Research on Intelligent Recognition Model and Method for Information Targets Based on Big Data-Driven Approach

With the continuous development of remote sensing building extraction technology, Current deep learning methods face limitations in building extraction from remote sensing images, particularly in preserving edge details and detecting small targets within complex backgrounds. The proposed MX-UNet is a network architecture integrating Swin Transformer with U-Net. demonstrating superior performance in remote sensing image segmentation, especially for building extraction tasks. The model achieves multi-scale feature fusion through cross-attention modules, effectively balancing global context capture and local detail preservation. Experiments on the Vaihingen dataset of aerial images from the International Society for Photogrammetry and Remote Sensing(ISPRS) demonstrate that our method achieves effective land cover classification, The big data-driven intelligent recognition model demonstrates scalable accuracy, with performance positively correlated to training data volum-larger datasets yield stronger recognition capability, with classification accuracy surpassing existing mainstream methods.