Phase Correlation Subpixel Matching Method Based on Maximum Kernel Density Estimation

Abstract

The invention relates to a phase related-sub-pixel matching method based on maximum-kernel-density estimation. The method is specifically as follows: 1. calculating an inter-image cross-power spectrum Q(u,v) is calculated; 2. separating a real part and an imaginary part of the Q(u.v) and applying phase edge filtering denoising to solve a phase angle matrix Psi(u,v); 3. unwrapping the phase angle matrix so as to obtain a corresponding plane matrix Psi unwrap(u,v); 4. using an MKDE method to perform fitting on the plane matrix so as to obtain plane equation parameters x0 and y0 and according to a proportion relation of the phase angle and an airspace pixel, obtaining a sub-pixel matching result. Compared with the prior art, the matching result is excellent in stability so that effects of gross errors on the plane fitting of the angle phase in the phase angle matrix can be prevented effectively so that the matching precision is improved and at the same time a pixel locking effect of the sub-pixel matching result is reduced significantly.

Description

technical field [0001] The invention relates to the field of remote sensing images, in particular to a phase correlation sub-pixel matching method based on maximum kernel density estimation. Background technique [0002] Sub-pixel precise matching of remote sensing images is one of the key technologies in the field of photogrammetry and remote sensing. Through sub-pixel matching between images, data such as stereo disparity, object displacement, and surface deformation field can be accurately obtained, which has extremely important applications in the fields of DEM generation, image fusion, and surface deformation monitoring. [0003] The sub-pixel matching method based on phase correlation has the characteristics of good anti-noise performance, little influence by image gray level, and high matching accuracy, and has received extensive attention in recent years. Among them, Hoge et al. used a method based on singular value decomposition (SVD) to decompose the phase matrix ...

AI Technical Summary

Problems solved by technology

[0004] However, there are still three problems in the existing phase correlation sub-pixel matching method: ① The phase matrix obtained by phase correlation usually contains large gross errors, and the traditional least squares fitting method is difficult to accurately solve the phase angle; ② image The oversampling of the phase matrix, the singular value decomposition of the phase matrix and other methods all have the disadvantages of high computational complexity and poor resistance to gross errors; ③ sub-pixel matching results will produce pixel locking effect (pixel locking effect), which directly affects the matching accuracy

Images