The Research Of Building Structural Damage Assessment Based On Airborne Oblique Images

Earthquake is one of the most serious disasters in the world. It seriously threatens to people’s personal safety and property safety because of its unexpectedness, short duration and heavy destruction. So rapid damage assessment after disaster events is helpful to implement emergency rescue. The traditional ground-based mapping method is too slow, and restricted by geography site. High resolution remote satellite images and the traditional aerial image can provide image data with high precision, but there are still many restrictions. Especially the structural damage assessment of buildings, it cannot provide the facade perspective. Oblique aerial photography technology alleviate the phenomenon, because it provides the East, South, West, North four directions besides a vertical perspective. It is possible that the structural damage assessment of buildings. So this paper presents structural damage assessment based on object-oriented Technology with oblique aerial images and carries out the following work:(1) Summarizing the development history, system composition, photographic principle, data characteristics and main application of oblique aerial photography technology. On the basis of this, summarizing the types of building after earthquake and listed the analysis model of damage characteristics.(2) The segmentation technique of oblique airborne images is researched. And aiming at the problem of incomplete vegetation segmentation made of the ingle threshold segmentation algorithm (optimal entropy), this paper proposes an improved segmentation algorithm with custom vegetation index.(3) Several classification methods are analyzed based on object-oriented classification of image objects and the fuzzy classification method is selected to classify the image objects. Structural damage assessment of buildings is achieved used fuzzy classification theory based on the spectrum, image texture, shape and other attribute information.(4) The overall accuracy, user accuracy, producer’s accuracy and Kappa coefficient related to the confusion matrix are used for quantitative analysis of the fuzzy classification results.Studies show that the improved of optimal entropy is more effective than the traditional method for vegetation extraction. Feature models are selected to research for fuzzy classification with image objects based on multi-scale segmentation. And the overall accuracy, user accuracy, producer’s accuracy and Kappa coefficient are used for quantitative analysis of the classification results,it is proved to be feasible.