Abstract: Color and moisture content of vegetable soybean are two important parameters in determining quality of vegetable soybean. In this paper,a multiple model fusion method were proposed to improve the prediction accuracies for color and moisture content of soybean during the drying process.Four feature extraction methods,namely,mean value,entropy value,relative divergence and standard deviation features,were used for extracting features from hyperspectral image,and then four partial least squares regression( PLSR) sub- models for color and moisture content prediction were developed using each feature alone,the multiple models fusion method,which was weighted average over the prediction results from the four sub- models,finally was obtained to improve the prediction accuracies. Compared to PLSR sub- models,the multiple model fusion method achieved consistently better results,with improvements of 4.3% and 7.7% for color and moisture content prediction,respectively. The paired t- test for root- mean- square error of prediction at 5% level of significance showed that the multiple model fusion method was superior over the PLSR sub- models. Hence,this multiple model fusion method provided a simple and robust means for improving color and moisture content prediction of soybean during drying process.