Semiautomated building extraction based on CSG model-image fitting

Building extraction based on pre-established models has been recognized as a promising idea for acquiring 3D data for buildings from aerial images. This paper proposes a novel building extraction method developed from the concept of fitting CSG (Constructive Solid Geometry) primitives to aerial images. To be practicable, this method adopts a semiautomatic procedure, carrying out high-level tasks (building detection, model selection, and attribution) interactively by the operator and performing optimal model-image fitting automatically with a least-squares fitting algorithm. Buildings, represented by CSG models, can be reconstructed part by part after fitting each parameterized CSG primitive to the edge pixels of aerial images. Reconstructed building parts can then be combined using CSG Boolean set operators. Consequently, a building is represented by a CSG tree in which each node links two branches of combined parts. This paper demonstrates ten examples of building extraction from aerial photos taken at a scale of 1:5,000 and scanned at a pixel size of 25 μm. All of the tests were performed in the prototypal system implemented in a CAD-based environment cooperated with a number of specially designed programs. The process time for each primitive is about 20 seconds and the successful rate of model-image fitting was about 90 percent. Evaluated with some check points, the fitting accuracy was about 0.3 m horizontally and 1 m vertically. The test results are encouraging and promote the theory of model-based building extraction.