Research On Key Technologies Of 3D Building Reconstruction Combining Morphology With Feature Point Detection

The rising technology of 3D reconstruction has great practical values, but manual operations are needed to add vertices and edges during present methods of 3D building reconstruction. Thus, this paper focuses on the technology of 3D building reconstruction combining morphology feature point detection, which makes it possible to reconstruct buildings automatically. The main contents of this paper are as follows:(1) An adaptive median filter using local texture information in images is proposed using double windows based on four differernt texture types and their standard deviations to restore the noise pixels and get rid of impulse noises, which provides line detection and corner detection with input images of high quality;(2) An improved vanishing point estimation algorithm based on the principles of perspective is proposed to modify false vanishing points according to the morphology features of buildings and the inherent defect of MS AC so that all the detected lines are classified correctly.(3) A method based on the morphology features of the buildings, combining the information of lines, corners and vanishing points, to detect the vertices and edges of the buildings is proposed to detect the vertices and edges of buildings in shape of cubes automatically, and the SIFT algorithm is applied to estimate the fundamental matrix between different image pairs to find out the correspondence of points.(4) An image sequence based 3D reconstruction system for buildings is delevoped on the algorithms above to meet the specific requirements of photogrammetry and experiments are conducted to veridate the algorithms above.