Automated Detection and Severity Levels Grading of Rice Spot Disease for Precision Crop Management

The presence of rice spot disease presents a substantial risk to the overall production of rice worldwide, which highlights the need for an effective and precise evaluation of disease severity in order to implement prompt interventions. This study introduces a new methodology that utilizes machine learning algorithms, notably Convolutional Neural Networks (CNN) and Random Forest (RF), for the automated categorization of severity levels in rice spot disease. A curated dataset consisting of 9,000 high-resolution photos of rice plants was compiled, including a wide range of disease severity levels, ranging from healthy plants to those badly affected by diseases. The hybrid model, which combines elements of CNN and RF, demonstrated exceptional performance with an overall accuracy of 95.49%. This surpasses the accuracy achieved by the separate CNN and RF models. The integration of deep learning with ensemble learning presents a resilient approach for the timely identification of diseases and the accurate management of crops. The comparative study reveals that the hybrid model exhibits greater performance, whereas the feature importance analysis provides insights into the crucial visual qualities that inform future research. This study signifies a crucial advancement in the field of precision agriculture, as it facilitates the ability to make well-informed decisions that protect crop yields and improve global food security.