A Portable Plant Physiological Feature Image Processing Technique for Groundnuts Rosette Disease Diagnosis

Abstract

Groundnut is one of the five most important oilseeds produced in the world. Compared to other major staple crops, groundnuts have received relatively little research worldwide, particularly in less developed countries where it is critical for food security and livelihoods. This crop is susceptible to many diseases, causing decline in productivity and quality, all of which affects agricultural economy. Traditional detection methods by farmers and researchers are time consuming, costly and less accurate necessitating a better and more reliable automatic solution for detecting groundnut foliar leaf diseases.  Five major foliar diseases of groundnuts include groundnut rosette, early and late leaf spots, Bacterial wilt and rust. This model was trained with and without stepwise resizing and simultaneously validated using cross-entropy loss. The dataset used for training and validation purposes was manually created. To evaluate the performance of our model, various performance metrics such as accuracy, sensitivity, F1 score, and precision were applied achieving a perfect precision value of 100% at high confidence threshold of 0.964 and F1-score of 0.8 at high confidence threshold of 0.454 demonstrating the model's balanced effectiveness in identifying ground rosette disease. The Groundnuts Rosette Disease Diagnosis Using Plant Physiological Feature Image Processing Technique model achieved reliable accuracy with a limited dataset of 9 groundnut rosette scale rates.