Papaya Fruit Type Classification using LBP Features Extraction and Naive Bayes Classifier

Abstract

This research proposes the classification method of papaya types based on leaf images using the Naïve Bayes classifier and LBP feature extraction. Papaya leaves are used because they have a unique pattern and texture from their leaf bones, besides leaf-based classification can be done before papaya trees produce fruit. In the preprocessing process, three stages are carried out in which conversion to grayscale, image adjustment and resize, to produce a good LBP feature extraction. The resize process is useful to reduce computational time during the training and testing process, where this process is done at the end of the preprocessing process to get a better pixel value. Image adjustment is used to sharpen the papaya leaf bone which is the main pattern of the papaya leaf. At the feature extraction stage, an image zoning process is carried out, by dividing the image into nine zones, to produce nine LBP features for each image. In the implementation phase, a total of 150 papaya leaf images were used which consisted of 125 training images and 25 testing images. Based on the results of the classification using the Naïve Bayes classifier by using nine zones each with 128-pixel cell size and image adjustment resulting 96% accuracy. The results of this accuracy are better than using cell sizes 32 and 64 and without image adjustment.