Saputra, Wanvy Arifha (2017) Penentuan Otomatis Seeded Region Growing Pada Region Watershed Untuk Segmentasi Citra Ikan Tuna. Masters thesis, Institut Teknologi Sepuluh Nopember.
Preview |
Text
5115201026-Master-Theses.pdf - Published Version Download (3MB) | Preview |
Abstract
Seeded region growing memiliki dua parameter utama, yaitu inisialisasi seed dan threshold. Kesalahan dalam penentuan parameter tersebut akan menyebabkan hasil kegagalan segmentasi dalam citra. Penentuan parameter tersebut dapat dilakukan dengan manual atau otomatis. Namun, pemilihan secara manual tidak dapat dilakukan pada basis realtime, sedangkan untuk pemilihan parameter secara otomatis harus efektif dalam menentukan seed dan threshold. Hal tersebut disebabkan penentuan secara otomatis memiliki kesalahan segmentasi yang lebih besar daripada manual.
Penelitian ini bertujuan untuk mengusulkan penentuan otomatis seeded region growing pada region watershed untuk segmentasi citra ikan tuna. Citra ikan tuna akan diproses kedalam ruang warna HSI (hue, saturasi, intesitas) dan untuk hue akan diproses dibentuk dengan region watershed. Setelah itu dilakukan perhitungan density suatu region, yang kemudian diurutkan density tersebut dan diambil density yang tertinggi. Region yang memiliki density tertinggi tersebut akan diambil berdasarkan intesitas gray level tertinggi, kemudian threshold didapatkan dari selisih jumlah rata-rata intesitas region yang digunakan dengan rata-rata intesitas region yang ditinggal. Hasil segmentasi akan diukur dengan menggunakan RAE (relative foreground area error), MAE (missclassification error) dan MHD (modified Hausdroff distance) yang dibandingkan dengan groundtruth.
Penentuan secara otomatis seeded region growing pada region watershed telah berhasil dilakukan. Penggunaan watershed dapat membuat kontur tertutup dan ruang warna HSI dapat mengatasi sebaran cahaya yang tidak merata pada citra. Penggunaan region watershed diruang warna HSI dapat membentuk region yang lebih sedikit, hal ini dapat meningkatkan efesiensi waktu untuk menentukan seed dan threshold. Hasil segmentasi citra ikan tuna berhasil dilakukan dengan dibuktikan nilai rata-rata RAE, ME dan MHD secara berurut pada data kategori 1 sebesar 6.77%, 1.78% dan 0.18%, sedangkan untuk pada data kategori 2 sebesar 3.44% , 1.30% dan 0.66%
========================================================
=============================
Seeded region growing has two key parameters, namely the initialization seed and threshold. Failure to determination of these parameters will cause error results in image segmentation. Determination of these parameters can do manually or automatically. However, the selection can not do manually on a realtime system. As for the selection of parameters should automatically be effective in determining seed and threshold, because automatically segmentation is have greater the error result than manually.
This research propose method to automatically determination seeded region growing in the watershed region for image segmentation of tuna. The image of the tuna will be processed into HSI (hue, saturation, intensity) color space and for the hue to be processed formed by the watershed region. After the calculation of density of the region, the region sorted and retrieved by highest density. Region which has the highest density will be taken by the highest intensity of gray level, then the threshold was obtained from the difference between the amount of the average intensity of regions used by the average intensity of regions neighbor. Segmentation results will be measured using RAE (relative foreground area error), MAE (missclassification error) and the MHD (modified Hausdroff distance) compared with Groundtruth.
Determining automatically seeded region growing in the watershed region has been successfully carried out. The use of the watershed can create closed contour and HSI color space can overcome uneven distribution of light in the image. The use of the watershed region in HSI color space can form a little of region, it can increase the efficiency of time to determine seed and threshold. The results of image segmentation tuna proved successful with an average value of RAE, ME and MHD sequentially in the data category 1 by 6.77%, 1.78% and 0.18%. As for the data category 2 at 3.44%, 1.30% and 0.66%.
| Item Type: | Thesis (Masters) |
|---|---|
| Uncontrolled Keywords: | seeded region growing; watershed; segmentasi; ikan tuna; segmentation; image of tuna |
| Subjects: | Z Bibliography. Library Science. Information Resources > ZA Information resources > ZA4050 Electronic information resources |
| Divisions: | Faculty of Information Technology > Informatics Engineering > 55101-(S2) Master Thesis |
| Depositing User: | - WANVY ARIFHA SAPUTRA |
| Date Deposited: | 22 Mar 2017 01:43 |
| Last Modified: | 06 Mar 2019 02:14 |
| URI: | http://repository.its.ac.id/id/eprint/2602 |
Actions (login required)
 |
View Item |