Mutual learning that can be seen as a derivative method of knowledge distillation takes advantages of the collective capabilities of multiple neural networks to promote the precision and stability of classification outcomes, and achieved great success in the field of computer vision. Recently, mutual learning has been exploited in the task of medical image classification. However, medical images belonging to different classes may be very similar, which may affect the classification performance of the mutual learning based methods. In this paper, we propose a novel mutual learning model, named DualDoctor, especially for the classification of medical images. DualDoctor is built by embedding the channel and spatial feature-based dynamic knowledge transfer module and the logit-based class alignment module into the vanilla mutual learning framework. The dynamic knowledge transfer module makes two sub-models sharing the critical information at the feature level, while the logit-based class alignment module enforces two sub-models to absorb the class correlation knowledge from each other at the logit level. Therefore, DualDoctor can adapt well to the characteristics of medical images so as to make more accurate classifications. The evaluation results on two public datasets, BUSI and ISIC2018, indicate that the proposed DualDoctor model performs better than the comparison methods in medical image classification tasks.

Due to the amplified climate effect of the Arctic, the volume of Arctic sea ice and the amount of multiyear ice have been continuously decreasing, and the melting time of summer snow has gradually get earlier, which greatly affecting the exchange of water and heat between the surface and atmosphere in the Arctic, thereby accelerating the warming of the Arctic which have shown a significant impact on global climate change. Taking the FY3B (FengYun-3B)/MWRI (MicroWave Radiometer Imager) as an example, this paper studies how to use the brightness temperature of the Microwave radiometer to effectively determine the sea ice Type. Based on the differences in radiation characteristics of different sea ice types, we use the contrast ratio and dynamic threshold method to retrieve the classification of the Arctic sea ice. Comparing the results with The Global Sea Ice Type products, it is shown that the misjudgment of multiyear ice on offshore areas have significant reduction in our results, and the multiyear ice coverage is basically consistent between the two types of data. The method used in this study can effective retrieve sea ice types in the Arctic.

Remote sensing image change detection (CD) is an important technology for monitoring ground object change. Although Transformer-based CD methods have been proposed and achieved good results, however, there does exist one open problem: Transformer-based methods are weak for localizing information acquisition, easily ignore detailed information, and are of high computational complexity. Also, the variation of target sizes challenges the generalization of networks. To address these issues, we propose a siamese network named as CBSASNet for remote sensing CD, in which channel bias split attention is employed to recover the information in the change region and the cross-temporal fusion module is utilized to highlight the information of change regions through the optimized single-temporal image features. The experimental result indicates that CBSASNet does not only outperform 16 state-of-the-art works, but also its modules complement each other in the ablation testing.

Breast cancer has become one of the most common cancers in the world, and it is also the most lethal cancer in women. As a non-invasive imaging modality, ultrasonography can diagnose the degree of breast lesions and be used for large-scale screening. However, since the lesions in breast ultrasound(BUS) images are morphologically diverse, accompanied by relatively low contrast and complex textures, BUS image recognition faces greater challenges than natural images. In this study, We propose a novel network architecture that combines convolutional neural network(CNN) with vision transformer(ViT) to aggregate local feature details and long-range feature dependencies. Moreover, in order to perform multi-scale feature fusion, we introduce cross attention between the deep feature map and the shallow feature map in the network block to carry out the interaction between the deep feature and the shallow feature information. To verify the effectiveness of the model, we constructed a large-scale dataset and conducted extensive experiments. The results show that our method achieves an accuracy of 85.33%, under the comparable parameter complexity, which outperforms most convolutional neural networks(CNNs) and vision transformers (ViTs).

Snow on sea ice is a sensitive indicator of climate change and is considered a key reason for amplified warming in the Arctic. The passive microwave remote sensing is an important method for Snow Depth (SD) observation. Therefore, understanding the influence of SD on observed brightness temperature (TB) is necessary for SD retrieval using microwave radiometer. Taking the FengYun-3D/Microwave Radiation Imager (FY3D/MWRI) as an example, this paper simulates the influence of SD on observed TBs from different channels using the MicroWave MODel (MWMOD) in the Arctic. Several conditions are considered for both first year ice (FYI) and multi-year ice (MYI), including various sea ice concentrations and water vapor contents in the atmosphere. The correlation between SD and TB is also briefly analyzed. It provides a theoretical reference for the retrieval of SD using microwave radiometer.

The significant wave height (SWH) is an important factor in the study of polar ocean and the forecast of polar marine environment. This study evaluated the SWH observed by the microwave altimeter (ALT) on board the HaiYang-2B (HY2B) satellite, using Cryosat2/SIRAL (SAR/Interferometric Radar Altimeter) and Icesat2/ATLAS (Advanced Topography Laser Altimeter System) data Comparison results show that the bias and the standard deviation (STD) of the SWH difference between HY2B/ALT and Cryosat2/SIRAL are −0.05±0.26m, and the difference between HY2B/ALT and Icesat2/ATLAS are 0.02±0.41m. The SWH from HY2B/ALT shows good consistency with that from Cryosat2/SIRAL and Icesat2/ATLAS. It can effectively reflect the real wave conditions in the Arctic

Thin ice thickness retrieval algorithm was developed by using FY-3D/MWRI brightness temperature in the Arctic. The algorithm was based on the fitting function models between polarization ratios of MWRI brightness temperature from 36GHz and 89GHz channels$(PR_{89}$and$PR_{36})$and the thermal ice thickness. Thermal ice thickness was obtained by using MODIS sea ice surface temperature data (MYD29) and ERA5 reanalyze meteorological data through the thermodynamic balance function of sea ice surface. The linear fitting model and the exponential fitting model were used to fit$\text{PR}_{89}$and$\text{PR}_{36}$respectively to thermal ice thickness. The root mean square errors of the thin ice thickness fitted by$\text{PR}_{89}$and$\text{PR}_{36}$were 0.003m and 0.0019m, respectively, and the corresponding standard deviations were 0.0505m and 0.0496m.

