CN105761272B - Pure material quantity determines method in a kind of imaging spectral mixed pixel - Google Patents
Pure material quantity determines method in a kind of imaging spectral mixed pixel Download PDFInfo
- Publication number
- CN105761272B CN105761272B CN201610149573.XA CN201610149573A CN105761272B CN 105761272 B CN105761272 B CN 105761272B CN 201610149573 A CN201610149573 A CN 201610149573A CN 105761272 B CN105761272 B CN 105761272B
- Authority
- CN
- China
- Prior art keywords
- spectrum
- matrix
- pure material
- mixed pixel
- imaging
- 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
- 238000000034 method Methods 0.000 title claims abstract description 50
- 230000003595 spectral effect Effects 0.000 title claims abstract description 34
- 239000000463 material Substances 0.000 title claims abstract description 33
- 238000003384 imaging method Methods 0.000 title claims abstract description 32
- 238000001228 spectrum Methods 0.000 claims abstract description 66
- 239000011159 matrix material Substances 0.000 claims abstract description 44
- 238000000354 decomposition reaction Methods 0.000 claims abstract description 9
- 238000012545 processing Methods 0.000 claims abstract description 8
- 230000004069 differentiation Effects 0.000 claims abstract description 7
- 238000004611 spectroscopical analysis Methods 0.000 claims abstract description 6
- 238000002835 absorbance Methods 0.000 claims description 8
- 230000008569 process Effects 0.000 claims description 7
- 238000002310 reflectometry Methods 0.000 claims description 5
- 239000013598 vector Substances 0.000 claims description 5
- 230000001174 ascending effect Effects 0.000 claims description 2
- 238000009499 grossing Methods 0.000 claims description 2
- 238000004458 analytical method Methods 0.000 abstract 1
- 238000002329 infrared spectrum Methods 0.000 abstract 1
- 238000004422 calculation algorithm Methods 0.000 description 26
- 229920000877 Melamine resin Polymers 0.000 description 13
- JDSHMPZPIAZGSV-UHFFFAOYSA-N melamine Chemical compound NC1=NC(N)=NC(N)=N1 JDSHMPZPIAZGSV-UHFFFAOYSA-N 0.000 description 13
- 238000002474 experimental method Methods 0.000 description 10
- 239000000126 substance Substances 0.000 description 8
- 235000010627 Phaseolus vulgaris Nutrition 0.000 description 6
- 244000046052 Phaseolus vulgaris Species 0.000 description 6
- 238000001514 detection method Methods 0.000 description 4
- 239000000203 mixture Substances 0.000 description 4
- 238000000513 principal component analysis Methods 0.000 description 4
- 238000013329 compounding Methods 0.000 description 3
- 238000002156 mixing Methods 0.000 description 3
- 238000010183 spectrum analysis Methods 0.000 description 3
- IJGRMHOSHXDMSA-UHFFFAOYSA-N Atomic nitrogen Chemical compound N#N IJGRMHOSHXDMSA-UHFFFAOYSA-N 0.000 description 2
- 230000008859 change Effects 0.000 description 2
- 230000007423 decrease Effects 0.000 description 2
- 238000010586 diagram Methods 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 238000007689 inspection Methods 0.000 description 2
- 238000005259 measurement Methods 0.000 description 2
- 238000012360 testing method Methods 0.000 description 2
- 238000012152 algorithmic method Methods 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 230000000052 comparative effect Effects 0.000 description 1
- 150000001875 compounds Chemical class 0.000 description 1
- 238000007796 conventional method Methods 0.000 description 1
- 238000001816 cooling Methods 0.000 description 1
- 230000003247 decreasing effect Effects 0.000 description 1
- 230000007812 deficiency Effects 0.000 description 1
- 238000000151 deposition Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 235000013305 food Nutrition 0.000 description 1
- 239000007788 liquid Substances 0.000 description 1
- 230000015654 memory Effects 0.000 description 1
- 229910052757 nitrogen Inorganic materials 0.000 description 1
- 239000005416 organic matter Substances 0.000 description 1
- 102000004169 proteins and genes Human genes 0.000 description 1
- 108090000623 proteins and genes Proteins 0.000 description 1
- 238000000985 reflectance spectrum Methods 0.000 description 1
- 230000008439 repair process Effects 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 230000035945 sensitivity Effects 0.000 description 1
- 241000894007 species Species 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 238000012795 verification Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/0002—Inspection of images, e.g. flaw detection
- G06T7/0004—Industrial image inspection
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T5/00—Image enhancement or restoration
