Palm oil bruise detection of texture and shape features experimental comparison on support vector machine and naïve bayes / Nurbaity Sabri … [et al.]

Palm oil is one of the largest and significant contributions to the Malaysia economy. It is important to improve the quality of this product as defects on palm oil fruit may affect the production of palm oil. Bruise is one of the defects on palm oil fruit where it is unavoidable during the field material activities. It will increase the number of Free Fatty Acid (FFA) and reduce number of palm oil quality. We proposed using a combination of four (4) textures Grey Level Co-occurrence Matrix (GLCM) and six (6) shape features to detect bruise and non-bruise. A comparison between two classifiers named Support Vector Machine (SVM) and Naïve Bayes has been done using the same features. The experiment shows Naïve Bayes classifier achieve 97.5% accuracy compared to SVM with the combination of two types of features. Further study will be done to classify the palm oil bruise into more stages.