Automation of Leaf Disease Prediction Framework based on Machine Learning and Deep Learning in different Crop Species

Abstract

The phenomenon of plant leaf illness has fatalistic influences on production and assists in promoting safety in products concerning agriculture. Owing to the reason that the leaf is major vulnerable portion of a plant it is influenced easily upon comparison with the other parts. Plant disease detection by means of numerous automatic models is advantageous as it minimizes immense work of big farm monitoring, and also at the early stage assists in detection of the disease symptoms. Several frameworks have been designed for plant disease.  But the error rate and overhead incurred in plant leaf disease diagnosis was not focused. In this research, characterize and investigate the plant leaf disease prediction using three novel proposed frameworks using Plant Village dataset. They are Gaussian Distributive Czekanowskis Region-based Deming Regression (GDCR-DM), Covariance Kalman Geometric Graph-based Bernoulli Classifier (CKGG-BC) and Independent Gaussian Gray Level Gaussian Gray Level Geometric Neural Network Classifier (IGGL-GNNC). The GDCR-DM framework identifies the plant leaf disease in an accurate manner with minimum false positive rate employing Czekanowski's dice and Deming regression function. Next, to reduce noise involved during preprocessing and segmentation, CKGG-BC framework is proposed that with the aid of Covariance Kalman filter function and Multiple Kernel Learning Classifier not only reduces the error but also enhances true positive rate considerably. Finally, IGGL-GNNC framework is designed that Sine Cosine Position Update using geometric functions for realistic disease prediction. The different parameters were evaluated for distinct plant leaf images obtained from Plant Village Image dataset.