CN108596104A - Wheat powdery mildew remote sensing monitoring method with disease characteristic preprocessing function - Google Patents
Wheat powdery mildew remote sensing monitoring method with disease characteristic preprocessing function Download PDFInfo
- Publication number
- CN108596104A CN108596104A CN201810383173.4A CN201810383173A CN108596104A CN 108596104 A CN108596104 A CN 108596104A CN 201810383173 A CN201810383173 A CN 201810383173A CN 108596104 A CN108596104 A CN 108596104A
- Authority
- CN
- China
- Prior art keywords
- powdery mildew
- feature
- sample
- primary election
- wheat
- 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.)
- Granted
Links
- 238000012544 monitoring process Methods 0.000 title claims abstract description 108
- 241000221785 Erysiphales Species 0.000 title claims abstract description 93
- 241000209140 Triticum Species 0.000 title claims abstract description 81
- 235000021307 Triticum Nutrition 0.000 title claims abstract description 81
- 201000010099 disease Diseases 0.000 title claims abstract description 60
- 208000037265 diseases, disorders, signs and symptoms Diseases 0.000 title claims abstract description 60
- 238000000034 method Methods 0.000 title claims abstract description 52
- 238000007781 pre-processing Methods 0.000 title claims abstract description 17
- 230000002068 genetic effect Effects 0.000 claims abstract description 15
- 238000012545 processing Methods 0.000 claims abstract description 14
- 238000012706 support-vector machine Methods 0.000 claims abstract description 14
- 238000005516 engineering process Methods 0.000 claims abstract description 11
- 238000012512 characterization method Methods 0.000 claims description 47
- HUTDUHSNJYTCAR-UHFFFAOYSA-N ancymidol Chemical group C1=CC(OC)=CC=C1C(O)(C=1C=NC=NC=1)C1CC1 HUTDUHSNJYTCAR-UHFFFAOYSA-N 0.000 claims description 24
- 238000005457 optimization Methods 0.000 claims description 23
- 239000011159 matrix material Substances 0.000 claims description 16
- 238000009826 distribution Methods 0.000 claims description 15
- 238000012216 screening Methods 0.000 claims description 14
- 238000000605 extraction Methods 0.000 claims description 12
- 238000002310 reflectometry Methods 0.000 claims description 9
- 238000011160 research Methods 0.000 claims description 9
- 238000012549 training Methods 0.000 claims description 9
- 239000002689 soil Substances 0.000 claims description 8
- 238000012937 correction Methods 0.000 claims description 6
- 230000005855 radiation Effects 0.000 claims description 6
- 238000007476 Maximum Likelihood Methods 0.000 claims description 4
- 239000000843 powder Substances 0.000 claims description 4
- 241001269238 Data Species 0.000 claims description 3
- 238000004364 calculation method Methods 0.000 claims description 3
- 238000005520 cutting process Methods 0.000 claims description 3
- 239000000284 extract Substances 0.000 claims description 3
- 230000035772 mutation Effects 0.000 claims description 3
- 241000208340 Araliaceae Species 0.000 claims 1
- 235000005035 Panax pseudoginseng ssp. pseudoginseng Nutrition 0.000 claims 1
- 235000003140 Panax quinquefolius Nutrition 0.000 claims 1
- 235000008434 ginseng Nutrition 0.000 claims 1
- 238000010606 normalization Methods 0.000 claims 1
- 230000007547 defect Effects 0.000 abstract 1
- 238000004422 calculation algorithm Methods 0.000 description 42
- 238000011835 investigation Methods 0.000 description 7
- 241000238631 Hexapoda Species 0.000 description 4
- 230000000694 effects Effects 0.000 description 4
- 235000013399 edible fruits Nutrition 0.000 description 3
- 238000005259 measurement Methods 0.000 description 2
- 238000012795 verification Methods 0.000 description 2
- 241001124076 Aphididae Species 0.000 description 1
- 241000196324 Embryophyta Species 0.000 description 1
- 241000125167 Rhopalosiphum padi Species 0.000 description 1
- 241000607479 Yersinia pestis Species 0.000 description 1
- 238000004458 analytical method Methods 0.000 description 1
- 238000013459 approach Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000015572 biosynthetic process Effects 0.000 description 1
- 238000009395 breeding Methods 0.000 description 1
- 230000001488 breeding effect Effects 0.000 description 1
- 235000013339 cereals Nutrition 0.000 description 1
- 238000007796 conventional method Methods 0.000 description 1
- 238000010219 correlation analysis Methods 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 230000007613 environmental effect Effects 0.000 description 1
- 239000004744 fabric Substances 0.000 description 1
- 238000005286 illumination Methods 0.000 description 1
- 238000007689 inspection Methods 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 239000000463 material Substances 0.000 description 1
- 238000012821 model calculation Methods 0.000 description 1
- 230000002265 prevention Effects 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/10—Terrestrial scenes
- G06V20/188—Vegetation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
- G06F18/2413—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on distances to training or reference patterns
- G06F18/24133—Distances to prototypes
- G06F18/24143—Distances to neighbourhood prototypes, e.g. restricted Coulomb energy networks [RCEN]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/29—Graphical models, e.g. Bayesian networks
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- Bioinformatics & Computational Biology (AREA)
- General Engineering & Computer Science (AREA)
- Artificial Intelligence (AREA)
- Life Sciences & Earth Sciences (AREA)
- Health & Medical Sciences (AREA)
- General Health & Medical Sciences (AREA)
- Multimedia (AREA)
- Image Processing (AREA)
Abstract
The invention relates to a remote sensing monitoring method for wheat powdery mildew with a disease characteristic preprocessing function, which overcomes the defects of high redundancy and poor monitoring precision of wheat disease characteristics compared with the prior art. The invention comprises the following steps: acquiring and preprocessing remote sensing data; extracting characteristic variables; processing the characteristic variables; constructing and optimizing a powdery mildew monitoring model; and obtaining a remote sensing monitoring result of wheat powdery mildew. The invention combines two characteristic selection technologies of relief and mRMR with a support vector machine optimized by a genetic method to form effective remote sensing monitoring on powdery mildew of regional scale.
Description
Technical field
It is specifically a kind of with Disease Characters preprocessing function the present invention relates to technical field of remote sensing image processing
Wheat powdery mildew remote-sensing monitoring method.
Background technology
Wheat powdery mildew is one of the Major Diseases during Wheat Production, can all be occurred in the entire breeding time of wheat,
The serious underproduction and quality is caused to reduce, the aggrieved rear general underproduction reaches 20% or more between 5% to 10% when serious.And
When effectively detect the generation of wheat powdery mildew and be of great significance to the yield and quality for improving wheat.
