vi
ABSTRAK
Bayi yang baru lahir sulit dikenali melalui wajah. Untuk mencegah kasus bayi
tertukar , dapat digunakan ciri unik berupa telapak kaki bayi untuk membedakan bayi
satu dengan lainnya. Teknologi jaringan syaraf tiruan diharapkan mampu mengenali
citra cap telapak kaki bayi, sehingga dapat membantu proses identifikasi bayi. Citra
cap telapak kaki yang berekstensi *.jpg terlebih dahulu difilter menggunakan
highpass filtering untuk mempertajam detail citra, dan terakhir dideteksi tepi Canny
untuk menandai bagian yang menjadi detail citra. Hasil deteksi tepi berupa citra biner
yang kemudian matriks citra biner ini digunakan untuk dilatih dan diuji
menggunakan metode SOM Kohonen. Gambar yang dilatih berupa 10 gambar cap
telapak kaki bayi asli dan 20 lainnya adalah citra asli yang telah diberi noise. Hasil
akhir berupa identifikasi telapak kaki bayi berdasarkan hasil pelatihan. Hasil
pengujian terhadap citra yang dilatih menunjukkan tingkat akurasi pengenalan
sebesar 90 % dan persentase akurasi pengenalan untuk citra yang tidak dilatih sebesar
66,7 %.
Kata kunci: jaringan syaraf tiruan, SOM Kohonen, pengenalan pola, telapak kaki bayi, deteksi tepi Canny, highpass filter
vii
Implementation of Self Organizing Map Kohonen Neural Network in Recognizing Baby’s Sole Of Foot
ABSTRACT
A newborn babies are difficult to identify through the face. In order to prevent the
cases of swapped babies, an approach on using the soles part of baby's feet is believed
to be a solution to distinguish one baby from another. By using the ANN technology,
it will help the recognition process of baby's foot image of the stamp, which will
result in the identification of newly born babies. The image of the stamp foot will be
in a *.jpg extension, in which it is first filtered using the highpass filtering to sharpen
up the image details. Then it is followed by Canny edge detection to mark some parts
of the image detail. The result of the edge detection process will be in the form of a
binary image. That binary image holds matrix value which later used in the training
and testing process of SOM Kohonen Method. The trained images will consist of the
10 initial baby’s foot stamp image along with the 20 noise version images. The result
of the identification process is based on the result of training. In the testing section, it
is found that a trained image has the accuracy of 90% in identifying babie’s feet while
the original image has only 66.7% of recognition level.
Keywords: artificial neural network, SOM Kohonen, pattern recognition, baby’s sole of foot, Canny edge detection, highpass filter