Hyperspectral anomaly detection method based on spectrum-preserving sparse auto-encoder

Abstract

The invention discloses a hyperspectral image anomaly detection method based on a spectrum-preserving sparse auto-encoder. The hyperspectral image anomaly detection method comprises the following steps: solving a wavelet coefficient of hyperspectral data; constructing a wavelet self-encoding network of linear unmixing constraint, replacing an inner product of an encoding layer with a spectral angular distance, selecting a Relu function as an activation function of the encoding layer, introducing a normalization layer and a dropout layer, adding a penalty term and a regularization term into a loss function, and constructing a network based on the spectral-preserving sparse self-encoder; and inputting the hyperspectral data to be measured into the sparse auto-encoder network, setting networkparameters, performing unmixing operation on the network parameters to obtain background end metadata, and calculating a reconstruction error to obtain required abnormal target data. Rapid and accurate abnormal target detection can be carried out on the hyperspectral image.

Description

technical field [0001] The invention belongs to the technical field of remote sensing image processing, in particular to a hyperspectral anomaly detection method based on a spectrum-preserving sparse autoencoder. Background technique [0002] Hyperspectral image target detection is an important content of hyperspectral data processing, and it plays an important role in various fields of human society. According to whether the target signal is known or not, hyperspectral target detection can be divided into supervised methods and unsupervised methods. The unsupervised method, that is, the hyperspectral image anomaly detection method, can detect abnormalities different from background pixels without prior information. The goal has been extensively studied by domestic and foreign personnel. Hyperspectral images have the following characteristics: [0003] (1) There are many spectral bands, and the bands are basically continuous. [0004] (2) The image contains not only rich ...

AI Technical Summary

Problems solved by technology

[0005] (3) The amount of hyperspectral data is large and the data redundancy is large

[0006] (4) Low spatial resolution and signal-to-noise ratio

[0010] However, the above method still has the following problems: (1) the number of calculation iterations is large; (2) the detection calculation is large; (3) the operation and convergence speed are slow

Images