Timely and accurate rice spatial distribution maps play a vital role in food security and social stability. Early-season rice mapping is of great significance for yield estimation, crop insurance, and national food policymaking. Taking Tongjiang City in Heilongjiang Province with strong spatial heterogeneity as study area, a hierarchical K-Means binary automatic rice classification method based on phenological feature optimization (PFO-HKMAR) is proposed, using Google Earth Engine platform and Sentinel-1/2, and Landsat 7/8 data. First, a SAR backscattering intensity time series is reconstructed and used to construct and optimize polarization characteristics. A new SAR index named VH-sum is built, which is defined as the summation of VH backscattering intensity for specific time periods based on the temporal changes in VH polarization characteristics of different land cover types. Then comes feature selection, optimization, and reconstruction of optical data. Finally, the PFO-HKMAR classification method is established based on Simple Non-Iterative Clustering. PFO-HKMAR can achieve early-season rice mapping one month before harvest, with overall accuracy, Kappa, and F1 score reaching 0.9114, 0.8240 and 0.9120, respectively (F1 score is greater than 0.9). Compared with the two crop distribution datasets in Northeast China and ARM-SARFS, overall accuracy, Kappa, and F1 scores of PFO-HKMAR are improved by 0.0507–0.1957, 0.1029–0.3945, and 0.0611–0.1791, respectively. The results show that PFO-HKMAR can be promoted in Northeast China to enable early-season rice mapping, and provide valuable and timely information to different stakeholders and decision makers.

Rapid and accurate assessment of building damage is crucial for effective post-disaster emergency response. The use of pre-disaster and post-disaster satellite imagery is a common approach for detecting building damage. This task involves two essential subtasks: building localization and damage classification. In building localization, the imbalance between buildings and background, along with low recall rates, often leads to boundary deviations, which negatively impact the accuracy of subsequent damage classification. In damage classification, features from both pre-disaster and post-disaster images, combined with localization results, are used; however, variations in imaging modalities and insufficient feature extraction from temporal images can introduce interference and reduce classification performance. To address these challenges, we propose a novel two-stage network, referred to as DDNet. In the first stage, the building localization network utilizes differential upsampling connections to enhance detailed feature acquisition and employs a unified focal loss to mitigate class imbalance between buildings and background, thereby balancing precision and recall. In the second stage, a joint attention module is introduced to effectively mine features from pre-disaster and post-disaster images, leading to improved classification accuracy. Finally, a connected component analysis algorithm is applied to convert pixel-level detection results into building-level damage outputs. On the xBD dataset, the proposed framework achieves a total F1 score of 79.56%, an F1 localization score of 86.38%, and an F1 damage classification score of 76.64%.

Monitoring grassland growth in large areas usually needs multiple images from different sensors or on different dates to cover the study area completely. Images from different sensors or on different dates need consistency correction to eliminate the sharp differences between images. The main contribution of this study is to promote a method for consistency correction of images on different days by constructing a linear regression equation of land cover types and the classification pixel mean. Taking a prefecture-level area in China as a test area, the consistency corrected images were applied for monitoring grassland growth. The results showed the following. First, compared with the normal correction equation constructed for two images, taking all features into account, the coefficient of determination of the equation corrected by the land cover types was improved, and the root mean square error was also significantly reduced. Secondly, the areas of consistency in the corrected image were improved compared with the original image, with an improvement rate of 21% for images from the same sensor and 25% for images from different sensors. The pixel average was much closer to the benchmark images, indicating that the corrected image was more consistent than the original image. Thirdly, when applied for monitoring grassland growth, consistency correction can solve the problem of misjudging grassland degradation. Grassland that was judged to be degraded using direct imagery, in fact, showed stable growth after consistency correction, and this type accounted for 7.33% of the regional grassland area. The seasonal characteristics of grass growth in the region were also obtained by monitoring the growth of grass in the region throughout the year. The application test showed that an effective image consistency correction method can improve the accuracy of grassland growth monitoring across a large area.

Unmanned aerial vehicle (UAV) remote sensing provides a flexible and efficient approach to early forest fire monitoring and warning. However, small fire spots and smoke plumes in UAV imagery vary substantially in texture, morphology, and spatial distribution across scales. Interactions between complex forest backgrounds and target features cause information loss during successive feature compression and misaligned spatial and semantic responses during cross-layer fusion. Limited onboard computing and storage further constrain model complexity. To address these challenges, we propose the lightweight feature alignment and multi-scale enhancement network (FAME-Net), which alleviates feature degradation and response mismatches during encoding and fusion while reducing computational redundancy in multi-scale prediction. At the feature encoding stage, the lightweight multi-scale enhanced backbone (LMEB) combines lightweight downsampling with multi-scale context modeling to reduce computational overhead and preserve fine details of small fire spots and weakly textured smoke. Its hierarchical features are fed into the alignment-aware multi-scale feature fusion neck (AMFF-Neck), which coordinates spatial and semantic responses across scales through cross-layer interactions, reducing fusion bias caused by scale discrepancies and background-induced coupling. The resulting features are processed by the lightweight shared convolution detection head (LSCD), which performs scale alignment, shared convolution, and learnable scale compensation for efficient multi-scale prediction. FAME-Net achieves mAP50 values of 71.5% and 65.3% on the integrated forest fire dataset and the Yuxi Wildfire dataset, respectively, with 1.7 M parameters and 3.9 GFLOPs, indicating a favorable accuracy–efficiency trade-off. On an NVIDIA Jetson Orin Nano Super (8 GB), it achieves inference speeds of 36.2 and 31.6 FPS on the two datasets, respectively, further supporting its potential for resource-constrained UAV deployment.