403 samples of compound feed were collected to study the rapid detection methods of hyperspectral imaging used to detect the nutrition components of the compound feed.Visible/infrared reflectance spectroscopy information of samples was collected by hyperspectral imager and leveragestudents residuals were used to eliminate outliers.Sample set was divided by the method of CG,SPXY and KS according to the proportion of 3∶1.Combined with different spectral pretreatment methods of MC,AS,FD,SD,OSC,MSC,SNV,Detrend and their combinations,the optimal optical wave length was selected by correlation index.Partial least squares (PLS) stoichiometric methods were used to establish the quantitative analysis model of crude protein,crude ash,moisture,total phosphorus,calcium content in compound feed based on hyperspectral image technology.Through validation,the validation set decision coefficient R2V of crude protein,root mean square error RMSEP,and relative analysis error RPDV was 0.777 8,2.6155%,and 2.114 3,respectively.When the R2V of crude ash was 0.775 8,RMSEP and RPDV was 1.0611% and 2.120 4.When the R2V of water was 0.631 4,RMSEP and RPDV was 1.6003% and 1.937 1.When the R2V of total phosphorus was 0.467 2,RMSEP and RPDV was 0.1916% and 1.357 0.When the R2V of calcium was 0.440 6,RMSEP and RPDV was 0.1755% and 1.310 5.Comparing those models,the effect of the optimal model of crude protein and crude ash established by the hyperspectral image technology was found to estimate performance better.Both of them can be used in the actual quantitative analysis.The quantitative analysis model of water prediction accuracy is still not ideal and needs to be further optimized.The quantitative analysis model of calcium and total phosphorus prediction ability is poor and cannot be used for quantitative analysis.