Although traditional ground investigation method investigation result is preferable, need to expend a large amount of manpower and materials, be not suitable for big
The research in region.Many scholars are monitored prediction using meteorological data to crop disease and insect.Wang Hejun etc. utilizes temperature, illumination
The prediction model of wheat powdery mildew is established with rainfall.Stansbury etc. is using warm-wet index model to wheat India bunt
Disease is predicted.Dutta etc. establishes the prediction model of aphid using meteorologic parameters such as air themperature and humidity.But by
It can be influenced by landform, mankind's activity in meteorological data, can not accurately obtain continuous spatial information, and the length of wheat
Gesture information is also to react the important factor of Wheat Diseases And Insect Pests situation, so there are one based on meteorological crop disease and insect monitoring and forecasting
Fixed limitation.
Since remote sensing can obtain continuous spatial information, while it can also reflect the growing way situation of crop, so one
A little scholars have carried out a series of research using remote sensing to crop disease and insect.Have in the prior art and is combined using environmental satellite data
Wavelet analysis and supporting vector machine model establish the monitoring model of wheat powdery mildew;Wheat powdery mildew is supervised using PRI
It surveys;Monitoring model is established to wheat powdery mildew based on Adaboost models and mRMR algorithms;Using each of environment sing data inverting
Kind vegetation index passes through logistic and returns the Occurrence forecast for realizing wheat powdery mildew;Winter wheat is extracted using environment sing data
The growing way factor, factor of the habitat, the prediction model of wheat aphid is established in conjunction with Method Using Relevance Vector Machine.
And the remote sensing features index of current most of scholar's primary study image crop disease, the selection for characteristic variable
Be generally only in method by simple correlation analysis, T examine etc. primary election feature is screened, although the feature filtered out with
The correlation of wheat diseases is big, but the redundancy between feature can lead to the reduction of model accuracy.
Therefore, the remote sensing prison how developed a kind of feature higher based on disease correlation and minimum redundancy and realized
Survey method has become the technical issues of urgent need solves.
Invention content
It is high, monitoring accuracy difference scarce that the purpose of the present invention is to solve wheat diseases feature redundancies in the prior art
It falls into, provides and a kind of solve the above problems with the wheat powdery mildew remote-sensing monitoring method of Disease Characters preprocessing function.
To achieve the goals above, technical scheme is as follows:
A kind of wheat powdery mildew remote-sensing monitoring method with Disease Characters preprocessing function, includes the following steps:
11) acquisition and pretreatment of remotely-sensed data, obtains the remotely-sensed data in research area, and is located in advance to remotely-sensed data
Reason, and pass through the cultivated area of Maximum likelihood classification extraction wheat;
12) extraction of characteristic variable, obtain wheat field survey data sample, using pretreated remote sensing image come into
The extraction of feature needed for the monitoring of row wheat powdery mildew;
13) processing of characteristic variable is calculated the weight of wheat powdery mildew feature using relief technologies, is sieved by threshold value
After choosing, using the mRMR choices of technology and target category correlation maximum and the character subset of mutual redundancy minimum is as white
The input variable of powder disease monitoring model;
14) structure of powdery mildew monitoring model and optimization build the powdery mildew monitoring model based on support vector machines,
And powdery mildew monitoring model is optimized;
15) acquisition of wheat powdery mildew remote sensing monitoring result is preserved from remote sensing image by the extraction characteristic variable of pixel
To the input variable as powdery mildew monitoring model in matrix A, the geographical coordinate for extracting each pixel is saved in matrix B, will
Matrix A is input in the powdery mildew monitoring model after optimization, obtains the matrix result C of research area's wheat powdery mildew monitoring situation,
Monitoring result drafting pattern is obtained the wheat powdery mildew monitoring knot of survey region by associate(d) matrix result C and geographical coordinate matrix B
Fruit spatial distribution map.
The acquisition and pretreatment of the remotely-sensed data include the following steps:
21) remotely-sensed data in wheat powdery mildew area is obtained, and carries out image radiation calibration and atmospheric correction processing, image
Radiation calibration formula is as follows:
L (λ)=GainDN+Bais,
Wherein, L (λ) is radiance value, and Gain is gain coefficient, and Bais is biasing coefficient, and DN is observation gray value;
22) the FLAASH atmospheric corrections module in ENVI5.1 remote sensing processing softwares is used to switch to the radiance of image
Reflectivity;
23) image is carried out cutting the image for obtaining wanted survey region;
24) according to the normalized differential vegetation index of survey region, near infrared reflectivity data, in conjunction with the maximum in ENVI5.1
Likelihood classification extracts the wheat planting area of survey region.
The extraction of the characteristic variable includes the following steps:
31) wheat field survey data sample is obtained comprising n sample point, each sample point remembered by manual tag
For healthy sample or disease sample;
32) the blue, green, red of remotely-sensed data, this four reflectivity datas of near-infrared and broadband vegetation index is chosen to make
For the primary election characterization factor of powdery mildew monitoring model, primary election characterization factor constitutes remote sensing image primary election feature set, wherein broadband
Vegetation index includes ratio vegetation index, triangle vegetation index, green wave band normalized differential vegetation index, enhancement mode meta file, returns
One changes vegetation index, optimization soil adjusts vegetation index, soil adjusts vegetation index, improved triangle vegetation index, improved
Simple ratio index, renormalization vegetation index.
The processing of the characteristic variable includes the following steps:
41) the primary election characterization factor for being directed to powdery mildew monitoring model calculates feature weight using relief methods;
From remote sensing image primary election feature set choose powdery mildew monitoring model primary election characterization factor, with this primary election feature because
Sub- σ carries out the calculating of feature weight, until all primary election characterization factors have carried out feature power in remote sensing image primary election feature set
The calculating of weight;
42) weight threshold is set, the average value of primary election characterization factor σ weights is compared with weight threshold;If more than
Primary election characterization factor σ is then included into screening characteristic set by weight threshold;If being less than weight threshold, primary election characterization factor σ is given up
It abandons;
43) dimension-reduction treatment is carried out to the feature for screening characteristic set by mRMR methods, is filtered out in relief methods
It screens in characteristic set using related between feature and disease classification, between feature and feature in mutual information measurement character subset
Degree, filters out n characteristic variable.
The structure of the powdery mildew monitoring model and optimization include the following steps:
51) powdery mildew monitoring model is built based on support vector machines, using n characteristic variable as sample set;
52) regression function minimum is handled, and support vector machines is by finding φ and b so that regression function is according to f (x)=ωT
The structural risk minimization of φ (x)+b;
Introduce non-negative slack variable ξiAnd penalty factor, problem representation to be optimized are:
Wherein, C is constant, ξiWithControl the upper bound and the lower bound of output constraint;
53) penalty factor and nuclear parameter are optimized using genetic method, to realize the optimization of powdery mildew monitoring model.
