Method for registration on remote-sensing image based on improved SIFT feature

Abstract

The invention provides a research on registration on a remote-sensing image based on an improved SIFT feature. According to the characteristics of large data volume and high overlapping rate of the remote-sensing image and defects of matching errors and the matching efficiency caused by complex factors including illumination and geometric difference etc., an improved method of SIFT feature descriptors is deeply researched: constructing the feature descriptors by employing a feature point circular area; performing division of a local area and calculation of a gradient histogram by employing an adaptive quantization strategy, re-determining the principal direction of each descriptor by employing an interpolation method, and improving the feature descriptors of an SIFT algorithm; and employing a novel SIFT feature matching algorithm based on block matching. According to the method, the time loss for feature extraction and matching is reduced through removal of non-overlapped areas, and the uniqueness and the robustness of the feature point descriptors can be guaranteed.

Description

technical field [0001] The invention relates to the field of remote sensing image processing, in particular to a registration method on remote sensing images based on improved SIFT features. Background technique [0002] At present, remote sensing image processing has been widely used in many fields such as environmental monitoring, terrain survey, military investigation, disaster warning, etc., among which image registration is a key step, and the accuracy and real-time performance of registration have great influence on these applications. influences. Therefore, many registration methods have been proposed to solve this problem, which are divided into two categories: region-based methods and feature-based methods. Since region-based methods are quite sensitive to geometric and nonlinear gray distortions in remote sensing images, and remote sensing images usually require a large amount of data processing, feature-based image registration methods are more suitable than regi...

AI Technical Summary

Problems solved by technology

[0003] The problem to be solved by the present invention lies in the large amount of image data acquired by the remote sensing platform and the characteristics of high overlap. It often brings problems such as low precision and low efficiency. The present invention proposes an improved SIFT descriptor construction method, and on this basis, applies a new SIFT matching method based on the idea of ​​​​region segmentation, which reduces the dimensionality of feature points. At the same time, it fully considers that the pixels with different distances from the feature points in the neighborhood have different influences on the feature point descriptors.

Images