Weakly supervised medical image segmentation seeks to reduce reliance on exhaustive pixel-wise annotations by leveraging coarse or indirect supervision, thereby lowering annotation costs while maintaining segmentation quality. Existing methods rely on coarse heuristics, lacking spatial fidelity and semantic localization in complex medical imagery. To address this limitation, we propose a Gaze-based Weakly Supervised Medical Image Segmentation (GWMIS) framework that exploits gaze annotations as rich, low-cost semantic priors. GWMIS consists of two streams: a pseudo-label generation stream and a cross-modal fusion segmentation stream. The former fuses gaze-derived Gaussian priors with Segment Anything Model (SAM) for structure-consistent pseudo masks; the latter aligns multi-scale visual and gaze features via dual-encoder crossattention for anatomically faithful segmentation. Evaluations on colonic polyp (3,784 images, five public datasets) and brain tumor segmentation (3,929 MRI slices, LGG dataset) demonstrate superior performance over state-of-the-art weakly supervised approaches.

This paper describes the composition and test results of an advanced Ka/X dual-band (35 GHz/9.6 GHz) airborne synthetic aperture radar (DASAR) system in China. The system was developed by the Aerospace Information Research Institute, Chinese Academy of Sciences (AIRCAS), with dual antennas coplanar arrangement and integrated with a pod. Ka-band radar has a cross-track interferometry mode, with a baseline length of 0.32m, spatial resolution 0.1m*0.1m (range*azimuth), and elevation accuracy 0.3m; X-band is fully polarimetric synthetic aperture radar (SAR) with spatial resolution 0.3m*0.3m (range*azimuth). DASAR is a highly integrated system with strong platform applicability. It is designed for dual-band observation, with potential for high-precision multi-source remote sensing (MSRS) applications.

Accurate crop type map over large area at high resolution plays an important role in national food production security. However, there are not so far large-area crop cover data product, mainly because the traditional classification methods are highly dependent on the complicated process, including data pre-process, feature-selection and various classifier decision, which results in high labor-cost and time-consumption. To overcome these challenges, we creatively introduce a deep learning strategy to produce a large-area autumn crop type map. Using a convolution neural network model to train standard GF-1 medium resolution images (16m) from Liaoning Province, China, we generated a basic model. Then, we applied the basic model to predict the autumn crop within Liaoning region, and further extensively predicted the autumn crop (rice and corn) within Jilin Province. And both provinces received satisfied overall accuracy with 89.9% for Liaoning and 84.1% for Jilin. It can conclude that the deep learning strategy is an effective solution to large-area crop mapping, which would pave a way to develop the large-area crop type product annually to meet with the requirement of the policy-making, research and agriculture management.

Deep learning (DL) method has been a state-of-art method for land cover mapping. However, DL is so data-hungry which makes it difficult to obtain the large amount label data feed into the process to ensure high performance. In this paper, we utilized the historical data which is the crop distribution map extracted by support vector machine algorithm as training dataset which was produced by traditional classification and was used to train a convolution neural network as a basic model, called as RSNet. Then, the basic model was fine-tuned which is usually used in this deep learning model training to improve the classification accuracy. The F1 scores of rice and maize were 78.91 % and 72.94%, respectively. The result verified that the migration ability of RSNet model is strong to produce the current year crop map using the historical labeled samples. From the fine-tuning model, higher overall classification accuracy of crop classification was achieved as 90.36%. This classification strategy brings us a bright future to map the crop distribution supported by the historical labelled dataset, which solves the data-hungry issues.

It is significant to extract cropland mapping accurately and rapidly in many fields. In southern China, restrictions on complex planting structure and fragmented patches, long rainy season all pose challenges to remote sensing. Deep Learning has advantages in such complex classification problem. The combination of medium and high resolution data is an effective way to solve the problem of data acquisition and scale refinement in fragmented agricultural landscape area. In this paper, a modified Pyramid Scene Parsing Network which could integrate the medium-resolution data and the high-resolution data (HM_PSP) was proposed focusing on different scales features. The experimental results showed the applicability of the HM_PSP, which also had better visual results, the 3.42 % and 1.58% improvement of OA for the test sets. By comparing the results of HM_PSP and M_PSP (only trained by the medium resolution data), the enhanced effect after adding high-resolution data, as well as the generalization ability of spatial and temporal generalization ability of the model had been verified.