The calculating of the feature weight is as follows:
61) a sample point A is randomly selected in n actual sample point;
62) assume that sample point A is healthy sample, two nearest samples are found around sample point A, one is arest neighbors
Healthy sample H, one is arest neighbors disease sample M;
63) calculate sample point A and arest neighbors health sample H between primary election characterization factor difference sigmaAH;
64) calculate sample point A and arest neighbors disease sample M between primary election characterization factor difference sigmaAM;
If 65) σAH< σAM, then illustrate that primary election characterization factor σ is conducive to distinguish similar and inhomogeneous arest neighbors, increase just
Select the weight of characterization factor σ;
If 66) σAH> σAM, then illustrate that primary election characterization factor σ is unfavorable for distinguishing similar and inhomogeneous arest neighbors, reduce just
Select the weight of characterization factor σ;
67) 61) step m times is repeated, and calculates the average value of m primary election characterization factor σ weight.
It is described that the feature progress dimension-reduction treatment for screening characteristic set is included the following steps by mRMR methods:
71) two stochastic variables x and y are given, the probability density function for corresponding to continuous variable is p (x), p (y), p
(x, y) p (x), p (y), p (x, y), then the mutual information expression between X and Y is as follows:
72) maximum correlation in mutual information calculating sifting characteristic set S between disease classification c, expression formula are utilized
It is as follows:
73) feature x in mutual information calculating sifting characteristic set S is utilizediAnd xjBetween redundancy, calculation expression is such as
Under:
74) mutual information difference criterion is utilized, the maximum feature of difference is selected, obtains a spy of minimal redundancy maximal correlation
This feature is included into the set of preferred features needed for modeling by sign;
75) 72) step n times are repeated in remaining sample set, finally obtain the preferred feature needed for n modeling.
Described being optimized to penalty factor and nuclear parameter using genetic method is included the following steps:
81) using the preferred feature needed for n modeling as sample data, training sample is formed;
82) average relative error of the output valve and desired value of training sample is calculated;
83) training sample selected, intersected, mutation operator;
84) judge whether the maximum genetic algebra for meeting initial setting up, obtained when meeting condition optimal penalty factor and
Nuclear parameter.
Advantageous effect
A kind of wheat powdery mildew remote-sensing monitoring method with Disease Characters preprocessing function of the present invention, with the prior art
Compared to by being combined two kinds of Feature Selections of relief and mRMR with the support vector machines by genetic method optimization, formed
Effective remote sensing monitoring is carried out to the powdery mildew of regional scale.
The present invention is by conjunction with Feature Dimension Reduction is carried out, passing through and setting certain threshold value relief algorithms and mRMR algorithms
The weights of the good feature of discrimination are improved, obtain optimal character subset.It is good at solution global optimum using genetic algorithm to ask
The characteristics of topic, strong robustness, carrys out optimization monitoring model, to improve the monitoring accuracy of model.
Description of the drawings
Fig. 1 is the method precedence diagram of the present invention;
Fig. 2 a are weight shared by each feature of 14 features of primary election in the present invention after the calculating of relief methods
The weights figure of size;
Fig. 2 b be in the present invention weight with 500 be radix, number shared by each phase characteristic in 14 features of primary election
Weights figure;
Fig. 3 a are the wheat powdery mildew monitoring result spatial distribution map using relief algorithm combinations GSSVM;
Fig. 3 b are the monitoring result spatial distribution map of the feature and GASVM after being screened by relief algorithms;
Fig. 3 c are the monitoring result spatial distribution map of the feature and GSSVM after being screened by relief+mRMR algorithms;
Fig. 3 d are that the monitoring result of the feature and GASVM after the screening of the present invention by relief+mRMR algorithms is empty
Between distribution map;
Fig. 4 a are the monitoring result local distribution figure of the feature and GSSVM after being screened by relief+mRMR algorithms;
Fig. 4 b are the monitoring result of the feature and GASVM after the screening of the present invention by relief+mRMR algorithms
Local distribution figure.
Specific implementation mode
The effect of to make to structure feature of the invention and being reached, has a better understanding and awareness, to preferable
Embodiment and attached drawing cooperation detailed description, are described as follows:
As shown in Figure 1, a kind of wheat powdery mildew remote sensing monitoring with Disease Characters preprocessing function of the present invention
Method, it is high using relief methods operational efficiency, by calculating the weight of image feature, the feature strong to classification capacity assign compared with
High weight.In view of relief algorithms do not consider existing correlation between image feature, cannot remove between image feature
Redundancy, utilize the redundancy between mRMR methods removal image feature herein, obtain that there is between feature minimum redundancy
And the character subset with maximum correlation between feature and target.By two kinds of feature selection approach of relief combinations mRMR,
Filter out feature higher with disease correlation and minimum redundancy.It includes the following steps:
The first step, the acquisition and pretreatment of remotely-sensed data.The remotely-sensed data in research area is obtained, and remotely-sensed data is carried out pre-
The cultivated area of wheat is extracted in processing by Maximum likelihood classification.It is as follows:
(1) remotely-sensed data in wheat powdery mildew area is obtained, and carries out image radiation calibration and atmospheric correction processing, image
Radiation calibration formula is as follows:
L (λ)=GainDN+Bais,
Wherein, L (λ) is radiance value, and Gain is gain coefficient, and Bais is biasing coefficient, and DN is observation gray value.
(2) the FLAASH atmospheric corrections module in ENVI5.1 remote sensing processing softwares is used to switch to the radiance of image
Reflectivity.
(3) image is carried out cutting the image for obtaining the region to be studied.
(4) according to the normalized differential vegetation index of survey region, near infrared reflectivity data, in conjunction with the maximum in ENVI5.1
Likelihood classification extracts the wheat planting area of survey region.
Second step, the extraction of characteristic variable.Wheat field survey data sample is obtained, pretreated remote sensing image is utilized
To carry out the extraction of the feature needed for wheat powdery mildew monitoring.It is as follows:
(1) wheat field survey data sample is obtained comprising n sample point, manual tag has been each sample point
Healthy sample or disease sample.
In practical applications, field survey data can be investigated in Wheat in Grain Filling Stage and be obtained, and disease is carried out using 5 investigation methods
Investigation, i.e., take the sample prescription of 1m × 1m in each points for investigation, in investigation dot center global positioning system (global position
System, GPS) it is positioned, 5 symmetrical points are uniformly chosen in sample prescription, 20 plants of wheats is each clicked and is investigated, and record sample
Powdery mildew incidence in side.Severity uses improved wheat powdery mildew in agricultural industry criteria (NY/T613-2002)
" 0~9 grade of method records disease severity, and wheat is uniformly divided into 9 sections by when investigation from top to bottom, is marked according to classification
Standard is classified, and disease index (disesase index, DI) is then calculated.In the present invention, it is contemplated that grade excessively exists
The occurrence degree of disease is only divided into health and two types occurs by the difficulty in monitoring.
