Abstract:Comprehensive, precise, and real-time monitoring of farmland systems is the foundation for driving the digital and intelligent transformation of agriculture. With the development of space-air-ground integrated monitoring networks, leveraging the complementary features of multi-source remote sensing platforms and artificial intelligence algorithms, farmland monitoring is transitioning from a static, two-dimensional (2D), and phenomenologically qualitative approach to a dynamic, three-dimensional (3D), and mechanistically quantitative paradigm. Treating the farmland as an open complex giant system, a stereoscopic monitoring framework was constructed based on the coupling mechanisms of the five core elements: water-soil-biology-atmosphere-human. On this basis, the research progress in intelligent extraction methods for spatial boundaries and agricultural infrastructure was systematically reviewed, including field parcels, rural roads, irrigation and drainage facilities, and shelterbelts. Furthermore, focusing on the four subsystems of water, soil, atmosphere, and biology, an in-depth analysis of advances in retrieval technologies for key variables was provided, such as soil moisture, soil organic matter, greenhouse gases, and land surface/canopy temperature, alongside crop identification, growth monitoring, and yield estimation. Finally, from the dimensions of multi-source heterogeneous data synergy, algorithmic innovation, system-level multi-element coupling, and digital twin construction, the current technical bottlenecks were discussed, and future development trends were outlined.