CN112330622B - Hyperspectral image band selection method based on ground feature maximum discrimination - Google Patents
Hyperspectral image band selection method based on ground feature maximum discrimination Download PDFInfo
- Publication number
- CN112330622B CN112330622B CN202011188704.8A CN202011188704A CN112330622B CN 112330622 B CN112330622 B CN 112330622B CN 202011188704 A CN202011188704 A CN 202011188704A CN 112330622 B CN112330622 B CN 112330622B
- Authority
- CN
- China
- Prior art keywords
- band
- feature
- discrimination
- ground
- wave
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/60—Analysis of geometric attributes
- G06T7/62—Analysis of geometric attributes of area, perimeter, diameter or volume
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/10—Image acquisition modality
- G06T2207/10032—Satellite or aerial image; Remote sensing
- G06T2207/10036—Multispectral image; Hyperspectral image
Abstract
The invention discloses a hyperspectral image band selection method based on maximum ground feature discrimination, and belongs to the field of remote sensing image processing. The method takes a high-dimensional polyhedron volume formed by different surface feature spectrums as an objective function, and the higher the volume value is, the maximum discrimination between surface features is indicated. Therefore, the wave band combination which enables the volume to obtain the maximum value is selected, the ground objects can be maximally distinguished, and therefore the classification precision of the ground objects of the hyperspectral image is improved. According to the iterative optimization searching method, efficient waveband set searching can be achieved, and waveband combinations enabling different surface feature spectrums to have high discrimination are obtained finally. The method can obtain good ground feature classification precision while reducing the calculation complexity, and has important significance in the field of hyperspectral data processing.
Description
Technical Field
The invention belongs to the technical field of remote sensing image processing, and particularly relates to a hyperspectral image band selection method based on ground feature maximum discrimination.
Background
The imaging spectrometer simultaneously images hundreds of wave bands of the ground object with extremely high spectral resolution to obtain a hyperspectral (data) image with abundant spectral information, thereby realizing revolutionary combination of space dimensional information and spectral dimensional information. Due to the adoption of the hyperspectral remote sensing technology, substances which are not detectable in broadband remote sensing can be detected in hyperspectral remote sensing. The hyperspectral remote sensing has extremely wide application in various fields such as environment assessment, resource investigation, agriculture and forestry and the like. However, the increase of the number of the wave bands of the hyperspectral image increases the computational complexity on one hand, and also brings about a "dimensionality disaster" effect on the other hand, namely, the application effect is reduced on the contrary along with the increase of the number of the wave bands.
Disclosure of Invention
The invention aims to provide a hyperspectral image band selection method based on the maximum distinguishing degree of ground objects.
In order to achieve the purpose, the invention adopts the technical scheme that:
a hyperspectral image band selection method based on feature maximum discrimination is used for selecting a specific band from hyperspectral images with the total band number of L to realize maximum discrimination of features, wherein the hyperspectral images comprise m feature types; the method comprises the following steps:
step 1, inputting surface feature spectrum information of various surface features contained in an image, and setting the number n of specific wave bands to be selected, wherein n is more than or equal to m and is less than L;
step 2, randomly selecting n wave bands from the hyperspectral image as candidate wave bands;
step 3, calculating the feature discrimination under the candidate wave band, wherein the feature discrimination is characterized by the volume of a high-dimensional polyhedron formed by the feature spectrums of m types of features under the corresponding wave band;
step 4, selecting a wave band from the rest unselected wave bands, and replacing a certain wave band in the candidate wave bands by the wave band to obtain n replacing results;
step 5, respectively calculating the ground feature discrimination corresponding to each result in the n replacement results, and selecting the maximum value in the n ground feature discrimination;
and 6, comparing the maximum value obtained in the step 5 with the feature discrimination corresponding to the candidate band in the step 4, if the maximum value obtained in the step 5 is greater than the feature discrimination corresponding to the candidate band in the step 4, using the replacement result corresponding to the maximum value as a new candidate band, and repeating the steps 4-6, otherwise, the candidate band in the step 4 in the cycle is the selected specific band.
