Abstract: Plants are continuously exposed to a variety of biotic and abiotic stresses throughout their growth cycle in smart agriculture, including pathogen invasion, extreme temperature fluctuations, heavy metal contamination, drought, and salinity. These stresses disrupt physiological homeostasis, thereby causing damage to plant growth and yield. Therefore, it is often required to rapidly, non-destructively, and accurately monitor the plant physiological status in plant science and precision agriculture. Current diagnostic tests can rely primarily on histochemical reagents following the isolation and purification of plant extracts. However, these typically destructive measurements cannot real-time track the endogenous dynamic signals. Non-destructive genetically encoded molecular sensors are often limited to model plants and time-consuming transgenic procedures. Given that most crops are non-model plants, there is a great demand to detect the stress signaling patterns. Fortunately, the second near-infrared (NIR-II, 900-1 880 nm) fluorescence imaging can be expected to offer a transformative solution for plant observation. Compared with visible light and the conventional NIR-I window, NIR-II imaging has exhibited superior optical properties, including significantly reduced photon scattering by cell walls, minimal interference from chlorophyll autofluorescence, and millimeter-scale deep tissue penetration. This review aims to present the recent advances in NIR-II fluorescence imaging for plant information acquisition, including the system principles, probe materials, architectures, and agricultural applications. 1) The underlying photophysical principles were proposed to explain the long-wavelength photons in the NIR-II window. Rayleigh and Mie scattering were then minimized to enhance image contrast in the heterogeneous plant tissues. 2) Four types of fluorescent probes were systematically categorized: organic small-molecule dyes, single-walled carbon nanotubes (SWCNTs), quantum dots (QDs), and rare-earth-doped nanoparticles (RENPs). A comparison was also provided for the trade-offs among their brightness (quantum yield), biocompatibility, and functionalization strategies. 3) The architecture of NIR-II imaging systems was described to cover excitation sources, filtering schemes, and advanced indium gallium arsenide (InGaAs) cameras. In addition to system components, the summary was also performed on the operational principles and typical applications of major imaging modalities. Specifically, the wide-field fluorescence imaging was introduced for macroscopic observation. Three-dimensional confocal and light-sheet microscopy were also discussed for the high-resolution visualization of internal biological structures. 3) The cutting-edge applications of NIR-II imaging in plants were highlighted, such as the non-invasive visualization of vascular bundles and root system architecture in soil, in situ dynamic monitoring of signaling molecules (e.g., ROS, and nitric oxide) under stress, and the early diagnosis of plant diseases before visible symptoms. The biosafety of the probes was specifically assessed on physicochemical properties, such as the influence of the size and surface charge on the uptake, transport, and subcellular distribution over plant physical barriers. A standardized evaluation framework included the residue analysis and long-term ecological monitoring. 4) A technical roadmap was proposed for the engineering implementation of NIR-II technology in smart agriculture. Critical challenges were then identified, including high equipment costs and environmental background noise in field settings. The recommendations were given to integrate the closed-loop strategy with the probes, imaging devices, and algorithms. This approach prioritized the high-brightness probes, portable imaging hardware, and artificial intelligence-driven image processing. This finding can also provide a theoretical and technical foundation to implement NIR-II sensing in plant science, nanotechnology, optics, and digital farming.