Open AccessPaddy Plant Disease Recognition, Risk Analysis, and Classification Using Deep Convolution Neuro-Fuzzy Network
Author(s) -
V. Vinoth Kumar,
K. M. Karthick Raghunath,
N. Rajesh,
Muthukumaran Venkatesan,
Rose Bindu Joseph,
N. Thillaiarasu
Publication year - 2021
Publication title -
journal of mobile multimedia
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.229
H-Index - 12
eISSN - 1550-4654
pISSN - 1550-4646
DOI - 10.13052/jmm1550-4646.1829
Subject(s) - convolutional neural network , artificial intelligence , computer science , fuzzy logic , pattern recognition (psychology) , field (mathematics) , population , deep learning , paddy field , rice plant , machine learning , mathematics , agronomy , biology , demography , sociology , pure mathematics
A significant number of the world’s population is dependent on rice for survival. In addition to sugarcane and corn, rice is said to be the third most growing staple food in the world. As a consequence of intensive usage of man-made fertilizers, paddy plant diseases have also risen at a faster pace in current history. Exploring the possible disease spread and classifying to detect the consequent impact at an early stage will prevent the loss and improve rice production. The core task of this research is to recognize and quantify different kinds of infections (disease) affecting the paddy plant crop, such as brown spots, bacterial blight, and leaf blasts. Both detection and recognition are carried out based on the risk analysis of paddy crop leaf images. We suggest a Deep Convolutional Neuro-Fuzzy Method (DCNFM) that combines one of the advanced machine learning variant, namely deep convolutional neural networks (DCNNs) and uncertainty handler called fuzzy logic. The synthesis has the benefits of both fuzzy logic and DCNNs when dealing with unstructured data, extracting essential features from imprecise and ambiguous datasets. From the crop field, continuous image data are captured through image sensors and fed as a primary input to the proposed model to analyze the risk and then later to classify them for precise recognition/detection of the disease. The detection/recognition rate of the DCNFM is found to be 98.17% which is comparatively found to be effective in comparison with the traditional CNN model.