A Fast Stitching Method for Aerial Images with Large Tilt and Low Overlap

Abstract

The present invention proposes a fast mosaic method for large-inclination and low-overlap-rate aerial images, which can improve registration robustness of large-inclination and low-overlap rate images, reduce stitching errors, and improve wide-width image synthesis efficiency. The present invention is realized through the following technical solutions: first, the position distribution relationship of the image is calculated by using the imaging auxiliary data of the aerial image, the image position relationship matrix is ​​established, and the reference map is calculated; then the adjacency graph path between the image and the reference map is searched by the shortest path , calculate the overlapping area between adjacent images on the adjacency path, and extract the salient feature points of the overlapping area. Secondly, the feature points or regions are normalized into vertices, and multi-scale graphs are constructed by using point sets with different adjacent distances, and the spatial transformation matrix is ​​calculated by using graph matching; finally, the corresponding coordinates of the synthesized image pixels in the source image are calculated using the spatial transformation matrix And use bilinear interpolation to calculate the pixel value of the composite image to get the final spliced ​​image.

Description

technical field [0002] The invention relates to an image mosaic method related to photogrammetry, computer vision, image processing and computer graphics in the field of image processing, in particular to a method for quickly mosaicing aerial images with large inclination and low overlap rate based on salient key point map matching. Background technique [0004] Image mosaic (image mosaic) technology is to combine several overlapping images, which may be images obtained at different times and from different perspectives, into a large seamless high-resolution image. The image sequence is matched and aligned in space, and after resampling and synthesis, a complete, high-definition new image of a wide-view scene containing the information of each image sequence is formed. Early image stitching research has been used in photogrammetry, mainly integrating a large number of aerial or satellite images. In order to solve the limitations of the shooting angle of the lens, with the d...

AI Technical Summary

Problems solved by technology

[0020] The purpose of the present invention is to solve the problem that the image stitching method at the present stage does not consider the serious distortion of the aerial image caused by the imaging conditions of long distance, large inclination angle, and low overlap rate, the local characteristics of the overlapping area are relatively different, and there are few feature points to be matched. problem, as well as the problem of long image stitching time caused by the large frame and large number of aerial images, a method is provided to introduce imaging auxiliary data to calculate the effective overlapping area, extract key points through the detection of salient areas in the overlapping area, and use the key points to construct Multi-scale graph and graph-based matching methods improve the success probability and stitching speed of image stitching under special imaging conditions, and realize the fast stitching method of aerial images with large tilt and low overlap rate

Images