A hyperspectral remote sensing image segmentation method based on spectral distance of spectral curve

Abstract

The invention discloses a hyperspectral remote sensing image segmentation method based on a spectral curve spectral distance. Firstly, a neighborhood spectral curve spectral distance model of a targetpixel is constructed. The Hilbert transform is used to obtain the odd-even filtering results by convolution operation, and the square root of the sum of squares of the odd-even filtering results is used as the local energy to construct the edge feature enhancement model. The edge feature enhancement model based on neighborhood spectral features is obtained by combining the two methods, which is used to obtain the edge feature enhancement results. The result of edge feature enhancement is input into watershed segmentation algorithm as gradient data, And the watershed segmentation algorithm isoptimized to achieve high-precision segmentation of remote sensing images, which can effectively suppress the over-segmentation phenomenon of hyperspectral remote sensing images, and solve the technical problems that weak edges and pseudo-edges in hyperspectral remote sensing images affect the segmentation results of remote sensing images.

Description

technical field [0001] The invention belongs to the technical field of image segmentation, and in particular relates to a hyperspectral remote sensing image segmentation method based on spectral curve spectral distance. Background technique [0002] Hyperspectral data provides hundreds of narrow spectral bands, which can form a complete and continuous spectral response curve to record the spectral information of the target object. Compared with multispectral data, hyperspectral remote sensing images provide richer spectral information of ground objects and more obvious spectral features, so it is possible to finely classify and directly identify ground object coverage types from spectral space. Image segmentation is the key technology of remote sensing information acquisition and ground object recognition, which provides a new idea for information extraction of hyperspectral images, and its core is to realize the segmentation of hyperspectral remote sensing images. [0003]...

AI Technical Summary

Problems solved by technology

The types of ground objects in remote sensing images are complex, and there are many noises and false edges in the edge gradient map. When the watershed algorithm is combined with the traditional edge enhancement algorithm for image segmentation, the aforementioned reasons will affect the remote sensing image segmentation effect.

Images