Journal · 2024
Anomaly Detection of Building Structure from Incomplete Point Cloud Obtained by UAV
Abstract
Japan has been severely impacted by natural disasters including but not limited to earthquakes such as the Great Hanshin-Awaji Earthquake in 1995 and the Kumamoto Earthquake in 2016. These seismic events have underscored the high number of casualties that result from individuals becoming trapped in collapsed buildings or affected by fires, thereby accentuating the need for building-specific earthquake assessments. Although experts have performed detailed analyses using automated satellite imagery and UAV-captured photos for longer-term objectives such as secondary disaster prevention, reconstruction, and insurance claim verification, these require a substantial amount of time to implement. This paper introduces two methods that employ UAVs to rapidly detect anomalous building structures, allowing for the simultaneous observation of multiple buildings. First, we present a method for identifying building structures from incomplete three-dimensional point clouds acquired by unmanned aerial vehicles (UAVs) that move faster over the target area in a limited time for rapid assessment. The method efficiently identifies the structural characteristics of buildings by operating under certain geometric assumptions such as the angles between building sides being approximately 90 degrees and vertical consistency in building shape. We also present a method for identifying collapsed buildings by extracting features from point clouds in Fisher vector and normal histogram and using a machine-learning model for detection. In our evaluations, we have shown that by limiting observations to less than half of the building structure, the first method can successfully recognize the geometric shape of 70% of the undamaged buildings. In the experiments for the second method, both feature extraction methods achieved a Receiver Operating Characteristic Area Under the Curve (ROC-AUC) values greater than 0.99.
Research Note
Japan has suffered tremendous damage from earthquakes and other natural disasters, including the 1995 Great Hanshin-Awaji Earthquake and the 2016 Kumamoto Earthquake. In these significant earthquakes, the number of victims due to occupants trapped by collapse and fire was exceptionally high, indicating the need to focus on the building structure for earthquake assessment. Detailed surveys by surveyors, automated satellite imagery, and UAV imagery are being used for assessment. However, these are aimed at secondary disaster prevention, reconstruction, and insurance coverage and require much time. Therefore, this method detects anomalous building structures by simultaneously observing multiple buildings using a UAV. The building structure is represented by the point cloud itself, and a machine learning model is used to classify the building as either undamaged or collapsed based on its features. These features are a Fisher Vector for the point cloud distribution and a Normal Histogram for the direction and number of planes in the point cloud. It takes advantage of the difference in shape between undamaged and collapsed buildings and the fact that the building surface becomes sparse due to collapse. Evaluation experiments showed both features achieved a ROC-AUC of more than 0.99, which indicated the machine learning model's performance independent of the threshold value and showed robustness to noise. Also the results show a trade-off between the time required for feature extraction, classification accuracy, and the degree of collapse that can be handled.
Published Papers
- 原田歩, 廣森聡仁, 山口弘純, & 東野輝夫. (2021). LiDAR を用いた疎な観測による多角柱復元手法の提案. マルチメディア, 分散協調とモバイルシンポジウム 2021 論文集, 2021(1), 1400-1407.
- 原田歩, 廣森聡仁, & 山口弘純. (2021). 3D 都市モデルを対象とした点群データによる大まかな建築群復元手法の検討と評価. 研究報告高度交通システムとスマートコミュニティ (ITS), 2021(33), 1-8.
- Harada, Ayumu, Akihito Hiromori, and Hirozumi Yamaguchi. "Anomaly Detection of Building Structure from Incomplete Point Cloud Obtained by UAV." Journal of Information Processing 32 (2024): 520-532.
