CN106777845A - The method that sensitive parameter builds wheat leaf blade powdery mildew early monitoring model is extracted based on subwindow rearrangement method - Google Patents
The method that sensitive parameter builds wheat leaf blade powdery mildew early monitoring model is extracted based on subwindow rearrangement method Download PDFInfo
- Publication number
- CN106777845A CN106777845A CN201710172841.4A CN201710172841A CN106777845A CN 106777845 A CN106777845 A CN 106777845A CN 201710172841 A CN201710172841 A CN 201710172841A CN 106777845 A CN106777845 A CN 106777845A
- Authority
- CN
- China
- Prior art keywords
- index
- class
- model
- wheat
- powdery mildew
- 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.)
- Pending
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/20—Design optimisation, verification or simulation
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
- G01N21/00—Investigating or analysing materials by the use of optical means, i.e. using sub-millimetre waves, infrared, visible or ultraviolet light
- G01N21/17—Systems in which incident light is modified in accordance with the properties of the material investigated
- G01N21/25—Colour; Spectral properties, i.e. comparison of effect of material on the light at two or more different wavelengths or wavelength bands
Abstract
The method that sensitive parameter builds wheat leaf blade powdery mildew early monitoring model is extracted the invention provides based on subwindow rearrangement method (SPA), is comprised the following steps：1) hyper spectral reflectance of susceptible wheat leaf blade is obtained；2) sensitive band is extracted from the original wave band of hyper spectral reflectance using SPA algorithms；3) spectral index related to disease possibility in the existing research of selection, sensitivity spectrum index is extracted using SPA algorithms from the spectral index；4) offset minimum binary Fisher face is utilized, using the sensitive band or the sensitivity spectrum index as a point input variable, wheat powdery mildew early monitoring model is built；5) the wheat powdery mildew early monitoring model is tested with two sorting algorithms, and based on independent susceptible kind with staying cross-verification method evaluation model performance.To sum up, based on SPA extract wheat powdery mildew sensitivity spectrum feature is accurate and wave band is few, the simple accuracy high stability of wheat powdery mildew monitoring model that builds is good.
Description
Technical field
The present invention relates to plant disease early prediction technical field, and in particular to extract sensitive ginseng based on subwindow rearrangement method
The method that number builds wheat leaf blade powdery mildew early monitoring model.
Background technology
Powdery mildew (Blumeriagraminis Speer) is a kind of worldwide Major Diseases in Wheat Production, is also influence
One of Major Diseases of wheat yield.It is the anti-white powder of wheat that EARLY RECOGNITION, fast monitored, quantitative assessment are carried out to wheat powdery mildew
The core key technology of sick precise breeding, accurate dispenser, ecological safety and loss appraisal etc..Forefathers' monitoring crop disease is mainly led to
Cross destructive sampling and measuring or field investigation diseased plant calculates the incidence of disease, severity or disease index, often time and effort consuming, efficiency
Lowly, subjectivity is strong, repeatable poor；And after wheat infection powdery mildew, its internal physiological activity and mode of appearance can show
Show abnormal symptom (Xu Bingliang etc., 2011).The photosynthetic capacity of wheat weakens, and plant strain growth is limited, the pigment of blade interior
Content and moisture are reduced, and blade is gradually turned to be yellow withered.This for based on spectrum monitoring identification powdery mildew provide well
Physiological condition.Therefore early in 1989, Denmark scholar (Lorenzen et al.1989) has found that 400-1100nm light can be utilized
Spectrum information recognizes wheat powdery mildew because the spectral reflectivity in the area there were significant differences compared with check variety；Later, people
The spectral signature based on first derivative is started with to recognize disease (Miller et al.1991；Baret et al.1994)；
Switzerland scholar Hamed Hamid Muhammed (2003；2005) hyperspectral information based on 360-900nm, using feature-
Vector-based (FVBA) method for feature analysis calculates the relation of weight coefficient, quantitative analysis disease severity and weight coefficient,
So that it is determined that the sensitive characteristic wave bands scope of disease.Cedric Bravo (2003) have screened 4 ripples between 460-900nm
Section, the severity of quantitative inversion stripe rust evil after normalization, later problem group membership Dimitrios Moshou (2004) profit
Disease is distinguished with neural network, the disease monitoring degree of accuracy is further increased.Moshou (2005) merges EO-1 hyperion and fluorescence
Image-forming information, and disease is more accurately identified based on self-organizing map.It can be seen that forefathers are based primarily upon spectrum primary reflection rate, single order
Derivative, vegetation index and severity build quantitative relationship, in extracting sensitivity spectrum (region) and be generally according to correlation etc.
The later stage less sensitive parameter for clearly extracting instruction early disease feature simultaneously builds corresponding model.
With the development of machine learning method and Chemical Measurement, scientists propose some new Variable Selections and use
It is representational such as random forest (RandomForest) in the extraction of feature.Although based on rearrangement in random forests algorithm
Variable importance evaluation method thinks very highly, but random forest each variable only by random rearrangement once, therefore the value for obtaining
It is not reproducible；The presence of Random Forest model disturbance variable has shielding effect, may filter out optimal variables set
Come.
The content of the invention
In view of this, it is an object of the invention to provide based on subwindow rearrangement method (Subwindow
PermutationAnalysis, SPA) extract the method that sensitive parameter builds wheat leaf blade powdery mildew early monitoring model, the party
Method makes full use of the cooperative effect between wheat powdery mildew characteristic variable by using subwindow rearrangement method, makes the monitoring side
Method has the stability and accuracy of more excellent statistical significance and Geng Gao.
In order to realize foregoing invention purpose, the present invention provides following technical scheme：
Wheat leaf blade powdery mildew early monitoring mould is built the invention provides sensitive parameter is extracted based on subwindow rearrangement method
The method of type, comprises the following steps：
1) hyper spectral reflectance of susceptible wheat leaf blade is obtained；
2) sensitive band is extracted from the original wave band of hyper spectral reflectance using SPA algorithms；
3) spectral index related to disease possibility in the existing research of selection, using SPA algorithms from the spectral index
Extract sensitivity spectrum index；
4) utilize offset minimum binary-Fisher face, using the sensitive band or the sensitivity spectrum index as
Divide input variable, build wheat powdery mildew early monitoring model；
5) the wheat powdery mildew early monitoring model is tested with two sorting algorithms, and based on independent susceptible kind
Showed with a cross-verification method evaluation model is stayed；
The step 2) and step 3) between there is no the limitation of time sequencing.
