This paper introduces a fast automatic building extraction method by the fusion of spectral information and three dimensional information acquired by "Airborne Laser Image Mapping System"

which integrates the image scanner

laser scanner

inertial navigating system and global positioning system. This method involves the breaking and merging algorithm of laser ranging points

the separation of buildings and trees

the selection of initial seeds

the self adjustment of the spectral reference and threshold in region growing

and the reconstruction of whole building contour with edges of high reliability. To overcome the distortions of airborne scanning image

we extensively used two dimension/three dimension transformations and the concept of curve template. To deal with the blurring and brokeness of edges and strong noise

we applied a statistical decision algorithm and a sampling template. This method is experimented on the data of City Huhehaote.