Anti-money laundering (AML) detection is of vital importance in financial risk control. Although Graph Neural Networks (GNN) have yielded promising results, existing motif-based approaches primarily focus on node anomaly detection on simple graphs, which hinders the direct identification of anomalous edges in directed temporal transaction networks. Moreover, consecutive transaction relationships, termed dual-edge motifs, have rarely been considered in previous AML studies. To address these gaps, we propose the D-EMAML framework, which consists of: (1) Fast-Motif-Gen, a GPU-accelerated dual-edge motif graph generator with pruning; (2) D-EMGNN, an attention-enhanced heterogeneous GNN module that reduces motif-type information redundancy; (3) MELP, a label aggregation scheme projecting predictions from the motif graph to the original graph. Extensive experiments on real-world and synthetic datasets demonstrate significant improvements over representative baselines and validate the contribution of each component. To our knowledge, this is the first application of dual-edge motif graphs for GNN-based edge anomaly detection in AML.

In daily clinical practice, clinicians exhibit robustness in identifying polyps with both location and size variations. It is uncertain if deep segmentation models can achieve comparable robustness in automated colonoscopic analysis. To benchmark the model robustness, we focus on evaluating the segmentation models on the polyps with various attributes (e.g. location and size) and healthy samples. Based on the Latent Diffusion Model, we perform attribute editing on real polyps and build a new dataset named Polyp-E. Our synthetic dataset boasts exceptional realism, to the extent that clinical experts find it challenging to discern them from real data. We evaluate various existing polyp segmentation models on the proposed benchmark. The results reveal most of the models are highly sensitive to attribute variations. As a novel data augmentation technique, the proposed editing pipeline can improve both in-distribution and out-ofdistribution generalization ability. The code and datasets has been released at https://github.com/RunpuWei/Polyp-E-Benchmark.

Abstract Deep learning has achieved great success in academic benchmarks but fails to work effectively in the real world due to the potential dataset bias. The current learning methods are prone to inheriting or even amplifying the bias present in a training dataset and under-represent specific demographic groups. More recently, some dataset debiasing methods have been developed to address the above challenges based on the awareness of protected or sensitive attribute labels. However, the number of protected or sensitive attributes may be considerably large, making it laborious and costly to acquire sufficient manual annotation. To this end, we propose a prototype-based network to dynamically balance the learning of different subgroups for a given dataset. First, an object pattern embedding mechanism is presented to make the network focus on the foreground region. Then we design a prototype learning method to discover and extract the visual patterns from the training data in an unsupervised way. The number of prototypes is dynamic depending on the pattern structure of the feature space. We evaluate the proposed prototype-based network on three widely used polyp segmentation datasets with abundant qualitative and quantitative experiments. Experimental results show that our proposed method outperforms the CNN-based and transformer-based state-of-the-art methods in terms of both effectiveness and fairness metrics. Moreover, extensive ablation studies are conducted to show the effectiveness of each proposed component and various parameter values. Lastly, we analyze how the number of prototypes grows during the training process and visualize the associated subgroups for each learned prototype. The code and data will be released at https://github.com/zijinY/dynamic-prototype-debiasing .

The geostationary interferometric infrared sounder (GIIRS) onboard the Fengyun-4B (FY-4B) is the first operational geostationary hyperspectral infrared (IR) sounder. This study analyzes the first-year FY-4B/GIIRS on-orbit calibration performance by comparing it to the collocated IR atmospheric sounder interferometer (IASI) observations and radiative transfer (RT) simulations. The results reveal that the mid-wave IR (MWIR) channels had a slightly larger calibration bias compared to the long-wave IR (LWIR) channels. However, the operational FY-4B/GIIRS showed improved performance compared to the experimental FY-4A/GIIRS. Furthermore, this study also found that most channels exhibited negligible annual and weak diurnal variations in calibration bias. However, there was a significant degradation in the LWIR channels (<850 cm1) and the weak diurnal variation in the MWIR channels. Finally, the calibration performance of FY-4B/GIIRS demonstrates a reduced dependence on brightness temperature (BT), except for the channel at wavenumber 703.125 cm1. Overall, it concluded that FY-4B/GIIRS demonstrated consistent calibration stability and high accuracy, highlighting its capability for precise quantitative applications.