Further, the volume of the high-dimensional polyhedron is calculated according to the following formula:
wherein the content of the first and second substances,
s i is the terrestrial object spectrum of the ith terrestrial object containing only n corresponding wave bands,
volume of high-dimensional polyhedron, A = [ s ] 2 -s 1 ,s 3 -s 1 ,…,s m -s 1 ]Det (·) is a determinant operator, and superscript T represents the transpose of the matrix.
The invention has the following advantages:
(1) The invention provides the method for selecting the wave bands based on the maximum discrimination of the spectra of the ground objects, and the selected wave bands ensure that the spectra of different ground objects keep higher discrimination, so higher classification precision can be obtained.
(2) The method only needs to calculate the spectrums of a plurality of ground objects, does not need image participation and has high calculation efficiency.
Drawings
FIG. 1 is an overall flow chart of the present invention.
Detailed Description
The technical scheme of the invention is further explained by combining the drawings and the detailed implementation mode.
As shown in fig. 1, a hyperspectral image band selection method based on the maximum feature discrimination is used for selecting a specific band from a hyperspectral image with a total band number of L, wherein the hyperspectral image comprises m feature types; the method comprises the following steps:
step 1, inputting surface feature spectrum information of various surface features contained in an image, and setting the number n of specific wave bands to be selected, wherein n is more than or equal to m and is less than L;
step 2, randomly selecting n wave bands from the hyperspectral images, and arranging the wave bands in an ascending order to serve as candidate wave bands;
step 3, calculating the feature discrimination under the candidate wave band, wherein the feature discrimination is characterized by the volume of a high-dimensional polyhedron formed by the feature spectrums of m types of features under the corresponding wave band;
step 4, selecting a wave band from the rest unselected wave bands, and replacing a certain wave band in the candidate wave bands by the wave band to obtain n replacing results;
step 5, respectively calculating the feature discrimination corresponding to each of the n replacement results, and selecting the maximum value of the n feature discrimination;
and 6, comparing the maximum value obtained in the step 5 with the feature discrimination corresponding to the candidate band in the step 4, if the maximum value obtained in the step 5 is greater than the feature discrimination corresponding to the candidate band in the step 4, using the replacement result corresponding to the maximum value as a new candidate band, and repeating the steps 4-6, otherwise, the candidate band in the step 4 in the cycle is the selected specific band.
The volume of the high-dimensional polyhedron is calculated according to the following formula:
wherein the content of the first and second substances,
s i is the surface feature spectrum of the i-th surface feature containing only n corresponding wave bands,
volume of high-dimensional polyhedron, A = [ s ] 2 -s 1 ,s 3 -s 1 ,…,s m -s 1 ]Det (·) is a determinant operator, and superscript T represents the transpose of the matrix.
The method only needs to input the spectral information of the ground features, then randomly selects a specified number of wave bands from all the wave bands, and conducts iterative search until the obtained wave band combination has the maximum high-dimensional polyhedral volume, and at the moment, the wave band set corresponding to the ground feature spectrum has the maximum discrimination.
The principle of the method is as follows: and taking a high-dimensional polyhedron volume formed by different surface feature spectrums as an objective function, wherein the higher the volume value is, the maximum discrimination between the surface features is indicated. Therefore, the wave band combination which enables the volume to obtain the maximum value is selected, the ground objects can be maximally distinguished, and therefore the classification precision of the ground objects of the hyperspectral image is improved. By the iterative optimization searching method, efficient waveband set searching can be realized, and waveband combinations which enable different surface feature spectrums to have high discrimination are obtained finally. The method can improve the ground feature classification efficiency under the condition of keeping higher ground feature classification precision.
The effect of the present method can be further illustrated by the following tests:
1. test conditions were used.
The computer is configured with an Intel Core i7-3770 CPU 3.4Ghz and 64GB memory; the operating system is Windows 7 64 bit professional edition, and the software environment is MATLAB 2017.
