Study On Building Damage Detection Using High Resolution Synthetic Aperture Radar Imagery

Earthquake severely threatens the property and lives of people,caused especially by building damage induced by earthquake. Remote sensing technology can acquire the disaster information much faster and larger, compared with the traditional ground survey approach. Especially the Synthetic Aperture Radar(SAR), which can work regardless of the weather and sunlight, does be playing an increasing role in the application of acquirement and assessment of disaster information.This thesis researches on the methodology of building damage detection from high resolution SAR imagery in order to meet the needs of disaster mornitoring and evulation in the earthquake emerency stage, which includes as follows.Firstly, review of the existing methods of damage detection using SAR image. Then introduce the basic theory of radar, the available high resolution SAR system, and analyze the geometric features of buildings showing in SAR image in detail, which reflect that buildings have high intensity layover, bright line induced by double bounce backscattering mechanism, dark shadow features.Secondly, the reduction method of speckle noise, which is the special character of SAR image, is researched. For the research purpose of building damage detection in the paper, different window size of speckle reduction filter is tested, and is evaluated quantitatively through subjective and objective indices. The results shows that the Gamma filter performs better in speckle reduction and edge preserving.Then, the building damage features of COSMO-Sky Med image and ariborn SAR image are analyzed, which indicates that higher resolution SAR image has more advantage of building damage detection. The layover and shadow areas of un-damaged buildings, which are extracted by accumulating SAR intensity in range direction, and the collapsed area of damaged buildings. The texture features, such as homogeneity, dissimilarity and entropy, are selected, analyzed and compared. The results indicate that, the geometry features of buildings on SAR image are important clues to identify damage, and more than two geometry features can obviously improve the ability of damage detection.Last, estimating building damage assessment quantitatively is studied, using post-event SAR intensity image only. According to the analysis before, the high-intensity area which includes the double-bounce scattering line and the high intensity layover area, is closely connected with building damage. Based on the buildings’ high intensity area, a SAR Damage Index(SARDI) is proposed. Then the building damage of 58 districts in Yushu after Yushu earthquake is classified into three classes using SARDI,after extracting the high-intensity area by a maximum likelihood classifier using SAR synthetic color image in frequency domain. Results show that the classification precision is 89.3% and Kappa index is 0.79 compared with the classification result from post-event optical image, which verify the effectiveness of the SARDI this paper proposed.