Rice provides essential nutrition and energy to the global population, contributing more than twenty percent of total caloric intake. Choosing the appropriate rice amongst the available verities in the market has become a biggest challenge due to their identical physical appearance. Biological and chemical methods such as DNA analysis and alkaline testing are costly, time consuming, and not accessible to the average person, making them unsuitable for the identification of rice grain varieties. Further, few image processing techniques have been incorporated for identification of different rice varieties which seems to be less accurate for verities with same morphological features. This paper describes a hybrid approach for enhancing the accuracy of rice variety identification using machine learning and digital image processing techniques. A total of 1,00,000 rice grain images were selected, with 20,000 for each variety. With the use of image processing techniques, the images were preprocessed in preparation for feature extraction. Initially 12 morphological, 4 shape and 25 texture features were extracted using bounding box and region property methods of image processing on 5 different verities of rice, later these extracted features were fed into different supervised classification algorithms like Support Vector Machines (SVM), K - Nearest Neighbors (K-NN), Decision Tree (DT) and Naıve Bayes (NB). The algorithms with the highest average classification accuracy of 99.24 is achieved with K - nearest neighbors. According to the performance measurement results, the study was successful in identifying the different types of rice.