面向智慧城市公园开发监测：多模态信息融合与分析的LLM智能体

Monitoring the development of newly constructed urban parks is crucial for evaluating urban planning effectiveness and optimizing resource allocation within the urbanization process. Traditional remote sensing-based change detection methods, however, exhibit significant limitations in high-level and intelligent analysis, thus failing to meet the demands of contemporary urban planning and management. To address these challenges, an intelligent monitoring framework leveraging Large Language Model (LLM) agents for multi-modal information fusion and analysis is proposed. This framework aims to extract and analyze multi-level information pertinent to park development from various remote sensing data, encompassing land use changes, vegetation coverage, facility construction, and human activity patterns. By integrating image recognition, natural language processing, and knowledge graph technologies, LLM agents are capable of comprehending and interpreting complex geospatial data, identifying subtle change patterns, and predicting potential development trends. For instance, LLM agents can analyze construction progress from satellite imagery, evaluate public feedback on park development by combining news reports and social media data, and even infer future park utilization based on historical data. This approach transcends simple pixel-level change detection, enabling semantic-level understanding and reasoning to provide more insightful monitoring reports. Furthermore, the framework supports the generation of natural language analysis reports and recommendations, making complex geospatial information easily understandable for non-expert users and facilitating more effective decision-making in urban planning and resource allocation.