Urban building collapse detection by exploiting invariant moment features from very high resolution imagery

In this paper, a method combining spatial information and spectral information was proposed for detection of urban building collapse caused by earthquake disasters. Given the spectral similarity between collapsed and undamaged classes, three invariant moments, namely Hu's moments, Zernike moments, and wavelet moments were used in this study. These moments were calculated for each image object, which is produced by image segmentation. The obtained invariant moments images and bitemporal multispectral images were combined and used to extraction collapsed buildings through direct multitemporal classification. The One-Class Support Vector Machine (OCSVM), a recently developed classifier was used in the classification. The proposed method was evaluated using bitemporal Quickbird images acquired in Bam, Iran, which was hit by a Mw 6.6 earthquake on December 26, 2003. The results showed that the combined use of spectral and spatial features significantly improved the collapse detection accuracy, compared to that of using spectral information alone.