Analisis Klasifikasi Bangunan Industri Menggunakan Data Orthophoto dan NDSM Dengan Pendekatan Deep Learning (Studi Kasus: Kelurahan Kali Rungkut, Surabaya)

Raihan, Muhammad Anis (2023) Analisis Klasifikasi Bangunan Industri Menggunakan Data Orthophoto dan NDSM Dengan Pendekatan Deep Learning (Studi Kasus: Kelurahan Kali Rungkut, Surabaya). Masters thesis, Institut Teknologi Sepuluh Nopember.

Text 6016221007_Master_Thesis.pdf - Accepted Version Restricted to Repository staff only until 1 October 2025. Download (9MB) | Request a copy

Abstract

Informasi penting yang dibutuhkan adalah tutupan lahan area terbangun, salah satunya adalah kawasan industri, dikarenakan kawasan industri dapat meningkatkan keadaan ekonomi warga sekitar. Untuk mendapatkan informasi tersebut, dapat dilakukan segmentasi dan klasifikasi dengan motode yang akhir- akhir seringan digunakan yaitu adalah pendekatan deep learning. Segmentasi dan klasifikasi tapak bangunan yang dilakukan, menggunakan dapat menggunakan data orthophoto. Namun, memiliki tantangan bahwa keadaan bangunan yang sangat beragam baik dari bentuk dan ukuran, sehingga memerlukan data tambahan berupa data elevasi (NDSM) untuk memudahkan pengidentifikasianya. Dalam penelitian ini, berbasis Mask Region-based Convolutional Neural Network (Mask R-CNN) untuk segmentasi tapak bangunan dan metode Deep Neural Network (DNN) untuk klasifikasi jenis bangunan industri dan non industri. metode yang digunakan untuk segmentasi tapak bangunan menghabiskan waktu selama 44 jam 10 detik dengan menghasilkan akurasi presisi 88,49%; kelengkapan (recall) 66,82%; dan F1-score 76,15%. Dengan hasil segmentasi berdasarkan standar ketelitian geometri peta RBI pada Peraturan Badan Informasi Geospasial Nomor 6 Tahun 2018, masuk pada ketelitian peta skala 1:1000 kelas 3 dam 1;5000 Kelas 3. Lalu untuk model klasifikasi yang dilakukan, menghasilkan akurasi yang sangat baik dengan nilai rata – rata presisi, recall, dan F1-score berturut – turut sebesar 0,94 ; 0,90 ; dan 0,92. Sedangkan secara kualitatif, model klasifikasi secara visual sudah sangat baik walaupun masi terdapat beberapa bangunan yang salah dalam pengklasifikasiannya, yang mungkin disebabkan oleh nilai variabel yang mungkin sama.
====================================================================================================================================
The important information needed is the land cover of built-up areas, one of which is an industrial area because industrial areas can improve the economic situation of residents. To obtain this information, segmentation, and classification can be done with the most recently used method, namely the deep learning approach. Extraction and classification of building sites can be done using orthophoto data. However, it has the challenge that the state of the building is very diverse both in shape and size, so it requires additional data in the form of elevation data (NDSM) to facilitate its identification. In this study, a Mask-based Region-based Convolutional Neural Network (Mask R-CNN) for building footprint extraction and a Deep Neural Network (DNN) method for the classification of industrial and non-industrial building types. the method used for building footprint extraction took 44 hours and 10 seconds with precision accuracy of 88.49%; completeness (recall) of 66.82%; and F1-score of 76.15%. The extraction results are based on the RBI map geometry accuracy standards in the Geospatial Information Agency Regulation Number 6 of 2018, it is included in the accuracy of the 1:1000 Class 3 and 1: 5000 Class 1 scale map. Then the classification model carried out, produces very good accuracy with average values of precision, recall, and F1-score of 0.94; 0.90; and 0.92, respectively. While qualitatively, the classification model is visually perfect even though there are still some buildings that are wrong in the classification, which may be caused by variable values that may be the same.

Item Type:	Thesis (Masters)
Uncontrolled Keywords:	Orthophoto, NDSM, Deep Learning, Mask R-CNN, Deep Neural Network, Industri Orthophoto, NDSM, Deep Learning, Mask R-CNN, Deep Neural Network, Industry
Subjects:	G Geography. Anthropology. Recreation > G Geography (General) G Geography. Anthropology. Recreation > G Geography (General) > G70.212 ArcGIS. Geographic information systems. G Geography. Anthropology. Recreation > G Geography (General) > G70.217 Geospatial data G Geography. Anthropology. Recreation > G Geography (General) > G70.5.I4 Remote sensing
Divisions:	Faculty of Civil, Planning, and Geo Engineering (CIVPLAN) > Geomatics Engineering > 29101-(S2) Master Thesis
Depositing User:	Muhammad Anis Raihan
Date Deposited:	08 Aug 2023 12:48
Last Modified:	08 Aug 2023 12:48
URI:	http://repository.its.ac.id/id/eprint/104375

Actions (login required)

View Item