Spatial feature extraction method for enhancing two-dimensional singular spectrum analysis

Abstract

The invention discloses a spatial feature extraction method for enhancing two-dimensional singular spectrum analysis. The spatial feature extraction effect is enhanced by utilizing local similarity of pixels. On the basis of traditional 2DSSA, an improved adaptive window is provided to replace a fixed window, an embedding process is realized by combining neighborhood similarity of target pixels, and a low-rank trajectory matrix is obtained; singular value decomposition is carried out on the low-rank trajectory matrix, the obtained maximum singular value is larger than the maximum singular value obtained by a fixed window, the proportion of the maximum singular value in all singular values is higher, and the maximum singular value corresponds to more spatial information; and the space subcomponent corresponding to the maximum singular value is used as a feature for reconstruction, a final feature component is obtained, and influence components such as noise are removed while space structure information is enhanced by the component. According to the method, the problems of spatial scale diversity, noise and the like in the hyperspectral image can be solved, the defects of a traditional 2DSSA method are overcome, the feature extraction effect of the image is enhanced, and the ground feature classification precision of the image is improved.

Description

technical field [0001] The invention belongs to the technical field of remote sensing image processing, and in particular relates to a spatial feature extraction method for enhancing two-dimensional singular spectrum analysis. Background technique [0002] With the advancement of sensor and satellite technology, hyperspectral remote sensing has become one of the important earth observation technologies that have developed rapidly in the past 30 years. An important means of promoting national economic development and construction. Although hyperspectral images contain rich spectral information, there is a phenomenon of "same spectrum, same spectrum foreign matter", which will cause serious misclassification of "salt and pepper noise" in classified images. In addition, the wavelength band of hyperspectral images is narrow, and each channel receives less light energy, resulting in a low signal-to-noise ratio and severe noise in some bands, which brings challenges to image inte...

AI Technical Summary

Problems solved by technology

However, this method also has some disadvantages: for example, a series of operations will bring high computational complexity, and the shape connectivity of different ground objects can easily cause confusion

[0005] However, similar to other widely used space technologies, 2DSSA also has some disadvantages

First of all, the above algorithms usually use regular-shaped or fixed-size windows for processing, ignoring the complexity of the shape and scale of image features, that is, the area used for spatial feature extraction does not match the spatial structure of the image

In addition, 2DSSA usually selects the spatial trend component corresponding to the largest singular value as the reconstructed feature image, but other singular values ​​also correspond to some useful information, which results in the loss of part of the spatial information, mainly manifested in the reduction of the edge difference of the ground object and the texture feature. Rarely, the image will be too smooth and the classification accuracy will decrease, especially for small objects

Images