The Chinese early-morning-orbiting Fengyun-3E (FY-3E) satellite fills the 6-h initial observation window for data assimilation in numerical weather prediction (NWP). The low-light band (LLB) on the medium-resolution spectral imager low light (MERSI-LL) of FY-3E can detect extremely low radiances at night, significantly enhancing nighttime observation capabilities as well as elevating data assimilation quality by improving the nighttime cloud mask algorithm. However, severe and nonlinear stray light contamination affects most nocturnal FY-3E/MERSI-LL LLB images, particularly those from the Southern Hemisphere, hindering further visualization applications. The analysis concluded that the stray light is closely associated with the refraction and reflection of sunlight entering the MERSI-LL, solar zenith angle (SZA), and detector number. To obtain clear and enhanced images, this study designed a fully automated and adaptive stray light correction and enhancement algorithm for the nocturnal low-light images of FY-3E/MERSI-LL. Three typical stray-light-contaminated scenarios were categorized for all nighttime images. The restored results showed that after processing, the “fog” stray light and stripes were essentially removed, and the details became richer and more prominent, significantly improving the visual effect and usability of the images. This algorithm is simple, efficient, and highly applicable, and will be integrated into the processing system of the FY-3E satellite to support near real-time applications of LLB images. However, some strong or unusual stray light still affects the local continuity of the images. Future low-light imagers of FY-3 satellites will feature more sophisticated instruments to reduce incident stray light in their optical system.

Atmospheric motion vectors (AMVs) from satellite measurements serve as critical indicators of atmospheric dynamics, playing an essential role in enhancing the prediction precision of numerical weather prediction (NWP) models through data assimilation (DA). The implementation of finer satellite-derived vector products has the potential to significantly augment the accuracy of atmospheric flow field data in high-resolution regional NWP model simulations, thereby fulfilling the burgeoning requirements of operational weather nowcasting and forecasting. This study is focused on the development of mesoscale AMV (MAMV) products, which are distinguished by their exceptional quality and spatiotemporal resolution, leveraging data from the geostationary high-speed imager aboard the Fengyun-4B geostationary meteorological satellite (FY-4B/GHI). MAMVs of FY-4B/GHI feature an enhanced horizontal resolution of 3 km, enabling more accurate identification and monitoring of nongeostrophic flow patterns of mesoscale weather systems, as well as their fast-evolving dynamical structures and characteristics. Furthermore, a comparative analysis with radiosonde measurements highlights the precision of MAMV products, as evidenced by a speed bias (SB) of 0.37 m/s, a speed root mean square error (sRMSE) of 4.68 m/s, and a direction root mean square error (dRMSE) of 26.35°. The prospects of high-resolution satellite wind field data hold great potential for propelling scientific advancement and enriching our comprehension of atmospheric dynamics. This is particularly valuable in the context of typhoon monitoring and forecasting, where such data can lead to significant improvements in predictive capabilities.

Polyp segmentation is important in the early diagnosis and treatment of colorectal cancer. Since polyps vary in shape, size, color, and texture, accurate polyp segmentation is very challenging. One promising solution is to model the contextual relation for each pixel. However, previous methods only focus on learning the dependencies between the position within an individual image and ignore the contextual relation across different images. In this paper, we propose a memory-based feature enhancement module to capture the cross-image contextual relations. Specifically, we first present a polyp-centric representation. Then a semantic memory is designed to extract the polyp prototypes across different images. The feature at one position can be further enhanced by the contextual embeddings stored in the semantic memory. The enhanced feature is propagated into the features of the previous levels as the multi-scale guidance. The experimental results show that our method achieves better performance than other state-of-the-art methods.

