Please use this identifier to cite or link to this item:
http://repository.ipb.ac.id/handle/123456789/172170Full metadata record
| DC Field | Value | Language |
|---|---|---|
| dc.contributor.advisor | Siregar, Vincentius P. | - |
| dc.contributor.advisor | Gaol, Jonson Lumban | - |
| dc.contributor.advisor | Panjaitan, James Parlindungan | - |
| dc.contributor.author | Ulfah, Diyanah | - |
| dc.date.accessioned | 2026-01-20T06:37:10Z | - |
| dc.date.available | 2026-01-20T06:37:10Z | - |
| dc.date.issued | 2026 | - |
| dc.identifier.uri | http://repository.ipb.ac.id/handle/123456789/172170 | - |
| dc.description.abstract | Peta habitat bentik berperan penting dalam penilaian ekosistem laut sebagai alat komunikasi efektif dalam mengambil keputusan untuk pengelolaan lingkungan. Teknologi penginderaan jauh memungkinkan pemetaan pola lingkungan berskala besar dan hemat biaya dibandingkan dengan survei lapangan, sehingga citra pengindraan jauh banyak dimanfaatkan dalam pemetaan habitat bentik. Penelitian ini bertujuan mengklasifikasikan habitat bentik dengan pendekatan berbasis piksel dan object-based image analysis (OBIA) menggunakan citra Unmanned Aerial Vehicle (UAV) Multispektral dengan algoritma machine learning yaitu Bayes, Decision Tree (DT), K-Nearest Neighbor (KNN), Random Forest (RF), dan Support Vector Mechine (SVM). Kemudian mengetahui akurasi klasifikasi citra UAV Multispektral dengan metode klasifikasi beserta algoritma yang digunakan. Penelitian ini dilakukan di perairan Pulau Lancang, Kepulauan Seribu. Metode penelitian terdiri dari pengamatan habitat bentik, klasifikasi berbasis piksel, klasifikasi OBIA, dan uji akurasi. Metode klasifikasi OBIA dilakukan dengan segmentasi multi-resolusi untuk menghasilkan objek yang sesuai secara morfologis dan ekologis dengan fitur habitat yang ingin dipetakan. Skala segmentasi yang digunakan pada reef level (Level 1) yaitu segmentasi 150, dan pada wilayah perairan dangkal untuk menghasilkan habitat bentik (Level 2) dilakukan dengan skala segmentasi 30, 50, dan 70. Selanjutnya di klasifikasi dengan algoritma mechine learning yang telah disebutkan diatas. Habitat bentik diperairan Pulau Lancang berdasarkan klasifikasi OBIA dan pixel-based didominasi oleh kelas habitat Karang. Pendekatan OBIA menghasilkan nilai akurasi tertinggi pada skala segmentasi 70 dengan algoritma RF sebesar 87,35%, algoritma DT sebesar 86,53%, dan algoritma Bayes sebesar 83,26%. Pada algoritma KNN nilai akurasi tertinggi dihasilkan dari skala segmentasi 30 sebesar 86,12% dan algorima SVM menghasilkan nilai akurasi tertinggi pada skala segmentasi 50 sebesar 76,73%. Kemudian pendekatan berbasis piksel menghasilkan nilai OA tertinggi pada algoritma RF sebesar 80,03%, algoritma DT sebesar 77,27%, algoritma SVM sebesar 72,73%, algoritma Bayes sebesar 68,13%, dan algoritma KNN sebesar 67,42%. Nilai koefisien kappa yang dihasilkan klasifikasi OBIA masuk kedalam kategori Baik hingga Sangat Baik, kemudian pada pendekatan berbasis piksel nilai koefisien kappa yang dihasilkan berada pada kategori Moderat hingga Baik. Perbandingan uji akurasi dengan pendekatan OBIA dan pixel-based, diperoleh nilai akurasi klasifikasi menggunakan pendekatan OBIA lebih tinggi dibandingkan penggunakan pendekatan pixel based. | - |
| dc.description.abstract | Benthic habitat maps is very important in marine ecosystem assessments as an effective communication tool for coastal environmental management decision-making. Remote sensing technology enables large-scale and cost-effective environmental pattern mapping compared to field surveys, making remote sensing imagery widely utilized in benthic habitat mapping. This study aims to classify benthic habitats using pixel-based and object-based image analysis (OBIA) approaches based on Unmanned Aerial Vehicle (UAV) multispectral imagery, employing machine learning algorithms including Bayes, Decision Tree (DT), K-Nearest Neighbor (KNN), Random Forest (RF), and Support Vector Machine (SVM). In addition, this study evaluates the classification accuracy of UAV multispectral imagery using the applied classification methods and algorithms. This study was conducted in the Lancang Island water, Seribu Islands. The research methods consisted of benthic habitat observation, pixel-based classification, OBIA classification, and accuracy assessment. The OBIA method was performed using multi-resolution segmentation to generate objects that are morphologically and ecologically appropriate for the habitat features to be mapped. The segmentation scales used at the reef level (Level 1) was a scale of 150, and in shallow waters to delineate benthic habitats (Level 2), segmentation scales of 30, 50, and 70 were applied. These were subsequently classified using several machine learning algorithms as described above The benthic habitats in Lancang Island water which conducted, based on both OBIA and pixel-based classifications, are found dominated by coral habitat class. The OBIA approach yielded the highest accuracy values at a segmentation scale of 70 with the RF algorithm (87.35%), DT algorithm (86.53%), and Bayes algorithm (83.26%). The KNN algorithm produced its highest accuracy from a segmentation scale of 30 (86.12%), and the SVM algorithm yielded its highest accuracy at a segmentation scale of 50 (76.73%). Meanwhile, the pixel-based approach produced the highest Overall Accuracy (OA) with the RF algorithm (80.03%), DT algorithm (77.27%), SVM algorithm (72.73%), Bayes algorithm (68.13%), and KNN algorithm (67.42%). The Kappa coefficient from the OBIA classification was categorized as Good, while the pixel-based approach yielded Kappa coefficients ranging from Moderate to Good. A comparison of the accuracy tests between the OBIA and pixel-based approaches shows that classification using the OBIA approach is superior to using the pixel-based approach. | - |
| dc.description.sponsorship | null | - |
| dc.language.iso | id | - |
| dc.publisher | IPB University | id |
| dc.title | Pemetaan Habitat Bentik Perairan Dangkal Menggunakan Citra Multispectral Unmanned Aerial Vehicle (UAV) di Perairan Pulau Lancang, Kepulauan Seribu | id |
| dc.title.alternative | Benthic Habitat Mapping of Shallow Water using Multispectral Unmanned Aerial Vehicle (UAV) in The Lancang Island, Seribu District | - |
| dc.type | Tesis | - |
| dc.subject.keyword | Benthic habitat | id |
| dc.subject.keyword | UAV Multispectral | id |
| dc.subject.keyword | OBIA | id |
| dc.subject.keyword | Pixel based | id |
| dc.subject.keyword | Machine learning | id |
| Appears in Collections: | MT - Fisheries | |
Files in This Item:
| File | Description | Size | Format | |
|---|---|---|---|---|
| cover_C5502222006_35268efcaab3461f8d1bf61c487b1559.pdf | Cover | 5.08 MB | Adobe PDF | View/Open |
| fulltext_C5502222006_7ea3af3e726f41eb856126c987c9076d.pdf Restricted Access | Fulltext | 3.01 MB | Adobe PDF | View/Open |
| lampiran_C5502222006_8e97683dcd4a4bf29db87d771b89abe0.pdf Restricted Access | Lampiran | 5.08 MB | Adobe PDF | View/Open |
Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.