- G06T5/70—Denoising; Smoothing
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Quality & Reliability (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Investigating Or Analysing Materials By Optical Means (AREA)
Abstract
The present invention discloses pure material quantity in a kind of imaging spectral mixed pixel and determines method, and mixed pixel spectroscopic data is obtained by using imaging spectrometer；The original spectrum relatively low to signal-to-noise ratio is smoothly pre-processed；Similarity differentiation is carried out to obtaining pixel spectrum, the high original spectrum of similarity is subjected to average centralization processing；By simplifying Eigenvalues Decomposition by spectrum matrix projection to orthogonal subspaces；The sub-space feature value descending of acquisition is normalized to weight sequence；Weight sequence two-dimensional coordinate system is established, by a kind of geometric distance automatic discrimination mode, determines mixed pixel pure material number.Using method provided by the invention, mixed spectra pure material number can be efficiently and accurately detected in the case of without any prior information.The present invention be applicable not only to closely imaging spectral (such as：Micro-imaging spectrum) and remote sensing EO-1 hyperion, it can be used for determining for non-imaged spectrum (such as near-infrared spectrum analysis) darky system number of components.
Description
Technical field
The present invention relates to pure material number in imaging spectral analysis technical field, more particularly to a kind of imaging spectral mixed pixel
Amount determines method.
Background technology
Imaging spectral technology is with fastest developing speed since being the 1990s, most noticeable spectral analysis technique, utilizes
Imaging spectrometer analyzes compounding substances chemical composition property or content, has that speed is fast, does not destroy sample, operation letter
Singly, the features such as stability is good, efficient, can be widely applied to the fields such as remote sensing, agricultural and food inspection.Imaging spectrometer data is stood
Cube usually contains the spectrum of thousands of a pixel points, due to imaging spectrometer limited spatial resolution or due to measured object
For mixed compound, what is generally comprised in each pixel is not only a kind of pure material, so mixed pixel is formed, these mixing
The spectrum of pixel is mixed by several pure substance spectra, therefore efficiently and accurately determines that mixed pixel pure material quantity helps
Analyzed in mixture, and very important meaning is mixed with to research mixed pixel solution.
During spectral measurement is carried out to compounding substances, due to the spectral mixture information acquired in spectrometer often
Be caused by the spectral signal of different pure materials mixes, based on this, with reference to famous langbobier law, a kind of linear hybrid
Model is widely used in studying mixed pixel spectrum resolution.The basic thought of linear mixed model is the mixed pixel studied
Spectrum (vector) can be seen as being multiplied with pure material Abundances by pure substance spectra and accumulated result, in recent years, is based on
Spectral linear mixing model, many algorithms are developed for use in pure material quantity survey, these algorithms can be roughly divided into as
Lower three classes：
First kind algorithm is the algorithm based on principal component analysis (PCA).The basic principle of PCA is empty in original high dimensional data
Between find one group of suitable orthogonal base vectors, its dimension is less than the dimension of initial data, but can retain the main of former data
Information.PCA can effectively estimate the number of pure material, and have good repellence to noise signal, but shortcoming exists
In, when dimension of a vector space is larger, the computation complexity of data covariance matrix and Eigenvalue Decomposition and interior in algorithm
Depositing storage consumption will all increase sharply.
Second class method is the algorithm based on virtual dimension (VD), is shown by calculating existing for sample characteristics inspection signal
Work property, pure material number can be accurately determined using the number of the signal source finally detected as pure material number, algorithm, but
Be disadvantageous in that needs to set specific false alarm rate for different spectroscopic datas, and operational efficiency is not high.
