Combined Use of Spectral and Spatial Features for Building Extraction in Multi-Spectral Imagery

As essential character of urban region, building extraction and recognition has been applied broadly in urban mapping, urban planning and population census. Traditional manual plotting is time consuming and expensive, which therefore challenges for automatic or semi-automatic solutions. High-resolution multi-spectral remote sensing imagery provides both spectral and spatial information for acquiring urban features to update geographic information database. An advanced algorithm based on the combined use of spectral and spatial features will be developed and employed to recognize and extract buildings from multi-spectral imagery in this paper. Firstly, the imagery is spatially filtered to achieve more homogeneous regions. With the spectral and spatial features, an automatic and iterative region growing algorithm is employed to segment the imagery. A feature vector is developed to recognize the buildings from the final segmentation result. The result shows that this method can extract 69.8% of the buildings in the tested imagery.