CN102663420B - Hyperspectral image classification method based on wavelet packet transformation and grey prediction model - Google Patents
Hyperspectral image classification method based on wavelet packet transformation and grey prediction model Download PDFInfo
- Publication number
- CN102663420B CN102663420B CN201210078647.7A CN201210078647A CN102663420B CN 102663420 B CN102663420 B CN 102663420B CN 201210078647 A CN201210078647 A CN 201210078647A CN 102663420 B CN102663420 B CN 102663420B
- Authority
- CN
- China
- Prior art keywords
- centerdot
- overbar
- hyperspectral
- hyperspectral image
- result
- 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.)
- Expired - Fee Related
Links
Landscapes
- Investigating Or Analysing Materials By Optical Means (AREA)
- Analysing Materials By The Use Of Radiation (AREA)
Abstract
A novel hyperspectral image classification method based on wavelet packet transformation and a grey prediction model belongs to the hyperspectral image processing field. The method comprises the following steps: firstly, acquiring hyperspectral data to be processed; secondly, using the wavelet packet transformation to decompose a hyperspectral response curve of each pixel; thirdly, using the grey prediction model to process a decomposition result; fourthly, using a characteristic construction result to supervise and classify hyperspectral data; fifthly, outputting a hyperspectral image ground object classification result. The method is an automatic hyperspectral image classification method. By using the method, wave band correlation can be effectively removed; data redundancy can be reduced; a negative effect of a dimension disaster on classification precision can be avoided; an application range is wide.
Description
Technical field
The present invention relates to a kind of novel hyperspectral image classification method based on wavelet package transforms and grey forecasting model, belong to Hyperspectral imagery processing field.
Background technology
High-spectrum remote-sensing (Hyperspectral Remote Sensing) technology is fast-developing in the last thirty years remote sensing of the earth technology, no matter is at business, military affairs or civil area, and it all has important theory value and wide application prospect.High spectrum resolution remote sensing technique utilizes imaging spectrometer from target to be measured, to obtain the spectral response with narrow interval, can capture the indiscoverable feature of conventional remote sensing technology, thereby establish solid physical basis for quantitative test material composition.China is that a few stand-alone development goes out one of country of complete high spectrum resolution remote sensing technique in the world, China's researcher is carried out high spectrum resolution remote sensing technique applied research comprehensive, multi-level, wide-range at categories such as mineral prospecting, medical diagnosis, reconnaissance behind enemy lines, battlefield monitoring, vegetation measurement, city plannings in recent years, all reaps rich fruits.
Compared with conventional remote sensing image processing, the principal feature of high-spectrum remote sensing has the following aspects:
1) data volume is large.Immediately observe numeric field data amount be exponential growth to same ground, this efficiency to Processing Algorithm has higher requirements.
2) correlativity is strong.Between high spectrum image adjacent band, exist stronger correlativity, and correlativity between this wave band is more much better than than spatial coherence, the spectrum sensitivity that this relevant generation reason comprises between topographic shadowing and sensor adjacent band relevant between the natural spectrum being produced by substance spectra reflecting attribute, that produced by terrain slope overlapping.
3) additive noise.The radiation characteristic of spectrometer record has superposeed by the noise of the generations such as atmosphere, sensor apparatus, quantification treatment and data transmission, and it can regard this classical problem of signal noise silencing as, can basis signal treatment theory solve.
4) mixed pixel point.Because resolution is limited, what the spectrum that single pixel place obtains reflected is not necessarily a kind of characteristic of material, and may be that ground observes territory (Ground Instantaneous Field of View) locate the mixing of several different material spectrum immediately, the complexity of mixing situation depends on concrete ground characteristics.
5) from target to image spectrum, mechanism and the mechanism of response are very complicated.Even commaterial, its spectrum performance also has very large difference conventionally, has the phenomenon of the different spectrum of so-called jljl and same object different images.
In sum, the superiority of high-spectrum remote sensing is take its larger data volume and higher data dimension as cost, therefore conventional Remote Sensing Image Processing Technology will be difficult to be applicable to high-spectrum remote sensing process field, and some image processing methods for traditional remotely-sensed data and technology face the challenge.
Wavelet package transforms and grey forecasting model are a kind of classical ways of Dynamic Data Processing.