Three classes method is a kind of method based on minimal error, wherein representative is HySime methods.This is
A kind of unsupervised algorithm, algorithm to find to subspace projection by original signal by that can minimize the sub empty of projection error
Between, pure material number is just used as using the dimension of subspace.Algorithm fully automated can obtain pure material number, and can be effective gram
Noise jamming is taken, but deficiency is to need first to estimate noise matrix, and the result of noise matrix estimation can estimate pure material number
Meter produces a very large impact, and increasing with spectral band number, and the estimated accuracy of algorithm can be decreased obviously.
Above method can relatively accurately estimate pure material number, but there is also the following problem：1) it is most of
Algorithmic method needs that threshold value is manually set, and the selection of threshold value is difficult to determine when pure pixel is similar to the characteristic value of noise.2)
Some algorithm computation complexities are higher, it is necessary to consume a large amount of memories, do not possess the real-time of pure material quantity detection.3) some are calculated
Method can be good at identifying for general EO-1 hyperion, but decline for EO-1 hyperion arithmetic accuracy of wave band number when more.
The content of the invention
In view of this, a kind of method automatically determined it is a primary object of the present invention to provide pure material number is used to be imaged
Spectrum analysis, this method can solve existing pure material number and determine that method needs artificial setting threshold parameter, computation complexity
Height, operational efficiency be not high, spectral band number increases the technical problems such as accuracy of detection decline.
To reach above-mentioned purpose, the technical proposal of the invention is realized in this way：It is pure in a kind of imaging spectral mixed pixel
Amount of material determines method, as shown in Figure 1, this method step is as follows：
Step A, mixed pixel spectroscopic data is obtained by using imaging spectrometer；
Step B, original spectrum relatively low to signal-to-noise ratio is smoothly pre-processed；
Step C, similarity differentiation is carried out to obtaining pixel spectrum, the high pre-processed spectrum of similarity is subjected to average center
Change is handled；
Step D, by simplifying Eigenvalues Decomposition by spectrum matrix projection to orthogonal subspaces；
Step E, the orthogonal subspaces characteristic value descending of acquisition is normalized to weight sequence；
Step F, weight sequence two-dimensional coordinate system is established, by a kind of geometric distance automatic discrimination mode, determines mixing picture
First pure material quantity p.
Wherein, obtaining mixed pixel spectroscopic data by using imaging spectrometer in the step A is specially：Obtain institute
There are pixel point spectral reflectivity or absorbance to obtain original spectrum matrix Ym×n, wherein m is the number of all pixels, and n is spectrum
Wave band number.
Wherein, original spectrum relatively low to signal-to-noise ratio in the step B, which smoothly pre-process, is specially：
To reduce influence of the high-frequency noise to spectrum, by original spectrum matrix Ym×nIn every spectrum y1×nRow pretreatment,
Using Moving Window the disposal of gentle filter：
WhereinXiIt is smooth rear and smooth preceding spectral vector y respectively1×nIn each wave band reflectivity or absorbance,
WjBe moving window it is smooth in weight factor, for simplify smoothing process, take Wj=1,2t+1 are smoothly to count, general t take 1 or
Person 2.
It is pointed out that the preprocess method of the present invention is not limited only to the above method, other any denoisings, go
The preprocess method that the garbages such as background eliminate should all belong to protection scope of the present invention.
