Javanese Batik Image Classification using Self-Organizing Map

Abstract

Batik has been officially recognized as one of Indonesia’s cultural heritages by UNESCO, and Indonesia now has a batik day which is always celebrated every 2 <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">nd</sup> October since 2009. The diversity of Indonesian batik patterns makes it difficult to recognize. So, Batik becomes an object of research related to pattern recognition and classification. This study proposes a method for classifying batik motifs using a Self-organizing Map (SOM) on an Artificial Neural Network (ANN). This study aims to classify Javanese batik motifs through computation using artificial intelligence. The samples of the batik motifs used in this study were the Kawung, Megamendung, and Parang motifs. The amount of data used is 150 images, where the number of each motif is 50 images. In pre-processing, we convert all images to grayscale and then perform segmentation to anticipate images that are not suitable. Feature extraction is done through three algorithms, namely Gray Level Co-occurrence Matrix (GLCM), RGB (red, green, blue), and HSV (Hue, Saturation, Value). The features obtained will be divided into training and testing data. SOM is used as a classifier. The highest accuracy of 77% is obtained by using the HSV feature. When combining all the features for classification by using the same portion of data for training, we obtained an accuracy of 90%. This result showed the potential of the SOM algorithm when classifying a large number of batik patterns.