Research On The Extraction Method Of Urban Buildings Based On High Resolution Remote Sensing Image

With the development of science and technology and the continuous development and progress of satellite image technology,the types of remote sensing satellites continue to increase,and the performance is also constantly upgrading.The resolution of many satellite images has reached the level of sub-meter level,which provides more rich details for the research of ground object information.And China is a large country with abundant geographical information.Along with the development of the country and the construction of the city,the landform change is becoming more and more huge.Therefore,the efficient use of satellite images provided by the information on the future of urban construction and planning are of great significance.For urban areas,most of the landforms are covered by buildings,for the location of urban buildings and the state of the extraction of information,not only convenient for urban population planning is also convenient for such like the people's livelihood issue of urban pipeline layout and urban construction.Therefore,this paper is based on my laboratory project,combined with practical needs,is committed to finding an accurate,rapid and efficient urban building extraction method.Some methods for building pixel characteristics are widely used in the traditional extraction methods,the common problem is that the extraction effect is not ideal,the noise is too high and so on.In order to accurately extract the urban buildings,this paper designs a building classification strategy based on object oriented and nearest neighbor classification.First of all,for the extraction of homogeneous regions using object-oriented ideas for image segmentation,an improved watershed algorithm is proposed.And then extract the segmented building part by shape constraint,and according to the general size of the building to filter the building information whose extracted area is too small or too large.Then aiming at the problem that the spectral characteristics of adjacent roads and buildings are very similar,a multi-directional morphological filter algorithm is used to separate the two.Finally,based on the extracted building information and the separated non-building information to carry out building post-extraction based on object-oriented nearest neighbor classification.So that all the buildings in the image can be extractedThrough the analysis of the experimental results,the method selected in this paper can get a clear and accurate extraction of urban building information in high resolution satellite images within a certain time range,which can play a good role in urban management and planning construction at the same time can also provide strong technical support for the municipal management.