An intelligent classification model for peanut’s varieties by color and texture features

Abstract

India is the second largest producer of peanuts in the world, and they are available in different forms: Bold or Runner, Java or Spanish and Red Natal. The main peanuts varieties produced in India are Kadiri-2, Kadiri-3, BG-1, BG-2, Kuber, GAUG-1, GAUG-10, PG-1, T-28, T-64, Chandra, Chitra, Kaushal, Parkash, Amber, etc. Peanut is the prime crop for our country’s peasants to increase their agricultural income. However, our country’s international trade price of peanut is only 80% of the average market price. The automation level of testing peanut kernels’ quality is low because of workers fatigue, and most of the work is done by manpower leads costly. The peanuts are evaluated in many areas for sowing and oilseed processing; they must be identified quickly and accurately for selection of a correct variety and kernels’ quality. The proposed testing method based on image processing and computer vision is a new one which is undamaged, speedy with high distinguishing rate, repeatability and low cost and fatigue. In this paper, machine-learning classifiers (Multilayer Perceptron, Simple Logistic, Support Vector Machines, and Sequential Minimal Optimization and Logistic classifiers) are investigated to obtain the best predictive model for peanuts classification. The training and test sets are used to tune the model parameters during the training epochs by varying the complexity of the predictive models with K-fold cross-validation. After obtaining optimized models for each level of complexity, a dedicated validation set is used to validate predictive models. The developed computer vision system provided an overall accuracy rate for the best predictive model in discriminating peanuts variety are Random Forest (82.27%), Multilayer Perceptron(84.9%), and libSVM (86.07%)