Epidemiology of Rhizoctonia aerial blight in soybean and its autodetection through Machine Learning Technique (MLT)
Loading...
Date
2018-08
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
G.B. Pant University of Agriculture and Technology, Pantnagar - 263145 (Uttarakhand)
Abstract
Rhizoctonia Aerial blight (RAB) caused by Rhizoctonia solani is one of the most important diseases of soybean in Uttarakhand causes heavy loss of crop yield every year. Present investigation on RAB of soybean was undertaken in relation to its occurrence, pathogenicity, epidemiology, autodetection and management of the disease. The study was conducted in Kharif, 2017 at Norman E. Borlaug Crop Research Centre, Pantnagar Uttarakhand. Pantnagar is considered as a hotspot for this disease. Out of total sixteen cultivars studied against R.solani, the cultivars exhibited moderate resistance were PK-472 (24.43%) and PK- 262 (25.60%), while cultivars JS-7105, JS-72-280, JS-93-05, Bragg, KHsb-2 and NRC-7 were found moderately susceptible. Infection rate was found to be maximum for cultivar and minimum for cultivar. Area under disease progressive curve was also calculated.It was maximum for cultivar VLS-58 (424.19) and minimum for cultivar PK-262 (160.99). Infection rate was calculated on the basis of disease index. Maximum infection rate was found to be cultivar JS-7244 (0.317 unit/day) and minimum for Shivalik (0.0008). Disease progression of RAB was started in the third week of September. In initial phase, the disease progression was quite high month of October, however it slowly declined later on. September and October are suitable months for initiation, development and progression of RAB disease. Disease progression was maximum for full seed to beginning of maturity (51-34 percent) phase. The PDI of all varieties exhibited negative correlated with temperature, relative humidity, rainfall and wind speed but significant, it was positive only for bright sunshine hours and evaporation but non significance for weather data on same day of disease incidence. Same trend was follow by weather variables on one week before and two week before from disease incidence. Infection rate was positively correlated with all weather variables.The data on disease progression in relation to corresponding weather variables with maximum temperature, maximum relative humidity, rainfall and bright sunshine hours were subjected to step-wise multiple regression analysis being a significant contribution in the prediction of disease index. By using these prediction equations for different cultivars, it is now possible to predict the disease index in advance and it provides sufficient time for contingency plan with plant protection input to restrict and manage RAB growth and its development. The second and important part of present investigation was to use machine learning technique for autodetection of disease considering these 10 different algorithms namely logistic regression. Support vector
machine, VGG-16, VGG-16 (with data augumentation), RseNet-18, ResNet18 (with data augumentation), ResNet-18(with augmentation and large size increase, ResNet-18(with augmentation and a further increase in size), ResNet-34 and ResNet-34(with data augmentation ) were used. The classification accuracy 66.77%, 65.77%, 74.16%, 78.18%, 73.49%, 77.85%, 76.51%, 76.84%, 94.15% and 95.53%, respectively was provided by these algorithms. Automated method for an early detection of a plant disease is vital for precision crop management. Farmers can capture the images of the diseased plant with a simple Smartphone and the algorithm (classifier) classify and diagnosis different diseases, The modern communication and sensor technologies conjugate with robust pattern recognition algorithms for information extraction and classification allow the development and use of the integrated system to tackle disease problems. However the study will provide the opportunity for disease management by using advanced technologies of computer. I.T. and smart telephony with least interference of mankind.
Description
Keywords
null