Klasifikasi buah-buahan berdasarkan bentuk dan warna

ABDULLAH, Drs. Agus Harjoko, MSc.,PhD

2006 | Tesis | S2 Ilmu Komputer

Klasifikasi buah-buahan yang dilakukan secara manual tidak efisien dan kurang teliti, untuk jumlah buah-buahan yang cukup besar. Perkembangan teknologi informasi memungkinkan klasifikasi buah-buahan dengan bantuan komputer. Permasalahannya adalah bagaimana sistem komputer dapat mengenali dan mengklasifikasikan buah-buahan. Pada penelitian ini dilakukan klasifikasi buah-buahan berdasarkan kemiripan (similarity) bentuk dan warna. Klasifikasi buah-buahan dilakukan secara tidak langsung melalui penangkapan citra buah dan proses awal background subtract. Kemiripan diukur melalui pengukuran jarak vektor fitur. Pemilah yang digunakan adalah Nearest Mean Classifier atau Minimum Euclidean Distance Classifier. Ada 12 belas kategori buah yang diperkenalkan pada sistem klasifikasi buah-buahan. Ada tiga macam perlakuan pada sistem yaitu, sistem dilatih menggunakan satu sampel, dua sampel dan tiga sampel buah pada masing-masing kategori, Sistem diuji untuk melakukan klasifikasi buah-buahan yang telah diperkenalkan kategorinya. Sampai dengan tiga sampel buah pelatihan, sistem telah mampu melakukan klasifikasi terhadap 48 buah-buahan dengan tingkat keberhasilan 96,53%, atau mempunyai kehandalan yang baik.

Classifying fruits in a large number is not efficient and accurate enough if it is done by traditional way. The development of information technology gives computer capability to classify fruit accurately. Yet the next problem that emerges is how does computer recognize and classify fruits. Can computer system provide an accurate classification on fruit by identifying its shape and color? This research tries to classify fruits based on similarity of fruits’ shape and color. The classification is done by image capture of fruit, and then background subtract used to preprocessing. The similarity is measured by feature vector distance. Classifier that is used to do classification in this research is called Nearest Mean Classifier or Minimum Euclidean Distance Classifier. There are 12 categories of fruits that have been recognized to fruits’ classification system. In this case, the research divides three way to train system. First the system is trained by using one fruit for every category. Second, it is trained by using two fruits for every category, and the last it is trained by using three fruits for every category. The next step is that the system is tested to classify another fruits that have been recognized their category. The value of test is that the system able to classify 48 fruits with succeed rate and high accuracy that reaches to 96,53% after it has been trained by using three fruits for every category. This is very clue that the system is good to be used.

Kata Kunci : Citra Digital,Klasifikasi Buah,buahan,Bentuk dan Warna, Classification,Fruit, Shape, Color, Nearest Mean Classifier, Euclidean