Creation of Panoramic Aerial Photographs on the Basis of Multiband Blending

Abstract

A panoramic aerial photograph composed of a series of consecutive images obtained from an unmanned aerial vehicle helps identifying visual objects more accurately. Automatic combining of aerial photographs is a complex technical problem, especially in the presence of artefacts of the recording equipment and video-recording conditions. It is assumed that a sequence of combined images is known and determined by the flight trajectory. The proposed method for combining aerial photographs is based on illumination leveling and multiband blending. The possibilities of application of SURF, FAST, FAST-ER, and KAZE point descriptors for combining images are studied. It is shown that the point descriptors on the basis of nonlinear diffuse filtering yield the best results with regard to accuracy in the presence of global and/or local visual distortions of combined images.