METHOD AND A SYSTEM FOR BUILDING A THREE-DIMENSIONAL MODEL FROM SATELLITE IMAGES

摘要

The invention relates to a method for building a 3D model of an area of interest on the surface of a planet. The method comprises providing a plurality of 2D images from satellites, where each 2D image out of the plurality of 2D images at least partly covers the area of interest. Also, each subarea of the area of interest is covered by at least two images, taken from different angles, out of the plurality of 2D images. Bundle adjustments are performed for the plurality of 2D images and the area of interest is divided into parts. For each part of the area of interest at least one combination and preferably all possible combinations out of two images covering that part of the area of interest are taken. Said two images are taken from the at least two images, taken from at least two different angles, out of the plurality of 2D images. Further, for each part of the area, point matching correspondence is established for a set of points for each such combination of two images. The sets of points are combined if more than one set of points exists. Even further, for each part of the area, a 3D model of said part of the area of interest is built based on the combined set of points. The 3D models from each part of the area of interest are combined to a 3D model of the whole area of interest. The invention also relates to a system, a computer program and a computer program product.

主权项

1. A method for building a 3D model of an area of interest on the surface of a planet, the method comprising:
providing a plurality of 2D images from satellites, where each 2D image out of the plurality of 2D images at least partly covers the area of interest, and where each subarea of the area of interest is covered by at least two images, taken from different angles, out of the plurality of 2D images; performing bundle adjustments for the plurality of 2D images; dividing the area of interest into parts and for each part of the area of interest:
taking at least one combination and preferably all possible combinations out of two images covering that part of the area of interest, wherein said two images are taken from the at least two images, taken from at least two different angles, out of the plurality of 2D images;establishing point matching correspondence for a set of points for each such combination of two images;combining the sets of points if more than one set of points exists; andbuilding a 3D model of said part of the area of interest based on the combined set of points; and combining the 3D models from each part of the area of interest to a 3D model of the whole area of interest.