Efficient Alignment of Aerial Images Based on Virtual Forces

Getting a contemporary aerial overview of a disaster area is a foremost task in search and rescue operations. We previously have introduced a novel method for registering a large amount of aerial images when camera parameters are almost unknown and no reference images are available. This paper provides two new methods, which improve our method for image registration based on virtual forces. The goal is to improve the performance of creating a contemporary overview map of a disaster area assembling several images taken by unmanned aerial vehicles (UAVs) equipped with cameras. In this paper, two new methods are introduced: a method for rotation estimation and a method for scale estimation. Both methods use fast heuristic approximation approaches and statistical methods to provide high robustness. We discuss the methods in detail and compare them to our previous approach regarding robustness and calculation speed. We can show that the new methods significantly increase the performance of the image registration process.