Vision based navigation using road-intersection image

This paper suggests the estimation method of the UAV's navigation information using an intersection image. When the UAV takes pictures of terrain vertically in the low altitude, the difference between two branch's azimuths is constant regardless of the position and azimuth. According to this idea, the database consists of the latitude, longitude and branch's azimuth of intersections. And the captured image data obtained from an image processing are built up similarly. The difference of branch's azimuth is calculated by the data of database and an image. It is possible to identify the relation between intersections and corresponding points on the image plane by comparing to the pattern. It can be easily shown that the scale factor is the distance ratio between identified intersections and corresponding points, and the UAV's altitude is calculated by the scale factor and camera parameters. Finally, it is possible to estimate the latitude and longitude of the UAV as using the geometric relation between an identified intersection, point, and the center point of an image plane.