Wavelet package transforms is one of instrument of being used widely in current application mathematics and engineering discipline.Compared with Fourier transform, wavelet transformation is the partial transformation of space and frequency, thereby can effectively information extraction from signal.The fundamental purpose of signal analysis is to find a kind of simple and effective signal transformation method, and the important information that signal comprises can be displayed.From physical significance, wavelet package transforms can carry out multiple dimensioned refinement analysis to function or signal by calculation functions such as flexible and translations, inherit and developed the thought of short time discrete Fourier transform localization, overcome again window size not with shortcomings such as frequency change simultaneously, a time-frequency window with frequency shift is provided, has solved the indeterminable many difficult problems of Fourier transform.
If system has the incomplete or uncertain of the randomness of ambiguity, dynamic change of structural relation and achievement data, claim this system to there is grey, the system with grey is called gray system.Gray system is both to have contained Given information, contains again unknown message or the non-system that knows information.In gray system theory, utilize original data sequence less or inapt expression gray system behavioural characteristic to do to set up after generation converts, in order to describe the model of the continuous change procedure of the inner things of gray system, be called gray model.One of important content of research gray system is abstract the system of how unclear from one, Global Information deficiency and sets up a model, the factor that this model can make gray system to clear and definite, develops into the more Research foundation that provides of knowing by indefinite by knowing little about it.Gray system theory is the product that cybernatic viewpoint and method extend to society, economic field, is also automatically to control the result that science combines with mathematics of operations research method.
In hyperspectral data processing system, for computation complexity requirement, the calculated amount of sorting algorithm is more few better, and it is poor to reach the traditional algorithm nicety of grading of this requirement.On the other hand, for the requirement of nicety of grading, sorting algorithm will have the good robustness of difference classification scene, and it is high to reach the algorithm computation complexity of this requirement.Therefore, need to find a kind of algorithm that considers computation complexity and this two aspects balance of nicety of grading, make that its time complexity is low, robustness good.
Summary of the invention
For the problems referred to above, the object of the present invention is to provide a kind of novel hyperspectral image classification method based on wavelet package transforms and grey forecasting model, process thereby propose a kind of wavelet package transforms and grey forecasting model of using the method that high spectral response curve carries out latent structure to different pixels in high spectrum image and then completes classification according to the feature of structure.This technological invention calculated amount is little, has kept again the accuracy of classification simultaneously, is applied to and in hyperspectral data processing system, has good robustness.
Method flow involved in the present invention comprises the following steps: (1) obtains pending high-spectral data; (2) application wavelet package transforms decomposes the high spectral response curve of each pixel; (3) application grey forecasting model is processed decomposition result; (4) use characteristic structure result is to the high-spectral data classification that exercises supervision; (5) output classification hyperspectral imagery result.Below the each step of the method flow process is elaborated.
(1) obtain pending high-spectral data, to arbitrary high spectrum pixel X, be designated as X=(x
1..., x
i..., x
k), wherein K is the wave band sum of high-spectral data, x
i, i=1 ..., K represents the spectral response numerical value of i wave band.
(2) application wavelet package transforms decomposes the high spectral response curve of each pixel: given wavelet mother function ψ and the maximum degree of depth j that decomposes, ψ is to X=(x in application
1..., x
i..., x
k) carry out j layer wavelet package transforms, obtain altogether 2
jindividual component, comprising 1 approximate component A and 2
j-1 details component
the energy coefficient of their correspondences be respectively a and
and meet relational expression (1):
Wherein, j is the maximum degree of depth of decomposing.
(3) application grey forecasting model is processed decomposition result: the energy coefficient sequence of computational details component
1 rank accumulation and sequence
Wherein, j is the maximum degree of depth of decomposing.
Order
:
The energy coefficient a and the development coefficient b that retain approximate component, the latent structure result of arbitrary pixel is made up of energy coefficient a and development coefficient b.
(4) use characteristic structure result is to the high-spectral data classification that exercises supervision.
(5) output classification hyperspectral imagery result.
The present invention has the following advantages: for hyperspectral data processing system, latent structure result is not subject to the interference of other pixel, strong robustness, and space complexity is little, and time complexity and the sexual intercourse of sample points retention wire, nicety of grading is high, applied widely.
Embodiment
Further illustrate the application process of this technological invention below with example.
1) obtain pending high-spectral data:
This example adopts Washington D.C.Mall high-spectral data, size is 1280 × 307 pixels, wavelength coverage is 0.4~2.4 μ m, remove after water vapor absorption wave band and low signal-to-noise ratio wave band, retain 191 wave bands, and intercept wherein a size be the subgraph of 562 × 307 pixels, subgraph comprises 7 class targets altogether, respectively: roof, meadow, trees, path, street, water, shade.