Preferably, the step 1) in susceptible wheat leaf blade according to severity divide sample grade.
Preferably, the step 1) in severity (SL) be on disease leaf scab subiculum covering blade area account for blade
The ratio of the gross area；
The sample grade is 9 grades, respectively 0%, 1%, 5%, 10%, 20%, 40%, 60%, 80% and 100%
Wheat leaf blade.
Preferably, the step 2) in spectral signature include red border area domain, be particularly preferred as 400~1000nm.
Preferably, the step 3) in spectral index include 13 kinds of EO-1 hyperion vegetation indexs and 13 kinds of differential smoothing indexes.
Preferably, the step 3) in spectral index feature include that normalization color ratio index, red side vegetation stress refer to
First differential summation, Chlorophyll absorption ratio index, physiological reflex index, anthocyanidin reflection index, conversion in the range of number, yellow side
Chlorophyll absorption index, indigo plant in the range of first differential summation, it is red while in the range of single order in the range of first differential summation and blue side
The ratio of first differential summation in the range of when the normalized value of differential summation and the red first differential summation in the range of are with indigo plant.
Preferably, the step 4) in two sorting algorithms to be categorized as with healthy leaves being positive class, be negative with ill blade
Class；Actual healthy leaves and healthy leaves is predicted to for real class, if actual healthy leaves is to bear class and be predicted to just
Class is false positive class；Actual health blade is negative class and to be predicted to negative class be very negative class, if actual health blade is positive class be predicted
It is false negative class into negative class.
Preferably, the step 5) in evaluation index include that model susceptibility, model specificity, model recipient operation are bent
Area and model overall classification accuracy under line.
Preferably, the computational methods of the model susceptibility are shown in formula I；SusceptibilityWherein TP is
Real class；TN is very negative class.
Preferably, the computational methods of the model specificity are as shown in formula II；SpecificityWherein TN
It is very negative class；FP is false positive class.
Preferably, recipient's operation that area computation method is drawn according to grader under the model receiver operating curve
Area value under the receiver operating curve of curve negotiating integral and calculating.
Preferably, the computational methods of the model overall classification accuracy are as shown in formula III；Nicety of gradingWherein TP is real class；TN is very negative class；FN is false negative class；FP is false positive class；TN is very negative class.
What the present invention was provided extracts sensitive parameter structure wheat leaf blade powdery mildew early monitoring mould based on subwindow rearrangement method
The method of type, comprises the following steps：1) hyper spectral reflectance of susceptible wheat leaf blade is obtained；2) using SPA algorithms from EO-1 hyperion
Sensitive band is extracted in the original wave band of reflectivity；3) spectral index related to disease possibility in the existing research of selection, utilizes
SPA algorithms extract sensitivity spectrum index from the spectral index；4) offset minimum binary-Fisher face is utilized, by institute
Sensitive band or the sensitivity spectrum index are stated as a point input variable, wheat powdery mildew early monitoring model is built；5) two are used
Sorting algorithm is tested to the wheat powdery mildew early monitoring model, and based on independent susceptible kind with staying a cross-verification
Method evaluation model is showed；The step 2) and step 3) between there is no the limitation of time sequencing.The method that the present invention is provided passes through
Using subwindow rearrangement method, the cooperative effect between wheat powdery mildew characteristic variable can be made full use of, make the monitoring method
Stability with more excellent statistical significance and Geng Gao.The present invention provide Method Modeling and inspection precision can reach 85% with
On；Based on SPA extract spectral signature number is few, model is simply accurate, and the cross-verification of model and independent kind inspection
Effect is all very stable.
Figure of description
Fig. 1 is the wheat leaf blade picture of different severities in embodiment 1；
Fig. 2 is different stage wheat leaf blade severity and the correlation analysis of spectral reflectivity of catching an illness in embodiment 1；
Fig. 3 is the correlation analysis of the wheat leaf blade severity with spectral reflectivity of different cultivars in embodiment 1；
Fig. 4 is the COSS values of SPA methods calculating characteristic wave bands in embodiment 1；
Fig. 5 be embodiment 1 in use influence of the different number of characteristic wave bands to model PLS-LDA precision；
Fig. 6 is the corresponding COSS values of SPA methods each spectral index of calculating in embodiment 1
Fig. 7 sets up the PLS-LDA models of wheat leaf blade health status to use different number of spectral index in embodiment 1
Overall accuracy.
Specific embodiment
Extracted the invention provides based on subwindow rearrangement method (Subwindow Permutation Analysis, SPA)
The method that sensitive parameter builds wheat leaf blade powdery mildew early monitoring model, comprises the following steps：
1) hyper spectral reflectance of susceptible wheat leaf blade is obtained；
2) sensitive band is extracted from the original wave band of hyper spectral reflectance using SPA algorithms；
3) spectral index related to disease possibility in the existing research of selection, using SPA algorithms from the spectral index
Extract sensitivity spectrum index；
4) utilize offset minimum binary-Fisher face, using the sensitive band or the sensitivity spectrum index as
Divide input variable, build wheat powdery mildew early monitoring model；
5) the wheat powdery mildew early monitoring model is tested with two sorting algorithms, and based on independent susceptible kind
Showed with a cross-verification method evaluation model is stayed；
The step 2) and step 3) between there is no the limitation of time sequencing.
The present invention obtains the hyper spectral reflectance of susceptible wheat leaf blade.
In embodiments of the invention, the wheat breed includes susceptible type wheat breed " life selects No. six " (Vh) and middle sense type
Wheat breed " raising spoke wheat No. 4 " (Vm).
The present invention in order to obtain susceptible wheat plant, preferably by powdery mildew fungi in wheat jointing post incoulation to wheat
On, 8 cells of a row on the leeward induce the upwind side selection of the another side of row in experimental plot as row region of disease is induced
4 cells, surrounding is fenced up with plastic sheeting carries out isolation processing, used as normal control area.
