Detection of oil palm leaf disease based on color histogram and supervised classifier

The diseases of oil palm plants ordinarily emerge on the leaves, causing a decrease in the quality of the crop. It needs to address since the demand for high-grade quality palm oil increase continuously. Some automatic detection models of oil palm leaf disease have been developed, but they commonly give low accuracy due to the extracted features are not discriminatory enough. This study proposes a new detection method of oil palm leaf disease to distinguish two leaf classes: healthy and infected. The feature extraction is carried out in the RGB, L*a*b, HSI, and HSV color spaces by splitting the histogram of each color channel into 8 bins. It is applied to the segmented leaf areas produced by the k -means clustering. A total of 41 selected features are generated using the principal component analysis (PCA) and then fed into the artificial neural network (ANN) classifier. The proposed method is evaluated using a local dataset consisting of 300 leaf images (150 healthy and 150 infected) with 10-fold cross-validation. The evaluation produces sensitivity, specificity, and accuracy of 99.3%, 100%, and 99.67%, respectively.