Analisis Perbandingan Ekstraksi Tapak Bangunan dan Jalan Menggunakan Mask R-CNN dan Hybrid OBIA - Mask RCNN dengan Data UAV Fixed Wing (Studi Kasus: Desa Banturejo, Kabupaten Malang)

Irawan, Reyhan Dhihan (2024) Analisis Perbandingan Ekstraksi Tapak Bangunan dan Jalan Menggunakan Mask R-CNN dan Hybrid OBIA - Mask RCNN dengan Data UAV Fixed Wing (Studi Kasus: Desa Banturejo, Kabupaten Malang). Other thesis, Institut Teknologi Sepuluh Nopember.

[thumbnail of 5016201044-Undergraduate_Thesis.pdf] Text
5016201044-Undergraduate_Thesis.pdf - Accepted Version
Restricted to Repository staff only until 1 October 2026.

Download (8MB) | Request a copy

Abstract

Pembangunan bertujuan untuk meningkatkan pemerataan, meningkatkan pertumbuhan ekonomi, menyebarkan kemakmuran di masyarakat dan memastikan terjadi merata di seluruh wilayah. Lokasi bangunan dan jalan yang tersebar juga mempengaruhi tata kelola kota. Untuk mencapai tujuan itu diperlukan pembangunan yang berkelanjutan dan pemenuhan keberadaan peta skala besar. Usaha mewujudkan pembangunan desa dan peta skala besar yang cepat dan teliti dapat dilakukan dengan pembuatan peta sebaran tapak bangunan dan jalan di desa dekat potensi wisata. Namun, pembuatan peta sebaran tapak bangunan dan jalan sering kali dilakukan secara manual menggunakan proses digitasi dimana metode ini memerlukan waktu yang lama. Dengan kemajuan teknologi menggunakan ilmu fotogrametri, ekstraksi tapak bangunan dan jalan dapat dilakukan secara otomatis. Penelitian ini bertujuan untuk melakukan segmentasi, training sample, dan klasifikasi bangunan dan jalan menggunakan algoritma Mask R-CNN dipadukan digitasi dan hybrid OBIA dipadukan dengan Mask R-CNN. Adapun data yang digunakan yaitu data citra UAV Fixed Wing tahun 2023 pada Desa Banturejo, Kabupaten Malang. Didapatkan hasil berupa peta klasifikasi sebaran bangunan dengan hasil uji validasi dan akurasi menggunakan confusion matrix nilai presisi 77,94%, overall accuracy 64,74%, recall 51,54%, F1 Score 62,02%, luas total 180.595 m2 dan jumlah polygon bangunan terdeteksi sebanyak 1.447 dengan metode Mask R-CNN. Sedangkan pada hyvrid OBIA-Mask RCNN klasifikasi bangunan menghasilkan nilai presisi 35,95%, overall accuracy 22,58%, recall 9,21%, F1 Score 14,66%, luas total 201.932 m2 dan jumlah polygon bangunan terdeteksi sebanyak 572. Pada peta klasifikasi sebaran jalan dengan metode Mask R-CNN didapatkan panjang total jalan 15.532 m dan lebar rata – rata 8,37 m sedangkan hybrid OBIA-Mask RCNN menghasilkan panjang total 11.222 m dan lebar rata-rata 8,89 m yang dibandingkan dengan hasil digitasi nilai panjang total jalan 14.073 m dan lebar rata-rata jalan 4,12 m.
=================================================================================================================================
Development aims to improve equity, increase economic growth, spread prosperity in community and ensure that it occurs evenly throughout the region. The dispersed location of buildings and roads also affects urban governance. Achieving these goals requires sustainable development and fulfillment of large-scale maps. Efforts to realize rapid and thorough rural development and large-scale maps can be made by making maps the distribution of building sites and roads in villages near tourism potential. However, making a map of the distribution of building footprints and roads is often done manually using a digitization process where this method takes a long time. With technological advances using photogrammetry, extraction of building footprints and road can be done automatically. This research aims to segment, train samples, and classify buildings and roads using Mask R-CNN algorithm combined with digitation and hybrid OBIA combined with Mask R-CNN. The data used is UAV Fixed Wing image data 2023 in Banturejo Village, Malang Regency. The results obtained in the form of building distribution classification map with the results of validation and accuracy tests using confusion matrix precision value 77.94%, overall accuracy 64.74%, recall 51.54%, F1 Score 62.02%, total area 180,595 m2 and number of building polygons detected 1,447 with Mask R-CNN method. While in the OBIA-Mask RCNN hybrids the building classification produces precision value 35.95%, overall accuracy 22.58%, recall 9.21%, F1 Score 14.66%, total area 201,932 m2 and number of building polygons 572. On the road distribution classification map with Mask R-CNN method, total length of the road is 15,532 m and average width is 8.37 m while hybrid OBIA-Mask RCNN produces total length of 11,222 m and an average width of 8.89 m which is compared to the digitized value of the total length of the road 14,073 m and the average width of the road 4.12 m.

Item Type: Thesis (Other)
Uncontrolled Keywords: Bangunan, Fixed Wing, Jalan, Mask R-CNN, OBIA, Building, Road
Subjects: G Geography. Anthropology. Recreation > G Geography (General) > G155.A55 Tourism and city planning
G Geography. Anthropology. Recreation > G Geography (General) > G70.217 Geospatial data
Divisions: Faculty of Civil, Planning, and Geo Engineering (CIVPLAN) > Geomatics Engineering > 29202-(S1) Undergraduate Thesis
Depositing User: Reyhan Dhihan Irawan
Date Deposited: 25 Jul 2024 04:47
Last Modified: 25 Jul 2024 04:47
URI: http://repository.its.ac.id/id/eprint/108780

Actions (login required)

View Item View Item