Classification of Banana Leaf Disease Using Random Forest Based on LAB Color Features and Segmentation Area

Banana (Musa spp.) is an important commodity in tropical and subtropical countries, serving as a major source of income for farmers and a staple food for millions of people worldwide. As such, banana has a high export value, prompting increased production to meet the growing market demand. However, banana plants are susceptible to various pests and diseases, especially on the leaves, such as Cordana, Pestalotiopsis, and Sigatoka, which can hinder fruit production. In identifying these diseases, farmers usually use visual observation, which is often inaccurate because the diseases have similar characteristics. To overcome this problem, digital image processing technology and machine learning approaches can be applied to improve efficiency in disease detection on banana leaf. Therefore, this study aims to build a system for identifying three diseases on banana leaf by applying digital image processing, such as threshold segmentation method and color feature extraction using LAB and area of segmented objects, and performing information fusion on feature extraction to improve accuracy. The classification results showed an excellent accuracy rate, reaching 94.42%, with precision, recall, and F1-score values of 93.71%, 93.58%, and 93.64%, respectively. The system was successfully developed using MATLAB software, which allowed users to load image, perform segmentation, and view classification results easily.