Study On Seismic Damage Change Detecting Techneque Based On Remote Sensing Images

On the basis of former research, the paper give more detailed study to the critical technology of remote sensing images change detection in the earthquake damages assessment. The paper analyzed the remote sensing data source which can be used in earthquake damages estimation, and summarized the traditional methods of image registration, in consideration of the specialization of the seismic disaster detection, we presented a method, which can enormously improve the image registration precision and the phenomenon of the image deformation inconsistency in some degree. The method is that, to choose plenty of ground control points using the relativity in different windows, and to transform space coordinates using triangulate network small surfaces elements. At the same time, the paper created residential area vectors in experimental areas, and extracted serious damaged areas by means of overlapping processing, which saved much time when the image was used to detect seismic disaster. We experiment on the remote sensing images (ETM+, ERS-2SAR, IRS-1C) acquired from pre- and post-Zhangbei and Taiwan earthquake, which took place on Jan. 10th, 1998 and on Sep.21st, 1999, with earthquake degree 6.2 and 7.6, respectively. The paper systematically studied the techniques of seismic damages detection, which include the following methods: the simple qualitatively display of important ground objects, such as buildings, constructions, and lifeline projects, in earthquake destruction region; the difference and ratio methods, which can extract seismic destruction regions at the scale of pixel, and the method of identifying seismic damages on the basis of correlative classification of the pre-and post-earthquake neighbor statistical features. The innovative work in the paper is that, it presented a seismic disaster extraction method on the basis of several indirect features correlative classification, such as variance, kurtosis, skewness, information entropy. The paper appraised the detection result of the proportion of collapse under different quadratic features assemble in different windows by means of confusion matrix, and gained the best quadratic features assemble in different proportion of collapse. We compiled a seismic disaster change detection system software using ENVI/IDL, and processed the images included in the paper by this system, and obtained perfect results. Furthermore, the paper built seismic disaster samples database to manage the typical images of important ground objects damaged by the earthquakes.