Ultrahigh-definition panoramic image registration technology for manned spacecraft AIT process

Abstract

In order to solve the problem that feature points on the same small target at different positions are easily recognized as the same feature point by a traditional SIFT registration algorithm, so that feature point matching errors are caused, and a panoramic image splicing result is influenced, the invention provides an ultra-high-definition panoramic image registration technology for a manned spacecraft AIT process, and in order to solve the problem, the ultra-high-definition panoramic image registration technology for the manned spacecraft AIT process is provided by utilizing a perceptual hash algorithm during feature registration. The method comprises the following steps: performing coarse purification on a matching point pair by using a single site, PHA, then judging whether the matching point is on a specific target, and removing the matching point on the specific target, so as to solve the problem of registration errors during feature registration, and specifically comprises the following steps: S1, constructing a training sample set for specific target detection by using original data collected by a single site; s2, training a detection model based on a Haar feature cascade classifier; s3, determining image matching point pairs based on an SIFT method; s4, performing coarse purification on the matching point pairs by utilizing a perceptual hash algorithm; s5, removing matching point pairs on the specific target based on the detection model; and S6, constructing an image transformation model.

Description

technical field [0001] The invention relates to an ultra-high-definition panoramic image registration technology in the AIT process of a manned spacecraft, and belongs to the technical field of digital image processing. Background technique [0002] Image registration technology is the core content of panoramic image stitching. Panoramic image registration includes matching methods based on image grayscale, matching methods based on transform domain and matching methods based on image features. The feature-based registration method first selects the feature information from the image, and then uses the image feature as the standard to search and match the feature set in the overlapping area of ​​the image. The image registration algorithm based on scale-invariant feature transform (SIFT) not only maintains invariance to rotation, scale scaling, and brightness changes, but also maintains stability to viewing angle changes, affine transformations, and noise. Widely used in p...

AI Technical Summary

Problems solved by technology

Based on the traditional SIFT registration algorithm, it is easy to identify the feature points on the same small target in different positions as the same feature point, resulting in feature point matching errors and affecting the results of panoramic image stitching.

Images