In the present invention, the susceptible wheat leaf blade preferably divides sample grade according to severity.It is described to be in a bad way
Degree is the ratio that scab subiculum covering blade area accounts for the blade gross area on disease leaf.In the present invention, the sample grade preferably 9
Level, preferably specific respectively 0%, 1%, 5%, 10%, 20%, 40%, 60%, 80% and 100% wheat leaf blade.The present invention
In, the severity is edited according to Nationwide Agricultural Technology Promotion Service Center《Staple crops pestforecasting skill
Art specification application manual》Computational Methods are obtained.
In the present invention, method of the acquisition with defect information hyper spectral reflectance is preferably uses spectrophotometer wheat leaf blade
Spectral reflectivity.The present invention is not particularly limited to the spectrometer, using spectrometer well-known to those skilled in the art
.In the embodiment of the present invention, produced using the U.S. Analytical Spectral Device (ASD) company
FieldSpecPro FR2500 spectrometers.The parameter of the measure includes wavelength band, sampling interval and spectral resolution.Institute
When stating wave band for 350~1050nm, preferred 1.4nm is spaced, spectral resolution is preferably 3nm；The wave band be 1050~
During 2500nm, spectrum sample is preferably spaced 2nm, and spectral resolution is preferably 10nm.The position of the Blade measuring is preferably
In blade tip, leaf and phyllopodium.The position measures three times, the hyper spectral reflectance averaged as the blade.Obtain EO-1 hyperion
After reflectivity, the present invention is extracted and severity using subwindow rearrangement method from the vegetation index of original wave band or correlation
Significantly correlated sensitive band or spectral index.
In the present invention, wheat leaf blade original spectrum reflectivity and blade severity are preferably carried out into correlation research point
Analysis, obtains sensitive band scope of the powdery mildew in wheat leaf blade susceptible early stage.The correlation analysis be specifically respectively by 400~
Each wave band in the range of 1000nm is used as a variable Xi(i=1,2,3 ..., 601), using scab ratio as corresponding
Variable Y, calculates each variable XiLinear dependence and variable Y between, draws corresponding coefficient correlation and P values.Result of calculation
Calculated under Matlab environment, coefficient correlation is Pearson came (Pearson) type.The correlation analysis are preferably sick with blade
The significantly correlated sensitive band of feelings severity, it is described it is significantly correlated be P≤0.01.
In the present invention, the scope of the sensitive band preferably includes red light region and near infrared region.In middle sense wheat product
In kind, the sensitive band region is preferably 675~700nm and 711~1000nm.In susceptible wheat breed, the sensitivity
Band is preferably 497~518nm and 578~1000nm.
In the present invention, the method for the sensitive band of the extraction determines 400 preferably by subwindow rearrangement method (SPA)~
The statistical distribution of predicated error, determines whether the wave band has information before and after each band po sition is reset in 1000nm full band ranges
Variable, and acquisition COSS values are calculated, COSS values are higher, and the variable is more meaningful.By COSS values to the importance of each wave band
It is ranked up, selection COSS values wave band in the top is sensitive band.In the present invention, 20 wave band before the COSS values ranking
Respectively：554、569、550、739、747、749、755、814、631、757、433、428、732、740、746、786、761、
596th, 750 and 768nm, wherein the selected wave band number in red border area domain is at most, this is red with the preliminary stage that wheat powdery mildew infects
Border area domain is relevant to the correlation highest of associated biochemical component caused by disease and its change.Therefore it is described by SPA from original
The sensitive band extracted in wave band preferably includes red border area domain.
The present invention using the method with sensitive band is extracted above, using subwindow rearrangement method, from may be related to disease
Spectral index in extract sensitivity spectrum index.Specifically, record of the present invention according to prior art, screens related to disease
Spectral index；Calculate the spectral index relational with wheat leaf blade severity, extraction obtains aobvious with severity
Write related sensitivity spectrum index.
In the present invention, the spectral index includes 13 kinds of EO-1 hyperion vegetation indexs and 13 kinds of differential smoothing indexes.The light
Spectrum index has been used for chlorophyll, moisture, nitrogen, the photosynthetic efficiency of monitoring crop, and after disease occurs, these parameters easily become
Change.Mainly include EO-1 hyperion vegetation index and differential smoothing index.The hyperspectral index feature preferably includes narrow-band normalizing
Change vegetation index (NBNDVI), nitrogen reflection index (NRI), triangle vegetation index (TVI), photochemical reflectance index (PRI), physiology
Reflection index (PhRI), Chlorophyll absorption ratio index (CARI), conversion Chlorophyll absorption index (TCARI), modified leaf are green
Plain absorption index (MCARI), red side vegetation coerce index (RVSI), vegetation senescence index (PSRI), anthocyanidin reflection index
(ARI), the insensitive vegetation index of structure (SIPI) and normalization color ratio index (NPCI).The selected differential smoothing refers to
Number preferably include it is blue in scope maximum first differential value (Db), it is blue while in the range of first differential summation (SDb), yellow side scope it is maximum
First differential value (Dy), Huang in the range of first differential summation (SDy), it is red while scope maximum first differential value (Dr), red side model
The ratio of first differential summation in the range of when enclosing interior first differential summation (SDr), the red first differential summation in the range of with indigo plant
(SDr/SDb) ratio (SDr/SDy), the Huang Bian of first differential summation in the range of when, the red first differential summation in the range of is with Huang
In the range of first differential summation with indigo plant in the range of the ratio (SDy/SDb) of first differential summation, it is red while in the range of first differential
Summation with indigo plant in the range of normalized value (SDr-SDb)/(SDr+SDb) of first differential summation, it is red while in the range of first differential
First differential in the range of when summation is with yellow normalized value (SDr-SDy)/(SDr+SDy) of first differential summation in the range of, Huang
Normalized value (SDy-SDb)/(SDy+SDb), the Red edge position (REP) of first differential summation in the range of summation and blue side.
In the present invention, the spectral index includes following step with the computational methods of the correlation of wheat leaf blade severity
Suddenly：Using 13 kinds of described hyperspectral indexes and 13 kinds of differential smoothing indexes totally 26 kinds of indexes as variable Xj(j=1,2,
3 ..., 26), using scab ratio as corresponding variable Y, calculate each variable XjLinear dependence and variable Y between, obtains
Go out corresponding coefficient correlation and P values.The statistical function that result of calculation is carried under Matlab environment using software is calculated, related
Coefficient is Pearson came Pearson types.In the present invention, it is described it is significantly correlated be P value≤0.01.
