Automatic Plotting of Architectural Facades with Multispectral Images

Abstract

This paper describes an automatic approach in architectural photogrammetry for the documentation and analysis of historic facades. Multispectral close-range images were taken and evaluated for identification of masonry wall materials. Digital image processing and close-range photogrammetric techniques were used to identify, nondestructively recognize, and map the materials and damages of a historic facade. Two supervised classification algorithms were analyzed: Maximum likelihood and minimum distance. Although the maximum likelihood was somewhat slower, it provides much higher classification accuracy than the minimum distance classifier. Additional to the statistical data, thematic images were obtained semiautomatically for visualization, management, and decision-making purposes.