MUTHMAINAH, 22314207 (2026) UJI AKURASI EKSTRAKSI TAPAK BANGUNAN ANTARA PLUGIN MAPFLOW-QGIS DAN DATASET GOOGLE OPEN BUILDINGS UNTUK PEMBUATAN PETA DASAR PERTANAHAN (Studi: Kalurahan Margoagung, Kapanewon Seyegan, Kabupaten Sleman). Diploma thesis, Politeknik Agraria STPN.
![[img]](http://repository.stpn.ac.id/style/images/fileicons/text.png) |
Text
Muthmainah_22314207.pdf Download (1MB) |
Abstract
The utilization of Artificial Intelligence (AI) technology increasingly supports efficiency and automation in geospatial data extraction, including massive-scale building footprint mapping. The Land Base Map (Peta Dasar Pertanahan), particularly in accelerating the Complete Systematic Land Registration (PTSL) program, is a crucial element in supporting the legal certainty of land rights, spatial planning, and land administration. Building footprint extraction from aerial imagery, as a main component of base map updating, is generally performed through manual digitization which requires immense time and resources; however, it can now be automated using AI-based deep learning approaches. This study aims to evaluate and compare the accuracy and effectiveness of building footprint extraction between the Mapflow-QGIS plugin and Google Open Buildings (GOB) data as a consideration for creating the Land Base Map in Kalurahan Margoagung, Kapanewon Seyegan, Sleman Regency. This research employed a quantitative approach using an object-based accuracy assessment method, with reference data (ground truth) consisting of 575 building footprints. The manually digitized results were distributed across three regional typologies: dense settlements (305 polygons), sparse settlements (83 polygons), and vegetation-covered areas (187 polygons). Performance evaluation was conducted using Precision, Recall, and F1-Score metrics for thematic accuracy, as well as Intersection over Union (IoU) for geometric accuracy. The Mapflow-QGIS plugin demonstrated superior geometric accuracy with an IoU of 0.71, Precision of 0.94, Recall of 0.87, and F1-Score of 0.90. The GOB data recorded a slightly higher F1-Score of 0.91, with a Precision of 0.96 and Recall of 0.87; however, its IoU value only reached 0.51—marginally exceeding the minimum threshold of IoU > 0.5 and remaining 0.20 points lower than the Mapflow-QGIS plugin, a significant difference in the context of cadastral mapping precision. The relatively low IoU value of the GOB data is attributed to the disproportionate expansion of polygon areas (over-bounding), caused by the global model's tendency to incorporate non-building elements into the building footprint boundaries. Based on the overall evaluation results, the Mapflow-QGIS plugin is considered more suitable to be integrated into the automated workflow for updating cadastral-scale Land Base Maps within the Ministry of Agrarian Affairs and Spatial Planning/National Land Agency (ATR/BPN), although GOB data remains relevant for identifying settlement distributions on a broader scale. Keywords: Artificial Intelligence, Building Footprint Extraction, Mapflow-QGIS, Google Open Buildings, Cadastral Base Map, Spatial Accuracy Assessment.
| Item Type: | Thesis (Diploma) |
|---|---|
| Subjects: | H Social Sciences > HD Industries. Land use. Labor > HD29 Pertanahan |
| Divisions: | Prodi Diploma IV Pertanahan |
| Depositing User: | yosep ka perpus |
| Date Deposited: | 30 Sep 2026 04:55 |
| Last Modified: | 30 Sep 2026 04:55 |
| URI: | http://repository.stpn.ac.id/id/eprint/5136 |
Actions (login required)
 |
View Item |