In the present invention, the extraction preferably calculates the corresponding COSS values of each spectral index using SPA methods, according to COSS values
Each spectral index is ranked up, spectral index feature is extracted from COSS values spectral index in the top.It is described to pass through
The spectral signature that subwindow rearrangement method is extracted from spectral index includes normalization color ratio index (NPCI), the red side vegetation side of body
First differential summation (SDy), Chlorophyll absorption ratio index (CARI), physiological reflex refer in the range of urgent index (RVSI), yellow side
Number (PhRI), anthocyanidin reflection index (ARI), conversion Chlorophyll absorption index (TCARI), first differential summation in the range of blue side
(SDb) normalized value (SDr-SDb)/(SDr+ of first differential summation in the range of when, the red first differential summation in the range of is with indigo plant
SDb) and the red first differential summation in the range of with it is blue while in the range of first differential summation ratio (SDr/SDb).
After obtaining sensitive band or sensitivity spectrum index, the present invention utilizes offset minimum binary-Fisher face, by profit
The sensitive band or sensitivity spectrum index extracted with SPA build wheat powdery mildew in early days as the input variable of the analytic approach
Monitoring model.
In the present invention, the mechanism for building wheat powdery mildew early monitoring model is as follows：Assuming that sample X is n × p squares
Battle array, n represents the sample size in this research, and p to be represented and reset the characteristic variable that analysis method (SPA) is selected using subwindow；Sample
This classification is recorded in the vector Y of n × 1, and Y refers to blade health classification, i.e. infected leaves defines the positive class in position, in Matab codes
Represented with+1, healthy leaves is defined as negative class, is represented with -1 in Matlab codes.By the use of algorithm PLS-LDA as grader, obtain
To the mathematical relationship between X and Y, the not clear and definite functional relations of X and Y, therefore address is model.
The evaluation index of grader has various, and the calculating of these indexs mostlys come from confusion matrix table 1.Confusion matrix is used
In match stop result and the real information of reality, concrete class is represented in matrix per a line, each row represent prediction classification.
The confusion matrix of table 1
In the present invention, the classification of two sorting algorithm is positive class preferably with healthy leaves, is negative class with infected leaves；
If actual healthy leaves and being predicted to healthy leaves for real class, if actual healthy leaves is to bear class and be predicted to just
Class is false positive class；If actual health blade is negative class and to be predicted to negative class be very negative class, if actual health blade is positive class quilt
It is false negative class to predict into negative class.
After obtaining monitoring model, the present invention is with two sorting algorithms to the step 4) wheat powdery mildew that obtains identifies model
Test.
In the present invention, the evaluation index preferably includes recipient's operation of model susceptibility, model specificity, model
(ROC) TG-AUC (Area under the Curve ofROC, AUC) and model overall classification accuracy.
In the present invention, the computational methods of the model susceptibility are shown in formula I；SusceptibilityWherein TP
It is real class；TN is very negative class.It in grader by the actual sample predictions for positive classification is positive classification that the susceptibility of model refers to
Ability, numerical value can represent that model correctly judges the probability of positive classification between 0-100%
In the present invention, the computational methods of the model specificity are as shown in formula II；SpecificityWherein
TN is very negative class；FP is false positive class.It in grader by the actual sample predictions for negative classification is negative class that the specificity of model refers to
Other ability, numerical value can represent that model correctly judges the probability of negative classification between 0-100%
In the present invention, the ROC that the ROC curve that the model AUC computational methods are drawn according to grader passes through integral and calculating
TG-AUC value.The AUC of model refers to the part size below ROC curve, and ideal sort model its AUC is
1, between 0.5-1.0, larger AUC represents disaggregated model and possesses preferable performance usual its value.Using AUC as commenting
Price card is accurate because many times ROC curve can not clearly illustrate the effect of which grader more preferably, and as a number
Value, corresponding A UC bigger grader effect is more preferable；
In the present invention, the computational methods of the model overall classification accuracy are as shown in formula III；Nicety of gradingWherein TP is real class；TN is very negative class；FN is false negative class；FP is false positive class；TN is very negative class.
The overall classification accuracy of model represents the overall differentiation accuracy rate of the model, represents decision-making ability of the grader to whole sample.
In the present invention, the calculating is preferably carried out under MATLAB.
With staying a cross-verification and susceptible infections wheat breed evaluation model
It is exactly to utilize original data set to stay a cross-verification, and each sample is once checked, it is assumed that sample number is n, then
Be input to for X1 in the model of structure and obtain a classification for prediction as test set by first sample set X1, so carries out down
Go, it is at this moment corresponding with the classification calculating of prediction using the actual classification Y of sample until all n sample standard deviations are predicted once
Evaluation index and the model is evaluated.Equally, tested with susceptible species data, refer to by susceptible species data collection
X input build model in obtain one prediction classification, and using susceptible kind concrete class Y calculate it is corresponding evaluation refer to
Mark and the model is evaluated.
Sensitive parameter structure wheat leaf blade is extracted based on subwindow rearrangement method with reference to what embodiment was provided the present invention
The detailed description of the method for powdery mildew early monitoring model, but they can not be interpreted as the limit to the scope of the present invention
It is fixed.
Embodiment 1
In November, 2014~2015 year June year is in Agricultural University Of Nanjing's decorated archway teaching and scientific research base (118 ° of 15 ' E, 32 ° 1 '
N) carry out, experiment kind is susceptible type wheat breed " life selects No. six " (Vh) and middle sense type wheat breed " raising spoke wheat No. 4 " (Vm),
Plot area 6m2(3m × 2m), 3 repetitions.All cell fertilisings, management condition are identical, and nitrogenous fertilizer, phosphate fertilizer, potash fertilizer are respectively urine
Element, calcium superphosphate, potassium chloride.Experimental plot powdery mildew fungi goes in jointing post incoulation, in the east 8 cells of a row as induction,
In experimental plot west, upwind side selects 4 cells, and surrounding is fenced up with plastic sheeting carries out isolation processing, used as normal control
Area.
