Identification of Pekalongan Batik Images Using Backpropagation Method

Abstract

Batik is an Indonesian cultural heritage that was recognized by UNESCO in 2009. One of the famous batik producing regions in Indonesia is Pekalongan City, Central Java. Pekalongan Batik has a distinctive characteristic compared to other regions, namely from the aspect of color and motif. Pekalongan batik uses bright colors and flora motifs. However, there are some batik craftsmen who still use dark colors. Identify Pekalongan’s typical batik motifs become an obstacle for tourists that are visiting Indonesia. There needs to be an automatic identification system to recognize Pekalongan Batik motifs. The automatic identification system of batik images can contribute to the development of technology in the field of artificial intelligence. This research was conducted by collecting image data taken through observation, interviews and literature review. The image data obtained are four batik motifs. 5 images will be taken from each motif and will be implemented into the system using Matlab R2014a. The next process is feature extraction using the Gray Level Co-Occurrence Matrix (GLCM) method to get information in the batik image. The author then identifies image using the Backpropagation method to obtain the epoch value, learning rate, and accuracy value based on the identification process. The tested system obtained the highest accuracy of 91.2% with epoch 100 and learning rate 0.03 on Sogan batik, 89.6% accuracy with epoch 100 and learning rate 0.02 on Jlamprang batik. 87.2% accuracy was obtained from Cap Kombinasi batik and Tiga Negeri batik with epoch 100, the learning rate was 0.01 and 0.04 respectively. The results of identifying images of Pekalongan batik can be implemented in a more interactive application.