Abstract: This article summarizes a visible near- infrared( Vis- NIR) hyperspectral imaging technique was investigated for non- destructive determination of moisture composition of cabernet sauvignon grape produced in Helan mountain.The hyperspectral images of grape over the spectral region between 400 nm and 1000 nm were acquired for 136 grape samples and the difference between raw wavelength and pretreatment wavelength( Average Smoothing,Gaussian Filter,Median Filter,Savitzky- Golay Smoothing,Normalize,MSC,SNV,Baseline Correction,Detrending) by partial least- squares regression( PLSR) were obtained.The efficient wavelengths were selected using weighted coefficients of principal component analysis( PCA),weightedβ- coefficients of partial least- squares regression( PLSR),SPA and CARS.Then the 4 characteristics wavelength of PLSR model of the moisture of cabernet sauvignon grape were built and the CARS method of PLSR model was the best. On this basis,the MLR,PCR and PLSR model of raw wavelength and optimal wavelength for the prediction of grape moisture were established.The results suggested that MSC method was considered to the optimal preprocessing method. The characteristic wavelength extracted by CARS method preformed better than MLR and PCR methods,and the correlation coefficient of prediction and the root mean square prediction error( RMSEP) were 0.806,0.144,respectively.Hence,It's feasible to determinate the moisture of cabernet sauvignon grape using these optimal wavelengths by hyperspectral imaging technique.