Wheat leaf blade hyper spectral reflectance is determined using the life of the U.S. Analytical Spectral Device (ASD) company
FieldSpecPro FR2500 spectrometers (350~2500nm of wavelength band, 25 ° of the angle of visual field, wherein 350~1050nm light of product
Spectrum sampling interval 1.4nm, spectral resolution 3nm；At intervals of 2nm, spectral resolution is 10nm to 1050~2500nm spectrum Bian samples,
Such as table 2).The wheat plant blade of test is cut in loading large size valve bag respectively by various position leaves, is put into and is had ice cube
Laboratory is transported in refrigeration back in bubble chamber.The single leaf spectrum blade for being carried with instrument indoors is pressed from both sides and determines every spectral reflectance of blade
Rate.Single leaf spectrum blade folder carries active light source, can be determined under closed environment, and stable operation, measurement error is small.Every blade
In measurement blade tip, leaf and three positions of phyllopodium, each position measures three times, the spectra measurement averaged as the blade.
Spectral response feature according to wheat powdery mildew, this research with reference to all kinds of spectral indexes and crop disease monitoring ground
Applicable cases in studying carefully, have selected 13 kinds of EO-1 hyperion vegetation indexs (table 2) and 13 kinds of differential smoothing index (table 3) (Gong et
Al., 2002), maximum first differential value, first differential summation, ratio and normalized value in the range of mainly red, yellow, blue side.
Table 2 has selected 13 kinds of EO-1 hyperion vegetation indexs
The selected differential smoothing index of 3 13 kinds of table
The structure of wheat leaf blade health status monitoring model and evaluation
Model construction：Offset minimum binary-the linear discriminant analysis (PLS-LDA) proposed using Barker.Before powdery mildew
In the health status monitoring research of phase wheat leaf blade, blade is divided into healthy leaves and the class of infected leaves two, healthy leaves refers to
There is not any scab in blade surface, referred to as positive class；Infected leaves refers to that blade surface just starts scab occur and the state of an illness is tight
Severe (scab ratio below 5%<7.5%) class, is referred to as born.
Model testing and evaluation：
Using two sorting algorithm testing models, and overall merit is come with susceptibility, specificity, AUC and overall classification accuracy
The performance of model.Four kinds of situations that two sorting technique expression models of application are predicted herein：If example is healthy leaves (negative class)
And healthy leaves (negative class) is predicted to, class (True Positive, TP) is referred to as really born, if example is to bear class and be predicted
Into positive class, referred to as false positive class (False Positive, FP).Correspondingly, if example is negative class and is predicted to positive class, claim
It is real class (True Negative, TN), if example is negative class is predicted to positive class, referred to as false positive class (False
Negative,FN)。
Data separate and analysis：
Herein using middle sense kind " raising spoke wheat No. 4 " data as test data set, sample size n sample range=168, wherein susceptible leaf
Piece (positive class) have 77, healthy leaves (negative class) have 91；Using susceptible kind " life selects No. 6 " data as individual authentication data
Collection, blade sample size n sample range=167, wherein infected leaves (positive class) have 81, healthy leaves (negative class) have 86, specific distribution
As shown in table 4.
The distributed number of each scab ratio of the susceptible early stage wheat leaf blade of the powdery mildew of table 4
As a result
The wheat leaf blade original spectrum reflectivity of two kinds is carried out with blade severity in being tested with 2014-2015
Correlation research is analyzed, and explores sensitive band scope of the powdery mildew in wheat leaf blade different susceptible periods.Figure it is seen that
Early stage is infected in powdery mildew, red light region and near infrared region are more first responded to severity, susceptible kind " life selects No. 6 "
Spectra of The Leaves is significantly correlated (P=0.01) with severity at 497~518nm and 578~1000nm, and middle sense kind " is raised
The Spectra of The Leaves of spoke wheat No. 4 " is significantly correlated (P=0.01) with severity at 675~700nm and 711~1000nm, says
There are sensitivity differences in response of the wheat breed of bright different resistances to spectrum.
Wheat sense powdery mildew monitoring feature based on spectral index is extracted
Spectral response feature according to wheat powdery mildew, this research with reference to all kinds of spectral indexes and crop disease monitoring ground
Applicable cases in studying carefully, selected 13 kinds of EO-1 hyperion vegetation indexs and 13 kinds of differential smoothing indexes, and quantitative analysis these refer to
Number and the relation (Fig. 3) of blade severity, as a result show：Except NRI and (SDr-SDy)/(SDr+SDy) be in a bad way
Correlation between degree significantly beyond (P-Value>0.01), other 24 kinds of spectral indexes are in a bad way with wheat leaf blade
Spend significantly correlated (P-Value≤0.01).
It is identical that the correlation height of different spectral indexes and severity shows trend between different cultivars, and same
Plant spectral index and show different between different cultivars from the correlation of severity, i.e., feel in the spectral index ratio of susceptible kind
The spectral index of kind is more sensitive to the blade severity of wheat powdery mildew.If vegetation index CARI is to susceptible kind
The wheat powdery mildew severity correlation of " life selects No. 6 " is (R=0.73, P=0.000) higher, and centering sense kind " raises spoke
The wheat powdery mildew severity correlation of wheat No. 4 " is relatively low (R=0.43, P=0.000), thus, it is believed that different resistances
The powdery mildew state of an illness information of wheat breed has differences to the sensitiveness of same spectral index.
1) the sensitive features wave band of wheat powdery mildew is extracted based on SPA methods
Each band po sition in 400~1000nm full band ranges is calculated herein by SPA subwindow rearrangement methods to reset
The statistical distribution of front and rear predicated error, determines whether the wave band is have information variable, and by COSS values to the weight of each wave band
The property wanted is ranked up (Fig. 4).20 wave band is respectively before COSS value rankings：554、569、550、739、747、749、755、814、
631st, 757,433,428,732,740,746,786,761,596,750 and 768nm, wherein the selected wave band number in red border area domain
At most, this is relevant to the correlation highest of the state of an illness with the red border area domain of preliminary stage that wheat powdery mildew infects.
Using totally according to the model overall accuracy in the case of two kinds of training and cross validation, to further determine that different numbers
Influence (Fig. 5) of the characteristic wave bands to model PLS-LDA precision.Result shows, totally according under training mode, precision curve totality
Dullness, overall accuracy is maintained within a certain range and fuctuation within a narrow range after wave band number reaches 5, and precision reaches most when wave band number is 16
Greatly；And the then precision highest when wave band number reaches 12 under cross validation pattern, therefore select what preceding 12 wave bands were extracted as SPA
Initial characteristicses wave band collection S1, S1={ 428,433,550,554,569,631,739,747,749,755,757 and 814nm }.
