Memetic salp swarm optimization algorithm based feature selection approach for crop disease detection system

The detection of disease development in plants becomes very crucial because of its adverse effect on the quality and productivity of agriculture. The automatic disease detection in plants using image processing and machine learning is beneficial due to its fast computing and practicability for continuous monitoring of a large farm. This paper presents an automatic disease detection system using image segmentation, feature extraction, optimization, and classification algorithms. This paper proposes a memetic salp swarm optimization algorithm (MSSOA), which is transformed into binary MSSOA to search for the optimal number of features that give the best classification accuracy. The performance of the proposed algorithm for feature selection is compared with five metaheuristic feature selection (BSSA, BPSO, BMFO, BCOA, IBHHO) algorithms against the UCI benchmark datasets. The obtained results indicate the proposed algorithm outperforms the other algorithms in obtaining good classification accuracy and reducing the feature size. The proposed algorithm is implemented for automatic disease detection of maize, rice, and grape plant and achieved a classification accuracy of 90.6%, 67.9%, and 91.6% and best classification accuracy of 93.6%, 79.1%, and 95%, respectively.