Information Extraction Of Damaged Town Buildings Caused By Earthquake With The Object-oriented Method From High Spacial Resolution Image

With the rapid development of remote sensing technology, remote sensing platform to increase, high-resolution commercial remote sensing satellite was successfully launched and running, images continue to lower costs, remote sensing technology has been provided the means of fast and efficient for the earthquake disaster investigation and damage assessment, as well provided the important information for disaster relief and mitigation decision-making. Domestic and foreign scholars have done a lot of exploration and pilot study, and have achieved many results of practical value in this area. But on the whole, the application is still relatively low level. Aerial remote sensing techniques of seismic damage investigation still stay in the artificial visual interpretation of the level, long period, low efficiency, and higher costs. The traditional pixel-oriented approaches show a lot of limitations for classification of high resolution satellite images, and they can not take full advantages of image information. Object-oriented image classification method based on image features such as the object spectrum, texture, shape and context information which used to classify, with a higher classification accuracy, According to the characteristics of high-spatial resolution remote sensing images, the paper presents a new object-oriented scheme of damaged building information extraction in the 5.12 Earthquake from high-spatial resolution remote sensing image in Wenchuan County.1) The multi-scale segmentation technology of remote sensing image has been studied and realized. On the basis of studying various image segmentation algorithms, the construction of image object of scale space has been realized mainly, and the image object network structure of multi-scale layer has been studied formed. At the same time, the strategy of multi-scale segmentation has been analyzed and summarized.2) The analysis of image object features has been studied, and the quantitative expression model of the spectrum, texture, shape and other geological features of image objects has been researched and established. The features selection of image object has been analyzed.3) The objected-oriented remote sensing image classification technology has been researched and realized. The principle of fuzzy classification has been analyzed. By building knowledge-based fuzzy systems, the system was introduced object-oriented remote sensing image classification to promote classification accuracy for the classification of features to provide more bases.4) Through the analysis of the secondary features of the building damage of high-resolution remote sensing images, the paper presents a new object-oriented scheme of damaged building information extraction in the 5.12 Earthquake from high- spatial resolution remote sensing image in Wenchuan County. The experimental results indicate that the new object-oriented information extraction technique has high accuracy compared to the traditional classification methods.