet) accomplish pixel-wise...

A high-precision feature extraction model is crucial for change detection. In the past, many deep learning-based supervised detection methods learned to recognize patterns from a large number of labelled bi-temporal images, whereas labelling remote sensing images very expensive and often time-consuming. Therefore, we propose coarse-to-fine semi-supervised method based on consistency regularization (C2F-SemiCD), which includes network with multi-scale attention mechanism(C2FNet) update...