BS<sup>3</sup>LNet: A New Blind-Spot Self-Supervised Learning Network for Hyperspectral Anomaly Detection

Gao, Lianru; Wang, Degang; Zhuang, Lina; Sun, Xingming; Huang, Min

doi:10.1109/tgrs.2023.3246565

Cited by 87 publications

(10 citation statements)

References 58 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In this context machine learning techniques allow for notable progress in a large range of application of remote sensing. Examples include a Siamese Transformer Network designed for HS image target detection [10] and some techniques for HS image denoising and anomaly detection [11], [12], [13], [14]. These advances highlight the growing ability of machine learning strategies to refine the understanding and interpretation of remote sensing data.…”

Section: Introductionmentioning

confidence: 99%

Multisource Remote Sensing Data Visualization Using Machine Learning

Plajer,

Băicoianu,

Majercsik

et al. 2024

IEEE Trans. Geosci. Remote Sensing

View full text Add to dashboard Cite

With the availability of several remotely sensed data sources, the problem of efficiently visualizing the information from multisource data for improved Earth observation becomes an intriguing and challenging subject. Multispectral and hyperspectral images encompass a wealth of spectral data that standard RGB monitors cannot replicate directly. Thus, it is important to elaborate methods for accurately representing this information on conventional displays. These images, with tens to hundreds of spectral bands, contain relevant data about specific wavelengths that RGB channels cannot capture. Traditional visualization methods often use only a limited amount of the available spectral information, resulting in a significant loss of information. However, recent advances in artificial intelligence models have provided superior visualization techniques. These AI-based methods allow for more realistic and visually appealing representations, which are important for the information interpretation and direct identification of areas of interest. The main goal of our study is to process aggregated datasets from various sources using a fully connected neural network (FCNN), while considering visualization as a secondary objective. Given that our data come from a variety of sources, a significant emphasis in our study was placed on the preprocessing stage. In order to achieve a consistent visualization across datasets from different sources, proper preprocessing by standardization or normalization procedures is essential. Our research comprises numerous experiments to demonstrate the effectiveness of the proposed technique for image visualization.

show abstract

Section: Introductionmentioning

confidence: 99%

Multisource Remote Sensing Data Visualization Using Machine Learning

Plajer,

Băicoianu,

Majercsik

et al. 2024

IEEE Trans. Geosci. Remote Sensing

View full text Add to dashboard Cite

show abstract

“…Convolutional neural networks (CNNs) have facilitated the progress of different computer visual tasks [1][2][3][4][5][6], including object detection [7][8][9][10] in remote sensing images (RSIs). However, accurate bounding box (BB)-level annotations [11][12][13] are difficult to obtain.…”

Section: Introductionmentioning

confidence: 99%

Weakly Supervised Object Detection for Remote Sensing Images via Progressive Image-Level and Instance-Level Feature Refinement

Zheng,

Wu,

et al. 2024

Remote Sensing

View full text Add to dashboard Cite

Weakly supervised object detection (WSOD) aims to predict a set of bounding boxes and corresponding category labels for instances with only image-level supervisions. Compared with fully supervised object detection, WSOD in remote sensing images (RSIs) is much more challenging due to the vast foreground-related context regions. In this paper, we propose a progressive image-level and instance-level feature refinement network to address the problems of missing detection and part domination for WSOD in RSIs. Firstly, we propose a multi-label attention mining loss (MAML)-guided image-level feature refinement branch to effectively allocate the computational resources towards the most informative part of images. With the supervision of MAML, all latent instances in images are emphasized. However, image-level feature refinement further expands responsive gaps between the informative part and other sub-optimal informative ones, which results in exacerbating the problem of part domination. In order to alleviate the above-mentioned limitation, we further construct an instance-level feature refinement branch to re-balance the contributions of different adjacent candidate bounding boxes according to the detection task. An instance selection loss (ISL) is proposed to progressively boost the representation of salient regions by exploring supervision from the network itself. Finally, we integrate the image-level and instance-level feature refinement branches into a complete network and the proposed MAML and ISL functions are merged with class classification and box regression to optimize the whole WSOD network in an end-to-end training fashion. We conduct experiments on two popular WSOD datasets, NWPU VHR-10.v2 and DIOR. All the experimental results demonstrate that our method achieves a competitive performance compared with other state-of-the-art approaches.

show abstract

“…H YPERSPECTRAL images (HSIs) contain tens or hundreds of spectral bands per pixel [1], [2], which provide abundant spectral information and have the potential to help us identify the different substances appearing in the scenes [3]- [6]. Based on it, hyperspectral imaging first attracts interest in remote sensing [7], [8] and medical imaging [9], then quickly spreads into many other fields.…”

Section: Introductionmentioning

confidence: 99%

Multishot Compressive Hyperspectral Imaging Based on Tensor Fibered Rank Minimization and Its Primal-Dual Algorithm

Xie,

Kang,

Dian

et al. 2024

IEEE J. Sel. Top. Appl. Earth Observations Remote Sensing

View full text Add to dashboard Cite

Coded aperture snapshot spectral imaging (CASSI) compresses tens to hundreds of spectral bands of hyperspectral image (HSI) to a two-dimensional (2D) compressive measurement. For spatially or spectrally rich scenes, the compressive measurement provided by a single snapshot CASSI may not be sufficient. By taking multiple snapshots of the same scene, multi-shot CASSI leads to a less ill-posed inverse reconstruction problem, making the CASSI system more suitable for spatially or spectrally rich HSI. Considering the strong spectral correlation of HSI and directional characteristics of mask shifting of multishot CASSI, the mode-1 tensor fibered rank (TFR) minimization is presented for its reconstruction in this paper. Specifically, the mode-1 TFR is derived from the tensor singular value decomposition (t-SVD) to the mode-1 t-SVD, and the mode-1 TFR minimization is reduced to a mode-1 tensor nuclear norm minimization problem, to achieve more accurate HSI characterization in multi-shot CASSI reconstruction. The primal-dual algorithm (PDA) is applied to solve the objective optimization problem, which is flexible. Experimental results on CAVE, Cuperite and Urban datasets demonstrate the effectiveness of the proposed method.

show abstract

BS³LNet: A New Blind-Spot Self-Supervised Learning Network for Hyperspectral Anomaly Detection

Cited by 87 publications

References 58 publications

Multisource Remote Sensing Data Visualization Using Machine Learning

Multisource Remote Sensing Data Visualization Using Machine Learning

Weakly Supervised Object Detection for Remote Sensing Images via Progressive Image-Level and Instance-Level Feature Refinement

Multishot Compressive Hyperspectral Imaging Based on Tensor Fibered Rank Minimization and Its Primal-Dual Algorithm

Contact Info

Product

Resources

About

BS3LNet: A New Blind-Spot Self-Supervised Learning Network for Hyperspectral Anomaly Detection

Cited by 87 publications

References 58 publications

Multisource Remote Sensing Data Visualization Using Machine Learning

Multisource Remote Sensing Data Visualization Using Machine Learning

Weakly Supervised Object Detection for Remote Sensing Images via Progressive Image-Level and Instance-Level Feature Refinement

Multishot Compressive Hyperspectral Imaging Based on Tensor Fibered Rank Minimization and Its Primal-Dual Algorithm

Contact Info

Product

Resources

About

BS³LNet: A New Blind-Spot Self-Supervised Learning Network for Hyperspectral Anomaly Detection