Hyperspectral anomaly detection method for component projection optimization separation

Abstract

The invention provides a hyperspectral anomaly detection method for component projection optimization separation. The hyperspectral anomaly detection method comprises the following steps: performing superpixel segmentation on a hyperspectral image by using an entropy rate-based superpixel segmentation algorithm; calculating an average value of pixel points contained in each superpixel to serve asa spectral vector of the superpixel; calculating a Mahalanobis distance from a pixel point contained in each super-pixel to the spectral vector, and summing discrete values representing the super-pixel; solving local abnormal factors one by one by taking the superpixel as a unit; calculating the product of the discrete value of each super-pixel and the reciprocal of the local abnormal factor of each super-pixel, and selecting a part of super-pixels with a smaller product as a background set to construct an estimated background set; setting a component projection and separation optimization filtering function, and solving an optimal filtering vector; multiplying the optimal solution by the hyperspectral image pixel by pixel to obtain a detection result. Image information is interpreted froma super-pixel level, an estimated background set is obtained by using local abnormal factors, and a hyperspectral image abnormal target detection result is obtained by combining optimized filtering.

Description

technical field [0001] The invention belongs to the technical field of computer image processing, and relates to a method for detecting an abnormality of an image target, in particular to a method for detecting an abnormality in a hyperspectrum by optimizing and separating component projections. Background technique [0002] Hyperspectral images contain multiple bands. Compared with grayscale images and RGB images, hyperspectral images contain more spectral information. Abnormal features with obvious distinctions are detected. Although with the maturity of hyperspectral imaging technology and the reduction of cost, hyperspectral images are used more and more, but there are still some limitations in anomaly detection technology for hyperspectral images. [0003] 1) Although the hyperspectral image has a very fine spectral resolution, its spatial resolution is generally low, so the target objects with detection often exist in the image in the form of extremely small pixels or...

AI Technical Summary

Problems solved by technology

Although hyperspectral images are being used more and more with the maturity of hyperspectral imaging technology and the reduction of cost, there are still some limitations in anomaly detection technology for hyperspectral images.

[0003] 1) Although the hyperspectral image has a very fine spectral resolution, its spatial resolution is generally low, so the target objects with detection often exist in the image in the form of very small pixels or even sub-pixels

[0004] 2) Traditional model-based detection methods still have the problem of insufficient background suppression

[0005] 3) Due to the absence of any prior information and the phenomenon of "same object with different spectrum, different object with same spectrum" phenomenon that may exist in hyperspectral images, some algorithms cannot correctly detect the target of interest when processing hyperspectral images, that is, the prominent effect on target information is not enough obvious

The method in "BASO: A Background-Anomaly Component Projection and Separation Optimized Filter for Anomaly Detection in Hyperspectral Images" still has misclassification of potential background and anomaly categories in practical applications, and further improvement is urgently needed

Images