KLASIFIKASI CITRA DENGAN METODE TRANSFORMASI WAVELET PADA LIMA JENIS BIJI-BIJIAN

Abstract

Up until now, seeds classification has been done manually by human eyes. But human eyes cannot work for 24 hours. Using a software which can classify seeds image into its class by texture analysis with Wavelet Packet Transformation, can be solved for this problem. In this research, several kinds of wavelet are used, there are: Daubechies-2, Daubechies-3, and Coiflet-1. Texture analysis begins with decomposition to get wavelet coefficients then entropies of 25 images which will be included in databases were counted. Next process is calculating differences entropies between the images which will be classified wilth the image in database. The last process is searching the shortest distance using Euclidean equation which will show that the tested image included in one of the classes in the database. The highest recognition level is at level 1 with the recognition level average is 82% for every seed, while the lowest recognition level is at level 4 with recognition level average is 20%. Seeds with the highest recognition level are rice and corns with the recognition level average is 100%, otherwise seeds with the lowest recognition level is peanuts. The recognition level for rice and corns are higher than others seeds because their unique textures, different with other three kinds of seeds. Peanuts have the lowest recognition level, because its texture is similar with other two seeds, i.e. soy beans and green bean, while soy beans and green bean have a very different texture if compared with other seeds.