Complicated and regionally representative diurnal cycle characteristics of clouds may introduce some errors in the cloud mask (CLM) algorithm of the Geostationary (GEO) meteorological satellite imaging system, which are very difficult to be assessed by using analogous products of fixed-passing polar-orbiting satellites. In this investigation, the diurnal cycle of the performance of the CLM algorithm of the Advanced Geosynchronous Radiation Imager onboard the China Fengyun-4A satellite (FY-4A/AGRI) is validated by using manually observed cloud covers (CC) at 25 ground-based stations in China. The results indicate that the CCs calculated by the FY-4A/AGRI CLM algorithm are overestimated at 11:00 BJT (Beijing Time) and 14:00 BJT (around noon) and underestimated at 08:00 BJT and 20:00 BJT (in the morning and evening) at most stations. In summer, compared with other seasons, the CCs obtained from the FY-4A/AGRI over northern China and the Tibetan Plateau are much better, consistent with the manual observations, but the situation is the opposite in southern China. The CC results retrieved at the vegetation surface by FY-4A/AGRI, however, show the best and stable performance. Because of that, the two independent cloud tests induce most of the overestimations, and some sensitivity experiments for the CLM algorithm are conducted. The results show that the best improvement effect is achieved after only closing one cloud test using the$3.8\mu \text{m}$band. Many extremely overestimated CC samples (about 56.3%) are eliminated. After that, the FY-4A/AGRI CLM product is more reasonable compared with the corresponding infrared and visible imageries.

Recently, deep convolutional neural networks (CNNs) have made important progress in cloud detection with powerful representation learning capability and yield significant performance. However, most existing CNN-based cloud detection methods still face serious challenges because of the variable geometry of clouds and the complexity of underlying surfaces. It is attributed that they only use the fixed grid to extract contextual information, which lacks internal mechanisms to handle the geometric transformations of clouds. To tackle this problem, we propose a deformable convolutional cloud detection network with an encoder-decoder architecture, named DCNet, which can enhance the adaptability of a model to cloud variations. Specifically, we introduce deformable convolution blocks at the encoder to capture saliency spatial contexts adaptively based on the morphological characteristics of clouds and generate high-level semantic representations. After this, we incorporate skip-connection mechanisms into the decoder that integrate low-level spatial contexts as guidance to recover high-level semantic pixel localization and export precise cloud-detection results. Extensive experiments on the GF-1 wide field-of-view (WFV) Satellite Imagery demonstrate that DCNet outperforms several state-of-the-art methods. A public reference implementation of our proposed model in PyTorch is available athttps://github.com/NiAn-creator/deformableCloudDetection.git.

This study mainly focuses on the diurnal and long-term variation characteristics of the calibration performance of the geostationary (GEO) meteorological satellite Fengyun-4A/ Advanced Geostationary Radiation Imager (FY-4A/AGRI) infrared (IR) channels from June 1, 2017 to December 31, 2020. An improved algorithm is developed in this research based on thein situobservations from ocean drifters, which can capture the diurnal variations of the calibration performance of FY-4A/AGRI IR channels. The results suggest that there are significant and nonnegligible diurnal variations of the uncertainties and biases for the brightness temperature (TB) observed by the FY-4A/AGRI, especially the IR channels at 3.71, 8.61, 10.83, and$12.07 \mu \text{m}$. Among them, the calibration performance from 16:00 to 18:00 UTC (around the time of local midnight at the subsatellite point of FY-4A) is worse at$3.71 \mu \text{m}$, while TB biases from 04:00 to 16:00 UTC are relatively large at$12.07 \mu \text{m}$. Moreover, after the operational calibration and update in June 2020, the long-term TB biases at channel 08 ($3.71 \mu \text{m}$) obviously decrease from about 5–2.5 K, and the TB biases at channel 14 ($13.54 \mu \text{m}$) increase from about 0.5–3 K, implying a possible positive or negative impact of the calibration update on the calibration performance of IR channels. Overall, this method based on thein situdrifter data can well monitor the on-orbit calibration performance of GEO satellite imaging sensor’s IR channels such as the FY-4A/AGRI, allowing an objective assessment on the effect of calibration update events.