(2) the blue, green, red of remotely-sensed data, this four reflectivity datas of near-infrared and broadband vegetation index is chosen to make
For the primary election characterization factor of powdery mildew monitoring model, primary election characterization factor constitutes remote sensing image primary election feature set.Wherein, broadband
Vegetation index includes ratio vegetation index, triangle vegetation index, green wave band normalized differential vegetation index, enhancement mode meta file, returns
One changes vegetation index, optimization soil adjusts vegetation index, soil adjusts vegetation index, improved triangle vegetation index, improved
Simple ratio index, renormalization vegetation index.
Third walks, the processing of characteristic variable.Since primary election intrinsic dimensionality is larger, meter can be increased by being directly used in the structure of model
Calculation amount, and unscreened feature can there is a situation where it is small or even unrelated with the occurrence degree correlation of disease, between feature
There is also redundancies, will have a direct impact on the precision of model, therefore combine feature of the relief and mRMR technologies to wheat powdery mildew
Carry out dimensionality reduction.
The weight that wheat powdery mildew feature is calculated using relief technologies is set certain threshold value, is screened by threshold value
Afterwards, using the mRMR choices of technology and target category correlation maximum and the character subset of mutual redundancy minimum is as white powder
The input variable of sick monitoring model.It is as follows:
(1) the primary election characterization factor for being directed to powdery mildew monitoring model calculates feature weight using relief methods.
From remote sensing image primary election feature set choose powdery mildew monitoring model primary election characterization factor, with this primary election feature because
Sub- σ carries out the calculating of feature weight, until all primary election characterization factors have carried out feature power in remote sensing image primary election feature set
The calculating of weight.
Wherein, the calculating of feature weight is as follows:
A, a sample point A is randomly selected in n actual sample point;
B, assume that the sample point A randomly selected is healthy sample, two nearest samples of searching around sample point A, one
A is arest neighbors health sample H, and one is arest neighbors disease sample M.
Here, if the sample point A randomly selected is disease sample, distinguishes and carry out reversed distinguish i.e. by same thinking
It can.
For example, it is assumed that the sample point A randomly selected is disease sample, then an arest neighbors disease is found around sample point A
The difference of primary election characterization factor is σ between evil sample H, sample point A and arest neighbors disease sample HAH;It is found around sample point A
The difference of primary election characterization factor is σ between one arest neighbors health sample M, sample point A and arest neighbors health sample MAM。
C, calculate sample point A and arest neighbors health sample H between primary election characterization factor difference sigmaAH;
D, calculate sample point A and arest neighbors disease sample M between primary election characterization factor difference sigmaAM;
If E, σAH< σAM, then illustrate that primary election characterization factor σ is conducive to distinguish similar and inhomogeneous arest neighbors, increase just
Select the weight of characterization factor σ;
If F, σAH> σAM, then illustrate that primary election characterization factor σ is unfavorable for distinguishing similar and inhomogeneous arest neighbors, reduce just
Select the weight of characterization factor σ.
Comparison in this way, the size of correlation by evaluating characteristic to the separating capacity of short distance sample,
If the distance delta of sample A and sample H in this featureAHLess than the distance in this feature on sample A and sample M
σAM, then illustrate that this feature is beneficial to distinguishing similar and inhomogeneous arest neighbors, then increase the weight of this feature;Conversely, such as
Distance big σs of the fruit sample A and sample H in this featureAHIn the distance delta of sample A and sample M in this featureAM, illustrate this feature
Negative effect is played to distinguishing similar and inhomogeneous arest neighbors, then reduces the weight of this feature.The weight of this feature factor is bigger,
Indicate that the classification capacity of this feature is stronger, the weight of this feature factor is smaller, indicates that this feature classification capacity is weaker.
G, A is repeated) step (a sample point A is randomly selected in n actual sample point) m times, and calculate m primary election spy
The average value for levying factor sigma weight, to obtain the primary election characterization factor, (the blue, green, red of remotely-sensed data, near-infrared this four are anti-
Penetrate rate data and broadband vegetation index) average value.
(2) weight threshold is set, the average value of primary election characterization factor σ weights is compared with weight threshold;If more than
Primary election characterization factor σ is then included into screening characteristic set by weight threshold;If being less than weight threshold, primary election characterization factor σ is given up
It abandons.Screening characteristic set is the characteristic set after being screened to all primary election characterization factors, to select some influence powers
Greatly, the big feature of correlation.
(3) dimension-reduction treatment is carried out to the feature for screening characteristic set by mRMR methods, is filtered out in relief methods
It screens in characteristic set using related between feature and disease classification, between feature and feature in mutual information measurement character subset
Degree filters out n required characteristic variables.
MRMR methods are the characteristic feature dimension-reduction algorithms based on information theory, using mRMR methods from by relief technologies
It finds out in the feature of screening with disease classification with maximum correlation and the n feature with minimum redundancy between each other.Its
It is as follows:
A, give two stochastic variables x and y, correspond to continuous variable probability density function be p (x), p (y), p (x,
Y) p (x), p (y), p (x, y), then the mutual information expression between X and Y is as follows:
B, using the maximum correlation in mutual information calculating sifting characteristic set S between disease classification c, expression formula is such as
Under:
C, feature x in mutual information calculating sifting characteristic set S is utilizediAnd xjBetween redundancy, calculation expression is such as
Under:
D, using mutual information difference criterion, the maximum feature of difference is selected, obtains a feature of minimal redundancy maximal correlation,
This feature is included into the set of preferred features needed for modeling;
E, repeat B in remaining sample set) step (using in mutual information calculating sifting characteristic set S with disease
Maximum correlation between classification c) n times, finally obtain the preferred feature needed for n modeling.
Here, relief algorithms operational efficiency is high, by calculating the weight of feature, the feature strong to classification capacity assign compared with
High weight, but relief algorithms do not consider existing correlation between feature, so the redundancy between feature cannot be removed
Property.MRMR algorithms can obtain between feature with the feature with maximum correlation between minimum redundancy and feature and target
Subset, but weight size can not be calculated.Therefore by conjunction with Feature Dimension Reduction is carried out, leading to relief algorithms and mRMR algorithms
It crosses and sets certain threshold value to improve the weights of the good feature of discrimination, obtain optimal character subset.
As shown in Figure 2 a, in Fig. 2 a abscissa be primary election 14 features (the blue, green, red of remotely-sensed data, near-infrared this four
A reflectivity data, triangle vegetation index, green wave band normalized differential vegetation index, enhancement mode meta file, is returned at ratio vegetation index
One changes vegetation index, optimization soil adjusts vegetation index, soil adjusts vegetation index, improved triangle vegetation index, improved
Simple ratio index, renormalization vegetation index) each feature serial number, the weight size that ordinate is characterized.For another example Fig. 2 b
It is shown, indicate that by weight be radix with 500, number shared by each phase characteristic.
