Vegetable and fruit crops are susceptible to multiple diseases that can affect yield and the quality of the produce. The current research aims to detect and assess emerging diseases in multiple crops including tomato, sweet pepper, potato, apple, corn, and grape through a new digital tool which will enhance crop health monitoring. This tool is capable of classifying plant diseases irrespective of crop. To enhance detection accuracy, all images were resized and then segmented using the GrabCut method to isolate the leaf regions. The segmented images were subsequently converted to HSV color space and grayscale before feature extraction was performed. A total of 96 features were extracted, including color histograms, Local Binary Pattern (LBP), and Gray-Level Co-occurrence Matrix (GLCM). The Extreme Gradient Boosting (XGBoost) algorithm was employed for disease classification. This study utilized 27,691 images from the PlantVillage dataset, and the proposed model achieved a classification accuracy of 97.20%, 100.0%, 97.68%, 97.01%, 87.94%, and 97.91% for tomato, sweet pepper, potato, apple, corn, and grape respectively.
Read more