The inter-calibration of the brightness temperature between FY-3B/MWRI and Aqua/AMSR-E is studied by 8858 files of MWRI L1 and 9327 files of AMSR-E L2A from November 18, 2010 to September 30, 2011 with the spatial coverage on north of 60°N. According to the polar projection, time-space matching and linear fitting, the inter-calibration parameters of the MWRI and AMSR-E are achieved. From the scatters of each channel and the statistic analysis, it shows that there are some deviation between the two sensors but it has obvious consistency on overall trend and the correlation coefficients of the TB in H and V polarization among all channel are more than 0.9. Using the linear regression analysis, the slopes and intercepts of the fitting equations for 10.7~89.0GHz H/V ascending and descending TB data on every month of the data sets were achieved. After inter-calibration the statistic parameters between the two sensors are obviously optimized.

Sea ice is a sensitive indicator of climate change. This study focuses on retrieving sea ice concentrations from the brightness temperatures recorded by the Microwave Radiation Imager (MWRI) on board the FengYun (FY)-3B satellite. After cross calibration with the Advanced Microwave Scanning Radiometer–EOS (AMSR-E) data from July to September 2011, MWRI brightness temperatures are used to calculate the sea ice concentrations based on the Arctic Radiation and Turbulence Interaction Study Sea Ice (ASI) algorithm with different tie points’ combinations. The combination corresponding to the most similar results to AMSR-E sea ice products is selected as the new tie points to retrieve the sea ice concentrations in the Arctic. After comparing with the products of AMWR-E and MWRI, the sea ice concentrations from this work are validated by Aqua/Moderate Resolution Imaging Spectroradiometer (MODIS) calibrated radiances data.

Sea surface temperature (SST) significantly affects the processes of air-sea interactions and thus forms an important indicator of climate changes. In SST predictions, the approach of artificial neural networks (ANNs) is data-driven, unlike that of the numerical models, which are physics-based. In this paper, Operational Sea Surface Temperature and Ice Analysis (OSTIA) dataset was used for training ANN models and verifying prediction results. To reduce the prediction error caused by SST variations, the authors propose to separate SST time series data into climatological monthly mean and monthly anomaly datasets, and construct two neural network models. The combination of these two models gives the final SST prediction results. This method was used to predict the SST in the South China Sea. The average bias and standard deviation between the predicted SST and OSTIA SST are -0.02 °C and 0.37 °C, respectively. The results indicate that the proposed training method gives good prediction accuracy.

Given the high albedo and low thermal conductivity, snow on sea ice is regarded as one of the key reasons for the amplification of the warming in polar regions. This study focused on the retrieval of snow depth on sea ice from brightness temperatures of the Microwave Radiation Imager (MWRI) on board the FengYun (FY)-3B satellite. After cross calibrated to the Advanced Microwave Scanning Radiometer-EOS (AMSR-E) Level 2A data, the MWRI brightness temperatures were applied to calculate the snow depths on sea ice according to the proportional relationship between the snow depth and the surface scattering in 18.7 and 36.5 GHz. The results were compared with the snow depths from two data sets: IceBridge ICDIS4 data set and the AMSR-E Level 3 Sea Ice products. It is proved that the method taken in this study is feasible and the results are reasonable.

The accuracy of satellite-derived sea surface temperature (SST) fields has drawn a great deal of attention, while little attention has been focused on the spatial precision. Very limited information of the uncertainty in pixel-to-pixel scale in the retrievals is provided by satellite in-situ matchups, which are widely separated in space and time because of cloud cover and the lack of in-situ observation. But the main contribution to the uncertainty in satellite retrievals results from atmospheric contamination - the spatial scale of which is, in general, large compared with the pixel separation of infrared sensors, hence the pixel-to-pixel uncertainty is often smaller than the accuracy determined from in-situ matchups, this makes selection of satellite-derived datasets for the study of submesoscale processes, for which the spatial structure of the upper ocean is significant, problematic. In this study, we develop an approach to evaluate the spatial fidelity of Advanced Very High Resolution Radiometer (AVHRR) SST fields, the spatial power spectra density (PSD) of which are compared with an in-situ measurement in Sargasso Sea. We found AVHRR spectra have elevated energy at the submesoscale, and the spectra tend to level off at smaller scale results from the instrument noise, the level of which is estimated based on the spectra comparison.

Accurate and high-quality depth maps are required in lots of 3D applications, such as multi-view rendering, 3D reconstruction and 3DTV. However, the resolution of captured depth image is much lower than that of its corresponding color image, which affects its application performance. In this paper, we propose a novel depth map super-resolution (SR) method by taking view synthesis quality into account. The proposed approach mainly includes two technical contributions. First, since the captured low-resolution (LR) depth map may be corrupted by noise and occlusion, we propose a credibility based multi-view depth maps fusion strategy, which considers the view synthesis quality and interview correlation, to refine the LR depth map. Second, we propose a view synthesis quality based trilateral depth-map up-sampling method, which considers depth smoothness, texture similarity and view synthesis quality in the up-sampling filter. Experimental results demonstrate that the proposed method outperforms state-of-the-art depth SR methods for both super-resolved depth maps and synthesized views. Furthermore, the proposed method is robust to noise and achieves promising results under noise-corruption conditions.