It can be seen that the size of weight shared by 14 features from Fig. 2 a, in conjunction with Fig. 2 b, when setting weight threshold as 2500,
Primary election feature of 8 features of the condition of satisfaction as mRMR algorithms can be filtered out, is then obtained by mRMR algorithms optimal
Three characteristic variables choose the maximum three characteristic variable conducts of weight that relief algorithms obtain as first group of characteristic variable
Second group of characteristic variable.
4th step, the structure of powdery mildew monitoring model and optimization.Build the powdery mildew monitoring based on support vector machines
Model, and powdery mildew monitoring model is optimized (the support vector machines GASVM for forming genetic algorithm optimization).
Wherein, the structure of powdery mildew monitoring model and optimization include the following steps:
(1) powdery mildew monitoring model is built based on support vector machines, using n characteristic variable as sample set.
(2) regression function minimum is handled, and support vector machines is by finding φ and b so that regression function is according to f (x)=ωT
The structural risk minimization of φ (x)+b;
Introduce non-negative slack variable ξiAnd penalty factor, problem representation to be optimized are:
Wherein:C is constant, ξiWithControl the upper bound and the lower bound of output constraint.
(3) penalty factor and nuclear parameter are optimized using genetic method, to realize the optimization of powdery mildew monitoring model.
Penalty factor and nuclear parameter are optimized using conventional method using genetic method, specifically include following steps:
A, using n modeling needed for preferred feature be used as sample data, formation training sample, can also this distinguish verification
Sample, to verify the reliability of model;
B, the average relative error of the output valve and desired value of training sample is calculated;
C, training sample selected, intersected, mutation operator;
D, judge whether the maximum genetic algebra for meeting initial setting up, optimal penalty factor and core are obtained when meeting condition
Parameter.
5th step, the acquisition of wheat powdery mildew remote sensing monitoring result, by the extraction characteristic variable of pixel from remote sensing image
It preserves into matrix A the input variable as powdery mildew monitoring model, the geographical coordinate for extracting each pixel and is saved in matrix B
In, matrix A is input in the powdery mildew monitoring model after optimization, the matrix knot of research area's wheat powdery mildew monitoring situation is obtained
Monitoring result drafting pattern is obtained the wheat powdery mildew prison of survey region by fruit C, associate(d) matrix result C and geographical coordinate matrix B
Survey result space distribution map.
In practical applications, realistic model can preferably be embodied by being verified to model using independent sample data
Precision.6 kinds of models of field survey data pair of present invention combination Hebei Shijiazhuang City research area's on May 24th, 2014 carry out
Assessment.
2 kinds of feature selecting algorithms (relief+mRMR algorithms) are listed in table 1, and in conjunction with 3 kinds of modeling methods, (SVM, grid are searched
Rope method optimization support vector machines GSSVM, genetic algorithm optimization support vector machines GASVM) establish powdery mildew monitoring model
User's precision, cartographic accuracy, overall accuracy and Kappa coefficients.
1 SVM, GSSVM, GASVM modeling method verification result contrast table of table
From table 1 it follows that the overall accuracy and Kappa coefficients of SVM monitoring models are monitored less than GSSVM and GASVM
Model, the overall accuracy and Kappa coefficients of relief+mRMR-SVM monitoring models are carried compared with relief-SVM monitoring models
Height, but only 64% and 0.286;In 2 GSSVM monitoring models, monitoring model that relief+mRMR algorithms are established it is total
Body precision and Kappa coefficients are 78.5% and 0.571, the monitoring model established higher than relief algorithms;It is supervised in 2 GASVM
It surveys in model, the monitoring model precision and Kappa coefficients that relief+mRMR algorithms are established are 85.7% and 0.714, are higher than
The monitoring model that the feature that relief algorithms filter out is established, can deduce and filter out the feature with disease correlation maximum
While removal feature between redundancy can effectively improve the precision of model.
Overall precision, user's precision and the cartographic accuracy of the above 3 kinds of model method institute established models of comparison can be seen that
Relief algorithm+mRMR algorithms carry out the precision of Feature Selection and the GASVM monitoring models (relief+mRMR-GASVM) built
The overall accuracy of highest, overall accuracy ratio GSSVM and SVM institutes established model is higher by 21.7% and 7.2%, and user's precision and drawing
Precision reaches 85.7%.It these results suggest that relief+mRMR algorithms institute established model is better than the mould that relief algorithms are established
Type, the monitoring model that the SVM (GASVM) based on genetic algorithm optimization is established is better than not optimized SVM and is based on grid search
The SVM (GSSVM) of method optimization, the powdery mildew monitoring model that relief+mRMR algorithm combinations GASVM is established are (of the present invention
Relief+mRMR-GASVM the precision of model can) be improved.
Equally, by taking Hebei Shijiazhuang City studies the remote sensing image of area's on May 26th, 2014 as an example.As shown in Figure 3a, it is
The wheat powdery mildew monitoring result spatial distribution map of relief algorithm combinations GSSVM;As shown in Figure 3b, it is to pass through relief
The monitoring result of feature and GASVM after algorithm screening;Fig. 3 c be by relief+mRMR algorithms screen after feature with
The monitoring result of GSSVM;Fig. 3 d are the monitoring result of the feature and GASVM after being screened by relief+mRMR algorithms.Fig. 3 a-
In Fig. 3 d, white area part is non-wheat planting district, and black region part is region of disease.
Onset area accounts in the percentage of total area, and Fig. 3 a are 44.5%, Fig. 3 b are 62.1%, Fig. 3 c are 62.4%,
Fig. 3 d are 60.4%.Fig. 3 a onset areas compared with its excess-three width figure are less, seriously occur with the wheat powdery mildew of on-site inspection
Deviation is larger.Compare from Fig. 3 a, Fig. 3 b, Fig. 3 c, Fig. 3 d it can be found that the result powdery mildew region of disease that Fig. 3 a and Fig. 3 b are monitored
Scattered distribution, and wheat powdery mildew is caused by cloth powdery mildew, its feature is that reproduction speed is fast, and propagation area is wide, one
As scattered will not occur, so the characteristics of monitoring result and wheat powdery mildew of Fig. 3 a and Fig. 3 b, runs counter to, and Fig. 3 c and Fig. 3 d
Monitoring result more meets wheat powdery mildew occurrence characteristic, confidence level higher.Fig. 3 c and Fig. 3 d are generally performing out consistency, but
It has differences in detail.
As shown in figures 4 a and 4b, Fig. 4 a are the monitoring knot of the feature and GSSVM after being screened by relief+mRMR algorithms
Fruit local distribution figure, Fig. 4 b are the local distribution of the monitoring result of the feature and GASVM after being screened by relief+mRMR algorithms
Figure.