Cloud top properties (CTPs) are important satellite products. However, the failures of CTPs derived from the Advanced Geostationary Radiation Imager of FengYun-4A (FY-4A/AGRI) have been occasionally reported by users. To this end, the feasibility of the operational CTP algorithm has been reviewed. First, this study reveals the slight differences in brightness temperature (BT) maps of thermal infrared (TIR) bands between two different imagers. Further analyses found that the nonretrieved pixels of CTP products from FY-4A/AGRI usually have a negative value down to −13 K in 10.8–13.5-$\mu \text{m}$BT difference (BTD$_{10.8-13.5\,\mu \text {m}}$), and the joint distribution related to BTD$_{10.8-13.5\,\mu \text {m}}$and BTD$_{10.8-12\,\mu \text {m}}$is well separated from those successfully retrieved ones. These findings are confirmed by the simulations of the radiative transfer forward model (approaching −12 K or lower) and the cross-validation between the products from AGRI/FY-4A and the Infrared Atmospheric Sounding Interferometer of Meteorological Operational Satellite Program-Satellite B. In essence, the bias at the 13.5-$\mu \text{m}$band is mainly affected by the relatively low accuracy and stability at the low-temperature end. Based on these findings, we have proposed a novel method to estimate calibration-related measurement biases of TIR bands and track their in-orbit stability. The statistical study based on the FY-4A/AGRI observations reveals significant daily and diurnal variations in performance and provides an insight into its stability at the TIR band ($13.5~\mu \text{m}$).

GaoFen-4 (GF-4) and Himawari-8 (H8) imagery data were utilized to demonstrate and validate the impact of enhanced spatiotemporal imaging resolution when tracking Super Typhoon Nepartak (2016). The GF-4 remote sensing satellite is the China’s first civilian high-resolution geostationary optical satellite, which has been launched at the end of December 2015.A classical TV-L1 optical flow (OF) algorithm and H8 cloud-top products were also presented to derive cloud-tracking motions, rotating centers, and geometric centers of Super Typhoon Nepartak to investigate the inner-core dynamics and structures of the typhoon. The typhoon positions of the rotating centers derived from the lowest velocities using GF-4 showed good agreement with centers derived using the H8 cloud-top pressure and height threshold-based method. The OF method failed to retrieve the typhoon center using H8 imagery data due to the relatively coarse spatiotemporal resolution. Conversely, refined features and shifts in the track of Typhoon Nepartak were apparent in the GF-4 imagery data as compared to H8. These findings illustrate the significant impact of a satellite imaging system with a higher spatiotemporal resolution when investigating the dynamics features of typhoon.

Random forests (RFs), an advanced machine learning (ML) method, was used here to develop a robust and rapid quantitative precipitation estimates (QPEs) algorithm for the new-generation geostationary satellite of Himawari-8. In this algorithm, the global precipitation measurement (GPM) product has been employed to train QPE prediction model. The real-time multiband infrared brightness temperature from Himawari-8, combined with the spatiotemporally matched numerical weather prediction (NWP) data from the global forecast system, have been used as predictor variables for QPE. Among the variables used in RF learning model, total precipitable water and$K$-index from NWP data have the highest rankings, indicating the importance of atmospheric environment for QPE. To enhance the accuracy of RF models or to optimize model training, a sample-balance technique has been utilized to adjust the ratios of samples in nonprecipitation/precipitation classification and quantitative precipitation regression data sets. Further sensitivity and validation analyses help determine the optimal RF classification and regression models for predicting nonprecipitation/precipitation pixel and rain rate. The selected RF classification model is found to predict precipitation area with an accuracy of 0.87. For predicted QPE product, the mean-absolute-error and root-mean-square error of RF regression model are 0.51 and 2.0 mm/h, respectively. Overall, the RF ML algorithm has a higher detection rate over homogenous ocean surface as compared with over land. Meanwhile, this RF algorithm tends to underestimate rain rate, especially in the presence of heavy rainfall. Despite this, it still produces a reasonable pattern of rainfall area and intensity, which are highly consistent with GPM observations.

