%A  NIM. 08650062 AGUNG NUR HIDAYAT
%T PERBANDINGAN METODE DETEKSI TEPI
DALAM KASUS PENGENALAN POLA GOLONGAN DARAH
MENGGUNAKAN ALGORITMA RUNUT BALIK
( BACK PROPAGATION )

%X Before performing a blood transfusion someone needs to know their blood type.
For regular medical checks blood type would be very easy to distinguish whether the
results of the test showed blood group A, B, AB or O. The process of blood type
checking the depends on how much accuracy medical blood checks. To keep the results
of the checks can be valid, needs to be a system that can help people to recognize the
type of blood type. A way to recognize blood type using the computer by using pattern
recognition methods and training of each blood type characteristics through the image.
Domain from this research is the image of blood checks. This image obtained
from scanning checks of blood groups from PMI Yogyakarta as many as 54 images for
training and recognition processes. The image is used with extension .bmp with the size
of 400 x 200 pixels. Before image recognition, preprocessing is performed which is
conversion of color images to grayscale images. The next process is edge detection
using sobel operator or prewitt operator. The use of two operators aims to determine
the optimal operator transform to the case of blood recognition. After the process of
edge detection, converted to a binary image in order to be processed by feature
extraction. The final step is the application of artificial neural network with
backpropagation algorithm for the hidden layer activation function is the sigmoid
activation for bipolar and output is linear.
Architectural optimal neural network are the three hidden layers with each
hidden layer has three node. The optimal value for the mean squared error parameter is
1e-1 or 0.1, epoch 1000 and learning rate 0.01. In this research, sobel operator is
better than the prewitt operator in making the recognition of blood type. When viewed
from the difference between the process, the prewitt operator slightly faster than Sobel
operator with a gap of 0.000052 sec. From 39 training data and 14 test data percent of
data obtained successful results of the recognition of blood group type of 92,86%.
Keywords: Backpropagation, Blood, Prewitt, Sobel


%D 2012
%I UIN SUNAN KALIJAGA
%L digilib7912