ABSTRAKSI: Tugas akhir ini membahas tentang bagaimana mengidentifikasi jenis jerawat dengan bantuan komputer, yang di dalamnya telah terdapat suatu aplikasi untuk melakukan identifikasi tersebut. Gangguan yang sering menyerang kulit manusia adalah jerawat. Jerawat dapat disebabkan oleh berbagai faktor, antara lain yaitu iklim, siklus hormonal yang tidak stabil, kebiasaan hidup kurang sehat, faktor keturunan, dan bakteri.
Pada tugas akhir ini dibuat suatu aplikasi untuk mengidentifikasi jenis jerawat melalui citra jerawat berdasarkan segmentasi warna dengan Block Overlapping dan analisis tekstur dengan deteksi BLOB (Binary Large Object) menggunakan identifikasi Jaringan Saraf Tiruan-Learning Vector Quantization. Cara untuk menganalisa kinerja sistem adalah dengan membandingkan kebenaran data output dengan data input dalam mengidentifikasi jenis jerawat.
Dari hasil pengujian diperoleh hasil akurasi untuk masing-masing jenis jerawat dan kulit normal adalah Blackhead sebesar 55%, Conglobata sebesar 66,67%, Cyst and Nodul sebesar 44%, Fulminans sebesar 60%, Normal sebesar 80%, Papula sebesar 60%, Pustula sebesar 53,33%, dan Whitehead 30%. Akurasi tersebut diperoleh dari pengujian 135 citra uji masing-masing berukuran 640x480 pixel, ukuran Block Overlapping 80 x 80, overlap 0%, hidden layer 100, epoch 500. Waktu komputasi rata-rata dalam mengidentifikasi jenis penyakit kulit adalah selama 8,304 detik.
Kata Kunci : jerawat, segmentasi warna dengan Block Overlapping, Binary Large Object (BLOB), Jaringan Saraf Tiruan-Learning Vector QuantizationABSTRACT: This final project is about how to identify types of acne using computers, in which there has been an application to identify them. The most disorders that can attack human’s skin is acne. Acne can be caused by many factors, such as climate, unstable hormonal siclus, unheathliving habits, descend factor and bacteria.
In this final project, we made an application to identify the type of acne through acne’s image based on color segmetation by Block Overlapping, texture analysis by Binary Large Object (BLOB), and detection using Learning Vector Quantization - Artificial Neural Network. The way to analyze system performance is to compare the truth of the output data and the input data in identifying the types of acne.
From the results of testing were obtained the accuracy for each type of acne is blackhead by 55%, conglobata by 66,67%, cyst and nodul by 44%, fulminans by 60%, normal by 80%, papula by 60%, pustula by 53,33% and whitehead by 30%. Accuracy was obtained from 135 test images that each size of image has 640x480 pixel, size of Block Overlapping 80x80, overlap 0%, hidden layer 100, and epoch 500. Average computing time in identifying the type of acne is during 8,304 seconds.
Keyword: the image of acne, color segmentation with Block Overlapping, texture analysis with BLOB detection, Learning Vector Quantization-Artificial Neural Network