Hyperspectral image intelligent unmixing method based on unsupervised training

A hyperspectral image, unsupervised technology, applied in the direction of neural learning methods, instruments, biological neural network models, etc., can solve problems such as the reduction of spatial accuracy and affect the application of hyperspectral data, and achieve the effect of excellent unmixing accuracy

Pending Publication Date: 2021-11-19
BEIJING RES INST OF URANIUM GEOLOGY
View PDF0 Cites 1 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

However, due to the influence of optical devices, the spatial accuracy of hyperspectral data is lower than that of more spectral data. The pixels corresponding to ground objects in hyperspectral images are mostly spectral mixtures of several different substances, which affects the application of hyperspectral data.

Method used

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
View more

Image

Smart Image Click on the blue labels to locate them in the text.
Viewing Examples
Smart Image
  • Hyperspectral image intelligent unmixing method based on unsupervised training
  • Hyperspectral image intelligent unmixing method based on unsupervised training
  • Hyperspectral image intelligent unmixing method based on unsupervised training

Examples

Experimental program
Comparison scheme
Effect test

Embodiment Construction

[0041] The present invention will be further described below by means of the accompanying drawings and specific embodiments.

[0042] Step 1: Build a deep neural network stacked by three autoencoders.

[0043] (1) Build the first autoencoder

[0044] The autoencoder includes two parts: encoding and decoding. The input of the first autoencoder is the hyperspectral matrix X, and the sigmoid activation function is used to construct the encoder according to the encoding formula (1) and decoding formula (2). l=1 means the first self-encoder, a l is the encoded output value of the first autoencoder, is the decoded output value of the first autoencoder, and are the weights of the encoding layer and decoding layer of the autoencoder respectively, set (*) T means transpose, Random numbers are used for initial assignment. The basic formula of autoencoder network is as follows:

[0045] alpha 1 = sigmoid(w e1 *X) (1)

[0046]

[0047] (2) Batch normalization before th...

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to view more

PUM

No PUM Login to view more

Abstract

The invention belongs to the field of hyperspectral image application, and particularly relates to a hyperspectral image intelligent unmixing method based on unsupervised training. Comprising establishing a deep neural network stacked by three auto-encoders, establishing a network full-connection output layer, establishing a loss function of the whole network, starting network training, performing back propagation layer by layer to update network parameters, and obtaining an end member spectrum and an abundance matrix. According to the method, the end member and component information of the mixed pixel of the hyperspectral image can be successfully identified, and the end member and the component of the image are identified through unsupervised training without a large amount of label data.

Description

technical field [0001] The invention belongs to the application field of hyperspectral images, and in particular relates to an intelligent unmixing method for hyperspectral images. Background technique [0002] With the development of remote sensing technology, hyperspectral images are more and more widely used, and the gradual improvement of spectral resolution can obtain multi-dimensional information characteristics of ground objects and increase the cognition of ground objects. However, due to the influence of optical devices, the spatial accuracy of hyperspectral data is lower than that of more spectral data. The pixels corresponding to ground objects in hyperspectral images are mostly spectral mixtures of several different substances, which affects the application of hyperspectral data. The decomposition of mixed pixels of hyperspectral images, that is, spectral unmixing, includes two main parts: endmember extraction and abundance calculation. Identifying the ground obj...

Claims

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to view more

Application Information

Patent Timeline
no application Login to view more
Patent Type & Authority Applications(China)
IPC IPC(8): G06K9/00G06N3/04G06N3/08
Inventor 朱玲秦凯崔鑫李明
Owner BEIJING RES INST OF URANIUM GEOLOGY
Who we serve
  • R&D Engineer
  • R&D Manager
  • IP Professional
Why Eureka
  • Industry Leading Data Capabilities
  • Powerful AI technology
  • Patent DNA Extraction
Social media
Try Eureka
PatSnap group products