Deep learning methods have recently been successfully explored for hyperspectral image (HSI) classification. However, training a deep-learning classifier notoriously requires hundreds or thousands of labeled samples. In this paper, deep few-shot method is proposed to address the small sample size problem HSI There are three novel strategies in algorithm. First, spectral–spatial features extracted reduce labeling uncertainty via residual 3-D convolutional neural network. Second, network...

In this paper we present a real-time approach to stitch large-scale aerial images incrementally. A monocular SLAM system is used estimate camera position and attitude, meanwhile 3D point cloud map generated. When GPS information available, the estimated trajectory transformed WGS84 coordinates after time synchronized automatically. Therefore, output orthoimage retains global without ground control points. The final image fused visualized instantaneously with proposed adaptive weighted...

This paper explores and discusses the revolutionary applications of digital twin technology in space environments its profound impact on future exploration activities. Originating from a proposal by National Aeronautics Space Administration (NASA) 2002, aims to enhance safety reliability missions creating precise virtual models. As has evolved, have successfully expanded beyond aerospace include Industry 4.0, healthcare, urban management, demonstrating remarkable cross-industry adaptability...

The deep learning methods have recently been successfully explored for hyperspectral image classification. However, it may not perform well when training samples are scarce. A transfer method is proposed to improve the classification performance in situation of limited samples. First, a Siamese network composed two convolutional neural networks designed local descriptors extraction. Subsequently, pretrained model reused knowledge tasks by feeding features extracted from each band into...

Optical remote sensing videos, as a new source of data that has emerged in recent years, have significant potential applications, especially national defense. In this paper, tracking pipeline named TDNet (tracking while detecting based on neural network) is proposed for optical videos correlation filter and deep networks. The used to simultaneously track ships planes videos. There are many target methods general video data, but they suffer some difficulties with low resolution those...

Deep neural networks have recently been successfully explored to extract deep features for hyperspectral image classification. Recurrent (RNNs) are an important branch of the learning family, which widely used sequence analysis. Indeed, RNNs model dependencies between different spectral bands image, inspired by observation that pixels can be considered as sequences. A disadvantage such methods is they don't consider effect neighborhood on final class label. In this letter, a RNN proposed...

The Permanently Shadowed Regions (PSRs) of the lunar south pole have never been directly sampled. To explore and discover resources, Chinese exploration mission is scheduled to land in direct sunlight near PSR, where sampling analysis will be carried out. selection sites for landing one key steps mission. main factors affecting site are distribution PSRs, surface slopes, rock distribution, light intensity, maximum temperature. In this paper, analyzed based on multi-source remote sensing...

Impact craters are the most prominent features on surface of Moon, Mars, and Mercury. They play an essential role in constructing lunar bases, dating Mars Mercury, exploration other celestial bodies. The traditional crater detection algorithms (CDA) mainly based manual interpretation which is combined with classical image processing techniques. CDAs are, however, inefficient for detecting smaller or overlapped impact craters. In this paper, we propose a Split-Attention Networks...

Detecting changes in multisource heterogeneous images is a great challenge for unsupervised change detection methods. Image-translation-based methods, which transform two to be homogeneous comparison, have become mainstream approach. However, most of them primarily rely on information from unchanged regions, resulting networks that cannot fully capture the connection between representations. Moreover, lack priori and sufficient training data makes vulnerable interference changed pixels. In...

Solar radiation is the excitation source that affects weather in atmosphere of earth, and some solar activities such as flares coronal mass ejections are often accompanied by radio bursts. The spectrum bursts helpful for astronomers to explore mechanism With development progress observation methods, Sun can be done at almost all times day. How quickly automatically identify small proportion burst data from huge corpus has become an important research direction. innovation this study enhance...

Inspired by simultaneous localization and mapping (SLAM) style workflow, this article presented an online sequential structure from motion (SfM) solution for high-frequency video large baseline high-resolution aerial images with high efficiency novel precision. First, as traditional SLAM systems are not good in processing low overlap images, based on our hierarchical feature matching paradigm multihomography BoW, we proposed a robust tracking method where the relative pose its scale...

Air quality is a significant environmental issue among the Chinese people and even global population, it affects both human health Earth's long-term sustainability. In this study, we proposed multiperspective, high-dimensional spatiotemporal data visualization interactive analysis method, studied analyzed relationship between air several influencing factors, including meteorology, economics. Six methods were integrated in each specifically designed improved for purposes. To reveal...

Lunar craters and rilles are significant topographic features on the lunar surface that will play an essential role in future research space energy resources geological evolution. However, previous studies have shown low efficiency detecting impact poor accuracy rilles. There is no complete automated identification method for to explore further. In this paper, we propose a new specific deep-learning called high-resolution global–local networks (HR-GLNet) discover simultaneously. Based GLNet...

Crater extraction and recognition is an important research content of deep space planetary science. Traditional crater detection algorithms (CDAs) are mainly based on feature construction which relies high-quality data. With the application learning in image semantic segmentation, new ideas have been brought to meteorite extraction. Many artificial intelligence proposed greatly simplifies process improves accuracy. However, with improvement accuracy, convolution kernels become more more,...

In the task of classifying images cracks in underwater dams, symmetry serves as a crucial geometric feature that aids distinguishing from other structural elements. Nevertheless, asymmetry distribution positive and negative samples within dam crack image dataset results long-tail problem. This asymmetry, coupled with subtle nature features, leads to inadequate extraction by existing convolutional neural networks, thereby reducing classification accuracy. To address these issues, this paper...

In this paper, we focus on the multi-target tracking (MOT) task in satellite videos. To achieve efficient and accurate tracking, propose a transformer-distillation-based end-to-end joint detection (JDT) method. Specifically, (1) considering that targets videos usually have small scales are shot from bird's-eye view, pixel-wise transformer-based feature distillation module through which useful object representations learned via using strong teacher network; (2) videos, such as airplanes,...