Higher dimensional space directional projection end member extraction method

Abstract

The invention relates to a higher dimensional space directional projection end member extraction method. The method comprises the following steps of: (1) inputting an image and number of end members; (2) initializing a basis vector corresponding to the x axis of a projection plane; (3) calculating a basis vector corresponding to the y axis of the projection plane; (4) calculating coordinates of each pixel in the projection plane and displaying projection points in a visual window according to the coordinates; (5) ticking off a new end member in the visual window; (6) calculating a basis vector corresponding to the x axis of the projection plane according to the new end member; (7) judging whether the number of the end members meets a requirement; and (8) outputting all the end members, namely outputting all formulas as the end members. The method has the beneficial effects that: projected data are quantitatively analyzed; a data point which is most possibly next end member is marked to guide operating personnel to make a choice; the operating personnel is also allowed to freely select according to actual conditions; and difficulty in selection is reduced on the basis of not changing a vertex selection principle, so that the end members are extracted more intuitively, quantitatively and automatically.

Description

technical field [0001] The invention relates to the field of high-dimensional space directional projection image analysis, in particular to a high-dimensional space directional projection endmember extraction method. Background technique [0002] The endmember extraction of hyperspectral images is based on the geometric interpretation of the linear spectral mixture model. It is believed that the mixed pixels are distributed in the high-dimensional space simplex with the endmember as the vertex, and the projection of the simplex in the low-dimensional space must be a convex set. And the vertices of the projection convex set must be the projection of the vertices of the simplex. The N-dimensional visualization technology projects all the pixel points of the hyperspectral image from the high-dimensional feature space onto a two-dimensional plane, and allows users to manually select those points that may be vertices in the projected points. By continuously randomly generating th...

AI Technical Summary

Problems solved by technology

However, the N-dimensional visualization technology has the following problems: 1. The projection plane in N-dimensional visualization is randomly generated, which leads to the lack of quantitative evaluation basis for vertex selection; 2. It mainly depends on the manual interpretation of the operator, which is relatively random and difficult to improve the accuracy; 3. The theoretical level of operators is relatively high, and it is difficult to popularize it to users who have practical application requirements but do not have a lot of professional knowledge.

Images