Wherein, it is specially to obtaining the progress similarity differentiation of pixel spectrum in the step C：
C1, ask for pretreated imaging spectral matrix Xm×nAveraged spectrum
Wherein x [i] is i-th pretreated spectrum；
C2, by pretreated spectrum matrix Xm×nEvery spectrum is selected one by one matches SAM with average spectrum progress spectral modeling
Similarity differentiates：
Wherein xilFor from Xm×nIn reflectance value or absorbance at i-th spectrum, l-th of wavelength for selecting；
If C3, Xm×nSimilarity SAM values are less than 0.01 between every spectrum of matrix, to Xm×nCarry out at average centralization
Reason, obtains matrix M, and average centralization processing is as follows：
If SAM values are not less than 0.01, step C3 is skipped, and make M=Xm×n。
Wherein, the step D specifically comprises the following steps：
Step D1, singular value decomposition is carried out to Metzler matrix：
M=S Λ DT (5)
In formula：S is m × r rank matrix with orthogonal rows S-1=ST；D is n × r rank row orthogonal matrixes D-1=DT, and Λ is r × r ranks
Diagonal matrix；
Step D2, C is maden=MTM, to CnMatrix carries out Eigenvalues Decomposition：
Cn=MTM=D Λ STSΛDT=D Λ2DT (6)
Since D is n × r rank row orthogonal matrixes, it is possible to as one group of row orthonormal basis, by CnMatrix projection
To D orthogonal basis, can obtain：
DTCn=(DTD)Λ2DT=Λ2DT (7)。
Wherein, step E, weight sequence geometry differentiates, comprises the following steps：
The weight sequence geometric distance method of discrimination is that the characteristic value for aligning the movie queen that trades carries out important factor and secondary
Factor discriminant classification, the step implementation process are as shown in Figure 2.
The sub-space feature value descending of acquisition is normalized to weight sequence in the step E is specially：The weight
Sequence geometric distance method of discrimination is that the characteristic value for aligning the movie queen that trades carries out important factor and secondary factor discriminant classification, described
Weight sequence geometric distance method of discrimination comprises the following steps：
Step E1, descending normalized is carried out to obtaining eigenvalue matrix Λ：
Wherein sort () represents descending arrangement computing, [λ1 λ2 … λr] it is characterized the number on value matrix Λ leading diagonals
Value, W are weight matrix, w1,w2…wrRepresent weighted value；
Step E2, weight sequence coordinate [A is established1,A2…Ar]=[(S1,w1),(S2,w2)…(Sr,wr)], by A1With Ar
Between line L1It is as follows as boundary straight line, the straight line formula：
L1:a1x+b1y+c1=0 (10)
Wherein (S1,S2…Sr) it is that any one group of ascending order arranges array, such as (1,2 ... r)；a1、b1、c1Represent linear equation L1
Coefficient；
Step E3, weight sequence of points [A is calculated1,A2…Ar] arrive L1DistanceIt is defined as follows：
Step E4, frontier distance is searchedMaximum：
Step E5, sequences of the p more than 4 is differentiated for first time, again by A2Line L between Ar2It is straight as border
Line, then
if floor(p/2)>2
L2:a2x+b2y+c2=0 (13)
Wherein floor () represents downward rounding operation；a2、b2、c2Represent linear equation L2Coefficient；
Differentiate sequences of the p less than or equal to 4 for first time, skip E5-E7；
Step E6, weight sequence of points [A is calculated2,A3…Ar] arrive L2Distance
Step E7, frontier distance is searchedMaximum, find out corresponding sequential value：
The present invention compared with prior art the advantages of be：
(1) present invention is in a creative way by the geometric distance of the weight sequence of points after spectrum matrix rectangular projection to boundary straight line
As pure material number judgment basis, this method solve most of algorithms at present needs rule of thumb artificially to set threshold parameter
The shortcomings that, relative to the degree of automation of conventional algorithm with higher.
(2) present invention spectrum matrix high to similarity can first carry out average centralization processing, strengthen to a certain extent
Whether the spectral information intensity of micro substance, can more accurately judge in compounding substances containing micro- compared to conventional algorithm
Quantity of material or similar substance.
(3) this method obtains weight sequence using characteristic value is simplified, and passes through a kind of geometric distance differentiation side of precise and high efficiency
Formula determines pure material number in spectral mixture, goes out covariance and correlation matrix, this method compared to conventional method demand
Calculation amount smaller, has the operational efficiency of higher.
(4) this method carries out original spectrum matrix unified rectangular projection, solves some algorithm for spectral band number
Increase the problem of pure material number estimated accuracy declines.
Brief description of the drawings
Fig. 1 is the flow chart of the method for the present invention；
Fig. 2 weight sequences Geometrically split differentiates implementation process schematic diagram；
Fig. 3 is the average spectrogram that experiment obtains sample various concentrations；
Fig. 4 is 100% melamine and 20% melamine right to examin weight sequence chart；
Fig. 5 is geometric distance differentiation and its close-up schematic view of 20% melamine weight sequence of points；
Fig. 6 proposition methods of the present invention and Hysime, VD Riming time of algorithm comparative result.
