Aerial image mosaicking with online calibration - A feasibility study

Kurzfassung

Compiling individual aerial images to a larger mosaic image is important for many remote sensing tasks, e.g. mapping. The standard way to address this problem is to orthorectify the image and later compile it together. For the orthorectification the orientation and the location of the camera has to be measured accurately and a Digital Elevation Model (DEM) is needed or ground control points have to be set manually. In this paper we present a feasibility study on an approach which works purely on original camera images without any GPS and/or IMU. The intern and extern camera parameters and the 3D feature points are calculated based on Structure from Motion (SfM). The ground surface is assumed to be flat and a plane is fitted on the 3D feature points. A virtual camera is calculated then which is perpendicular to the plane and the images are transformed into each other by a homography. We test this approach on image sequences captured by a standard DSLR camera.