HOML-SL: IoT Based Early Disease Detection and Prediction for Sugarcane Leaf using Hybrid Optimal Machine Learning Technique

Abstract

Sugarcane is the mostly cultivated crop in Indian agriculture. The quantity and quality of sugarcane production is decreasing day by day due to several diseases. Because of fungi, bacterial, pathogenic, and viral infections often affect crop production with diseases that affect farmers ’economic incomes. In order to increase the quality of the crop, early detection is required. An image processing practice plays a critical role to predict the disease in various plants. In this paper we propose a hybrid optimal machine learning technique (HOML-SL) for detect disease in sugarcane leaf. First, we introduce a non-linear cluster based optimization (NCO) algorithm for segmentation which segments the diseased area from the original sugarcane leaf. Second, we develop an optimal feature selection process using cross layer optimization (CLO) algorithm which select optimal features among multiple features. Finally, we illustrate a moth flame based capsule neural network (MFO-CNN) algorithm to classify the sugarcane leaf disease (such as mosaic virus, red rot, grassy shoot and ring spot). The performance of proposed HOML-SL disease detection can compare with the accessible state-of-art detection techniques in terms of accuracy, precession, F-measure and Recall.