In addition, this 12 characteristic wave bands be widely distributed in 400~500nm, 501~600nm, 601~700nm, 701~
In each subspaces of 800nm and 801~900nm, with reference to the method for Subspace partition, one is selected according to COSS values in each subspace
The wave band of individual highest scoring further constitutes one group of characteristic wave bands collection S2, and wheat is diagnosed in early stage to test this group of characteristic wave bands collection
Effect during blade health status, S2={ 433,554,631,739 and 814nm }.
2) the sensitivity spectrum index of wheat sense powdery mildew is extracted based on SPA
Using above-mentioned same method, the corresponding COSS values (Fig. 6) of each spectral index, Ran Houcong are calculated based on SPA methods
Optimal spectral index combination is extracted in existing spectral index, the monitoring model of wheat leaf blade health status is set up.Result shows
Show：10 spectral index before ranking：NPCI、RVSI、SDy、CARI、PhRI、ARI、TCARI、SDb、(SDr-SDb)/(SDr+
) and SDr/SDb SDb
It is same according to training with the case of two kinds of cross validation, to evaluate and setting up small using different number of spectral index totally
The overall accuracy (Fig. 7) of the PLS-LDA models of wheat blade health status.Result shows, totally according under training mode, spectral index
The monitoring accuracy that number reaches model after 4 is maintained within a certain range and fuctuation within a narrow range, and the precision under cross validation pattern becomes
Change and be similar to according under training mode with total, therefore, the maximum preceding 4 kinds of spectral indexes of selection COSS values are to wheat leaf blade health status
Diagnosed, respectively NPCI (normalization color ratio index), RVSI (red to coerce index in vegetation), (differential is total while yellow for SDy
With) and CARI (chlorophyll absorption ratio index).
3) structure and evaluation comparison of wheat leaf blade health status monitoring model
Using the early time data collection of two kinds of wheat leaf blades of different resistant varieties as research object, extracted based on the above
Spectral signature as input variable, wheat leaf blade health status monitoring model is built based on PLS-LDA methods respectively, using staying
The early stage of one cross-verification (Leave-One-Out Cross Validation, LOOCV) and susceptible wheat breed " life selects No. 6 "
Data are tested (Independent Variety Validation, IVV) to model, testing result such as table 5 and the institute of table 6
Show.
Table 5
Table 6
Be can be seen that from table 5 and table 6：
(1) when the sensitive band extracted based on SPA carries out discriminant analysis to wheat leaf blade health status, model and check
Precision can reach more than 90%, but monitoring model is very poor to the test effect of independent kind, illustrate that sensitive band is only suitable for using
In the diagnosis of single variety wheat leaf blade health status, and mode input variables number up to 316, excessive variable increased
The complexity of model.
(2) based on SPA extract sensitivity spectrum feature is accurate and wave band is few, the powdery mildew monitoring model that builds is simple, it is accurate
It is really and stable.
As seen from the above embodiment, it is anti-from spectrum respectively that what the present invention was provided resets analytical technology (SPA) by subwindow
Diagnosis identification monitoring of the extraction sensitive features to wheat leaf blade health status in rate, spectral index is penetrated to be analyzed.Using SPA
The characteristic wave bands that technology is extracted are only 12, the classification essence of the wheat leaf blade health status monitoring model built based on PLS-LDA
It is 85.12% to spend, and cross-verification precision is 84.52%, and the testing accuracy of independent susceptible kind is 84.43%；Equally, from spectrum
During index level analysis, the nicety of grading of the PLS-LDA monitoring models that 4 spectral indexes extracted using SPA technologies are built is
82.14%, cross-verification precision is 80.95%, and the testing accuracy of independent susceptible kind is 85.63%, and this also show SPA skills
The characteristic wave bands of art selection and its model accuracy of structure are good, and stability is high.
The above is only the preferred embodiment of the present invention, it is noted that for the ordinary skill people of the art
For member, under the premise without departing from the principles of the invention, some improvements and modifications can also be made, these improvements and modifications also should
It is considered as protection scope of the present invention.
Claims (10)
1. the method that sensitive parameter structure wheat leaf blade powdery mildew early monitoring model is extracted based on subwindow rearrangement method, including with
Lower step：
1) hyper spectral reflectance of susceptible wheat leaf blade is obtained；
2) sensitive band is extracted from the original wave band of hyper spectral reflectance using SPA algorithms；
3) spectral index related to disease possibility in the existing research of selection, is extracted using SPA algorithms from the spectral index
Sensitivity spectrum index；
4) offset minimum binary-Fisher face is utilized, the sensitive band or the sensitivity spectrum index is defeated as dividing
Enter variable, build wheat powdery mildew early monitoring model；
5) the wheat powdery mildew early monitoring model is tested with two sorting algorithms, and based on independent susceptible kind with staying
One cross-verification method evaluation model is showed；
The step 2) and step 3) between there is no the limitation of time sequencing.
2. method according to claim 1, it is characterised in that the step 1) in susceptible wheat leaf blade according to being in a bad way
Degree divides sample state of an illness grade.
3. method according to claim 2, it is characterised in that the severity is scab subiculum covering on disease leaf
Blade area accounts for the ratio of the blade gross area；
The sample grade is 9 grades of severity SL, respectively 0%, 1%, 5%, 10%, 20%, 40%, 60%, 80%
Wheat leaf blade with 100%.
4. method according to claim 2, it is characterised in that the step 2) in sensitive band include red border area domain.
5. method according to claim 3, it is characterised in that the step 3) in spectral index include that 13 kinds of EO-1 hyperions are planted
By index and 13 kinds of differential smoothing indexes.
6. method according to claim 5, it is characterised in that the step 3) in spectral index feature include normalization color
Plain ratio index, it is red while vegetation coerce index, it is yellow while in the range of first differential summation, Chlorophyll absorption ratio index, physiology it is anti-
Penetrate index, anthocyanidin reflection index, conversion Chlorophyll absorption index, indigo plant in the range of first differential summation, it is red while in the range of one
Rank differential summation with indigo plant in the range of the normalized value of first differential summation and it is red while in the range of first differential summation and Lan Bianfan
Enclose the ratio of interior first differential summation.