Embodiment
It is by the following examples and referring to the drawings, right for the object, technical solutions and advantages of the present invention are more clearly understood
The present invention is further described.
For the practical effect of the verification present invention, experiment is ground for the adulterated problem of feedstuff of getting worse
Study carefully.Illegal businessman produces the high illusion of feedstuff protein content for puppet, is often added in feedstuff such as dregs of beans
Illegal such as melamine etc. of organic matter containing high protein, since light yellow or yellowish-brown is presented in dregs of beans appearance, melamine is white
Lenticular, the color of dregs of beans can cover the presence of melamine, and the content of melamine adulterated is very low, therefore by apparent
The difficulty of feature recognition dregs of beans doping is larger.The present invention is detected its pure material species using micro-imaging spectral technique.
Feedstuff dregs of beans is chosen in experiment and 99% melamine, spectral detection use Perkin Elmer companies of the U.S.
400 near-infrared microscopical imaging spectrometers of FITR Microscope Spotlight.Experiment is equipped with melamine and pure dregs of beans
Mass ratio is 100%, 20%, 1.5% sample, and adulterated sample uses centrifugal blender mixed at high speed five minutes, is taken out a small amount of
Sample carries out tabletting with tablet press machine, wherein each concentration is equipped with two tabletting samples.Experiment pre- thermospectrometry apparatus 30 minutes first, will
Liquid nitrogen adds cooling detector and is used for improving instrumental sensitivity, and setting experiment parameter is：Reflectance spectrum, sky are gathered under imaging pattern
Between resolution ratio be 25 × 25 μm, spectral resolution 4cm-1, sweep speed 1.0cm/s, scanning wave band totally 761 wave bands, institute
There is experiment to take identical scan method, and carried out under identical experiment parameter.
Doping sample is measured using micro-imaging spectrum, experiment selects same region, measures melamine and mixes
The spectrum that concentration is 100%, 20%, 1.5% is closed, each concentration gradient sample pixel point number is 1024.
Step A, imaging spectrometer data is obtained：
Obtain imaging spectral absorbance data and obtain original spectrum matrix Y1024×761, wherein 1024 be the number of all pixels
Mesh, 761 be the wave band number of spectrum.Fig. 2 is the average spectrum under experiment acquisition sample 100%, 20%, 1.5% 3 concentration
Figure；
Step B, imaging spectral reflectivity data pre-processes
To reduce influence of the high-frequency noise to spectrum, by original spectrum matrix Y1024×7615 the disposal of gentle filter are carried out,
New spectrum matrix X is generated after processing1024×761。
Step C, spectrum matrix similarity differentiates
C1, the imaging samples spectral absorbance data X for asking for various concentrations respectively1024×761Averaged spectrum, such as Fig. 3 institutes
Show.
C2, by pretreated spectrum matrix Y1024×761A spectrum average spectrum progress SAM similarity is selected at random to sentence
Not：
C3, differentiate that 100%, 20%, 1.5% melamine sample SAM values of discovery are all higher than 0.01, skips average centralization
Processing procedure.
Step D, spectrum matrix rectangular projection
Step D1, respectively to the X of various concentrations1024×761Matrix carries out singular value decomposition.
Step D2, to the C of various concentrationsnMatrix carries out Eigenvalues Decomposition, obtains the different diagonal characteristic value Λ of master.
Step E, weight sequence geometry differentiates
Step E1, the processing of sort (Λ) descending is carried out to obtaining eigenvalue matrix Λ, and is normalized.
Step E2, weight sequence coordinate [A is established1,A2…Ar]=[(S1,w1),(S2,w2)…(Sr,wr)], divide shown in Fig. 4
Not Wei 100%, 20% melamine pattern detection weight sequence chart, calculate boundary straight line L1。
Step E3, weight sequence of points [A is calculated1,A2…Ar] arrive L1DistanceFig. 5 show 20% melamine
Sample is to straight line L1Geometric distance schematic diagram.
