A Hybrid Deep Learning Framework Approach for the Detection of Different Varieties of Grain Types

This study was conducted for the detection of the types of grain which germinate in India. Every class of grain has different and unique kind of proteins, carbohydrates and nutrients. The utilization of grains highly depends on their type. The main motive of the pabulum industry today is to fulfil the consumers’ demand. We propose a hybrid deep learning framework composed of the ensemble of CNNs for feature extraction and an integrated Random Forest model for classification. A distinct type of 13 grain types have been classified—Chickpeas, Lentils, Peanuts, Soybeans, Fava Beans, Finger Millets, Fonio, Japanese Millet, Kodo Millet, Barley, Oats, Rice and Wheat. Our proposed framework outperformed (classification accuracy 96.12%) the state of art algorithms for detection of grain types. Index Terms—— Grain, SVM (Support Vector Machine), Deep Learning, CNN (Convolution neural network), RF (Random Forest), KNN (K-Nearest Neighbor).