Comparing Fig. 4 a and Fig. 4 b it can be found that same, in Fig. 4 a and Fig. 4 b, white area part is non-wheat planting district,
Black region part is region of disease.The region for being divided into health in Fig. 4 d is divided into morbidity in fig.4.In Fig. 4 b wheat health with
It is distributed relatively uniform between morbidity plot, and the region that Fig. 4 a then show monoblock mostly is all diseased region, only a small number ofly
Block shows to be uniformly distributed, compare two width local distribution figures can be found that by relief+mRMR algorithms screen after feature with
The monitoring model goodness of fit of GSSVM is not high, although monitoring overall trend be actually consistent, in detail section separating capacity
The monitoring model of feature and GASVM after being screened compared to relief+mRMR algorithms is weaker, and relief+mRMR-GASVM models exist
It is still applicable in the monitoring of part.
In conclusion relief+mRMR-GASVM models of the present invention are thin in overall trend, characteristics of incidence, part
It is all more consistent with actual conditions on section, the onset state of wheat powdery mildew can be really reflected, daily life is can be used for
It generates in the demand monitored in real time to wheat powdery mildew in living, by accurately obtaining the onset state of powdery mildew, spatial distribution is special
Sign carrys out planned offer evidence for prevention and cure, improves the yield of wheat.
The basic principles, main features and advantages of the present invention have been shown and described above.The technology of the industry
Personnel are it should be appreciated that the present invention is not limited to the above embodiments, and what is described in the above embodiment and the description is only the present invention
Principle, various changes and improvements may be made to the invention without departing from the spirit and scope of the present invention, these variation and
Improvement is both fallen in the range of claimed invention.The present invention claims protection domain by appended claims and its
Equivalent defines.
Claims (8)
1. a kind of wheat powdery mildew remote-sensing monitoring method with Disease Characters preprocessing function, which is characterized in that including following
Step:
11) acquisition and pretreatment of remotely-sensed data, obtains the remotely-sensed data in research area, and is pre-processed to remotely-sensed data, and
The cultivated area of wheat is extracted by Maximum likelihood classification;
12) extraction of characteristic variable is obtained wheat field survey data sample, is carried out using pretreated remote sensing image small
The extraction of feature needed for wheat powdery mildew monitoring;
13) processing of characteristic variable is calculated the weight of wheat powdery mildew feature using relief technologies, is screened by threshold value
Afterwards, using the mRMR choices of technology and target category correlation maximum and the character subset of mutual redundancy minimum is as white powder
The input variable of sick monitoring model;
14) structure of powdery mildew monitoring model and optimization build the powdery mildew monitoring model based on support vector machines, and right
Powdery mildew monitoring model optimizes;
15) acquisition of wheat powdery mildew remote sensing monitoring result, the extraction characteristic variable by pixel from remote sensing image are preserved to square
Input variable in battle array A as powdery mildew monitoring model, the geographical coordinate for extracting each pixel is saved in matrix B, by matrix A
It is input in the powdery mildew monitoring model after optimization, the matrix result C of research area's wheat powdery mildew monitoring situation is obtained, in conjunction with square
Monitoring result drafting pattern is obtained the wheat powdery mildew monitoring result space of survey region by battle array result C and geographical coordinate matrix B
Distribution map.
2. a kind of wheat powdery mildew remote-sensing monitoring method with Disease Characters preprocessing function according to claim 1,
It is characterized in that, the acquisition and pretreatment of the remotely-sensed data include the following steps:
21) remotely-sensed data in wheat powdery mildew area is obtained, and carries out image radiation calibration and atmospheric correction processing, image radiation
It is as follows to calibrate formula:
L (λ)=GainDN+Bais,
Wherein, L (λ) is radiance value, and Gain is gain coefficient, and Bais is biasing coefficient, and DN is observation gray value;
22) the FLAASH atmospheric corrections module in ENVI5.1 remote sensing processing softwares is used to switch to reflect by the radiance of image
Rate;
23) image is carried out cutting the image for obtaining the region to be studied;
24) according to the normalized differential vegetation index of survey region, near infrared reflectivity data, in conjunction with the maximum likelihood in ENVI5.1
Classification extracts the wheat planting area of survey region.
3. a kind of wheat powdery mildew remote-sensing monitoring method with Disease Characters preprocessing function according to claim 1,
It is characterized in that, the extraction of the characteristic variable includes the following steps:
31) obtain wheat field survey data sample comprising n sample point, each sample point manual tag be denoted as it is strong
Health sample or disease sample;
32) the blue, green, red of remotely-sensed data, this four reflectivity datas of near-infrared and broadband vegetation index are chosen as white
The primary election characterization factor of powder disease monitoring model, primary election characterization factor constitute remote sensing image primary election feature set, wherein broadband vegetation
Index includes ratio vegetation index, triangle vegetation index, green wave band normalized differential vegetation index, enhancement mode meta file, normalization
Vegetation index, optimization soil adjust vegetation index, soil adjusts vegetation index, improved triangle vegetation index, improved simple
Ratio index, renormalization vegetation index.
4. a kind of wheat powdery mildew remote-sensing monitoring method with Disease Characters preprocessing function according to claim 1,
It is characterized in that, the processing of the characteristic variable includes the following steps:
41) the primary election characterization factor for being directed to powdery mildew monitoring model calculates feature weight using relief methods;
The primary election characterization factor that powdery mildew monitoring model is chosen from remote sensing image primary election feature set, with this primary election characterization factor σ
The calculating of feature weight is carried out, until all primary election characterization factors have carried out feature weight in remote sensing image primary election feature set
It calculates;
42) weight threshold is set, the average value of primary election characterization factor σ weights is compared with weight threshold;If more than weight
Primary election characterization factor σ is then included into screening characteristic set by threshold value;If being less than weight threshold, primary election characterization factor σ is given up;
43) dimension-reduction treatment is carried out to the feature for screening characteristic set by mRMR methods, in the screening that relief methods filter out
The degree of correlation in character subset between feature and disease classification, between feature and feature is weighed using mutual information in characteristic set,
Filter out n characteristic variable.
5. a kind of wheat powdery mildew remote-sensing monitoring method with Disease Characters preprocessing function according to claim 1,
It is characterized in that, the structure of the powdery mildew monitoring model and optimization include the following steps:
51) powdery mildew monitoring model is built based on support vector machines, using n characteristic variable as sample set;
52) regression function minimum is handled, and support vector machines is by finding φ and b so that regression function is according to f (x)=ωTφ(x)
The structural risk minimization of+b;
Introduce non-negative slack variable ξiAnd penalty factor, problem representation to be optimized are:
Wherein, C is constant, ξiWithControl the upper bound and the lower bound of output constraint;
53) penalty factor and nuclear parameter are optimized using genetic method, to realize the optimization of powdery mildew monitoring model.