2) application wavelet package transforms decomposes the high spectral response curve of each pixel:
Setting wavelet mother function ψ is Haar small echo, the maximum degree of depth j=3 that decomposes.By X=(x
1..., x
i..., x
k) carry out, after 3 layers of WAVELET PACKET DECOMPOSITION, can obtaining 8 components, comprising 1 approximate component and 7 details components.The energy coefficient of remembering their correspondences is respectively a and d
1..., d
7.
3) application grey forecasting model is processed decomposition result:
The energy coefficient sequence d of computational details component
1..., d
71 rank accumulation and sequence
Order
?
B is called development coefficient,
it is grey action.
The energy coefficient a and the development coefficient b that retain approximate component, the latent structure result of arbitrary high spectrum pixel is made up of energy coefficient a and development coefficient b.
4) use characteristic structure result is to the example classification that exercises supervision.
5) output example classification results.
The present invention, through the concrete enforcement of analogue system, can effectively avoid the interference of dimension disaster and Hao Si phenomenon, completes the supervised classification of high spectrum image under the condition that retains target principal character.The present invention is for hyperspectral data processing system, and latent structure result is not subject to the interference of other pixel, strong robustness, and space complexity is little, and time complexity and the sexual intercourse of sample points retention wire, nicety of grading is high, applied widely.
Claims (1)
1. the hyperspectral image classification method based on wavelet package transforms and grey forecasting model, is characterized in that the method comprises the steps:
(1) obtain pending high-spectral data, to arbitrary high spectrum pixel X, be designated as X=(x
1..., x
i..., x
k), wherein K is the wave band sum of high-spectral data, x
i, i=1 ..., K represents the spectral response numerical value of i wave band;
(2) application wavelet package transforms decomposes the high spectral response curve of each pixel: given wavelet mother function ψ and the maximum degree of depth j that decomposes, ψ is to X=(x in application
1..., x
i..., x
k) carry out j layer wavelet package transforms, obtain altogether 2
jindividual component, comprising 1 approximate component A and 2
j-1 details component
the energy coefficient of their correspondences be respectively a and
and meet relational expression (1):
Wherein, j is the maximum degree of depth of decomposing;
(3) application grey forecasting model is processed decomposition result: the energy coefficient sequence of computational details component
1 rank accumulation and sequence
Wherein, j is the maximum degree of depth of decomposing;
Order
:
The energy coefficient a and the development coefficient b that retain approximate component, the latent structure result of arbitrary pixel is made up of energy coefficient a and development coefficient b;
(4) use characteristic structure result is to the high-spectral data classification that exercises supervision;
(5) output classification hyperspectral imagery result.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201210078647.7A CN102663420B (en) | 2012-03-22 | 2012-03-22 | Hyperspectral image classification method based on wavelet packet transformation and grey prediction model |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201210078647.7A CN102663420B (en) | 2012-03-22 | 2012-03-22 | Hyperspectral image classification method based on wavelet packet transformation and grey prediction model |
Publications (2)
Publication Number | Publication Date |
---|---|
CN102663420A CN102663420A (en) | 2012-09-12 |
CN102663420B true CN102663420B (en) | 2014-05-14 |
Family
ID=46772904
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201210078647.7A Expired - Fee Related CN102663420B (en) | 2012-03-22 | 2012-03-22 | Hyperspectral image classification method based on wavelet packet transformation and grey prediction model |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN102663420B (en) |
Families Citing this family (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN104182997B (en) * | 2014-08-15 | 2017-05-10 | 浙江科技学院 | Hyperspectral image compression method and application |
CN105117593A (en) * | 2015-08-18 | 2015-12-02 | 河海大学 | Wavelet transform and particle swarm optimized grey model-based short-term wind speed forecasting method |
CN107038502A (en) * | 2017-04-18 | 2017-08-11 | 国网安徽省电力公司芜湖供电公司 | Consider the improvement wavelet packet electricity demand forecasting method of Seasonal Characteristics |
