Research On Classification Of Building Damage Information Based On High Resolution Remote Sensing Image Data

Earthquakes, tsunamis and other natural disasters always cause heavy damage to people’s lives and property. The information of the building’s damage degree is an important indicator of post-earthquake hazard assessment. The acquisition of the information of the building’s damage degree is crucial to the post-earthquake assistants and can reduce the losses of the earthquake. With the rapid development of remote sensing technology high-resolution Multi-temporal, massive remote sensing data provide a reliable assessment of the data source for disaster relief and disaster losses. In particularly, the rapid development of UAV technology, UAV airborne remote sensing has been widely used in obtaining the post-earthquake disaster information. The post-earthquake seismic characteristics of the sea and land is different, the building damage classification and extraction of information has important practical value and practical significance.The data sources of the sea earthquake of the building damage information classification and extraction was selected the World-view2 high-resolution satellite imagery Sendai City before and after the earthquake in Japan, March 11, 2011. Combined with the feature of Sendai actual land types, the multi-scale segmentation image before the earthquake, by setting the different characteristics in different segmentation layer threshold to realize the different feature information extraction, classification result in vector data output and by adding up the image of the earthquake on board after segmentation, analysis on the characteristics of different threshold set to extract the arable land, water, building, then the earthquake unclassified image segmentation to extract according to dimension 45 other feature information. Finally, I analysis the two phase extraction results, and then defied the buildings to reduce, building did not change, increase class building, and then output the detection vector data and make the data transfer matrix calculated in ArcGIS software, we can obtain the buildings collapse area. Classification results accuracy evaluation is that before the earthquake overall classification precision reaches 88.8%, after the overall accuracy of 85.7%.The data used to classify the land building of the earthquake damaged information is the uav aerial images and World-view2 high-resolution satellite imagery of Longtoushan Town. Ludian County. Zhaotong City. Yunnan Province before and after the earthquake on August 4, 2014.To extract the building earthquake damage information- is divided into four types include basic intact, moderate damage, severe damage,and completely collapsed using the method of object-oriented information extraction and change detection.Select appropriate parameters for the object-oriented segmentation, build a set of rule according to the spectrum, texture and shape features of different ground object and establish a characteristic space,and to classify the disaster area situation based on fuzzy function classification combined with appropriate classification of the nearby. Firstly, segmented the two-times image dimension based on spectral mean, Blue_ratio, Area features and so on. Then I established proper feature space with the neighboring classification method, and made the fine classification according to the grades of four types of damage to buildings, Classification results accuracy evaluation is that before the earthquake overall classification precision reaches88.16%, after the overall accuracy of 91.56%. And then output the detection vector data and make the data transfer matrix calculated. The results show that it can satisfy the rapid assessment requirements to extract the four types of building earthquake damage information used the method of object-oriented earthquake damage information extraction based on the optimal scale.It shows that the method to classify and extract the remote sensing information of the damage to buildings after the earthquake and tsunami is relatively simple, the mainly reason is the debris of the damage buildings was taken away by the sea, the images of the buildings after the earthquake is in good condition and basic condition; but the debris of the damage buildings by the land earthquake was scattered in the streets, land or the undamaged buildings nearby, that caused some difficulties to extract the building earthquake damage information. Based on two experimental area in the structure of the earthquake damage information extraction,using the object-oriented image change detection classification method,,and through the selection of building earthquake damage information characteristics by high-resolution remote sensing image, implements a contoured by front facies classification results vector interested in earthquake area change detection technique,the establishment of a set of rules and the setting of threshold that can meet the requirement of the disaster loss evaluation by fine classify of the building earthquake damage information.