Hyperspectral anomaly detection method based on attention self-encoding network

Abstract

The invention discloses a hyperspectral anomaly detection method based on an attention self-encoding network. The hyperspectral anomaly detection method comprises the following steps: 1, making a training data set; 2, distinguishing contributions of different wave bands to anomaly detection by adopting a channel attention mechanism; 3, increasing the separability of abnormal pixels and backgroundpixels by adopting a spatial attention mechanism; 4, extracting spatial-spectral characteristics by adopting a coding network; 5, weakening the network expression capability by adopting a memory module; 6, reconstructing input data through a decoder; and 7, determining the abnormal degree of each pixel on the reconstructed image by using an RX algorithm. According to the method, the contribution of the useful wave band to the final anomaly detection result is effectively increased, and the interference of the wave band with the low signal-to-noise ratio to anomaly detection is reduced, so thatthe separability of an anomaly region and a background region is improved, the anomaly detection precision is effectively promoted, and the false alarm rate is effectively reduced while the detectionprecision is improved.

Description

technical field [0001] The invention belongs to the technical field of hyperspectral image processing, and relates to a hyperspectral anomaly detection method based on an attention self-encoding network. Background technique [0002] A hyperspectral remote sensing image is a three-dimensional data cube, in which two dimensions express the spatial relationship, and the other dimension expresses the reflection or radiation intensity of ground objects in different bands. Because of its rich spectral information and the spatial position relationship of ground objects, it is widely It is used in fields such as battlefield reconnaissance, food safety, and environmental monitoring. Hyperspectral anomaly detection has attracted the attention of a large number of scholars in recent years because it does not require prior information and meets actual needs. [0003] Existing hyperspectral anomaly detection algorithms are roughly divided into the following four categories: statistical ...

AI Technical Summary

Problems solved by technology

[0005] The purpose of the present invention is to provide a hyperspectral anomaly detection method based on the attention self-encoding network, which solves the hyperspectral anomaly detection algorithm of the prior art, which does not make reasonable use of the spatial spectrum information of the hyperspectral image, and the depth network is for abnormal pixels The reconstruction ability is too strong, which leads to the problem of insufficient accuracy of anomaly detection

Images