CN109557031A (en) * | 2018-11-06 | 2019-04-02 | 核工业北京地质研究院 | A kind of rock core EO-1 hyperion Information extraction method |
Citations (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101697231A (en) * | 2009-10-29 | 2010-04-21 | 西北工业大学 | Wavelet transformation and multi-channel PCNN-based hyperspectral image fusion method |
Family Cites Families (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US7805005B2 (en) * | 2005-08-02 | 2010-09-28 | The United States Of America As Represented By The Secretary Of The Army | Efficient imagery exploitation employing wavelet-based feature indices |
-
2012
- 2012-03-22 CN CN201210078647.7A patent/CN102663420B/en not_active Expired - Fee Related
Patent Citations (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101697231A (en) * | 2009-10-29 | 2010-04-21 | 西北工业大学 | Wavelet transformation and multi-channel PCNN-based hyperspectral image fusion method |
Non-Patent Citations (2)
Title |
---|
Hyperspectral Image Classification Using Wavelet Packet Analysis and Gray Prediction Model;Jihao Yin等;《2010 International Conference on image Analysis and Signal Processing》;20100411;第322-326页 * |
Jihao Yin等.Hyperspectral Image Classification Using Wavelet Packet Analysis and Gray Prediction Model.《2010 International Conference on image Analysis and Signal Processing》.2010,第322-326页. |
Also Published As
Publication number | Publication date |
---|---|
CN102663420A (en) | 2012-09-12 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
Wang et al. | Capability of Sentinel-2 MSI data for monitoring and mapping of soil salinity in dry and wet seasons in the Ebinur Lake region, Xinjiang, China | |
Ghosh et al. | Aboveground biomass estimates of tropical mangrove forest using Sentinel-1 SAR coherence data-The superiority of deep learning over a semi-empirical model | |
CN110287457B (en) | Corn biomass inversion measurement method based on satellite radar remote sensing data | |
Zhang et al. | Soil moisture content retrieval from Landsat 8 data using ensemble learning | |
Singh et al. | Estimation of soil moisture applying modified dubois model to Sentinel-1; a regional study from central India | |
Mega et al. | Gauge adjusted global satellite mapping of precipitation (GSMaP_Gauge) | |
CN102663420B (en) | Hyperspectral image classification method based on wavelet packet transformation and grey prediction model | |
CN108766203B (en) | Compact polarization rice mapping method and system | |
Palandro et al. | Detection of changes in coral reef communities using Landsat-5 TM and Landsat-7 ETM+ data | |
Yan et al. | Inland water mapping based on GA-LinkNet from CyGNSS data | |
Liu et al. | Soil salinity inversion in Hetao Irrigation district using microwave radar | |
CN112285709B (en) | Atmospheric ozone remote sensing laser radar data fusion method based on deep learning | |
Zhao et al. | A new approach for forest height inversion using X-band single-pass InSAR coherence data | |
Xie et al. | Reconstruction of subsurface temperature field in the south China Sea from satellite observations based on an attention U-net model | |
Prasad et al. | Remote sensing satellite data and spectral indices: an initial evaluation for the sustainable development of an urban area | |
Shao et al. | Sea state parameters retrieval from cross-polarization Gaofen-3 SAR data | |
Ashphaq et al. | Preliminary examination of influence of Chlorophyll, Total Suspended Material, and Turbidity on Satellite Derived-Bathymetry estimation in coastal turbid water | |
CN107832798A (en) | Polarimetric SAR Image object detection method based on NSCT ladder pessimistic concurrency controls | |
Chandola | Polarimetric SAR interferometry for forest aboveground biomass estimation | |
Zoljoodi et al. | Seasonal and interannual cycles of total phytoplankton phenology metrics in the Persian Gulf using ocean color remote sensing | |
Pauly et al. | Modelling the distribution and ecology of Trichosolen blooms on coral reefs worldwide | |
Xu et al. | A comparative study on the applicability and effectiveness of NSVI and NDVI for estimating fractional vegetation cover based on multi-source remote sensing image | |
Al-Mutairi et al. | Spatial enhancement of DEM using interpolation methods: a case study of Kuwait’s coastal zones | |
Kang et al. | Direct aerosol optical depth retrievals using MODIS reflectance data and machine learning over East Asia | |
Goerner et al. | Monitoring of the Ecuadorian mountain rainforest with remote sensing |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | Publication | ||
PB01 | Publication | ||
C10 | Entry into substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
C14 | Grant of patent or utility model | ||
GR01 | Patent grant | ||
CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20140514 Termination date: 20150322 |
|
EXPY | Termination of patent right or utility model |