The Orbiting Carbon Observatory‐2 (OCO‐2) collects solar‐induced chlorophyll fluorescence (SIF) at high spatial resolution along orbits ( _{oco2_orbit}), but its discontinuous spatial coverage precludes its full potential for understanding the mechanistic SIF‐photosynthesis relationship. This study developed a spatially contiguous global OCO‐2 SIF product at 0.05° and 16‐day resolutions ( _{oco2_005}) using machine learning constrained by physiological understandings. This was achieved by stratifying biomes and times for training and predictions, which accounts for varying plant physiological properties in space and time. _{oco2_005} accurately preserved the spatiotemporal variations of _{oco2_orbit} across the globe. Validation of _{oco2_005} with Chlorophyll Fluorescence Imaging Spectrometer airborne measurements revealed striking consistency (R² = 0.72; regression slope = 0.96). Further, without time and biome stratification, (1) _{oco2_005} of croplands, deciduous temperate, and needleleaf forests would be underestimated during the peak season, (2) _{oco2_005} of needleleaf forests would be overestimated during autumn, and (3) the capability of _{oco2_005} to detect drought would be diminished.

Landsat-like moderate resolution remote sensing images are widely used in land use and land cover (LULC) classification. Limited by coarser resolutions, most of the traditional LULC classifications that are based on moderate resolution remote sensing images focus on the spectral features of a single pixel. Inspired by the spatial evaluation methods in landscape ecology, this study proposed a new method to extract neighborhood characteristics around a pixel for moderate resolution images. 3 landscape-metric-like indexes, i.e., mean index, standard deviation index, and distance weighted value index, were defined as adjacent region features to include the surrounding environmental characteristics. The effects of the adjacent region features and the different feature set configurations on improving the LULC classification were evaluated by a series of well-controlled LULC classification experiments using K nearest neighbor (KNN) and support vector machine (SVM) classifiers on a Landsat 8 Operational Land Imager (OLI) image. When the adjacent region features were added, the overall accuracies of both the classifiers were higher than when only spectral features were used. For the KNN and SVM classifiers that used only spectral features, the overall accuracies of the LULC classification were 85.45% and 88.87%, respectively, and the accuracies were improved to 94.52% and 96.97%. The classification accuracies of all the LULC types improved. Highly heterogeneous LULC types that are easily misclassified achieved greater improvements. As comparisons, the grey-level co-occurrence matrix (GLCM) and convolutional neural network (CNN) approaches were also implemented on the same dataset. The results revealed that the new method outperformed GLCM and CNN approaches and can significantly improve the classification performance that is based on moderate resolution data.

Background and Objectives: In this article, we propose an image segmentation model based on Chan-Vese (CV) for image segmentation. By taking into account the local features of the image, the new proposed model can successfully segment images with intensity nonuniformity.Materials and Methods: We quantitatively compare our method with other two state-of-the-art algorithms, namely, CV model and local binary fitting (LBF) model in segmenting synthetic MR images with the ground truth from BrainWeb; the data can be available at: https://www.mni/mcgill.ca/brainweb/. For segmenting the missing and weak boundaries, to deal with the intensity inhomogeneity, based on the LBF model, we introduced the convex total variation regularization term, for explicit smoothing of the level set function ø. The evolution equation will be solved through the level set method of calculus of variations.Results: In the experimental processing, we use some real images and magnetic resonance imaging brain images as the experimental images, to validate the stabilization of algorithm. The experimental results on comprehensive and sincerity images show the outstanding of our proposed model with reference to stabilization and availability.Conclusions: We propose a new segmentation local information of an image and introduce a new regularization functional is to keep the level set function smooth. Finally, various experimental results on real and low-contrast image, showing which is a powerful type of images, including some that would be difficult to segment with gradient-based methods. In addition, the advantages of the proposed model are better than CV model and the LBF model. Our new model can effectively segment a real image.

