Building extraction refers to the automatic identification and separation of buildings from the background in remote sensing images. It plays a significant role in urban planning, land management, and disaster monitoring. Deep-learning methods have shown advantages in building extraction, but they still face challenges such as variations in building types, object occlusions, and complex backgrounds. To address these issues, SDSNet, a deep convolutional network that incorporates global multi-scale feature extraction and cross-level feature fusion, is proposed. SDSNet consists of three modules: semantic information extraction (SIE), multi-level merge (MLM), and semantic information fusion (SIF). The SIE module extracts contextual information and improves recognition of multi-scale buildings. The MLM module filters irrelevant details guided by high-level semantic information, aiding in the restoration of edge details for buildings. The SIF module combines filtered detail information with extracted semantic information for refined building extraction. A series of experiments conducted on two distinct public datasets for building extraction consistently demonstrate that SDSNet outperforms the state-of-the-art deep-learning models for building extraction tasks. On the WHU building dataset, the overall accuracy (OA) and intersection over union (IoU) achieved impressive scores of 98.86% and 90.17%, respectively. Meanwhile, on the Massachusetts dataset, SDSNet achieved OA and IoU scores of 94.05% and 71.6%, respectively. SDSNet exhibits a unique advantage in recovering fine details along building edges, enabling automated and intelligent building extraction. This capability effectively supports urban planning, resource management, and disaster monitoring.

Accurately delineating soil elements from satellite imagery is fundamental for regional geological mapping and survey. However, vegetation cover and complex geomorphological conditions often obscure diagnostic surface information, weakening the visibility of key geological features. Additionally, long-term tectonic deformation and weathering processes reshape the spatial organization of soil elements, resulting in substantial within-class variability, inter-class spectral overlap, and fragmented structural patterns—all of which hinder reliable segmentation performance for conventional deep learning approaches. To mitigate these challenges, this study introduces a Reinforced Feature and Multiscale Feature Fusion Network (RFMFFNet) tailored for semantic interpretation of soil elements. The model incorporates a rectangular calibration attention (RCA) module into a ResNet101 backbone to recalibrate feature responses in critical regions, thereby improving scale adaptability and the preservation of fine geological structures. A complementary multiscale feature fusion (MFF) component is further designed by combining sparse self-attention with pyramid pooling, enabling richer context aggregation while reducing computational redundancy. Comprehensive experiments on the Landsat-8 and Sentinel-2 datasets verify the effectiveness of the proposed framework. RFMFFNet consistently achieves superior segmentation performance compared with several mainstream deep learning models. On the Landsat-8 dataset, the oPA and mIoU increase by 2.4% and 2.6%, respectively; on the Sentinel-2 dataset, the corresponding improvements reach 4.3% and 4.1%.