Hyperspectral target detection method and system based on feedback abundance constraint, storage medium and processor

Abstract

The invention discloses a hyperspectral target detection method and a system based on a feedback abundance constraint, a storage medium and a processor. The method is characterized by calculating a spectral characteristic d of a target ground object and an autocorrelation matrix R; calculating abundance information f(ri) corresponding to each pixel ri in a hyperspectral image; extracting a space characteristic ff(ri) of the target ground object, feeding the extracted ff(ri) back to and fusing to the spectral characteristic of the hyperspectral image and updating; carrying out linear normalization processing on the abundance information f(ri) and defining a spectral autocorrelation matrix as R'; carrying out target ground object detection so as to acquire target ground object abundance information fc(ri); and carrying out binarization processing so as to acquire the target ground object to be detected. In the invention, when an autocorrelation matrix is used to restrain the pixel, an influence of a foreign matter cospectrum phenomenon can be reduced, which is good for detection of the target ground object; simultaneously, space information is integrated into the spectral characteristic so as to increase detection precision and solve a problem in the prior art.

Description

technical field [0001] The present invention relates to a constraint-based hyperspectral target detection technology, in particular to a feedback abundance constraint-based hyperspectral target detection method and system, a storage medium and a processor. Background technique [0002] Hyperspectral images use dozens to hundreds of fine spectral bands to image target areas, and can simultaneously acquire rich spectral and spatial features of ground objects, and have broad application prospects in environmental monitoring, precision agriculture, military security and other fields. As the core task of hyperspectral technology, hyperspectral target detection is based on the similarity standard between the target and ground objects in terms of spectral characteristics. Currently, according to the known background and target, typical target detection algorithms include Orthogonal Subspace Projection (OSP) method, Constrained Energy Minimization (CEM) method, TCIMF algorithm, RXD ...

AI Technical Summary

Problems solved by technology

But its disadvantage is that on the one hand, it is sensitive to the known target features. When the phenomenon of "same-spectrum foreign objects" exists, the detection effect is relatively poor due to the existence of objects similar to the target in the image. On the other hand, when using all pixels When the background is suppressed, the signal of the target ground object is also weakened to a certain extent. This method has little effect on small targets, but for large target ground objects, the detection effect will be affected to a certain extent. Detection accuracy is not high

Images