Aerial Image Stitching Based on Fusion of Geographic Coordinates and Image Features

Abstract

With the rapid development of unmanned aerial vehicle technology, unmanned aerial vehicles have been widely used to obtain ground remote sensing images. Due to the limited field of view of a single remote sensing image, it is necessary to stitch the images to obtain a large-scale scene image to apply to various practical applications. This paper introduces a two-stage stitching algorithm combining geographic coordinate information and image features. The stitching of sequence images is divided into intra-line stitching and inter-line stitching. Firstly, we use geographic coordinates and image features to calculate transformation parameters between a row of images, respectively, and select an appropriate transformation matrix to stitch a single row of images based on the rotation parameters in the homography matrix. Then, we also calculate the transformation matrices between different rows of images based on the geographic coordinate information and image features, and select the appropriate transformation matrix based on the comparison of the rotation parameters in different transformation matrices. Stitching experiments are carried out for various scenarios. Compared with traditional stitching methods, the proposed method integrates information from different sources, has higher reliability, and can adapt to stitching of various types of scenarios.