DETEKSI CEPAT BANGUNAN DARI ORTOFOTO UNTUK VISUALISASI TIGA DIMENSI WILAYAH RENDAMAN TSUNAMI

• Main Author: Fikri Yuzar, Reza
• Format: Final Project
• Language: Indonesia
• Subjects: Geologi, hidrologi & meteorologi
• Online Access: https://digilib.itb.ac.id/gdl/view/55485
• Institution: Institut Teknologi Bandung

Palabuhanratu village, which is located on the southern coast of the island of Java and adjacent to a segment of tectonic plates, causes the Palabuhanratu area to be potentially affected by a tsunami. The three-dimensional city model from Palabuhanratu Village can be used as a basis for visualizing the tsunami hazard. To form a three-dimensional city model, building polygons and elevation data are needed. The manual digitization process to obtain building polygons is quite a time-consuming process. In this study, automated extraction of building footprint from orthophoto using machine learning was carried out and used the extraction results to create a three-dimensional model of the city. To visualize the tsunami hazard, a three-dimensional model of the city is superimposed on the model of the tsunami inundation area. Machine learning model training was conducted by providing training samples in the form of roof classification on orthophoto. The model is used to extract building footprint which in the next stage are used with elevation data from the digital elevation model to form a three-dimensional city model. From the research results, the accuracy of building trace extraction using machine learning is 83.75% and can identify 4288 of a total of 6418 building polygons as a result of manual digitization.