6. a kind of wheat powdery mildew remote-sensing monitoring method with Disease Characters preprocessing function according to claim 4,
It is characterized in that, the calculating of the feature weight is as follows:
61) a sample point A is randomly selected in n actual sample point;
62) assume that sample point A is healthy sample, two nearest samples are found around sample point A, one is arest neighbors health
Sample H, one is arest neighbors disease sample M;
63) calculate sample point A and arest neighbors health sample H between primary election characterization factor difference sigmaAH;
64) calculate sample point A and arest neighbors disease sample M between primary election characterization factor difference sigmaAM;
If 65) σAH< σAM, then illustrate that primary election characterization factor σ is conducive to distinguish similar and inhomogeneous arest neighbors, it is special to increase primary election
Levy the weight of factor sigma;
If 66) σAH> σAM, then illustrate that primary election characterization factor σ is unfavorable for distinguishing similar and inhomogeneous arest neighbors, it is special to reduce primary election
Levy the weight of factor sigma;
67) 61) step m times is repeated, and calculates the average value of m primary election characterization factor σ weight.
7. a kind of wheat powdery mildew remote-sensing monitoring method with Disease Characters preprocessing function according to claim 4,
It is characterized in that, described include the following steps the feature progress dimension-reduction treatment for screening characteristic set by mRMR methods:
71) two stochastic variables x and y are given, the probability density function for corresponding to continuous variable is p (x), p (y), p (x, y) p
(x), p (y), p (x, y), then the mutual information expression between X and Y is as follows:
72) utilize the maximum correlation in mutual information calculating sifting characteristic set S between disease classification c, expression formula as follows:
73) feature x in mutual information calculating sifting characteristic set S is utilizediAnd xjBetween redundancy, calculation expression is as follows:
74) mutual information difference criterion is utilized, the maximum feature of difference is selected, obtains a feature of minimal redundancy maximal correlation, it will
This feature is included into the set of preferred features needed for modeling;
75) 72) step n times are repeated in remaining sample set, finally obtain the preferred feature needed for n modeling.
8. a kind of wheat powdery mildew remote-sensing monitoring method with Disease Characters preprocessing function according to claim 5,
It is characterized in that, described being optimized to penalty factor and nuclear parameter using genetic method is included the following steps:
81) using the preferred feature needed for n modeling as sample data, training sample is formed;
82) average relative error of the output valve and desired value of training sample is calculated;
83) training sample selected, intersected, mutation operator;
84) judge whether the maximum genetic algebra for meeting initial setting up, optimal penalty factor and core ginseng are obtained when meeting condition
Number.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810383173.4A CN108596104B (en) | 2018-04-26 | 2018-04-26 | Wheat powdery mildew remote sensing monitoring method with disease characteristic preprocessing function |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810383173.4A CN108596104B (en) | 2018-04-26 | 2018-04-26 | Wheat powdery mildew remote sensing monitoring method with disease characteristic preprocessing function |
Publications (2)
Publication Number | Publication Date |
---|---|
CN108596104A true CN108596104A (en) | 2018-09-28 |
CN108596104B CN108596104B (en) | 2021-01-05 |
Family
ID=63609419
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201810383173.4A Active CN108596104B (en) | 2018-04-26 | 2018-04-26 | Wheat powdery mildew remote sensing monitoring method with disease characteristic preprocessing function |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN108596104B (en) |
Cited By (13)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109711319A (en) * | 2018-12-24 | 2019-05-03 | 安徽高哲信息技术有限公司 | A kind of method and system that grain unsound grain image recognition sample database is established |
CN109948237A (en) * | 2019-03-15 | 2019-06-28 | 中国汽车技术研究中心有限公司 | A method of for predicting bicycle discharge amount |
CN110008905A (en) * | 2019-04-09 | 2019-07-12 | 安徽大学 | Regional scale stripe rust of wheat monitoring method based on the red side wave section of remote sensing image |
CN110188727A (en) * | 2019-06-05 | 2019-08-30 | 中煤航测遥感集团有限公司 | Marine oil spill amount evaluation method and device |
CN110309780A (en) * | 2019-07-01 | 2019-10-08 | 中国科学院遥感与数字地球研究所 | High resolution image houseclearing based on BFD-IGA-SVM model quickly supervises identification |
CN111639649A (en) * | 2020-05-26 | 2020-09-08 | 中国地质大学(武汉) | Method and system for identifying and encoding numbered musical notation image based on real-time image stream |
WO2020258973A1 (en) * | 2019-06-27 | 2020-12-30 | 岭南师范学院 | Human body feature parameter selection method based on improved rrelieff in combination with mrmr |
CN114049564A (en) * | 2021-10-22 | 2022-02-15 | 长江师范学院 | Pine wood nematode disease grade prediction model construction method based on hyperspectral remote sensing image |
CN114548249A (en) * | 2022-02-15 | 2022-05-27 | 中国农业科学院植物保护研究所 | Wheat powdery mildew occurrence degree prediction method based on machine learning |
CN115201210A (en) * | 2022-08-12 | 2022-10-18 | 河南省农业科学院农业经济与信息研究所 | Satellite data-based wheat yellow mosaic disease remote sensing monitoring method |
CN116680548A (en) * | 2023-08-03 | 2023-09-01 | 南京信息工程大学 | Time sequence drought causal analysis method for multi-source observation data |
CN117636185A (en) * | 2024-01-26 | 2024-03-01 | 安徽大学 | Pine wood nematode disease detecting system based on image processing |
CN114049564B (en) * | 2021-10-22 | 2024-07-02 | 长江师范学院 | Pine wood nematode disease level prediction model construction method based on hyperspectral remote sensing image |
Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102629305A (en) * | 2012-03-06 | 2012-08-08 | 上海大学 | Feature selection method facing to SNP (Single Nucleotide Polymorphism) data |
CN104794496A (en) * | 2015-05-05 | 2015-07-22 | 中国科学院遥感与数字地球研究所 | Remote sensing character optimization algorithm for improving mRMR (min-redundancy max-relevance) algorithm |
US20160165178A1 (en) * | 2014-12-03 | 2016-06-09 | King Abdulaziz City For Science And Technology | Super-resolution of dynamic scenes using sampling rate diversity |
CN106897703A (en) * | 2017-02-27 | 2017-06-27 | 辽宁工程技术大学 | Remote Image Classification based on AGA PKF SVM |
CN107103306A (en) * | 2017-05-22 | 2017-08-29 | 安徽大学 | Winter wheat powdery mildew remote sensing monitoring method based on wavelet analysis and support vector machine |
-
2018