Step E4, frontier distance is searchedThe corresponding sequential value of maximum.
Step E5, four concentration mixing samples differentiate that p is respectively 1,2,2 respectively less than 4 for the first time, skip step E5-E7.Most
Whole testing result is as shown in the table.
The it is proposed method WSGSDM of the present invention of form 1 and Hysime, virtual dimension VD algorithm operation results
After form show method of the invention and Hysime algorithms more popular at present, virtual dimension algorithm VD relatively
As a result, in table it can be seen that method testing result proposed by the present invention be significantly better than Hysime, VD algorithm operation result, due to
Hysime algorithms need first to carry out noise estimation to each wave band of original spectrum, and noise is by measurement error in actually measuring
Produced with factors such as instrument errors, single noise type can not be appropriately determined its noise end member, thus Hysime is estimated
End member number contain unnecessary noise end member and variation information, it is impossible to accurately estimate end member number, VD algorithms can compare
End member number is accurately estimated, but end member number is easily excessively estimated when concentration is relatively low, Fig. 6 show the fortune of three kinds of algorithms
The row time, it can be seen from the figure that algorithm operational efficiency proposed by the present invention is apparently higher than other two kinds of algorithms.
Above example is provided just for the sake of the description purpose of the present invention, and is not intended to limit the scope of the present invention.This
The scope of invention is defined by the following claims.The various equivalent substitutions that do not depart from spirit and principles of the present invention and make and repair
Change, should all cover within the scope of the present invention.
Claims (5)
1. pure material quantity determines method in a kind of imaging spectral mixed pixel, it is characterised in that：This method step is as follows：
Step A, mixed pixel spectroscopic data is obtained by using imaging spectrometer；
Step B, original spectrum relatively low to signal-to-noise ratio is smoothly pre-processed；
Step C, similarity differentiation is carried out to obtaining pixel spectrum, the high pre-processed spectrum of similarity is carried out at average centralization
Reason；
Step D, by simplifying Eigenvalues Decomposition by spectrum matrix projection to orthogonal subspaces；
Step E, the orthogonal subspaces characteristic value descending of acquisition is normalized to weight sequence；
Step F, weight sequence two-dimensional coordinate system is established, by a kind of geometric distance automatic discrimination mode, determines that mixed pixel is pure
Amount of material p；
The sub-space feature value descending of acquisition is normalized to weight sequence in the step E is specially：The weight sequence
Geometric distance method of discrimination is that the characteristic value for aligning the movie queen that trades carries out important factor and secondary factor discriminant classification, the weight
Sequence geometric distance method of discrimination comprises the following steps：
Step E1, descending normalized is carried out to obtaining eigenvalue matrix Λ：
S.t.j=1,2,3；W=[w1,w2...wr]
Wherein sort () represents descending arrangement computing, [λ1 λ2 ... λr] it is characterized the numerical value on value matrix Λ leading diagonals, W
It is weight matrix, w1,w2... wrRepresent weighted value；
The step F comprises the following steps：
Step F1, weight sequence coordinate [A is established1,A2... Ar]=[(S1,w1),(S2,w2)... (Sr,wr)], by A1With Ar it
Between line L1It is as follows as boundary straight line, the straight line formula：
L1:a1x+b1y+c1=0 (10)
Wherein (S1,S2... Sr) it is that any one group of ascending order arranges array, such as (1,2... r)；a1、b1、c1Represent linear equation L1
Coefficient；
Step F2, weight sequence of points [A is calculated1,A2K Ar] arrive L1DistanceIt is defined as follows：
Step F3, frontier distance is searchedMaximum：
Step F4, sequences of the p more than 4 is differentiated for first time, again by A2Line L between Ar2As boundary straight line, then
if floor(p/2)>2
L2:a2x+b2y+c2=0 (13)
Wherein floor () represents downward rounding operation；a2、b2、c2Represent linear equation L2Coefficient；
Differentiate sequences of the p less than or equal to 4 for first time, skip F4 F6；
Step F5, weight sequence of points [A is calculated2,A3... Ar] arrive L2Distance：
Step F6, frontier distance is searchedMaximum, find out corresponding sequential value：
2. pure material quantity determines method in a kind of imaging spectral mixed pixel according to claim 1, it is characterised in that：
Obtaining mixed pixel spectroscopic data by using imaging spectrometer in the step A is specially：Obtain all pixel point spectrum
Reflectivity or absorbance obtain original spectrum matrix Ym×n, wherein m is the number of all pixels, and n is the wave band number of spectrum.
