Edge large distortion image matching method

Abstract

The present invention provides an edge large distortion image matching method. The method comprises the steps of: establishing each-level pyramid image for an image to be matched and a reference image, and performing rough matching; employing an integrated radiation transformation model to perform model construction according to a rough matching result so as to eliminate parallax; employing a multi-level and multi-feature mode, performing segmentation and blocking processing of the image to be matched and the reference image according to a multi-level and multi-feature matching rule grid, andperforming multi-level and multi-feature matching for each block obtained after segmentation and blocking processing; performing segmentation and blocking processing of the image to be matched according to a gross error elimination rule grid, and employing the radiation transformation model to perform loop iteration for each grid in the gross error elimination rule grid for gross error elimination; and employing one control point of each grid after gross error elimination, employing an integrated cubic polynomial iteration model for gross error elimination again according to a camera objectivelens radial directional and tangential distortion for theory to obtain high-precision matching points and perform output according to a predetermined format.

Description

technical field [0001] The invention belongs to the field of photogrammetry and computer vision, and relates to a high-precision registration scheme for large edge distortion images based on image segmentation and multi-level multi-features, in particular to a matching method for large edge distortion images, which is used for image segmentation and block segmentation And multi-level multi-feature mode, high-precision automatic matching of control points is performed on large edge distortion images. Background technique [0002] At present, domestic high-resolution optical satellites all use three-mirror coaxial or triple-mirror off-axis optical systems. Due to the influence of radial and tangential distortion of the optical system, the geometric distortion of the edge of the image before the camera’s in-orbit geometric calibration is very large, and the middle The distortion is small. For example, the distortion of the panchromatic / multispectral camera image of the Gaofen-1...

AI Technical Summary

Problems solved by technology

[0004] 1) If large edge distortions are taken into account, too many mismatch points will be caused due to the large search range, and it is impossible to construct a strict geometric constraint model for gross error elimination;

[0005] 2) If the small distortion in the middle is taken into account, due to the small search range, it will cause the problem that points cannot be matched except the central area

Images