2. Test methods.
The method is adopted to select the wave bands, only the spectrum of the ground object needs to be input, and the output result is the selected wave band set. In order to verify the effect of the method, the method is compared with the original calculation effect when the wave band is not selected: the contrast group directly utilizes the original wave band to classify the ground features, and the ground feature classification of the method comprises the steps of firstly selecting the wave band and then classifying the wave band corresponding to the hyperspectral image. The classification method is an SVM method, a control group is completely the same as a training sample of the method (randomly selecting 15% from the ground objects to be classified), and the comparison analysis is mainly carried out according to the ground object classification precision and the calculation time.
3. Test contents and results.
The experiment is carried out by selecting Indian Pines public images acquired by an AVIRIS sensor in the United states, the size of the images is 145 multiplied by 145 pixels, the number of effective spectrums is 220, and the types of the ground objects participating in classification are 15.
The test result is as follows, classification processing is carried out by using a comparison method, the classification accuracy is 73.31%, and the time is consumed for 43.11s; the method is used for selecting 15 wave bands for processing, the classification accuracy is 80.82%, and the time is consumed for 4.75s.
Test results show that the wave band selected by the method can obtain higher ground object classification precision, and meanwhile, the calculation efficiency can be greatly improved.
In a word, the hyperspectral image band selection method takes the maximum distinguishing degree of the ground features as the target function to select the hyperspectral image band, reduces the calculation complexity, can obtain good ground feature classification precision, and has important significance in the field of hyperspectral data processing.
Claims (2)
1. A hyperspectral image wave band selection method based on the maximum distinguishing degree of ground objects is characterized by being used for selecting a specific wave band from a hyperspectral image with the total wave band number of L to realize the maximum distinguishing of the ground objects, wherein the hyperspectral image contains m ground object types; the method comprises the following steps:
step 1, inputting surface feature spectrum information of various surface features contained in an image, and setting the number n of specific wave bands to be selected, wherein n is more than or equal to m and is less than L;
step 2, randomly selecting n wave bands from the hyperspectral image as candidate wave bands;
step 3, calculating the land feature discrimination under the candidate wave band, wherein the land feature discrimination is characterized by the volume of a high-dimensional polyhedron formed by land feature spectrums of m types of land features under the corresponding wave band;
step 4, selecting a wave band from the rest unselected wave bands, and replacing a certain wave band in the candidate wave bands by the wave band to obtain n replacing results;
step 5, respectively calculating the ground feature discrimination corresponding to each result in the n replacement results, and selecting the maximum value in the n ground feature discrimination;
and 6, comparing the maximum value obtained in the step 5 with the feature discrimination corresponding to the candidate band in the step 4, if the maximum value obtained in the step 5 is greater than the feature discrimination corresponding to the candidate band in the step 4, using the replacement result corresponding to the maximum value as a new candidate band, and repeating the steps 4-6, otherwise, the candidate band in the step 4 in the cycle is the selected specific band.
2. The hyperspectral image band selection method based on the maximum feature discrimination of the ground object according to claim 1 is characterized in that the volume of the hyperspectral polyhedron is calculated according to the following formula:
wherein, the first and the second end of the pipe are connected with each other,
s i is the surface feature spectrum of the i-th surface feature containing only n corresponding wave bands,
volume of high-dimensional polyhedron, A = [ s ] 2 -s 1 ,s 3 -s 1 ,…,s m -s 1 ]Det (·) is a determinant operator, and superscript T represents the transpose of the matrix.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011188704.8A CN112330622B (en) | 2020-10-30 | 2020-10-30 | Hyperspectral image band selection method based on ground feature maximum discrimination |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011188704.8A CN112330622B (en) | 2020-10-30 | 2020-10-30 | Hyperspectral image band selection method based on ground feature maximum discrimination |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN112330622A CN112330622A (en) | 2021-02-05 |
