Abstract:
The purpose of this thesis was to develop an information retrieval system to aid the identification of Buddha amulets. The information will be processed based on the K-nearest neighborhood classification technique used to identify each type of Buddha amulet from a database. The texture analysis technique was applied in order to extract 10 features from images taken with a digital camera. Moreover, a white paper background was used in each image to insure a high standard of resolution, as most amulets are made from soil. It is necessary to photograph them at close range. The study used the characteristics of Buddha amulets to applied the feature extraction algorithms, which consisted of 10 feature extractions with texture analysis from Gray-level Co-occurrence Matrices (GLCM). The experiments were conducted on more than 1,200 images from 44 different kinds of Buddha amulets. The data set was at 880 Buddha amulet images. The system was tested using 440 images from the test data set. The result of these experimental tests yielded accuracy at 77.27% average precision and recall are 0.773 and 0.752, respectively. In conclusion, this developed system achieved all goals successfully.