7. method according to claim 1, it is characterised in that the step 5) in being categorized as in two sorting algorithms：With
Healthy leaves is positive class, is negative class with infected leaves；Actual healthy leaves and healthy leaves is predicted to for real class, such as fruit
It is false positive class that border healthy leaves is negative class and is predicted to positive class；If actual healthy leaves is to bear class and be predicted to negative class to be
Class is really born, if it is false negative class that actual healthy leaves is positive class to be predicted to bear class.
8. method according to claim 1, it is characterised in that the step 5) in the index evaluated include that model is sensitive
Area and model overall classification accuracy under degree, model specificity, model receiver operating curve.
9. method according to claim 8, it is characterised in that the computational methods of the model susceptibility are shown in formula I；Wherein TP is real class；TN is very negative class；
The computational methods of the model specificity are as shown in formula II；Wherein TN is very negative class；FP
It is false positive class.
10. method according to claim 8, it is characterised in that areal calculation side under the model receiver operating curve
Area value under the receiver operating curve that the receiver operating curve that method is drawn according to grader passes through integral and calculating；
The computational methods of the model overall classification accuracy are as shown in formula III；Wherein TP
It is real class；TN is very negative class；FN is false negative class；FP is false positive class；TN is very negative class.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710172841.4A CN106777845A (en) | 2017-03-22 | 2017-03-22 | The method that sensitive parameter builds wheat leaf blade powdery mildew early monitoring model is extracted based on subwindow rearrangement method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710172841.4A CN106777845A (en) | 2017-03-22 | 2017-03-22 | The method that sensitive parameter builds wheat leaf blade powdery mildew early monitoring model is extracted based on subwindow rearrangement method |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN106777845A true CN106777845A (en) | 2017-05-31 |
Family
ID=58966576
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201710172841.4A Pending CN106777845A (en) | 2017-03-22 | 2017-03-22 | The method that sensitive parameter builds wheat leaf blade powdery mildew early monitoring model is extracted based on subwindow rearrangement method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN106777845A (en) |
Cited By (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107330892A (en) * | 2017-07-24 | 2017-11-07 | 内蒙古工业大学 | A kind of sunflower disease recognition method based on random forest method |
| CN108304791A (en) * | 2018-01-23 | 2018-07-20 | 山东农业大学 | Seeds multispectral remote sensing recognition methods is easily obscured in a kind of mountain area based on cloud model |
| CN108510102A (en) * | 2018-02-07 | 2018-09-07 | 青岛农业大学 | A kind of water-fertilizer integral control method of irrigation using big data calculative strategy |
| CN109035231A (en) * | 2018-07-20 | 2018-12-18 | 安徽农业大学 | A kind of detection method and its system of the wheat scab based on deep-cycle |
| CN109142236A (en) * | 2018-09-13 | 2019-01-04 | 航天信德智图（北京）科技有限公司 | The withered masson pine identifying system of infection pine nematode based on high score satellite image |
| CN110082298A (en) * | 2019-05-15 | 2019-08-02 | 南京农业大学 | A kind of wheat breed head blight comprehensive resistance identification method based on high spectrum image |
| CN110132856A (en) * | 2019-05-18 | 2019-08-16 | 安徽大学 | Wheat scab catch an illness seed identification spectrum disease index construction and application |
| CN112528789A (en) * | 2020-12-02 | 2021-03-19 | 安徽大学 | Remote sensing monitoring method for wheat stripe rust for early and middle stage disease onset analysis |
| CN112881309A (en) * | 2021-02-06 | 2021-06-01 | 内蒙古农业大学 | Establishment method of potato leaf nitrogen detection model and detection method of potato leaf nitrogen |
| CN114092795A (en) * | 2020-07-31 | 2022-02-25 | 中国矿业大学(北京) | Crop disease grade evaluation method based on vegetation index normalization |
| CN116577313A (en) * | 2023-07-14 | 2023-08-11 | 中国农业科学院植物保护研究所 | Quantification method for latent infection amount of powdery mildew of wheat |
Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2004047609A3 (en) * | 2002-11-27 | 2004-07-29 | Visiopharm Aps | A method and a system for establishing a quantity measure for joint destruction |
| EP2704100A1 (en) * | 2012-08-29 | 2014-03-05 | GE Aviation Systems LLC | Method for simulating hyperspectral imagery |
| CN103868880A (en) * | 2014-01-24 | 2014-06-18 | 河南农业大学 | Wheat leaf nitrogen content monitoring method based on spectrum double-peak index and method for establishing monitoring model |
| CN104408307A (en) * | 2014-11-25 | 2015-03-11 | 河南农业大学 | Method for rapidly monitoring morbidity degree of in-field wheat powdery mildew and monitoring model establishment method thereof |
| CN105067532A (en) * | 2015-07-15 | 2015-11-18 | 浙江科技学院 | Method for identifying early-stage disease spots of sclerotinia sclerotiorum and botrytis of rape |
| CN105954281A (en) * | 2016-04-21 | 2016-09-21 | 南京农业大学 | Method for non-destructive identification of paddy moldy fungal colony |
| CN106290197A (en) * | 2016-09-06 | 2017-01-04 | 西北农林科技大学 | The estimation of rice leaf total nitrogen content EO-1 hyperion and estimation models construction method |
-
2017
- 2017-03-22 CN CN201710172841.4A patent/CN106777845A/en active Pending
Patent Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2004047609A3 (en) * | 2002-11-27 | 2004-07-29 | Visiopharm Aps | A method and a system for establishing a quantity measure for joint destruction |
| EP2704100A1 (en) * | 2012-08-29 | 2014-03-05 | GE Aviation Systems LLC | Method for simulating hyperspectral imagery |
| CN103868880A (en) * | 2014-01-24 | 2014-06-18 | 河南农业大学 | Wheat leaf nitrogen content monitoring method based on spectrum double-peak index and method for establishing monitoring model |
| CN104408307A (en) * | 2014-11-25 | 2015-03-11 | 河南农业大学 | Method for rapidly monitoring morbidity degree of in-field wheat powdery mildew and monitoring model establishment method thereof |
