SAR slant range image match automatic extraction method based on geometric model

Abstract

The invention provides a SAR slant range image match automatic extraction method based on a geometric model. According to the method, the problem that enough repeated characteristics cannot be extracted by a SIFT when large geometric distortion exists between slant range images can be overcome, and correct matching points can be effectively screened out based on the improved RANSAC algorithm. At first, the imaging geometrical relationship between two images is established through the carrier aircraft track and DEM so as to transform one image A onto coordinates of another image B to obtain an image C; then, matches between the image B and the image C are extracted by adopting scale invariant feature transform method (SIFT) on a sampled image; and finally, correct matches are screened out by adopting random sample consensus method (RANSAC), and match coordinates on the image C are transformed onto the image A to obtain matches of the image A and the image B.

Description

technical field [0001] The invention belongs to the field of remote sensing image processing, and relates to a method for automatically extracting homonymous points of SAR oblique distance images based on a geometric model. Background technique [0002] Under the condition of sparse control points, when the block adjustment method is used to stitch the SAR images of multiple airways, it is necessary to extract the same name points from the oblique range images with large viewing angle differences. In addition, homonym point extraction also has important applications in the fields of image registration, stereo image pairing, and target detection and recognition. The same-name point extraction can be divided into three stages: (1) feature detection and description: detect the point, line or surface features in the image, and calculate the structure or shape information of the feature neighborhood to form a descriptor vector; (2) feature matching: Calculate the distance betwee...

AI Technical Summary

Problems solved by technology

[0003] For the slant-range images of adjacent flight strips, different viewing angles and terrains lead to different geometric distortions at different positions of the image. Therefore, when SIFT is applied to the extraction of the same-named points of SAR slant-range images, there are two problems: one is the detected feature points It is not easy to match, and the second is that the correct matching point is not easy to filter

There are still local affine deformations between the transformed images, and SIFT can be used to extract the same-named points, but the extracted same-named points usually contain errors, and the traditional RANSAC algorithm can only filter out some correct ones, missing a large number of correct same-named points point, so when applying this method to SAR images, it is urgent to make corresponding improvements in combination with the characteristics of SAR slant range images

Images