Extracting damaged building information from single remote sensing images of post-earthquake

摘要

In the high resolution images, undamaged buildings take on natural texture features, but to the damaged or extensive damaged buildings, there are always some low grayscale blocks because of their coarsely damaged section. By using such statistical information as the number of holes in every region, or the ratio between the area of the region and the holes', et al., damaged buildings can be separated from undamaged buildings, thus automatic detection of damaged buildings can be reached. Based on these characteristics, in this work, a new method to detect the damage buildings automatically by using region structure and statistic information of the texture is presented. Also, in order to test its validity, 1-meter-resolution iKonos merged image of Bhuj earthquake, India, 2001, and grayscale aerial photos of Tangshan earthquake, China, 1976, are selected as two examples to detect the damaged buildings automatically. Satisfactory results are obtained.