•  
  •  
 

Abstract

Coffee is the second most traded commodity globally after oil and represents a major source of income in many countries. Kingdom of Saudi Arabia has a great interest in growing and expanding of coffee in line with the Saudi Green Initiative as implementation of Vision 2030 AD. This expansion is accompanied by a greater spread of insect pests and the emergence of diseases affecting coffee. Recently, machine learning technologies have been beneficial in the agricultural era in detection and classification of fruit and tree diseases. The study initially focuses on kind of disease that appears in the leafy area of the coffee plant, which is susceptible to many diseases such as Cercospora spp., magnesium deficiency, and others. Deep learning, convolutional neural networks (CNN), support vector machines (SVM), and other imaging and machine learning techniques are used in this paper to detect and classify leaf diseases. The JMuBEN and JMuBEN2 databases from Kenya were used in the first experiment, and the Fyfa Mountains database from the Jizan region was used in the second experiment to automatically detect and classify coffee tree leaf disease. In an SVM model, data preprocessing and data transformation methods are used to generate accurate information to train the model. Grid search is also used across the parameter grid to optimize the estimator parameters used in applying the model. The experimental results showed superior performance compared to many modern basic methods in terms of accuracy, reaching 100%. CNNs have also proven their effectiveness and accuracy in the fields of pattern recognition and image classification. As a result, a CNN model is introduced that takes advantage of transfer learning, which significantly reduces the model training time.

Download

Share

COinS