In this paper, we developed a unified and operational cloud mask algorithm for the new-generation geostationary (GEO) meteorological satellite imagers of the Advanced Geostationary Radiation Imager (AGRI) aboard Fengyun-4A (FY-4A) and the Advanced Himawari Imager (AHI) aboard Himawari-8 (H08). We investigated the all-round performance of the cloud mask algorithm. Spatiotemporally, the algorithm matches the official Collection-6 cloud mask products of a Moderate Resolution Imaging Spectroradiometer (MODIS) from both the Terra and Aqua platforms, which we employed as the benchmark for performing intercomparisons and validations. The robust cloud mask algorithm can show high consistency between FY-4A/AGRI and H08/AHI. The MODIS-based validation results suggest that cloudy scene identification is better than that observed for clear skies for both FY-4A/AGRI and H08/AHI; there is also a relatively low false-alarm ratio (FAR). Moreover, the algorithm is more reliable during daytime hours, with a hit rate (HR) of approximately 92% for both FY-4A/AGRI and H08/AHI. We found slightly higher accuracy in cloud-masking results over water than those over land. Furthermore, we found that more than 67% of the matched pixels for both advanced GEO imagers had no bias when taking MODIS as the benchmark. Overall, HR values were approximately 91.04% and 91.82% for FY-4A/AGRI and H08/AHI, respectively. These results confirm the high quality of the algorithm for retrieving real-time cloud mask products.

Near-infrared (NIR) channels, such as the 1.6- and 2.13-μm channels of Moderate Resolution Imaging Spectroradiometer (MODIS), play an important role in inferring cloud properties because of their sensitivity to cloud amount and particle size. Instead of the 2.13-μm channel, which has shown great success on MODIS, the central wavelength of the Visible Infrared Imaging Radiometer Suite (VIIRS) is shifted to 2.25 μm. This paper investigates the influences of NIR channels (i.e., 2.13 and $2.25 μm) on cloud optical and microphysical property retrievals and reveals the potential applications of the 2.25-μm channel to cloud thermodynamic phase and multilayer cloud detections by combining with the 1.6-μm channel. Rigorous radiative transfer simulations are performed to provide theoretical reflectance at the channels of interest, and MODIS and VIIRS observations are used for case studies. Our results indicate a minor influence of the 2.25-μm channel on cloud optical depth and effective particle size retrievals. In combination with the 1.6-μm channel, the 2.25-μm channel provides additional information indicating cloud phases. However, the 1.6- and 2.13-μm channels do not show any sensitivity to cloud phase. Furthermore, by considering the infrared-based cloud phase results, the 1.6- and 2.25-μm channel combination becomes possible to infer multilayer clouds. Case studies based on simultaneous MODIS and VIIRS observations demonstrate the capability of the 1.6-2.25-μm channel combination for determining cloud phase and multilayer clouds. Collocated satellite-based active lidar observations further validate these advantages of the 2.25-μmu channel over the original 2.13-μm channel.

