Study Of Building-based Damage Detection Due To Earthquake Using Multi-perspective Aerial Images

Natural disasters such as earthquakes is an extremely grave threat to human life and property.After an earthquake,the area,amount,rate,and type of the building damage are essential information for emergency response actions and rescue work,and also provide important decide-making information to government for post-disaster reconstruction.Remote sensing techniques play an important role for damage detection and assessment due to their non-contact,low cost,wide field of view,and fast response capacities.Therefore,damaged information extraction and detailed and accurate assessment using remote sensing data is practical significance.Our paper studies damaged building region detection and damaged grade determination for emergency response actions and post-disaster reconstruction application.Our paper makes the following contribution.(1)The damaged types of building is complex and diversity due to its stereoscopic object feature,our paper studies the damaged feature representation in the remote sensing images based on EMS-98 and different damaged form of building,and summarizes damaged factor that impacts the building damage grade.Our paper describes the feature of various damaged factor and build math representation model of damaged factor,which provides a good theory support for building damaged detection and determination.(2)According to the practical demand of emergency response actions and rescue work at an early stage of an earthquake,a novel method is proposed for detecting damaged building regions based on semantic scene change in a visual Bag-of-Words model.Semantic scene change can provide a new point of view since it can indicate the land-use variation at the semantic level.Pre-and post-disaster scene change and damaged information in building regions are represented by a uniform visual codebook frequency.The scene change of damaged and non-damaged building regions is discriminated using the Support Vector Machine(SVM)classifier.Differing from traditional methods based on change detection for damaged building regions,our method have better self-adaptive and accuracy of detection,which provides a new method for the demand of emergency response actions and rescue work after an earthquake.(3)In EMS-98,moderate damage involves building rooftop and facade damage problems.The detection of damaged rooftop areas is crucial to improve the accuracy of classification of building damaged types.In our paper,an approach for the automatic detection of damaged rooftops areas based on Visual Bag-of-Words Model is presented.First,the building rooftop is segmented into different superpixel areas.Then,the Visual Bag-of-Words model is employed to build semantic feature vectors for damaged or non-damaged part of each superpixel area.Finally,damaged and non-damaged parts of rooftop superpixel areas are discriminated using SVM.Compared with the traditional approaches,our approach can detect the various type damaged areas of rooftops,such as debris,spalling,rubble,and hole areas,which provides foundation for multi-level damaged detection of building.(4)In EMS-98,moderate damage involves building rooftop and facade damage problems.Although much works have been done on rooftop damaged detection,so far,very little works has addressed facade damaged related to moderate damage.In this letter,a novel approach for automatic detection of damaged facade based on local symmetry features and the Gini Index using oblique aerial images is presented.First,local symmetry points are detected in a sliding window.Then,we obtain histogram bins of local symmetrical points in the vertical and horizontal directions.Finally,damaged and undamaged of building facades are distinguished using Gini Index.An evaluation of experimental result shows that this method is feasible and effective for the detection of damaged facades,which provides foundation for multi-level damaged detection of building.(5)The detailed damaged information of building is vitally important for later reconstruction in disaster areas.In our paper,a framework for detecting the building damaged grades using multi-perspective aerial images is proposed.The geo-information infrastructure data in the pre-disaster and multi-perspective aerial images in the post-disaster are used as data sources,3D damaged feature of the image-based 3D point cloud and 2D damaged features of the images are extracted to build multi-dimension damaged factor,and the damage grades are classified using fuzzy neural network.Differing from traditional methods of building damaged grade determination,multi-dimension damaged factor proposed in our method can obtain more comprehensive and accurate damage information and damaged type of building,and fuzzy neural network can be used to better distinguish the damaged types of building.