Abstract: Digital acquisition technology constructs digital replicas of the physical world and provides a three-dimensional infrastructure for virtual-real collaboration in artificial intelligence, with extensive applications in autonomous driving, surveying and remote sensing, intelligent robots, smart cities and other fields. This paper systematically reviews the evolution of 3D digital acquisition technologies over the past 30 years, dividing the development into three stages: classical geometric computing, multi-source heterogeneous data, and deep learning-driven technologies. It analyzes technical systems including discrete point clouds, mesh surfaces, multi-sourced data collection, data registration, 3D mapping, real-time reconstruction, multi-view photogrammetry and neural rendering. Furthermore, this paper elaborates on the emerging stage of spatial intelligence integrating spatial computing and embodied intelligence. On this basis, this paper reviews existing technical frameworks and representative algorithms centered on popular research directions including physics simulation engines, bidirectional virtual-real alignment, and closed-loop perception-decision systems. Finally, we discuss development trends in content representation, virtual-real alignment and intelligent applications empowered by current multimodal large models, and provide systematic references for relevant researchers.