Change Detection Of Buildings In Urban Area With High-resolution Remote Sensing Images Based On Multi-scale Object

With the increasing popularity of high resolution and very high resolution remote sensing images,high resolution remote sensing image data has become an important data source for geographical conditions monitoring.At present,the rich surface detail information provided by high-resolution remote sensing images has satisfied the interpretation of certain small-scale features,greatly reducing the workload of field investigation.To deduce the workload of office work by manual,how to detect the ground object and change of ground object automatically and accurately has became a serious problem to be sovled in hige resolution remote sensing image processing.However,with the increasing of remote sensing image spatial resolution,the image complexity also increases.In the high-resolution images,one object is usually composed of multiple pixels.The increased purity of pixel reduces the impact of hybrid pixels on automatically interpretation,but it also increases the spectral difference within the object,so that the phenomenon of "same object with different spectrum,homologous spectrum with different objects" is more common.In addition,the difference of the same object in multi-temporal images is more obvious than the low-resolution remote sensing image.A higher challenge is posed for automatic interpretation of high-resolution remote sensing images.Buildings are important man-made objects which closely related to human activities,and are one of the important targets of human observation.In the urban planning,land use,disaster monitoring and disaster assessment,population estimation and forecasting,urban ecology and other urban researches,the building as the main research subject,its information extraction and change detection has become one of the current hotspot in the high-resolution remote sensing image processing research.In the study of high-resolution remote sensing image building change detection,it is difficult to extract buildings because of the complex and diverse characteristics of the building in the high-resolution remote sensing image.In addition,there is spatial shifting of the same building in different phases,resulting in large differences in buildings in high resolution remote sensing when the buildings are still not changed.Based on the comprehensive analysis of existing object-oriented change detection methods,this paper proposes an object-oriented method to detect the change information of high-resolution remote sensing image buildings.The main research work of this paper includes the following parts:(1)Construction of multi-temporal image objects based on matching the corresponding objectsAs to the problem of image object construction of multi-temporal remote sensing images in object-oriented change detection method,this paper puts forward the concept of "the corresponding objects",which is similar to "the corresponding points".The corresponding objects refer to the same object in different remote sensing images.The difference of remote sensing images of different phases not only comes from the change of objects,but also from the difference of images caused by sensors and the influence of the external environment when signal is received.In the case of buildings,the spatial shifting caused by the inconsistency of data acquisition conditions makes the same buildings have large differences in multi-temporal remote sensing images,which has caused great interference to the detection of buildings.Aiming at improving this problem,this paper proposes a multi-temporal image object construction method based on the matching of the corresponding objects,and abandons the way of constructing the same image object in different time-phase image in traditional object-oriented change detection.A method is proposed considering the spectrum,texture,geometric,spatial distance and spatial topological relation of image objects to determine the matching degree of multi-temporal image objects.Next,according to the matching degree,the heuristic search method of the corresponding objects is used to determine the relationship of objects in different images.(2)Change detection of urban built-up area based on the corresponding objectsThe built-up area is the actual unit of urban development activities and is indispensable information for urban research and analysis,and urban monitoring.At the same time,the built-up area is also the scene information of building in a large scale.Therefore,this paper puts forward the change detection method of urban built-up area based on the corresponding objects.Built-up area identification is an important part of the change detection.However,due to the diversity,type,shape,structure and arrangement of buildings in different built-up areas,the spectral and geometric characteristics of different built-up areas are obviously different.On the basis of fully observing and analyzing the characteristics of the high-resolution remote sensing images,this paper presents a texture feature description method for the texture of the built-up area.Then this feature is used as the bottom feature in the visual attention calculation model based on information theory.This paper proposes saliency index of built-up area to describe the built-up area in the high resolution remote sensing images.The index is used as an important feature to identify the built-up area in the change detection of the built-up area.The implement of built-up area extraction by constructing the image objects and the image object relationship of the different time phases and then comparing the characteristics and categories of matching objects.(3)Change detection of urban building based on multi-scale corresponding objectsThe building demolition and construction projects during urban development are usually carried out to the general area.In view of the characteristics of urban buildings demolition and construction,this paper proposes a strategy of step-by-step change detection,and proposes a method of urban building change detection based on multi-scale corresponding objects.In this paper,a multi-scale image objects is constructed in the process of building change detection.Firstly,the urban built-up area change detection method based on the corresponding objects is used to detect the change of the built-up area in the image to reduce the range of building change detection.Then,the change of the building is determined by the corresponding objects relationship on this small scale and the hierarchical relationship between the multi-scale objects,so as to realize the regional-object change detection.In order to accurately identify and extract the buildings,this paper proposes a high-resolution remote sensing image building extraction method based on the scene information,which utilizes the context relation between the building and the shadow,the building and the built-up area.