Monitoring grassland productivity dynamics is essential for understanding the impacts of climate variation and human activities. Solar-induced chlorophyll fluorescence (SIF) has been validated as an effective indicator of gross primary productivity. Satellite-derived vegetation indices (VIs) have long been used as key proxies for vegetation productivity. However, the ability of different VIs to represent grassland productivity in relation to SIF, as well as their spatiotemporal consistency with SIF at various scales, remains unclear. In this study, we systematically compared the performance of the Normalized Difference Vegetation Index (NDVI), the Enhanced Vegetation Index (EVI), and the Near-Infrared Reflectance of Vegetation (NIRv), using SIF as a benchmark in grassland areas of China. Utilizing TROPOMI SIF and MODIS VI datasets from 2018 to 2021, we analyzed the spatial and temporal consistency between VIs and SIF at a monthly scale and 0.05-degree resolution, employing Pearson correlation coefficients, paired-sample t-tests, and two-way Analysis of Variance (ANOVA). The results indicate that NIRv consistently demonstrates a higher capacity to capture variations in SIF compared to EVI and NDVI. In low-elevation areas with high-productivity grasslands, all three vegetation indices exhibit a stronger ability to represent vegetation productivity than in high-elevation areas with low-productivity vegetation types. These findings suggest that, at a monthly and regional spatiotemporal scale, NIRv can serve as a robust complement to SIF in monitoring vegetation productivity dynamics, particularly given the challenges in acquiring high-quality, long-term SIF data.

Satellite-retrieved solar-induced chlorophyll fluorescence (SIF) provides an advanced proxy for global vegetation productivity. Recently, new high-quality remote sensing SIF datasets and reanalysis products have significantly advanced the application of SIF. However, the lack of long-term, daily resolution datasets continues to limit the precise exploration of vegetation dynamics, primarily due to challenges in daily modeling accuracy, substantial data volume, and computational demands. In this study, supported by the Google Earth Engine (GEE) platform, we developed a data-driven approach based on the Moving Spatial–Temporal Window Sampling (MSTWS) strategy for reconstructing long-term daily SIF. By learning the relationship between high-spatial-resolution Orbiting Carbon Observatory (OCO)-3 SIF and MODIS surface reflectance, we established a spatially and temporally specific daily prediction model for each day of the year (DOY), reconstructing the long-term daily OCO-3 SIF (DOSIF) from 2001 to the present with a global contiguous distribution. The prediction framework demonstrated robust performance with an R² of 0.92 on the training set and 0.81 on the validation set, indicating strong predictive ability and resistance to overfitting. Systematic evaluation of the dataset showed that DOSIF accurately captures the expected spatiotemporal distribution patterns. Cross-sensor validation with independent airborne SIF measurements further enhanced the reliability of the DOSIF dataset.

Retrieving pedestrian images using natural language descriptions remains challenging due to the prevalence of imperfect annotations in real-world training data. Most existing methods rely on the strong assumption of perfectly aligned image–text pairs, largely ignoring the detrimental impact of annotation noise, which typically manifests as coarse-grained descriptions and erroneous correspondences. These imperfections severely degrade model performance and generalization. To address these issues, we propose a novel framework centered on two key innovations. First, we develop a probabilistic noise identification mechanism that employs a dual-channel Gaussian mixture model (GMM) to assess alignment consistency at both global and local feature levels. Second, for samples identified as noisy, we implement a description synthesis pipeline that leverages a multimodal large language model (MLLM) to generate refined descriptions. A dynamic semantic consistency module then filters these synthesized texts to ensure quality. Comprehensive evaluations on three benchmark datasets—CUHK-PEDES, ICFG-PEDES, and RSTPReid—demonstrate the superior performance of our method: ICFG-PEDES Rank-1 = 68.13%, Rank-5 = 83.39%, Rank-10 = 89.02%; RSTPReid Rank-1 = 66.31%, Rank-5 = 86.87%, Rank-10 = 92.01%; CUHK-PEDES Rank-1 = 75.98%, Rank-5 = 90.34%, Rank-10 = 94.32%. These results show consistent top-k improvements over prior methods and validate the effectiveness of the proposed noise-aware pseudo-text augmentation.