Based on the Visible and near infrared (Vis-NIR) hyperspectral imaging system to establish a rapid diagnosis and prediction models of water content in tomato leaves,the best proc essing and modeling methods for spectral data were chosen,and then the spectral response features of leaves mois ture content to different bands was discussed,Ultimately chemical imaging and visualization analysis were carried ou t on the distribution of water content in the leaves according to prediction models.Combined with thresholding approach the spectrum of region of interest (ROI) which gained from 110tomato leaves samples with differ ent growth periods were acquired,then a variety of pretreatment methods were used to optimize the(Mi croscopic hyperspectral imaging detection and mechanism of spectral respons e of peroxidase in tomato cell) spectrum,and the methods of β weight coefficient,Succe ssive Projections Algorithm(SPA),Uninformative Variable Elimination(UVE),Competitive Adaptive Weighting Algorithm(CARS) and the combination of UVE-SPA and CARS-SPA were analyzed to extract the feature wavelength from the optimized spectral data,the validity of modeling with Multiple Linear Regression(MLR),Principal Components Regression(PCR) and Partial Least Squares Regression (PLSR) for moisture content prediction were evaluated b y extracting wavelength,subsequently the optimal combined model was obtained,and the visuali zation of leaf water content and its distribution was realized with weight coefficient assignment method,at last the spectral response characteristics for water content in different bands were parsed.Conclusively,th e model of Baseline-CARS-MLR achieved the highest correlation coefficient (Rp) of 0.95for predicting with RMSEP of 0.042.It has certain advantage to quickly assess the moisture content of leaves based on high spectra l imaging,all the researches were to provide a theoretical basis for real-time evaluation of the water deficit of tomato plants and the development of intelligent irrigation technology.