A Study of Three-Dimensional Reconstruction Using High Overlap Aerial Photograph Sequence

The unmanned aerial vehicle (UAV), as a low-cost airborne instrument platform, enjoys a lot of advantages such as flexible, high resolution and so on. Currently, UAV is widely used in resource management and environmental surveillance as it can obtain a sequence of high overlap images by continuously shooting on specified areas, of which the overlap is up to 60% along the heading, and is up to 30% in the side. Stereo pair refers to the same area in two different images which are taken from different angles. Actually, the overlap area obtained from aerial photograph sequence contains a lot of stereo pairs, which can be used to the 3D reconstruction. In this paper, three-dimensional reconstruction using intersection of the overlap area obtained by UAV image sequence is systematically studied in theory. First, pixel-level matching algorithm of overlap area is carried. Second, coordinates of pixels in overlap area is calculated with condition equations. Third, applying the separation over point cloud constituted by matched pixels; in the end, the three dimensional algorithm of terrain and its features is studied.