Effectiveness of Background Segmentation Algorithm and Deep Learning Technique for Detecting Anthracnose (leaf blight) and Golovinomyces cichoracearum (powdery mildew) in Rubber Plant.

Rubber is a crucial crop in the Philippines, but it faces the problem of pests and diseases that can significantly reduce yields, impacting rubber exports. To address this problem, a study used image processing techniques to enhance the identification of diseases in rubber trees. The researchers used background segmentation and deep learning to improve the accuracy of identifying anthracnose (leaf blight) and powdery mildew in rubber trees. The study found that using a background segmentation algorithm significantly improved the accuracy of all three architectures, DenseNet, Inception, and MobileNet, for both anthracnose and powdery mildew identification. Specifically, the DenseNet architecture had the highest accuracy rate of 97% for powdery mildew. Meanwhile, the MobileNet architecture achieved an accuracy rate of 98% for powdery mildew and 90% for anthracnose. Overall, the results suggest that using background segmentation algorithms can enhance the accuracy of identifying plant diseases in rubber trees, which could boost the economic growth of rubber exports in the country. [ABSTRACT FROM AUTHOR]