The Medium Resolution Spectral Imager (MERSI) was installed as a key payload on the FengYun-3C (FY-3C) polar-orbit meteorological satellite, which was successfully launched on September 23, 2013. We used 14 months of continuous FY-3C/MERSI Level-1B data (from November 2013 to December 2014) to evaluate the on-orbit image spatial quality. Based on a polar ice block image, a sharp target modulation transfer function (MTF) estimation method is used to quantitatively estimate the MTF value at the Nyquist frequency, which is an index for the spatial quality. The results show very good stability in the first year and a relatively lower spatial quality (MTF approximately 0.15) along the FY-3C/MERSI scan direction. This lower spatial quality is primarily attributed to the known and fixed 27% overlapped scan mode of MERSI, which can significantly reduce the image contrast. By using this fixed overlapped proportion (27%), we develop a fast and robust image restoration algorithm based on the Gaussian elimination (GE) method with lower and upper triangular matrix decomposition (LU). The speed-up ratio of this GE with LU decomposition method can attain a value of 626.30 compared with the traditional GE method when it solves linear equations with 2048 MERSI scan pixels. After the image restoration process, significant enhancement in the image spatial quality along the scan direction for every band of FY-3C/MERSI can be found with an increased MTF value of approximately 0.30. However, we evaluate the possible effect of this restoration algorithm on the original digital number (DN) and reflectance values. We find a slight decrease in the total averaged DN (0.5) and reflectance (<; 0.5%, relative bias) values. The variation in DN or reflectance after the image restoration process exhibits a positive correlation with homogeneity of the original target. Moreover, a sensitivity study on the reflectance reveals that it has a more significant impact on the inhomogeneous pixel with a low DN value.

Practices of developing applications based on SOA principles need to be renovated to meet new requirements for large-scale data processing and to adapt advancements in IT infrastructures in cloud computing. Resource planners will have a significant role as they bridge the development and operation phases in application development. To accomplish these tasks, this paper presents architecture design methods, which reinforce designing and assessing architecture with an applied Design Structure Matrix (DSM). This approach can embrace development of both business logics of applications and IT resources to run them in the cloud and the methods are established through consulting practices on a datacenter.

Because the advances in research technologies and the geospatial data with which they deal are diverse and complex. The OGC Web Processing Service (WPS) has several areas of complexity. Basic request-response mechanisms need to contend with delays/failures, especially for mid-term or long-term actions. The asynchronicity in communication between a user and the corresponding Web processing service, or between two services, is significant. The Web Notification Service (WNS), an OGC notification and communication service by which a client may conduct asynchronous dialogues (message interchanges) with one or more other services, can be useful when satisfying a client request requires many collaborating services and/or when there are significant delays in satisfying the request. WNS-based asynchronous notification middleware has been implemented by augmenting the OGC Web Processing Service with message-based asynchronous notification to resolve WPS asynchronous communication and notification problems.

The geospatial data and service capabilities provided by data servers compliant with the Web service specifications of the Open Geospatial Consortium (OGC) are very important to the members of the science community of the Coordinated Enhanced Observing Period (CEOP) who are familiar with the Open-source Project for a Network Data Access Protocol (OPeNDAP). This paper discusses the implementation of middleware to translate the OPeNDAP and OGC protocols to make OGC-compliant data services available to the CEOP community. At the front end this middleware acts as an OPeNDAP server while at the back end it acts as an OGC WCS client. The middleware is being developed through a phased approach. In the first phase, the OGC Web Coverage Service (WCS) connector was developed. The second phase that we are working on is to develop the OGC Catalog Service for Web (CSW) connector which augments OPeNDAP with the catalog search capability. The optional third phase will develop the OGC Web Feature Service (WFS) connector.

With the widespread use of the Internet a large amount of geographical information is currently being stored and delivered over the Internet. Geographic references of Web page are information entities that are discovered from the context and can be mapped to some geographic locations. They are very useful for explaining real world information and for knowledge discovery. In the paper, a Geo-ontology is designed and developed to describe geographic references on the Semantic Web. Based on it, a tool is developed to georeferencing and the result is saved in OWL .The research is an attempt to step to Semantic Web and provide us with direction for future work.