Built Environment Reasoning from Remote Sensing Imagery Using Large Vision-Language Models

A new research paper submitted to arXiv investigates the application of large vision-language models (LVLMs) for smart city development and built environment analysis. The study leverages remote sensing imagery at multiple spatial scales as inputs for multimodal language modeling to characterize urban structures. Key tasks addressed include generating design suggestions, assessing constructability, identifying land-use patterns, and detecting potential risks. The authors evaluate the performance of state-of-the-art large language models, specifically InternVL and Qwen, focusing on their accuracy and reliability in producing built-environment-related recommendations. The findings highlight the significant potential of integrating remote sensing data with advanced AI models to assist in urban planning and decision-making processes. This work represents a technical advancement in using artificial intelligence for geospatial analysis and urban infrastructure management, demonstrating how multimodal AI can enhance the efficiency and precision of smart city initiatives.