Eureka AIR delivers breakthrough ideas for toughest innovation challenges, trusted by R&D personnel around the world.

Nearest Neighbor Hyperspectral Image Classification Method Based on Dictionary and Band Recombination

A technology of hyperspectral images and classification methods, applied in the field of spectral data classification, can solve the problems of low classification accuracy of hyperspectral images, incomprehensible combination of space-spectral information, and high time complexity, so as to increase the classification accuracy and time complexity. Low, the effect of improving the classification accuracy

Active Publication Date: 2018-03-06
XIDIAN UNIV
View PDF4 Cites 0 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

[0005] The purpose of the present invention is to solve the technical problems that the current single classifier is not high in classification accuracy of hyperspectral images, and the classification effect is not good; while the ordinary space-spectrum combination method has high time complexity and the combination of space-spectrum information is not tight. A low-time complexity, high-precision nearest neighbor hyperspectral image classification method based on dictionary and band recombination, which specifically includes the following steps:

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
  • Nearest Neighbor Hyperspectral Image Classification Method Based on Dictionary and Band Recombination
  • Nearest Neighbor Hyperspectral Image Classification Method Based on Dictionary and Band Recombination
  • Nearest Neighbor Hyperspectral Image Classification Method Based on Dictionary and Band Recombination

Examples

Experimental program
Comparison scheme
Effect test

Embodiment 1

[0041] The present invention is a space-spectrum combination method based on dictionary and band recombination for rapid classification of nearest neighbor hyperspectral images, see figure 1 , including the following steps:

[0042] Step 1 read hyperspectral remote sensing data image, hyperspectral remote sensing data image referred to as hyperspectral image.

[0043] Step 2: Preprocess the read hyperspectral image first, use the spatial neighborhood information of the hyperspectral image, take the L nearest neighbor pixels for each pixel of the hyperspectral image, and perform L-neighborhood mean processing to obtain the preprocessed hyperspectral image. In the hyperspectral initial image, there are a certain number of pixels polluted by noise. Without preprocessing, these pixels are very destructive to the classification accuracy. Therefore, before classification, the present invention utilizes the hyperspectral image Spatial information, noise pixels are processed.

[00...

Embodiment 2

[0050] The nearest neighbor hyperspectral image rapid classification method based on the space-spectrum combination of dictionary and band recombination is the same as embodiment 1, in step 2, each pixel in the hyperspectral image is carried out L-neighborhood mean value processing, used in the present invention The obtained hyperspectral data is the Indian Pines hyperspectral image in the AVIRIS data set, which contains a total of 200 bands, specifically:

[0051] It is assumed that the hyperspectral image has N wave bands in total, and in the experiments of the present invention, N=200. A certain pixel in the hyperspectral image is x i ={x i,1 ,...,x i,N}, with x i As the center, take the L neighboring pixels {x 1 ,...,x L}, in the experiments done by the present invention, L=24. where x 1 ={x 1,1 ,...,x 1,N}∈R N , R N is the N-dimensional feature space, x 2 ={x 2,1 ,...,x 2,N}∈R N , and so on, for x i Do 24-average processing for all L+1=25 pixels in the 5×5...

Embodiment 3

[0055] The nearest neighbor hyperspectral image classification method based on dictionary and band recombination is the same as in embodiment 1-2, in step 3, there are m pixels {x i,1 ,...,x i,m},make

[0056]

[0057] Perform the above average processing on all bands to get {x 1 ,...,x N}, N=200, for {x 1 ,...,x N} sort from small to large to obtain the hyperspectral image mean sequence, and then rearrange the hyperspectral bands according to this sequence, and divide the reorganized N bands into n equal parts in order to form n sub-bands. In the present invention, the Indian image is divided into 3 equal parts, that is, n=3. The first sub-band is bands 1-66 after band reorganization, the second sub-band is bands 67-133, and the third sub-band is bands 134-200.

[0058] In step 5, let sub-dictionary D=[x 1 ,x 2 ,...,x d ]exist dimensional feature space, and its category label is ω i ∈{1,2,...,C}. C is the number of categories. Let d be the number of atoms in ...

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 discloses a nearest neighbor hyperspectral image classification method based on dictionary and band recombination. The realization steps include: using the spatial information of the hyperspectral image to perform L-neighborhood mean value processing on each pixel; The point mean value band is reorganized and divided into n-equal sub-bands; some pixels are randomly selected to form a dictionary, and other pixels form a test sample set, and the dictionary and test sample set are divided into n sub-dictionaries and sub-test sample sets according to sub-bands and compared correspondingly The sub-dictionary is used for nearest neighbor discrimination to obtain n initial classification maps; n-KNN discrimination is performed to obtain the final result map. The invention solves the problems of low classification accuracy, poor classification effect, high time complexity of the ordinary space-spectrum combination method, and loose combination of space-spectrum information. ‑ Spectral combination method for classification, in a short period of time, a higher precision classification map was obtained. High precision and low time complexity.

Description

technical field [0001] The invention belongs to the technical field of spectral data classification, and mainly relates to the classification of hyperspectral remote sensing data, in particular to a space-spectrum combination method for rapid classification of nearest neighbor hyperspectral images based on dictionaries and band recombination. It can be used in cartography, marine remote sensing, vegetation survey, atmospheric research, agricultural remote sensing, environmental monitoring and other fields. Background technique [0002] Hyperspectral remote sensing technology has developed rapidly since the 1980s. Hyperspectral records the continuous spectral characteristics of ground objects with its rich band information, and has the possibility of recognizing more types of ground objects and classifying objects with higher accuracy. However, the high spectral dimension and spectral resolution of hyperspectral images bring great opportunities for ground object classificati...

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): G06K9/62G06K9/66
Inventor 慕彩红焦李成云智强熊涛刘红英冯婕田小林张文龙吴生财
Owner XIDIAN UNIV
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
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