Pattern Recognition of Balinese Carving Motif Using Learning Vector Quantization (LVQ)

Abstract

Bali is a world tourism destination with its cultural uniqueness, one of the Balinese cultural products that need to be maintained is the art of Balinese carvings in traditional buildings and sacred buildings, to inherit the culture it needs a management, documentation and dissemination of information by utilizing technology. Digital image processing and pattern recognition can be utilized to preserve arts and culture, the technology can be utilized to classify images into specific classes. Balinese carving is one of the carvings that have many variations, if these carvings are analyzed then required an appropriate method for feature extraction process to produce special features in the image. So they can be recognized and classified well and provide information that helps preserve Bali. The aim of this research is to get the right feature extraction method to recognize and classify Bali carving pattern image based on the accuracy of HOG feature extraction method with PCA trained using LVQ. The results of the test data obtained the best accuracy of HOG is 90% with cell size 32 × 32 and block size 2 × 2, PCA obtained 23.67% with threshold 0.01 and 0.001, from training input with learning rate = 0.001 and epoch = 1000.