Image matching method based on Zernike matrix

Abstract

The invention discloses an image matching method based on a Zernike matrix. The method comprises the following steps: S1, establishing N layers of template pyramid images; S2, calculating K matrixes of each layer of template images in the N layers of template pyramid images, and establishing an RCS table; S3, establishing N layers of object pyramid images; S4, calculating K matrixes of each object sub-image of the L-th layer of the N layers of the object pyramid images through searching the RCS table; S5, calculating correlation coefficients of K matrixes of the L-th layer of the template pyramid images and the sub-image of each object pyramid image in the L-th layer, and obtaining a matching point position (x<L>,y<L>) of the L-th layer of the object pyramid images; and S6, determining whether L is equal to 1, if so, entering S8, and if not, L=L-1, and entering S7; S7, calculating K matrixes of each object sub-image in the L-th layer within a scope of N<L>*N<L> by taking a point with the coordinates (2 x<L>,2y<L>) as a center in the L-th layer, and returning to S5; S8, calculating the correlation coefficients within a scope of N<Buttom>*N<Buttom> by taking a matching point with the coordinates (x<Buttom>,y<Buttom>) in the first layer as a center point, and obtaining a sub-pixel matching position with the coordinates (x,y); and S9, at the sub-pixel matching position with the coordinates (x,y), according to phase, calculating an image rotation angle alpha<^>.

Description

technical field [0001] The invention belongs to the fields of image processing and machine vision, and more specifically relates to an image matching method based on Zernike moments. Background technique [0002] Machine vision has the advantages of non-contact, non-damage, high efficiency and stability, and is widely used in electronic packaging equipment. Image matching is an important part of machine vision. Modern IC packaging equipment needs to accurately locate the position of the chip through the vision system, and its positioning accuracy can reach um or even nm level. And it often happens that the template image and the target image are rotated, and the classic image matching algorithm can only match the translation image. Therefore, it is an urgent technical problem to develop an image matching algorithm with high precision, good real-time performance, and the ability to match any rotation angle. [0003] Image matching methods can be divided into point matching...

AI Technical Summary

Problems solved by technology

But the cross-correlation algorithm cannot match the rotated image

The moment matching algorithm and the mutual information algorithm generally have a large amount of calculation.

Images