| CN105067532A (en) * | 2015-07-15 | 2015-11-18 | 浙江科技学院 | Method for identifying early-stage disease spots of sclerotinia sclerotiorum and botrytis of rape |
| CN105954281A (en) * | 2016-04-21 | 2016-09-21 | 南京农业大学 | Method for non-destructive identification of paddy moldy fungal colony |
| CN106290197A (en) * | 2016-09-06 | 2017-01-04 | 西北农林科技大学 | The estimation of rice leaf total nitrogen content EO-1 hyperion and estimation models construction method |
Non-Patent Citations (2)
| Title |
|---|
| 孙通等: "基于近红外光谱和子窗口重排分析的山茶油掺假检测", 《光学学报》 * |
| 邓平基: "《医学高等数学》", 30 June 2012, 天津大学出版社 * |
Cited By (14)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107330892A (en) * | 2017-07-24 | 2017-11-07 | 内蒙古工业大学 | A kind of sunflower disease recognition method based on random forest method |
| CN108304791A (en) * | 2018-01-23 | 2018-07-20 | 山东农业大学 | Seeds multispectral remote sensing recognition methods is easily obscured in a kind of mountain area based on cloud model |
| CN108510102A (en) * | 2018-02-07 | 2018-09-07 | 青岛农业大学 | A kind of water-fertilizer integral control method of irrigation using big data calculative strategy |
| CN109035231A (en) * | 2018-07-20 | 2018-12-18 | 安徽农业大学 | A kind of detection method and its system of the wheat scab based on deep-cycle |
| CN109142236A (en) * | 2018-09-13 | 2019-01-04 | 航天信德智图（北京）科技有限公司 | The withered masson pine identifying system of infection pine nematode based on high score satellite image |
| CN110082298B (en) * | 2019-05-15 | 2020-05-19 | 南京农业大学 | Hyperspectral image-based wheat variety gibberellic disease comprehensive resistance identification method |
| CN110082298A (en) * | 2019-05-15 | 2019-08-02 | 南京农业大学 | A kind of wheat breed head blight comprehensive resistance identification method based on high spectrum image |
| CN110132856A (en) * | 2019-05-18 | 2019-08-16 | 安徽大学 | Wheat scab catch an illness seed identification spectrum disease index construction and application |
| CN110132856B (en) * | 2019-05-18 | 2021-06-25 | 安徽大学 | Construction and application of spectrum disease index for identifying wheat scab infected seeds |
| CN114092795A (en) * | 2020-07-31 | 2022-02-25 | 中国矿业大学(北京) | Crop disease grade evaluation method based on vegetation index normalization |
| CN112528789A (en) * | 2020-12-02 | 2021-03-19 | 安徽大学 | Remote sensing monitoring method for wheat stripe rust for early and middle stage disease onset analysis |
| CN112881309A (en) * | 2021-02-06 | 2021-06-01 | 内蒙古农业大学 | Establishment method of potato leaf nitrogen detection model and detection method of potato leaf nitrogen |
| CN116577313A (en) * | 2023-07-14 | 2023-08-11 | 中国农业科学院植物保护研究所 | Quantification method for latent infection amount of powdery mildew of wheat |
| CN116577313B (en) * | 2023-07-14 | 2023-10-27 | 中国农业科学院植物保护研究所 | Quantification method for latent infection amount of powdery mildew of wheat |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN106777845A (en) | The method that sensitive parameter builds wheat leaf blade powdery mildew early monitoring model is extracted based on subwindow rearrangement method | |
| CN108875913B (en) | Tricholoma matsutake rapid nondestructive testing system and method based on convolutional neural network | |
| Delalieux et al. | Detection of biotic stress (Venturia inaequalis) in apple trees using hyperspectral data: Non-parametric statistical approaches and physiological implications | |
| Liu et al. | A disease index for efficiently detecting wheat fusarium head blight using sentinel-2 multispectral imagery | |
| CN101881726B (en) | Nondestructive detection method for comprehensive character living bodies of plant seedlings | |
| CN104374738B (en) | A kind of method for qualitative analysis improving identification result based on near-infrared | |
| Pilling et al. | High-throughput quantum cascade laser (QCL) spectral histopathology: a practical approach towards clinical translation | |
| Bandi et al. | Performance evaluation of various statistical classifiers in detecting the diseased citrus leaves | |
| CN106706546A (en) | Analysis method for artificial intelligence learning materials on basis of infrared and Raman spectrum data | |
| CN102788752A (en) | Non-destructive detection device and method of internal information of crops based on spectrum technology | |
| WO2021012898A1 (en) | Artificial intelligence-based agricultural insurance surveying method and related device | |
| Bendel et al. | Evaluating the suitability of hyper-and multispectral imaging to detect foliar symptoms of the grapevine trunk disease Esca in vineyards | |
| CN103389281A (en) | Pu'er tea clustering analysis method based on near-infrared spectroscopy | |
| Matteoli et al. | A spectroscopy-based approach for automated nondestructive maturity grading of peach fruits | |
| CN106290224A (en) | The detection method of bacon quality | |
| CN107219184A (en) | A kind of meat discrimination method and device traced to the source applied to the place of production | |
| CN105717066A (en) | Near-infrared spectrum recognition model based on weighting association coefficients | |
| CN103278467A (en) | Rapid nondestructive high-accuracy method with for identifying abundance degree of nitrogen element in plant leaf | |
| Liu et al. | Estimating leaf chlorophyll contents by combining multiple spectral indices with an artificial neural network | |
| CN110779875B (en) | Method for detecting moisture content of winter wheat ear based on hyperspectral technology | |
| CN112528789A (en) | Remote sensing monitoring method for wheat stripe rust for early and middle stage disease onset analysis | |
| Zhao et al. | Assessment of sugarcane yield potential across large numbers of genotypes using canopy reflectance measurements | |
| Mandal et al. | Spectral characterization and severity assessment of rice blast disease using univariate and multivariate models | |
| Zhang et al. | Evaluating maize grain quality by continuous wavelet analysis under normal and lodging circumstances | |
| CN108663334A (en) | The method for finding soil nutrient spectral signature wavelength based on multiple Classifiers Combination |
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 | ||
| RJ01 | Rejection of invention patent application after publication | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20170531 |