Potato Leaf Disease Classification Using Curvelet Empirical Wavelet and Principal Component Analysis

Potatoes are an important food. High potato yields can be obtained from a good monitoring process for potato leaf disease classification. Image decomposition is commonly used with feature extraction to enhance the performance of potato leaf disease classification. However, some image decomposition methods cannot provide high performance due to the loss of information during the decomposition process. In this paper, a Curvelet empirical wavelet transform-I with principal component analysis, the CEWT-I-PCA method, is introduced to extract appropriate feature vectors for classifying diseases in potato leaf images. The proposed method efficiently extracts feature vectors by keeping the major information of the potato leaves. Based on Mendeley data with 426 potato leaf images, the proposed method provides 98.37% more accuracy than the other methods. Our method can adaptively decompose major components of potato leaf disease, consequently improving the accuracy of the classification of potato leaf disease images.