Research on the Classification and Planning of Functional Areas in Smart Cities Based on Artificial Intelligence

This study presents an innovative approach for urban functional zoning and planning based on artificial intelligence (AI) techniques. Utilizing multi-modal remote sensing data and geographic information, this research proposes a deep learning framework to accurately classify and segment urban functional areas. The proposed method leverages a Transformer-based Spatio-Temporal Fusion Network (TBST) that integrates high-resolution satellite imagery and time-series population data. The TBST model comprises two feature extraction branches, ResMixer and PDNet, which respectively handle spatial and temporal features, followed by a Transformer-based adaptive fusion layer to enhance the extraction of informal settlements in urban areas. Experimental results demonstrate that the multimodal data fusion significantly improves the accuracy of urban functional area classification, achieving superior performance metrics compared to single-modal approaches. This method holds significant potential for advancing smart city planning by providing precise and up-to-date spatial distribution information, thus supporting urban policy-making and sustainable development.