| Earthquake disaster is one of the most fatal natural calamities, which seriously threatens peoples' life and properties. A destructive earthquake has a feature of truculency and gusty. It can destroy a city in several to several tens seconds. Unfortunately, limited by the science and technology and also beyond the people's knowledge, now people can't give the accurate forecasts to all destructive earthquakes. Therefore, making active earthquake counter-measures, quickly surveying the damage distribution and amount, quickly evaluating the damage, valid decision-making about salvation, are effective approaches to reduce damage of earthquake.Because of excellent technical predominance, remote sensing has been introduced into seismic protection and disaster reduction early in 1970s. With rapid development of technique systems of remote sensing, increase of remote sensing platforms, successfully launching of high resolution commercial satellites and its well working, descending cost of imagery, remote sensing has become a rapid and effective method to survey and evaluate earthquake damage, and also provides important information to earthquake relief and decision-making for damage reduce. Till now, the method of using remote sensing to detect earthquake damage can be classified into two big sorts: mono-temporal technique and multi-temporal technique. The mono-temporal technique uses one-time image obtained after earthquake with manual visual recognition to detect damaged elements, and the multi-temporal technique uses two images of different times before and after events by change detection to find earthquake damage. However, both of the two methods have many insufficiencies. The multi-temporal technique is conditioned by the heterogeneity of the two images, different parameters like resolution or incidence angle for every scene leading to a lower accuracy of the interpretation. And a large temporal gap between images always leads to the impossibility of damage detection for the new buildings which appear on the post event scene. The current mono-temporal technique can't meet the practical requirement of damage evaluation and relief because of its heavy workload, low speed and efficiency.During the course of this study, the author has noticed such a fact as following.In high-resolution remote sensing imagery after events, undamaged buildings show well-proportioned texture, but to extensive damaged buildings or collapsed buildings, many flecky low-gray-scale regions appear in their imagery positions for their rough damaged sections. At the same time, in the high resolution optical remote sensing image, city buildings always show high luminance, but vegetation, bare ground, street and water areas all show low luminance. Therefore, when using proper threshold to segment the image luminance, some holes will appear in these low-gray-scale regions. If we count out the mount of the holes in every region, and with the other two statistics: the ratio between hole area and region area, and the ratio between number of boundary points and full summation of points inside the region, then the computer can distinguish the damaged buildings from undamaged buildings, and reaches the purpose of automatic recognition. Based on such analysis, and with support of The National High Technology Research and Development Program of China (No. 2001BA601B040105) and the tenth Key Technologies R&D Program (2001AA13 6040), we proposes a new damage detection method which uses imagery texture and structure statistical information in high resolution remote sensing image after earthquakes to automatically extract building damage information. In order to test the efficiency of this method, an original model has been designed on MATLAB 6.1, and 1 m resolution iKonos image of the Bhuj, India earthquake of 2001, and aerial photograph of the Tangshan, China earthquake of 1976, as two examples to check this method. The results show that the results from using this automatic damage detection are basically accordant with those from manual image interpretat...
|
PDF Full Text Request