Object detection, as of one the most fundamental and challenging problems in computer vision, has received great attention recent years. Over past two decades, we have seen a rapid technological evolution object detection its profound impact on entire vision field. If consider today's technique revolution driven by deep learning, then, back 1990s, would see ingenious thinking long-term perspective design early vision. This article extensively reviews this fast-moving research field light...

Remote sensing image change detection (CD) is done to identify desired significant changes between bitemporal images. Given two co-registered images taken at different times, the illumination variations and misregistration errors overwhelm real object changes. Exploring relationships among spatial–temporal pixels may improve performances of CD methods. In our work, we propose a novel Siamese-based attention neural network. contrast previous methods that separately encode without referring...

Object detection, as of one the most fundamental and challenging problems in computer vision, has received great attention recent years. Over past two decades, we have seen a rapid technological evolution object detection its profound impact on entire vision field. If consider today's technique revolution driven by deep learning, then back 1990s, would see ingenious thinking long-term perspective design early vision. This paper extensively reviews this fast-moving research field light...

Super-resolution is an image processing technology that recovers a high-resolution from single or sequential low-resolution images. Recently deep convolutional neural networks (CNNs) have made huge breakthrough in many tasks including super-resolution. In this letter, we propose new single-image super-resolution algorithm named local-global combined (LGCNet) for remote sensing images based on the CNNs. Our LGCNet elaborately designed with its "multifork" structure to learn multilevel...

Cloud detection is one of the important tasks for remote sensing image processing. In this paper, a novel multilevel cloud method based on deep learning proposed images. First, simple linear iterative clustering (SLIC) improved to segment into good quality superpixels. Then, convolutional neural network (CNN) with two branches designed extract multiscale features from each superpixel and predict as three classes including thick cloud, thin noncloud. Finally, predictions all superpixels in...

Automatic ship detection on spaceborne optical images is a challenging task, which has attracted wide attention due to its extensive potential applications in maritime security and traffic control. Although some image methods have been proposed recent years, there are still three obstacles this task: 1) the inference of clouds strong waves; 2) difficulties detecting both inshore offshore ships; 3) high computational expenses. In paper, we propose novel method called SVD Networks (SVDNet),...

Modern change detection (CD) has achieved remarkable success by the powerful discriminative ability of deep convolutions. However, high-resolution remote sensing CD remains challenging due to complexity objects in scene. Objects with same semantic concept may show distinct spectral characteristics at different times and spatial locations. Most recent pipelines using pure convolutions are still struggling relate long-range concepts space-time. Non-local self-attention approaches promising...

This paper investigates an intriguing question in the remote sensing field: "can a machine generate humanlike language descriptions for image?" The automatic description of image (namely, captioning) is important but rarely studied task artificial intelligence. It more challenging as must not only capture ground elements different scales, also express their attributes well how these interact with each other. Despite difficulties, we have proposed captioning framework by leveraging techniques...

Training deep learning-based change detection (CD) models heavily relies on large labeled data sets. However, it is time-consuming and labor-intensive to collect large-scale bitemporal images that contain building change, due both its rarity sparsity. Contemporary methods tackle the insufficiency mainly focus transformation-based global image augmentation cost-sensitive algorithms. In this article, we propose a novel data-level solution, namely, Instance-level Augmentation (IAug), generate...

Deep learning methods have achieved considerable progress in remote sensing image building extraction. Most extraction are based on Convolutional Neural Networks (CNN). Recently, vision transformers provided a better perspective for modeling long-range context images, but usually suffer from high computational complexity and memory usage. In this paper, we explored the potential of using efficient We design an dual-pathway transformer structure that learns long-term dependency tokens both...

Convolutional neural networks have made a great breakthrough in recent remote sensing image super-resolution (SR) tasks. Most of these methods adopt upsampling layers at the end models to perform enlargement, which ignores feature extraction high-dimension space, and thus, limits SR performance. To address this problem, we propose new framework for images enhance high-dimensional representation after layers. We name proposed method as transformer-based enhancement network (TransENet), where...

Leveraging the extensive training data from SA-1B, Segment Anything Model (SAM) demonstrates remarkable generalization and zero-shot capabilities. However, as a category-agnostic instance segmentation method, SAM heavily relies on prior manual guidance, including points, boxes, coarse-grained masks. Furthermore, its performance in remote sensing image tasks remains largely unexplored unproven. In this paper, we aim to develop an automated approach for images, based foundational model...

