Robust automatic panoramic unmanned aerial vehicle image splicing method and device

Abstract

The invention discloses a robust automatic panoramic unmanned aerial vehicle image splicing method and device. In a traditional automatic panoramic image splicing method, assuming that a camera precisely rotates around an optical center, transformation between source images is a group of special homography matrixes, such ideal data is difficult to obtain actually, especially, remote sensing imagesobtained by an unmanned aerial vehicle cannot meet such ideal conditions, that is, the images may not be on the same plane (including parallax), and even may suffer from non-rigid changes, resultingin bad splicing results. In order to overcome the challenges, a non-rigid matching algorithm is introduced into the splicing system, accurate feature matching is generated on a remote sensing image, and in the registration stage, the invention further provides a new differential transformation and overall adjustment strategy, so that the splicing system is suitable for the panoramic splicing effect of the unmanned aerial vehicle image. Experimental results show that the visual effect of the method is superior to that of a traditional method and some latest methods.

Description

technical field [0001] The present invention relates to the field of image processing, in particular to the field of image stitching, and more specifically, to a robust automatic panoramic UAV image stitching method and device. Background technique [0002] Unmanned Aerial Vehicle (UAV) remote sensing technology is a low-altitude technology that has become an important means of information collection. It has been widely used in various ground survey applications due to its fast acquisition speed, convenient operation, good safety, and low investment. However, due to the low flight altitude and the limited focal length of the camera, UAV images have a small view of the scene, making it difficult to capture a relatively complete target area. In order to obtain a complete scene of the desired target, the present invention needs to combine multiple images of the same target through technical means. The purpose of image stitching is to combine overlapping parts of multiple imag...

AI Technical Summary

Problems solved by technology

However, for drone images, non-rigid matching is crucial, since these images usually contain some local deformations, which cannot be resolved by rigid matching

On the other hand, remote sensing images captured by UAVs are taken from low altitudes, and the obtained images cannot be approximated as a plane, and there is a certain amount of parallax

In conventional stitching systems, a single-plane perspective transformation (homography matrix) is used for optimization together with global adjustments, this approach leads to ghosting errors

Images