Automatic identification of acacia leaf diseases in plantation forests using wavelet energy and Shannon entropy

Abstract

Protecting plantation forests from pests and diseases are essential for keeping trees healthy and productive. Diagnosis of diseases is critical for disease management in plantation forests. For forest plantation with large concession areas, manual identification is time consuming and subjective due to inconsistency of investigator's decision. This paper proposes a method for automatic acacia leaf diseases identification through digital image processing using wavelet energy and Shannon entropy of sub-bands from the orthogonal discrete wavelet packet decomposition (DWPT). These features are used as input for the classifier. A support vector machine (SVM) is used to classify whether a leaf is health or suffering from some diseases. We have examined 1766 leaf samples containing five diseases: leaf spot, leaf blight, leaf curl, phyllode rust and anthracnose leaf spot. The experimental results show that the proposed method obtained accuracy of 91% in differentiating healthy leaves and acacia leaf diseases. The ROC curve of acacia leaf identification indicates that the system is reliable to distinguish the leaf diseases. This system will help to reduce the yield losses and can help surveyors, forest rangers or public users for gathering information, record observation and diseases identification in plantation forests quickly, accurately and automatically.