- 2018-04-26 CN CN201810383173.4A patent/CN108596104B/en active Active
Patent Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102629305A (en) * | 2012-03-06 | 2012-08-08 | 上海大学 | Feature selection method facing to SNP (Single Nucleotide Polymorphism) data |
US20160165178A1 (en) * | 2014-12-03 | 2016-06-09 | King Abdulaziz City For Science And Technology | Super-resolution of dynamic scenes using sampling rate diversity |
CN104794496A (en) * | 2015-05-05 | 2015-07-22 | 中国科学院遥感与数字地球研究所 | Remote sensing character optimization algorithm for improving mRMR (min-redundancy max-relevance) algorithm |
CN106897703A (en) * | 2017-02-27 | 2017-06-27 | 辽宁工程技术大学 | Remote Image Classification based on AGA PKF SVM |
CN107103306A (en) * | 2017-05-22 | 2017-08-29 | 安徽大学 | Winter wheat powdery mildew remote sensing monitoring method based on wavelet analysis and support vector machine |
Non-Patent Citations (2)
Title |
---|
曹路,欧阳效源: "基于遗传算法的支持向量机的参数优化", 《计算机与数学工程》 * |
王露,龚光红: "基于ReliefF-mRMR特征降维算法的多特征遥感图像分类", 《中国体视学与图像分析》 * |
Cited By (19)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109711319B (en) * | 2018-12-24 | 2023-04-07 | 安徽高哲信息技术有限公司 | Method and system for establishing imperfect grain image recognition sample library |
CN109711319A (en) * | 2018-12-24 | 2019-05-03 | 安徽高哲信息技术有限公司 | A kind of method and system that grain unsound grain image recognition sample database is established |
CN109948237A (en) * | 2019-03-15 | 2019-06-28 | 中国汽车技术研究中心有限公司 | A method of for predicting bicycle discharge amount |
CN110008905B (en) * | 2019-04-09 | 2021-02-12 | 安徽大学 | Regional scale wheat stripe rust monitoring method based on red-edge wave band of remote sensing image |
CN110008905A (en) * | 2019-04-09 | 2019-07-12 | 安徽大学 | Regional scale stripe rust of wheat monitoring method based on the red side wave section of remote sensing image |
CN110188727B (en) * | 2019-06-05 | 2021-11-05 | 中煤航测遥感集团有限公司 | Ocean oil spill quantity estimation method and device |
CN110188727A (en) * | 2019-06-05 | 2019-08-30 | 中煤航测遥感集团有限公司 | Marine oil spill amount evaluation method and device |
WO2020258973A1 (en) * | 2019-06-27 | 2020-12-30 | 岭南师范学院 | Human body feature parameter selection method based on improved rrelieff in combination with mrmr |
CN110309780A (en) * | 2019-07-01 | 2019-10-08 | 中国科学院遥感与数字地球研究所 | High resolution image houseclearing based on BFD-IGA-SVM model quickly supervises identification |
CN111639649B (en) * | 2020-05-26 | 2024-03-01 | 中国地质大学(武汉) | Method and system for identifying and encoding numbered musical notation image based on real-time image stream |
CN111639649A (en) * | 2020-05-26 | 2020-09-08 | 中国地质大学(武汉) | Method and system for identifying and encoding numbered musical notation image based on real-time image stream |
CN114049564A (en) * | 2021-10-22 | 2022-02-15 | 长江师范学院 | Pine wood nematode disease grade prediction model construction method based on hyperspectral remote sensing image |
CN114049564B (en) * | 2021-10-22 | 2024-07-02 | 长江师范学院 | Pine wood nematode disease level prediction model construction method based on hyperspectral remote sensing image |
CN114548249A (en) * | 2022-02-15 | 2022-05-27 | 中国农业科学院植物保护研究所 | Wheat powdery mildew occurrence degree prediction method based on machine learning |
CN115201210A (en) * | 2022-08-12 | 2022-10-18 | 河南省农业科学院农业经济与信息研究所 | Satellite data-based wheat yellow mosaic disease remote sensing monitoring method |
CN116680548A (en) * | 2023-08-03 | 2023-09-01 | 南京信息工程大学 | Time sequence drought causal analysis method for multi-source observation data |
CN116680548B (en) * | 2023-08-03 | 2023-10-13 | 南京信息工程大学 | Time sequence drought causal analysis method for multi-source observation data |
CN117636185A (en) * | 2024-01-26 | 2024-03-01 | 安徽大学 | Pine wood nematode disease detecting system based on image processing |
CN117636185B (en) * | 2024-01-26 | 2024-04-09 | 安徽大学 | Pine wood nematode disease detecting system based on image processing |
Also Published As
Publication number | Publication date |
---|---|
CN108596104B (en) | 2021-01-05 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN108596104A (en) | Wheat powdery mildew remote sensing monitoring method with disease characteristic preprocessing function | |
Huang et al. | Detecting sugarcane borer diseases using support vector machine | |
Zhou et al. | Strawberry maturity classification from UAV and near-ground imaging using deep learning | |
CN102495005B (en) | Method for diagnosing crop water deficit through hyperspectral image technology | |
CN110008905B (en) | Regional scale wheat stripe rust monitoring method based on red-edge wave band of remote sensing image | |
CN108830249B (en) | Winter wheat powdery mildew remote sensing monitoring method based on ASD hyperspectral data | |
CN107256451A (en) | Agriculture suitability evaluation analysis method of the agriculture big data based on GIS | |
CN108982377A (en) | Corn growth stage spectrum picture and chlorophyll content correlation and period division methods | |
Lin et al. | Intelligent greenhouse system based on remote sensing images and machine learning promotes the efficiency of agricultural economic growth | |
CN107330413A (en) | A kind of mother drug plants recognition methods based on remote sensing technology | |
CN102072885A (en) | Machine vision-based paddy neck blast infection degree grading method | |
CN107316116A (en) | A kind of leafy vegetable production prediction method | |
Diniz et al. | Forecasting frost risk in forest plantations by the combination of spatial data and machine learning algorithms | |
CN113657158A (en) | Google Earth Engine-based large-scale soybean planting region extraction algorithm | |
Zhao et al. | Study on the classification method of rice leaf blast levels based on fusion features and adaptive-weight immune particle swarm optimization extreme learning machine algorithm | |
US11170219B2 (en) | Systems and methods for improved landscape management | |
Cao et al. | Detecting asymptomatic infections of rice bacterial leaf blight using hyperspectral imaging and 3-dimensional convolutional neural network with spectral dilated convolution | |
CN108846370B (en) | Method for automatically analyzing severity of disease in middle and late stages of wheat powdery mildew | |
CN117077868A (en) | Model-based grain yield prediction system and method | |
CN116883874A (en) | Evaluation method and system for comprehensive growth vigor of cinnamomum camphora dwarf forest | |
CN116151454A (en) | Method and system for predicting yield of short-forest linalool essential oil by multispectral unmanned aerial vehicle | |
Shiranvand et al. | An analysis of dieback areas of Zagros oak forests using remote sensing data case study: Lorestan oak forest, Iran | |
Chiang et al. | A hybrid artificial neural network-based agri-economic model for predicting typhoon-induced losses | |
CN113112147A (en) | Citrus huanglongbing risk assessment method based on ground actual measurement and remote sensing cooperation | |
Simpson | Crop yield prediction using a CMAC neural network |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant |