Abstract: The grain separator is an important machine for separating paddy and brown rice, the precise control of the separating device is one of the decisive factors to improve the separation efficiency. A paddy and brown rice detection device is proposed to detect the types of particles at the discharge end of the grain separator. Focuses on the study of paddy and brown rice, Firstly, pictures of paddy and brown rice were collected to extract color, shape and texture features of paddy and brown rice. The single feature model and different feature combination model were constructed, and four classifiers, namely support vector machine （SVM）, K-nearest neighbor （KNN）, Bayes and Random Forest, were used for training and recognition. The experimental results show that compared with other feature combinations, the SVM classifier based on texture + shape + color has the highest recognition rate, and the recognition accuracy reaches 97.8%. The method can be applied to the testing device, and provides a theoretical basis for the precise control of the material separation device.