Use of Hyperspectral Data with Intensity Images for Automatic Building Modeling*

Geospatial databases are needed for many tasks in civilian and military applications. Automated building detection and description systems attempt to construct 3-D models using primarily PAN (panchromatic) images. These systems can make use of cues derived from other sensor modalities to make the task easier and more robust. The recent development of hyperspectral sensors such as HYDICE (HYperspectral Digital Imagery Collection Experiment) can provide reasonably accurate thematic maps. Such data, however, tends to be of lower resolution, have geometric distortions and camera models are needed to map points between the different sensors. We use the thematic map to provide cues for presence of buildings in the PAN images for accurate delineation. It is shown that such cues can not only greatly improve the efficiency of the automatic building detection system but also improve the quality of the results. Quantitative evaluations are given.