Urban green space zero-waste design faces challenges such as insufficient lifecycle waste reduction, inefficient interdisciplinary knowledge integration, disconnection between site data and design decision-making, and over-reliance on individual experience in design schemes. Traditional experience-driven design models are increasingly inadequate to support the demands for refined and intelligent transformation. To address these issues, this study proposes a Perception–Knowledge–Decision-Making framework driven by multi-source sensing data for the zero-waste design of urban green spaces. The framework collects multidimensional site data through drone imaging sensors, environmental sensors, and laser scanning technology. The study constructs a knowledge graph that semantically links four core entities: strategy, performance, case, and industrial-chain information. Stored in the Neo4j graph database, it enables dynamic knowledge association and efficient retrieval. An intelligent auxiliary decision-making tool is developed by combining knowledge graph-enhanced retrieval and intelligent agent technologies, thereby improving the capabilities for identifying, reasoning, and matching strategies for complex design problems. A simulation-based case study at a community garden in Shanghai was conducted to demonstrate the feasibility of the proposed framework, revealing its potential to improve waste resource utilization matching, design efficiency, and resource utilization effectiveness. This study presents a proof-of-concept framework and provides a theoretical pathway and methodological exploration for sustainable landscape design supported by cutting-edge sensing technologies.