GF-4 image registration method based on multi-feature mixing

Abstract

The invention puts a GF-4 image registration method, which has better robustness and adaptability, based on multi-feature mixing by aiming at automatic registration among GF-4 multispectral images. The method comprises the following steps that: 1) carrying out necessary preprocessing and image enhancement processing on a base image and a registration image; 2) according to topographic data elevation features, carrying out regional partitioning processing on an image overlapping area; 3) extracting and matching two features of two images to form an initial mixed point pair set; 4) setting a distance threshold value between feature points to optimize the initial mixed point pair set; 5) according to registration accuracy requirements, determining a final matching point pair set; 6) halving afinal matching point into a control point and a check point, establishing an affine transformation model for two-image registration according to the control point, and determining an affine transformation parameter; and 7) carrying out image resampling on the registration image, and evaluating the registration accuracy.

Description

technical field [0001] The invention relates to an automatic registration method of Gaofen-4 satellite images based on multi-feature mixing, which is used to realize automatic, fast and high-precision registration of images under an area array time-sharing imaging system, and belongs to the technical field of remote sensing image processing. Background technique [0002] Gaofen-4 (GF-4) satellite is my country's first high-resolution earth observation satellite in geostationary orbit. It was successfully launched in Xichang on December 29, 2015. Its successful launch and application will significantly enhance my country's space-based The ability to observe the earth is a strategic deployment to break through the key common technology of "optical remote sensing with high spatial resolution and high temporal resolution at the same time", and it has pioneering significance for the development of high-orbit high-resolution remote sensing satellite technology. With its unique load...

AI Technical Summary

Problems solved by technology

At the end of the 20th century, Harris and Stephens proposed the Harris operator, which is invariant to image rotation and grayscale changes, but does not have scale

In 2012, Tang Yonghe et al. proposed a fast image matching algorithm based on Harris operator to solve the problem of real-time and robustness in local feature matching.

However, due to factors such as low spatial resolution of GF-4 images, blurred local images, and great influence from the external environment, the existing algorithms still do not have ideal results in the process of feature extraction and matching of GF-4 images.

[0005] Harris algorithm is a corner detection algorithm, its feature points can be evenly distributed in the whole image, the registration effect is close to the global matching, and the gray level change of the image has little effect on it, but it does not have the characteristics of scale invariance. There are limitations in the application of remote sensing image automatic registration

The SIFT algorithm can maintain local invariance for image rotation, scale scaling, and even partial 3D viewing angle changes and illumination changes. It is a stable and adaptable local feature matching algorithm, but it does not have global characteristics. For larger images, there are obvious defects

Images