Application of machine learning technique for diagnosis of powdery mildew disease in wheat
Loading...
Date
2021-10
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
G.B. Pant University of Agriculture and Technology, Pantnagar - 263145 (Uttarakhand)
Abstract
Powdery mildew is one of the most common fungal disease of wheat caused by an obligate biotrophic pathogen Blumeria graminis f. sp. tritici. Present investigation was conducted in rabi season 2020-21 at Norman E. Borlaug Crop Research Center, Pantnagar, Uttarakhand. The study was undertaken to create a disease diagnosis model for powdery mildew in wheat. Ten different deep learning approaches namely VGG16 (without augmentation), VGG16 (with under sampling), VGG16 (with over sampling), ResNet50 (without augmentation), ResNet50 (with under sampling), ResNet50 (with over sampling), ResNet50 (with under sampling and augmentation), EfficientNetB3 (with augmentation), EfficientNetB5 (with augmentation) and EfficientNetB7 (with augmentation) were used to check the best model for disease diagnosis. The accuracies attained by these algorithms were 61.7%, 59 %, 77 %, 58-63 %, 55-61 %, 74.7 %, 74%, 74.8 %, 73.6 % and 75.1 %, respectively. Automatic computer system for detecting and classifying of diseases is very important for efficient management. The present study will provide the opportunity for disease management by using advanced learning technologies with least interference of mankind. The study was also conducted to check the influence of weather parameters with disease progress of powdery mildew. Infection rate and PDI were used to analyze the effect of weather variables. PDI was positively correlated with both maximum (r=0.82) and minimum temperature (r=0.61) and positively for bright sunshine hours (r=0.81) while with morning (r=-0.73) and evening relative humidity (r=-0.77), it was negatively correlated. Maximum temperature (r=-0.52) and sunshine hours (r=-0.51) showed a negative correlation with the rate of infection while a positive correlation was seen with morning (r= 0.54) and evening relative humidity (r= 0.61). Step-wise multiple regression analysis was done and a prediction equation was developed (R2=0.45).