Remote sensing image classification forms the foundation of various understanding tasks, serving a crucial function in remote interpretation. The recent advancements Convolutional Neural Networks (CNNs) and Transformers have markedly enhanced accuracy. Nonetheless, scene remains significant challenge, especially given complexity diversity scenarios variability spatiotemporal resolutions. capacity for whole-image can provide more precise semantic cues discrimination. In this paper, we...

Despite the success of deep learning-based change detection methods, their existing insufficiency in temporal (channel, spatial) and multi-scale alignment have rendered them insufficient capability mitigating external factors (illumination changes perspective differences, etc.) arising from different imaging conditions during detection. In this paper, a Bi-temporal Feature Alignment (BiFA) model is proposed to produce precise map lightweight manner by reducing impact irrelevant factors....

Remote Sensing Image Change Captioning (RSICC) aims to describe surface changes between multi-temporal remote sensing images in language, including the changed object categories, locations, and dynamics of changing objects (e.g., added or disappeared). This poses challenges spatial temporal modeling bi-temporal features. Despite previous methods progressing change perception, there are still weaknesses joint spatial-temporal modeling. To address this, this paper, we propose a novel RSCaMa...

Target detection is an important application in the hyperspectral image processing field, and several algorithms have been proposed past decades. Some traditional detectors are built based on statistical information of target background spectra, their performances tend to be affected by spectral quality. previous methods cope with this problem refining spectra make detector robust. In paper, instead doing similar this, we propose a new hierarchical method suppress backgrounds while...

Remote sensing images are widely used in various fields. However, they usually suffer from the poor contrast caused by haze. In this letter, we propose a simple, but effective, way to eliminate haze effect on remote images. Our work is based dark channel prior and common imaging model. order halo artifacts, use low-pass Gaussian filter refine coarse estimated atmospheric veil. We then redefine transmission, with aim of preventing color distortion recovered The main advantage proposed...

Ship detection in optical remote sensing imagery has drawn much attention recent years, especially with regards to the more challenging inshore ship detection. However, work on this subject relies heavily hand-crafted features that require carefully tuned parameters and complicated procedures. In letter, we utilize a fully convolutional network (FCN) tackle problem of design framework possesses simplified procedure robust performance. When tackling FCN, there are two major difficulties: 1)...

Generative adversarial network (GAN) has made great progress in recent natural image super-resolution tasks. The key to its success is the integration of a discriminator which trained classify whether input real high-resolution (HR) or generated one. Arguably, learning strong discriminative prior essential for generating high-quality images. However, remote sensing images, we discover, through extensive statistical analysis, that there are more low-frequency components than may lead...

Given a spectral library, sparse unmixing aims at finding the optimal subset of endmembers from it to model each pixel in hyperspectral scene. However, still remains challenging task due usually high mutual coherence library. In this paper, we exploit priori information image alleviate difficulty. It assumes that some materials library are known exist Such can be obtained via field investigation or data analysis. Then, propose novel incorporate into unmixing. Based on alternating direction...

Automatic oil tank detection plays a very important role for remote sensing image processing. To accomplish the task, hierarchical detector with deep surrounding features is proposed in this paper. The extracted by learning model aim at making tanks more easily to recognize, since appearance of circle and information not enough separate targets from complex background. method divided into three modules: 1) candidate selection; 2) feature extraction; 3) classification. First, modified ellipse...

Joint spectral and spatial information should be fully exploited in order to achieve accurate classification results for hyperspectral images. In this paper, we propose an ensemble framework, which combines different scales. The motivation of the proposed method derives from basic idea: by integrating many individual learners, learning can better generalization ability than a single learner. work, learners are obtained joint spectral-spatial features generated Specially, develop two...

Ensemble learning is an important group of machine techniques that aim to enhance the nonlinearity and generalization ability a system by aggregating multiple learners. We found ensemble show great potential for improving performance traditional hyperspectral target detection algorithms, while at present, there are few previous works have been done on this topic. To end, we propose based Constrained Energy Minimization (E-CEM) detector image detection. Classical algorithms like (CEM),...

Multispectral remote sensing images are often contaminated by haze, which causes low image quality. In this paper, a novel dehazing method based on deep convolutional neural network (CNN) with the residual structure is proposed for multispectral images. First, multiple CNN individuals connected in parallel and each individual used to learn regression from hazy clear image. Then, outputs of fused weight maps produce final result. designed network, individuals, mining multiscale haze features...