Feature Group Extraction for Rice Disease Prediction using VGG-16 Model

Rice plants have been the focus of significant attention from agricultural specialists. However., there has not been any comprehensive inquiry into the matter of plant diseases. It is crucial to accurately identify rice infections to prevent the major adverse effects of the disease on crop yield. The current diagnostic procedures for rice diseases are neither exact nor effective., and additional apparatus is sometimes required. It is essential to detect any sickness in its earliest stages to ensure the rice plants continue to develop normally and healthily. Manual detection of disease involves significant time and effort; hence., creating an automated system makes sense. This paper presents a machine learning technique for diagnosing rice leaf tissue illnesses. According to the findings., the three diseases known as leaf smut., bacterial leaf blight., and brown spot harm rice plants the most often. The data was presented as high-quality images of rice leaves with various types of damage superimposed on a white background. This paper proposes a novel feature-clustering based VGG-16 model for rice disease detection. Initially., the crop images are preprocessed with a Wiener filter and color enhancement techniques. Then., the features are grouped using a density-based technique known as DBSCAN. Finally., the VGG-16 model is used for classifying rice diseases. The decision tree approach achieved an accuracy of more than 97% on the test dataset with 10-fold cross-validation.