Using Image Features to Investigate Accuracy of Classification Techniques in Categorizing Visual Defect Types of Defective Yellow Soybean Seeds

Visual defects of yellow soybean seeds are one of the many factors influencing the soybean price. If those defects are found and classified correctly, the farmers or sellers can cope with those problems immediately. Therefore, this study investigated the accuracy of classification techniques using 14 image features. Those features were extracted utilizing image processing techniques. The obtained feature values were used for classification in four ways. From the classification processes using 10 techniques, the C5.0 with boosting method (trials = 80) provided up to 93.53% accuracy, with the highest accuracy being for six defect types (types 1, 3, 4, 5, 8, and 10). The results can be used as an alternative approach to identify the causes of damage in soybean seeds.