| CN112330622B true CN112330622B (en) | 2022-11-11 |
Family
ID=74296052
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202011188704.8A Active CN112330622B (en) | 2020-10-30 | 2020-10-30 | Hyperspectral image band selection method based on ground feature maximum discrimination |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112330622B (en) |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114417247B (en) * | 2022-01-19 | 2023-05-09 | 中国电子科技集团公司第五十四研究所 | Hyperspectral image band selection method based on subspace |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108764005A (en) * | 2018-01-31 | 2018-11-06 | 华侨大学 | A kind of high-spectrum remote sensing atural object space Spectral Characteristic extracting method and system |
| CN109238463A (en) * | 2018-08-22 | 2019-01-18 | 天津大学 | A kind of active EO-1 hyperion detection system of LED based low cost |
Family Cites Families (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106778680B (en) * | 2017-01-06 | 2018-06-26 | 杭州电子科技大学 | A kind of hyperspectral image band selection method and device based on critical bands extraction |
| CN107220662B (en) * | 2017-05-16 | 2019-07-19 | 西北工业大学 | Hyperspectral image band selection method based on global optimum's cluster |
| CN108596077A (en) * | 2018-04-20 | 2018-09-28 | 南阳理工学院 | The adaptive spectrum focus wave section selection method of high spectrum image |
| CN110766619A (en) * | 2019-09-19 | 2020-02-07 | 北京航空航天大学 | Unsupervised band selection algorithm based on band quality analysis |
| CN111597951A (en) * | 2020-05-12 | 2020-08-28 | 中国电子科技集团公司第五十四研究所 | Hyperspectral image feature selection method for target detection |
-
2020
- 2020-10-30 CN CN202011188704.8A patent/CN112330622B/en active Active
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108764005A (en) * | 2018-01-31 | 2018-11-06 | 华侨大学 | A kind of high-spectrum remote sensing atural object space Spectral Characteristic extracting method and system |
| CN109238463A (en) * | 2018-08-22 | 2019-01-18 | 天津大学 | A kind of active EO-1 hyperion detection system of LED based low cost |
Also Published As
| Publication number | Publication date |
|---|---|
| CN112330622A (en) | 2021-02-05 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN107451614B (en) | Hyperspectral classification method based on fusion of space coordinates and space spectrum features | |
| Geng et al. | A fast volume-gradient-based band selection method for hyperspectral image | |
| CN111881953B (en) | Remote sensing hyperspectral image classification method based on local binary pattern and KNN classifier | |
| Jia et al. | Cluster-space representation for hyperspectral data classification | |
| CN106503739A (en) | The target in hyperspectral remotely sensed image svm classifier method and system of combined spectral and textural characteristics | |
| CN110770752A (en) | Automatic pest counting method combining multi-scale feature fusion network with positioning model | |
| Huang et al. | Double nearest proportion feature extraction for hyperspectral-image classification | |
| CN109583469B (en) | K-means hyperspectral image band clustering method based on mutual information | |
| CN105913092B (en) | Figure canonical hyperspectral image band selection method based on sub-space learning | |
| Boggavarapu et al. | A new framework for hyperspectral image classification using Gabor embedded patch based convolution neural network | |
| CN109697465B (en) | Hyperspectral image classification method based on wavelet kernel local Fisher discriminant analysis | |
| CN104268556A (en) | Hyperspectral image classification method based on nuclear low-rank representing graph and spatial constraint | |
| CN108154094A (en) | The non-supervisory band selection method of high spectrum image divided based on subinterval | |
| Liu et al. | Classification of urban hyperspectral remote sensing imagery based on optimized spectral angle mapping | |
| Deng et al. | Citrus disease recognition based on weighted scalable vocabulary tree | |
| CN112330622B (en) | Hyperspectral image band selection method based on ground feature maximum discrimination | |
| CN115457311A (en) | Hyperspectral remote sensing image band selection method based on self-expression transfer learning | |
| CN111597951A (en) | Hyperspectral image feature selection method for target detection | |
| CN111680579A (en) | Remote sensing image classification method for adaptive weight multi-view metric learning | |
| CN107203779A (en) | The EO-1 hyperion dimension reduction method kept based on empty spectrum information | |
| CN115578632A (en) | Hyperspectral image classification method based on expansion convolution | |
| CN114863291A (en) | Hyperspectral image band selection method based on MCL and spectral difference measurement | |
| CN111079850B (en) | Depth-space spectrum combined hyperspectral image classification method of band significance | |
| CN106033545B (en) | Wave band selection method of determinant point process | |
| Li et al. | Spectral similarity-preserving hyperspectral band selection |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |