Looking for breakthrough ideas for innovation challenges? Try Patsnap Eureka!

A wrapper hyperspectral band selection method based on pixel clustering

A band selection and hyperspectral technology, which is applied in the field of image processing, can solve the problems of limited scope of use of supervised band selection methods, achieve the effects of optimizing clustering effects, improving band selection efficiency, and improving reliability

Active Publication Date: 2018-05-25
XIDIAN UNIV
View PDF3 Cites 0 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

These methods cannot be used when there are no labeled samples, making the scope of use of supervised band selection methods greatly limited

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
  • A wrapper hyperspectral band selection method based on pixel clustering
  • A wrapper hyperspectral band selection method based on pixel clustering
  • A wrapper hyperspectral band selection method based on pixel clustering

Examples

Experimental program
Comparison scheme
Effect test

Embodiment 1

[0026] The specific steps of the wrapper hyperspectral band selection method based on pixel clustering are as follows:

[0027] (1) Input the hyperspectral image to be selected by the band, assuming that the number of original bands of the image is P, convert the image into hyperspectral data in matrix form;

[0028] The hyperspectral image selected in this example is the Indiana hyperspectral image acquired by the airborne visible and infrared imaging spectrometer AVIRIS. The hyperspectral image contains 16 different types of ground objects, including soybeans, wheat, corn, Grassland, artificial buildings, roads, etc., these features are distributed in different areas of the map. The hyperspectral image contains 220 original bands, ie P=220, and the wavelength range is 0.4 μm-2.5 μm. The number of pixels in the hyperspectral image is 145*145, and the size of the hyperspectral image data converted into matrix form is 145*145*220. In order to facilitate subsequent data proces...

Embodiment 2

[0052] The wrapper formula hyperspectral band selection method based on pixel clustering is the same as in embodiment 1, wherein in step (3), a representative point is selected from the segmented superpixel block, and the specific steps include:

[0053] 3.1: In this example, 462 superpixel blocks are generated in step (2), assuming that there are j pixels in the i-th superpixel block, the spectral information of each point in the i-th superpixel block is recorded as where each element , k=1,2,...,j; both are a 220-dimensional feature vector, representing the spectral information of the point; then the average spectral value in the block is:

[0054]

[0055] 3.2: Calculate each pixel in the i-th superpixel block , k=1,2,...,j; and the average spectral value The correlation of j correlation values ​​{corr 1 , corr 2 ,..., corr j};

[0056] 3.3: Select the pixel with the greatest correlation in each superpixel block as the most representative pixel point, and selec...

Embodiment 3

[0059] The wrapper type hyperspectral band selection method based on pixel clustering is the same as that in embodiment 1-2, wherein in step (4), the initial clustering results of 462 representative points are used as initial labels, and then support vector machine (SVM) is used to perform initial clustering Class results are optimized and classified to obtain the final clustering results of 462 representative points; the specific steps include:

[0060] 4.1: Perform k-medoids clustering on the obtained 462 representative pixel points, set 462 representative points to gather into 25 classes in this embodiment, then the initial clustering result of 462 representative points is expressed as {C k}, k=1,2,...,462; where C k ∈{1,2,...,25}; the specific steps of k-medoids clustering of 462 representative points include:

[0061] 4.1a: Enter the number of clustering categories. In this example, the number of clustering categories is 25, and 25 representative points are randomly sele...

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 proposes a wrapper type hyperspectral band selection method based on pixel clustering. The specific operation steps are as follows: input the hyperspectral image to be selected and convert it into a matrix form; perform superpixel segmentation on the hyperspectral data to obtain superpixel blocks; use the correlation method to select a representative from each superpixel block point; use the unsupervised k-mediods method to cluster the obtained pixels first, and then use the svm classifier to further optimize the clustering effect to obtain the final clustering result; use the final clustering result, use the representative point as a labeled sample, and use the wrapper formula method to select bands. The invention solves the technical problem that the supervised band selection method cannot be used when there are no labeled samples. The invention has wide application range and good selection effect, and the supervised band selection method is used in the field of unsupervised band selection, which expands the application range of supervised band selection. It is used for data dimensionality reduction in hyperspectral image processing to facilitate subsequent data processing.

Description

technical field [0001] The invention belongs to the technical field of image processing, and further relates to an unsupervised hyperspectral band selection method in the field of hyperspectral imagery (Hyperspectral Imagery) band selection, specifically a wrapper hyperspectral band selection method based on pixel clustering, which is used for Data dimensionality reduction in hyperspectral image processing facilitates subsequent data processing. Background technique [0002] In recent years, with the development of remote sensing technology and imaging spectrometers, hyperspectral remote sensing can obtain spectral information of ground objects in continuous bands. Although rich spectral data information makes accurate target recognition possible; but at the same time, the phenomenon of data redundancy inevitably occurs due to the huge amount of data, which brings excessive data dimensionality, large amount of calculation, and problems to the subsequent hyperspectral image p...

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 Patents(China)
IPC IPC(8): G06K9/62
CPCG06F18/23213G06F18/2411
Inventor 曹向海焦李成姚利汪波棚杨淑媛刘红英马晶晶马文萍
Owner XIDIAN UNIV
Who we serve
  • R&D Engineer
  • R&D Manager
  • IP Professional
Why Patsnap Eureka
  • Industry Leading Data Capabilities
  • Powerful AI technology
  • Patent DNA Extraction
Social media
Patsnap Eureka Blog
Learn More
PatSnap group products

Browse by: Latest US Patents, China's latest patents, Technical Efficacy Thesaurus, Application Domain, Technology Topic, Popular Technical Reports.

© 2024 PatSnap. All rights reserved.Legal|Privacy policy|Modern Slavery Act Transparency Statement|Sitemap|About US| Contact US: help@patsnap.com