The protein content of milk will affect the quality of milk.The feasi bility of predicting the protein content of milk is studied by using the spectral feature information of hyperspectral image.In this paper,a prediction modeling method (CARS-SPA-BP) based on compet itive adaptive reweighted sampling (CARS) and successive projections algorithm (SPA) combined with multilaye r feedforward neural network (back propagation,BP) is proposed.In the experiment,250groups of hyperspectral data of five kinds of milk were collected by the visible/near infrared hyperspectral imaging syst em.Through the experimental comparison,the standardized method was used to preprocess the obta ined absorption spectrum,and then the CARS combined with SPA was used to select the characteristic wavelength,18chara cteristic wavelengths are obtained.Through experi ments,the determination coefficients R²c and R²p of training set and test set of CARS-SPA-BP model r each 0.971and 0.968respectively,and the root mean square error of calibration (RMSEC) and root mean square error of prediction (RMSEP) reach 0.033and 0.034,respectively.It is found that the prediction results of multila yer back propagation (BP) neural network model based on CARS and SPA are not significantly lower than that of full wavelength m odel,Therefore, the CARS combined with SPA for wavelength screening and BP neural network can basically complete the prediction of milk protein content.In order to verify the prediction ability of CARS-SPA-BP model,the t raditional partial least squares regression (PLSR) is used to model under the same data environment .The experimental results show that CARS-SPA-BP has significantly improved R²p and RMSEP compare d with PLSR.The results show that CARS-SPA-BP can make full use of the spectral characteristic s of milk to achieve high-precision detection of milk protein content.