Abstract: In order to identify grains by their texture features, a combined texture model was developed to extract texture features. This combined texture model includes two parts. One is to extract texture features based on co-occurrence matrix, whish is related to texture directionality, coarseness, contrast, and homogeneity; the other is to extract texture features by a multifractal model, with these features relating to texture structures. Three types of grains (CWRS wheat, oat, barley) were selected for classification and the mean classifying accuracy is 95%.