Research Of Building Extraction Based On High-resolution Remote Sensing Images

In high-resolution remote sensing images,the precise location and identification of building information is one of the important information sources for landscape analysis,population estimate,urban planning,environment surveying and disaster warning systems.The traditional building extraction methods with human intervention not only cost a lot of manpower and material resources,but also cannot reflect information about urban rapid development in real time.Hence with the development of high-resolution remote sensing images,building extraction currently is important in the application of high-resolution remote sensing images.At present,quite a few algorithms are available for building extraction by high-dimensional but low-level features(e.g.,textural-,contextual-,and morphological information).However,most of them still have some obvious disadvantages:(i)more complex computation,(ii)the ignorance of spectral information,(iii)restriction on the shape of the building,(iv)the higher resolution results in a larger data volume,which makes processing more difficult,and(v)the difficulty in efficiently balancing performance and complexity.In order to deal with this problem,this paper is trying to propose several algorithms for building extraction from high-resolution remote sensing images.Generally,main contents of this paper include the following parts:(i)Research of building extraction based on the improved morphological building index algorithm from high-resolution remote sensing imagesThe basic ideal of traditional morphological building index(MBI)algorithm is that the spectral-structural characteristic of building could described by white top-hat of reconstruction(W-TH).However,if there is shadow existed in the images or the luminance from building and area around buildings is closed,the luminance difference will be smaller and cause a high miss alarm rate for building extraction by MBI algorithm.Hence,the improved MBI algorithm was proposed by introducing the spectral feature of building and local image enhancement.The interference by shadow would be reduced.Besides,the post-processing was employed to refine building objects.The experimental results demonstrated the improved MBI algorithm achieved better building extraction.The overall accuracy is up to 86.06%,and the miss alarm rate which is caused by shadow would be reduced by the improved MBI.(ii)Research of building extraction based on the choice of the best spectral building index from high-resolution remote sensing imagesThe spectral feature is ignorance in the existing algorithms,which have more complex computation.Therefore,this paper proposed the Normalized Spectral Building Index(NSBI)and the Difference Spectral Building Index(DSBI)derived from existing indices for the building extraction based on the statistical properties.The model of spectral index and building spectral feature were analyzed.Then the optimal index was chosen to get the candidate building objects for different remotely sensed images.Furthermore,the post-processing was employed.The experimental results showed the proposed method could extract building effectively.The average overall accuracy is 86%.Besides,the running-time of the spectral building index is almost one-fifth from traditional MBI.