TY - GEN
T1 - Analysis of Damage to Buildings affected by the Tsunami in the Palu Coastal Area Using Deep Learning
AU - Meilano, Irwan
AU - Rahadian, Achmad Ikbal
AU - Suwardhi, Deni
AU - Suminar, Wulan
AU - Atmaja, Fiza Wira
AU - Pratama, Cecep
AU - Sunarti, Euis
AU - Haksama, Setya
N1 - Publisher Copyright:
© 2020 IEEE.
PY - 2020/12/7
Y1 - 2020/12/7
N2 - Assessing the building damage after a tsunami is the first step to quantitatively learn about the amount of damage it caused. Indonesia is an archipelagic country, with two-thirds of its territory consisting of water. It has the second-longest coastline in the world, increasing the potential for tsunami damage in Indonesian territory. In this study, an analysis of building damage due to the tsunamis was carried out and Palu was assigned as the study location. Palu's coastal area suffered a tsunami on September 28, 2018, caused by an earthquake with a magnitude of 7.5. The location and the number of buildings were generated through object detection using deep learning from high-resolution satellite imagery data. Object detection was carried out using pre-trained YOLOv3 models that are trained using 315 satellite images as data sets and produce a model with a loss value of 33.15. Object detection was carried out on satellite imagery before and after the tsunami and produced building distribution data with an accuracy of 76.78% and 74.20%, respectively. Comparisons of building data detected from the two satellite images were then analyzed using a tsunami height zone map to see the correlation between building damage and tsunami height. From spatial and correlations analysis, 1,547 damaged buildings were detected, giving the data a positive correlation type. Using the student's t-test, it was concluded that there was a significant correlation between building damage and tsunami height.
AB - Assessing the building damage after a tsunami is the first step to quantitatively learn about the amount of damage it caused. Indonesia is an archipelagic country, with two-thirds of its territory consisting of water. It has the second-longest coastline in the world, increasing the potential for tsunami damage in Indonesian territory. In this study, an analysis of building damage due to the tsunamis was carried out and Palu was assigned as the study location. Palu's coastal area suffered a tsunami on September 28, 2018, caused by an earthquake with a magnitude of 7.5. The location and the number of buildings were generated through object detection using deep learning from high-resolution satellite imagery data. Object detection was carried out using pre-trained YOLOv3 models that are trained using 315 satellite images as data sets and produce a model with a loss value of 33.15. Object detection was carried out on satellite imagery before and after the tsunami and produced building distribution data with an accuracy of 76.78% and 74.20%, respectively. Comparisons of building data detected from the two satellite images were then analyzed using a tsunami height zone map to see the correlation between building damage and tsunami height. From spatial and correlations analysis, 1,547 damaged buildings were detected, giving the data a positive correlation type. Using the student's t-test, it was concluded that there was a significant correlation between building damage and tsunami height.
KW - Palu
KW - Tsunami
KW - building damage
KW - deep learning
KW - satellite imagery
UR - http://www.scopus.com/inward/record.url?scp=85113393177&partnerID=8YFLogxK
U2 - 10.1109/AGERS51788.2020.9452780
DO - 10.1109/AGERS51788.2020.9452780
M3 - Conference contribution
AN - SCOPUS:85113393177
T3 - Proceeding - AGERS 2020: IEEE Asia-Pacific Conference on Geoscience, Electronics and Remote Sensing Technology: Understanding the Interaction of Land, Ocean and Atmosphere: Disaster Mitigation and Regional Resillience
SP - 95
EP - 97
BT - Proceeding - AGERS 2020
PB - Institute of Electrical and Electronics Engineers Inc.
T2 - 3rd IEEE Asia-Pacific Conference on Geoscience, Electronics and Remote Sensing Technology, AGERS 2020
Y2 - 7 December 2020 through 8 December 2020
ER -