In recent years, the use of deep learning in remote sensing domain has made it possible to automate mapping in large-scale. In this paper, we propose a transfer learning method which pre-train a convolutional neural network (CNN) with middle-resolution remote sensing data in 2016, and fine-tune it in following years with a spot of high-resolution remote sensing data in 2017. We used the fine-tuned model to mapping the early-rice in 25 countries which cost only 21 minutes, and yielded an overall accuracy of 81.68%. The result demonstrate that the convolutional neural network model can transfer in different time period with little adjustment in a very high accuracy.

In this paper, an innovative method of integrating hard and soft classification, extended support vector machines, is introduced to map impervious surface in urban-rural fringe area, where is experiencing dramatic land cover change. The proposed method incorporates the variance texture information as an effective indicator to represent the heterogeneity of urban-rural fringe and possesses the merits of hard and soft classification to map both pure pixels and mixed pixels of impervious surface. The better performance compared with conventional hard and soft method was demonstrated in urban-rural fringe area, Beijing.

This paper developed an approach to determine optimal parameters, C and s, for support vector domain description (SVDD) model to map specific land cover from integrating of training and window-based validation sets (WVS-SVDD). The validation set based on window-based approach made a tighten hypersphere because of compact constraint by the outlier pixels which were located closely to the target class in the feature space. The target land coves of wheat and bare land were considered to test the proposed method's performance. The overall accuracy for wheat reached as high as 94.37%. However, the underestimation of wheat, only 71.12% of the user's accuracy, attributed to the validation set covering a small portion of wheat spectra. The larger window sizes were tested to achieve more wheat pixels' samples for validation set. The results showed that wheat accuracy were improved along with window size increasing and the overall accuracies were higher than 88% The bare land as more heterogeneous land cover against wheat was selected to analyze the applicability and suitability of WVS-SVDD, SVDD classification was conducted and compared to the support vector machine (SVM) method as benchmark. The producer's and user's accuracies for bare land were over 80% at 2.4-m resolution scale and the overall accuracies were similar to those produced by the SVM at coarser spatial resolutions, highlighting the applicability of WVS-SVDD. Therefore, the developed method showed its advantages using the optimal parameters, C and s, for mapping homogeneous wheat and heterogeneous bare land, which exhibits great potentials to achieve highly accurate specific land cover.

This paper has proposed a new method that integrates the advantage of optical image for delineating land surface boundaries and the superiority of PolSAR data for obtaining corn information despite bad weather conditions. The comparison between the proposed method and both pixel- and object-based method was made to test their performance for corn classification. The analysis shows that the proposed method can significantly improve the overall accuracy and kappa value of the other two methods, which indicates that the proposed method was suitable for crop mapping using optical and Radarsat-2 PolSAR data.

The SVDD method, one of the most popular one-class classifiers, could use training samples of the interest class to derive accurate classification and thus is adopted in this paper. However, the penalty parameter C and the kernel width s should be tuned carefully to construct an optimal hypersphere. This research developed a single window flexible pace search method to select optimal parameters. First, 120 edge pixels were acquired from parcel boundary and PCA image. Then a 3*3 window was applied to the training samples to obtain the buffer training set. Then optimal parameters were select through the flexible pace search method. Under optimal parameters, the buffer training set yielded an accurate classification with an overall accuracy of 89.70%, which differed slightly with that derived from the SVM classification. Thus, we conclude that our proposed method could be used to select optimal parameters for the SVDD method.

To take advantage of conventional hard land use/cover change detection method (HLUCD) and soft land use/cover change detection method (SLUCD), we develop a soft and hard land use/cover change detection method (SHLUCD) for crop distribution mapping. Two HJ-1/CCD images, acquired on 6 October 2011 (T1) and 16 April 2012 (T2) which represented the period of sowing and jointing respectively, were utilized by SHLUCD to extract wheat area in study area. The results show that the crop distribution derived from the SHLUCD reflects reality more accurately than that from HLUCD and SLUCD. Crops distribution mapping derived from SHLUCD give lowest RMSE and bias and the highest R2than that from other two methods in all window size. Wheat distribution in typical area and mixed pixels zone could be identified by land use/cover change status and land change scope respectively through SHLUCD. Moreover, the theory and methods employed in developing the SHLUCD provide a new way for crop distribution mapping based on change detection technique.