Identification of Banana Types with the Least-Squares Support Vector Machine

Abstract

The classification of banana species is still done manually by banana farmers. This identification process has the disadvantage that it requires more manpower to sort, the level of perception of the type of fruit produced can be different because humans can experience fatigue, are not always consistent, and human judgments are also subjective. Thus, a tool is needed that can identify the type of banana fruit precisely and accurately. One of them is by creating a computer-based system using the statistical feature extraction method of digital images. By performing color feature extraction using Color Moments (RGBHSVYCbCr), then texture extraction using Gray-Level Co-occurence Matrix (GLCM), and using the Least-Squares Support Vector Machine (LS-SVM) method for classification of banana species. LS-SVM is a modification of SVM, which is used to improve classification performance. In the SVM algorithm, there is quadratic programming that is used to obtain the optimal solution in determining the Lagrange function; from the Lagrange function, it will be used in calculating the value of the weight and bias parameters. Quadratic programming is not efficient when applied to higher spatial dimensions because the computation will be very complex and very long. LS-SVM is better than standard SVM in terms of the calculation process, faster convergence, and higher precision. At the end of the experiment, the LS-SVM method succeeded in detecting the type of banana with a test accuracy value of 90%.