A novel and automatic framework for producing building damage map using post-event LiDAR data. (October 2019)

• Record Type: Journal Article
• Authors: Janalipour, Milad; Mohammadzadeh, Ali
• Is Part Of: International journal of disaster risk reduction. Volume 39(2019)
• Publication Date: 2019-10
• Subjects: Building damage map -- Post-event LiDAR data -- nDSM -- Inclination angle -- Automatic framework -- Texture extraction; Emergency management -- Periodicals; Risk management -- Periodicals; Disaster relief -- Periodicals; Hazard mitigation -- Periodicals; 363.34
• DOI: 10.1016/j.ijdrr.2019.101238
• Languages: English
• ISSNs: 2212-4209

Abstract

Abstract: A building damage map would help us to manage rescue operations, particularly after earthquakes. Light Detection and Ranging (LiDAR), as a data in remote sensing and photogrammetry sciences, has been recently used for producing building damage maps. However, there is no automatic building damage detection method based on post-event LiDAR data. For this reason, in this research, a novel and automatic building damage detection framework based on the post-event LiDAR data was proposed. In the proposed framework, at first three texture extraction strategies were employed to attain textural features. Then, the K-Means clustering algorithm was utilized to detect damaged buildings using the extracted textural features and produce a preliminary building damage map. Afterwards, for specifying pancaked buildings, normalized Digital Surface Model (nDSM) obtained from post-event LiDAR data was employed. Finally, for detecting inclined buildings, at first, the inclination angle of each building was extracted from post-event LiDAR data. Then, the inclined buildings were specified using the extracted inclination angle. The results showed that selecting an appropriate strategy for texture extraction greatly increases the accuracy of results. The obtained results indicated that the post-event texture features along with the K-Means clustering algorithm, the inclination angle and nDSM can help to reach acceptable results in building damage detection.