3. pure material quantity determines method in a kind of imaging spectral mixed pixel according to claim 2, it is characterised in that：
The original spectrum relatively low to signal-to-noise ratio, which smoothly pre-process, in the step B is specially：
To reduce influence of the high-frequency noise to spectrum, by original spectrum matrix Ym×nIn every spectrum y1×nRow pretreatment, uses
Moving Window the disposal of gentle filter：
WhereinXiIt is smooth rear and smooth preceding spectral vector y respectively1×nIn each wave band reflectivity or absorbance, WjIt is
Weight factor during moving window is smooth, to simplify smoothing process, takes Wj=1,2t+1 are smoothly to count, and t takes 1 or 2.
4. pure material quantity determines method in a kind of imaging spectral mixed pixel according to claim 2, it is characterised in that：
It is specially to obtaining the progress similarity differentiation of pixel spectrum in the step C：
C1, ask for pretreated imaging spectral matrix Xm×nAveraged spectrum
Wherein x [i] is i-th pretreated spectrum；
C2, by pretreated spectrum matrix Xm×nEvery spectrum and average spectrum is selected one by one to carry out spectral modeling to match SAM similar
Degree differentiates：
Wherein xilFor from Xm×nIn reflectance value or absorbance at i-th spectrum, l-th of wavelength for selecting；
If C3, Xm×nSimilarity SAM values are less than 0.01 between every spectrum of matrix, to Xm×nAverage centralization processing is carried out, is obtained
It is as follows to matrix M, average centralization processing：
If SAM values are not less than 0.01, step C3 is skipped, and make M=Xm×n。
5. pure material quantity determines method in a kind of imaging spectral mixed pixel according to claim 4, it is characterised in that：
The step D specifically comprises the following steps：
Step D1, singular value decomposition is carried out to Metzler matrix：
M=S Λ DT (5)
In formula：S is m × r rank matrix with orthogonal rows S-1=ST；D is n × r rank row orthogonal matrixes D-1=DT, and Λ is diagonal for r × r ranks
Battle array；
Step D2, C is maden=MTM, to CnMatrix carries out Eigenvalues Decomposition：
Cn=MTM=D Λ STSΛDT=D Λ2DT (6)
Since D is n × r rank row orthogonal matrixes, it is possible to as one group of row orthonormal basis, by CnMatrix projection to D just
Hand on base, can obtain：
DTCn=(DTD)Λ2DT=Λ2DT (7)。
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610149573.XA CN105761272B (en) | 2016-03-16 | 2016-03-16 | Pure material quantity determines method in a kind of imaging spectral mixed pixel |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610149573.XA CN105761272B (en) | 2016-03-16 | 2016-03-16 | Pure material quantity determines method in a kind of imaging spectral mixed pixel |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN105761272A CN105761272A (en) | 2016-07-13 |
| CN105761272B true CN105761272B (en) | 2018-05-15 |
Family
ID=56333297
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201610149573.XA Active CN105761272B (en) | 2016-03-16 | 2016-03-16 | Pure material quantity determines method in a kind of imaging spectral mixed pixel |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN105761272B (en) |
Families Citing this family (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106124050B (en) * | 2016-06-22 | 2017-12-15 | 电子科技大学 | A kind of high spectrum sub-pixel target Blind extracting detection method based on subspace |
| CN106228009B (en) * | 2016-07-20 | 2019-01-08 | 北京航空航天大学 | A kind of abundance estimation method of spectral mixture |
| CN107179294B (en) * | 2017-07-31 | 2019-09-03 | 西安文理学院 | A kind of near-infrared spectral analytical method based on subspace analysis |
| CN108776144B (en) * | 2018-06-12 | 2020-10-09 | 南京大学 | Remote sensing identification method for ocean oil spill emulsion based on group spectral characteristics |
