Variance Feature Based Rice Diseases Prediction Model Using NARX Recurrent Network and KNN

Toran Verma,Sipi Dubey

The research aims to implement rice disease prediction model using the recurrent network.Rice or paddy is the staple food in Asia. Diseases in the plant induced by abiotic factors which compromised; plant’s performance to produce or survive. These affect the social life and economy of the farmer/nation. This research is aspired to automate multi-diseases prediction of rice plant and prediction of vigorous disinfected plant simultaneously. The proposed technique is a combination of rice plant image features extraction and soft computing. In this proposed system, five categories of disease infected rice plant images along with disinfected images had considered. The different infected and disinfected plant reflect unique feature pattern. The unique feature pattern extracted after segmentation of images of all six considered rice plant category. These feature patterns are the hybrid features of color, entropy and wavelet coefficient. The ANOVA technique had applied in color features, and wavelet features to perform in-depth analysis and summarization by evaluating variances. These hybrid features had used to design Nonlinear Autoregressive model with Exogenous inputs (NARX) to predict rice plant disease category. The K-Nearest Neighbor technique applied on the predicted output of NARX to improve the disease prediction performance. The generated response gives prominence result for prediction with 93.9% accuracy for unknown feature pattern during the test. This model can adapt as plant diseases forecaster by considering all parameters which affect the plant's behavior.

Volume 11 | 05-Special Issue

Pages: 2253-2264