Panoramic image stitching and georeferencing of kite aerial photographs using XFeat

Abstract

Kite aerial photographs (KAPs) can serve as a viable option for the remote sensing of coastal areas where other unmanned aerial vehicles (UAVs) such as drones may have trouble maintaining their flight patterns due to the presence of strong wind. However, these KAPs lack any geographic information making it difficult to geolocalize these images by traditional methods. This paper demonstrates a workflow that processes these KAPs into panoramic images, and georeferences them onto publicly available satellite imagery. This was done by extracting matching features between a set of images using XFeat, stitching these images together by estimating their affine transformations, then warping them together. This same method was also applied to register the output panoramic images from the stitching process to a satellite image of the same area. Applying this method to the areas of Brgy. Bagong Silang, Paluan, and Calusa Island, Cagayancillo, we were able to georeference 80.29 ha and 109 ha of coastline imagery respectively.