| CN112419580A (en) * | 2020-10-27 | 2021-02-26 | 铁道警察学院 | Hyperspectral imaging-based banknote authenticity identification method |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101266296A (en) * | 2008-04-28 | 2008-09-17 | 北京航空航天大学 | High light spectrum small target detection method and apparatus |
| CN103955711A (en) * | 2014-05-20 | 2014-07-30 | 北京航空航天大学 | Mode recognition method in imaging spectrum object recognition analysis |
| CN105184302A (en) * | 2015-10-09 | 2015-12-23 | 北京航空航天大学 | High spectral image end member automatic extraction method |
-
2016
- 2016-03-16 CN CN201610149573.XA patent/CN105761272B/en active Active
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101266296A (en) * | 2008-04-28 | 2008-09-17 | 北京航空航天大学 | High light spectrum small target detection method and apparatus |
| CN103955711A (en) * | 2014-05-20 | 2014-07-30 | 北京航空航天大学 | Mode recognition method in imaging spectrum object recognition analysis |
| CN105184302A (en) * | 2015-10-09 | 2015-12-23 | 北京航空航天大学 | High spectral image end member automatic extraction method |
Non-Patent Citations (2)
| Title |
|---|
| 一种基于光谱奇异值检测的高光谱遥感小目标探测方法;李庆波 等;《光谱学与光谱分析》;20080831;第28卷(第8期);第1832-1836页 * |
| 基于正交投影的海面小目标检测技术;杨勇 等;《电子与信息学报》;20130131;第35卷(第1期);第24-28页 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN105761272A (en) | 2016-07-13 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN105761272B (en) | Pure material quantity determines method in a kind of imaging spectral mixed pixel | |
| Gavrilovic et al. | Blind color decomposition of histological images | |
| Zhang et al. | Hybrid detectors based on selective endmembers | |
| KR101913612B1 (en) | System and method for identifying complex tokens in an image | |
| Burazerović et al. | Detecting the adjacency effect in hyperspectral imagery with spectral unmixing techniques | |
| CN111044483A (en) | Method, system and medium for determining pigment in cream based on near infrared spectrum | |
| US8983183B2 (en) | Spatially varying log-chromaticity normals for use in an image process | |
| CN105954252A (en) | Qualitative detection method for illegal ingredient Sudan red in raw materials of feeds | |
| CN103729651A (en) | Hyperspectral remote sensing image classification method based on manifold neighbor measurement through local spectral angles | |
| CN115905881B (en) | Yellow pearl classification method and device, electronic equipment and storage medium | |
| CN107402192A (en) | A kind of method of quick analysis essence and flavoring agent quality stability | |
| CN106846295A (en) | The assay method and device of the soil organism | |
| US8934735B2 (en) | Oriented, spatio-spectral illumination constraints for use in an image progress | |
| US8428352B1 (en) | Post processing for improved generation of intrinsic images | |
| Begelman et al. | Blind decomposition of transmission light microscopic hyperspectral cube using sparse representation | |
| EP2776979B1 (en) | Post processing for improved generation of intrinsic images | |
| Dondi et al. | An interactive tool for speed up the analysis of UV images of Stradivari violins | |
| US8849018B2 (en) | Log-chromaticity clustering pipeline for use in an image process | |
| US8842907B2 (en) | Method for performing a multi-clustering merge for use in an image process | |
| CN106228009B (en) | A kind of abundance estimation method of spectral mixture | |
| CN113125376A (en) | Method, device, equipment and medium for identifying age limit of dried orange peel | |
| CN117288692B (en) | Method for detecting tannin content in brewing grains | |
| CN110274888B (en) | Smart phone based eriocheir sinensis internal edible quality nondestructive testing method | |
| US8811732B2 (en) | Weighted entropy minimization for optimizing a log-chromaticity normal for use in an image process | |
| Soori et al. | Chemical sensing of washing powder by multispectral imaging |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |