Optimasi Model Segmentasi Citra Metode Fuzzy Divergence Pada Citra Luka Kronis Menggunakan Algoritma Genetika
| Main Authors: | Rachmansyah, Ghenniy, Mahmudy, Wayan Firdaus, Perdana, Rizal Setya |
|---|---|
| Format: | Article info application/pdf eJournal |
| Bahasa: | eng |
| Terbitan: |
Fakultas Ilmu Komputer, Universitas Brawijaya
, 2016
|
| Online Access: |
http://jtiik.ub.ac.id/index.php/jtiik/article/view/163 http://jtiik.ub.ac.id/index.php/jtiik/article/view/163/pdf |
Daftar Isi:
- AbstrakLuka kronis merupakan masalah yang masih terbilang berat dalam penanganan, memerlukan ketekunan, biaya mahal, tenaga terlatih dan terampil. Proses pengkajian luka masih dilakukan secara manual, membutuhkan waktu yang cukup lama dan menghasilkan hasil yang lebih subyektif. Dengan adanya permasalahan tersebut, maka dibutuhkan sistem yang dapat membantu pengkajian luka dengan pendekatan citra digital atau dikenal dengan istilah digital planimetry. Fokus permasalahan yang diselesaikan hanya sebatas pada penggolongan komposisi jaringan luka dengan pendekatan segmentasi citra. Pada task segmentasi citra, algoritma yang digunakan yaitu fuzzy divergence yang dioptimasi menggunakan algoritma genetika untuk pemilihan nilai threshold optimal. Pada algoritma genetika, representasi kromosom berupa real-coded, proses reproduksi meliputi operasi extended intermediate crossover dan random mutation, serta metode seleksi elit dengan penambahan mekanisme random injection. Metode yang diusulkan dapat digunakan untuk mengoptimasi model segmentasi citra multilevel thresholding dengan meminimalkan nilai fuzzy divergence dengan parameter algoritma genetika; meliputi ukuran populasi sebesar 60, kombinasi ukuran cr dan mr secara berturut-turut 0.6 dan 0.4, dan ukuran generasi sebesar 100. Kemudian, berdasarkan evaluasi hasil segmentasi citra menggunakan Standar Deviasi (SD), distribusi Gamma menghasilkan hasil segmentasi yang lebih baik.Kata kunci: luka kronis, digital planimetry, segmentasi citra, fuzzy divergence, algoritma genetikaAbstractChronic wounds are a problem that is still difficult in wound management, require persistence, high cost for treatment, and trained-skilled personnel. In wound management, the assessment process are still performed manually, however it’s very time-consuming and produce more subjective outcomes. Given these problems, there is a need for a system that helps wound assessment with the approach in measuring wound size using digital images, known as digital planimetry. In this work, the focus only on wound tissue classification using image segmentation. In image segmentation, the algorithm used is fuzzy divergence that optimized by using genetic algorithm for selecting optimal threshold. For genetic algorithm, the representation of chromosomes is real-coded, then reproduction process using the extended intermediate crossover and random mutation, and elitism selection with the addition of random injection mechanism. The proposed method can use to optimize image segmentation multilevel thresholding by minimizing the value of fuzzy divergence with genetic algorithm parameters which includes the size of the population is 60, the combination of size Cr and Mr respectively 0.6 and 0.4, and the size of generation is 100. Then, based on the evaluation result of image segmentation using Standard Deviation (SD), found that Gamma distribution leads better segmentation as compared to others.Keywords: chronic wounds, digital planimetry, image segmentation, fuzzy divergence, genetic algorithm