CN108663334A - The method for finding soil nutrient spectral signature wavelength based on multiple Classifiers Combination - Google Patents
The method for finding soil nutrient spectral signature wavelength based on multiple Classifiers Combination Download PDFInfo
- Publication number
- CN108663334A CN108663334A CN201810705297.XA CN201810705297A CN108663334A CN 108663334 A CN108663334 A CN 108663334A CN 201810705297 A CN201810705297 A CN 201810705297A CN 108663334 A CN108663334 A CN 108663334A
- Authority
- CN
- China
- Prior art keywords
- wavelength
- multiple classifiers
- soil nutrient
- spectral signature
- spectral
- 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
Classifications
-
- 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
- G01N21/31—Investigating relative effect of material at wavelengths characteristic of specific elements or molecules, e.g. atomic absorption spectrometry
-
- 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
Landscapes
- Physics & Mathematics (AREA)
- Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Spectroscopy & Molecular Physics (AREA)
- Theoretical Computer Science (AREA)
- Chemical & Material Sciences (AREA)
- Biochemistry (AREA)
- General Health & Medical Sciences (AREA)
- Analytical Chemistry (AREA)
- Immunology (AREA)
- Pathology (AREA)
- Life Sciences & Earth Sciences (AREA)
- Computer Hardware Design (AREA)
- Evolutionary Computation (AREA)
- Geometry (AREA)
- General Engineering & Computer Science (AREA)
- Health & Medical Sciences (AREA)
- Investigating Or Analysing Materials By Optical Means (AREA)
Abstract
The present invention relates to spectroscopic data processing and soil monitoring technical fields, disclose a kind of method for finding soil nutrient spectral signature wavelength based on multiple Classifiers Combination, and step is：(1) soil spectrum is taken；(2) soil nutrient content value is obtained；(3) two or more algorithms are used to calculate characteristic wavelength；(4) Combining Multiple Classifiers screen characteristic wavelength；(5) model is established, prediction effect is analyzed.The present invention is based on spectral techniques, Combining Multiple Classifiers are introduced into spectral signature wavelength and extract field, the advantage of each characteristic wavelength algorithm can be given full play to, learn from other's strong points to offset one's weaknesses, by its effective integration, new approaches are provided to finding optimal spectral wavelength, basis is provided to establish more accurate soil nutrient content model.
Description
Technical field
The present invention relates to spectroscopic data processing and soil monitoring technical fields, are specifically based on multiple Classifiers Combination and find soil
The method of nutrient spectral signature wavelength.
Background technology
Soil is the important component that plant depends on for existence, soil nutrient is that can plant grow up healthy and sound basic, soil
Earth nutrient is insufficient, the growth of crop can be influenced, if soil nutrient is because of overfertilization, it will cause environmental pollution and the wasting of resources.
Soil nutrient content can be varied from factors such as environment, times, monitor soil nutrient content in real time, to grasping this area's soil
Earth situation, environmental problem important role, while also having to the healthy growth of plant, the rational application of fertilizer, environmental protection important
Meaning.Soil nutrient content is quickly detected currently based on spectral technique and has certain progress, usually acquires standard soil sample light
Spectrum and its nutrient content value, model is established by a series of chemometrics methods, to realize to unknown pedotheque nutrient
The prediction of content.
Although spectroscopic methodology slowdown monitoring soil nutrient content, full spectrum can be used for the foundation of soil nutrient model soon,
On the one hand precision of prediction can be influenced, on the one hand will increase run time.Composition of the soil is extremely complex, is wrapped in soil spectrum
Contained a large amount of material information, contained in these information noise information and with detection the incoherent garbage of target, they
Presence can reduce the predictive ability of model.Required Soil K+adsorption clarification of objective wavelength how is accurately found, is one
The major issue of a urgent need to resolve.
Invention content
The present invention is in order to solve the above technical problems, propose to find soil nutrient spectral signature wavelength based on multiple Classifiers Combination
Method, be achieved using following technical scheme.
Based on the method that multiple Classifiers Combination finds soil nutrient spectral signature wavelength, steps are as follows：
(1) pedotheque spectrum is obtained, spectroscopic data after selecting original spectral data according to actual demand or pre-process.
(2) pedotheque nutrient content value is obtained.
(3) spectroscopic data is calculated separately to the characteristic wavelength of each algorithm with various features wavelength algorithm.Characteristic wavelength is calculated
Method includes correlation coefficient process (CC) successive projection algorithm (SPA), without information variable elimination algorithm (UVE), genetic algorithm (GA) side
Method scheduling algorithm.Characteristic wavelength algorithm species number should be greater than 1.
(4) regard each characteristic wavelength algorithm as a single classifier, a variety of single classifiers are merged, according to practical need
It asks the suitable fusion method of selection to merge multi-categorizer, filters out characteristic wavelength.Fusion method includes intersection method, union
Method, ballot method and weighted voting algorithm.
Equipped with N number of grader, XnFor calculating characteristic wavelength the set n=1,2 ... ..., N of n-th of grader, fusion method
For the characteristic wavelength fusion results Y of intersection method and union method1、Y2See following formula
Y1=X1∪X2∪…∪Xn∪…∪XN
Y2=X1∩X2∩…∩Xn∩…∩XN
Ballot method is counted to each wavelength of spectrum, and each single classifier (i.e. a kind of characteristic wavelength algorithm) is right
The extraction of full spectral wavelength all as a ticket has been thrown to the wavelength, then counts the ballot number of each wavelength.Equipped with N number of classification
Device, XnFor the calculating characteristic wavelength set of n-th of grader, wherein n=1,2 ... ..., N；Full spectrum has M wavelength points, judges
M-th of wavelength points is B in the votes of n-th of gradernm, the ballot sum of m-th of wavelength points is Tm, wherein m=1,
2 ... ..., M.
T is selected according to actual demandmThe whole wavelength points such as=k (k=1,2 ... ..., N) are used as spectral signature wavelength, as
Fusion results.
Weighted voting algorithm is on ballot method basis, and the wavelength points more to votes select larger range of wavelength.
Nearest Neighbor with Weighted Voting selection characteristic wavelength is divided to two kinds of forms, and one is full spectrum adjacent wavelengths weightings, and one is the weightings of intersection adjacent wavelengths.
Equipped with N number of grader, XnFor the calculating characteristic wavelength set of n-th of grader, wherein n=1,2 ... ..., N；Full spectrum has M
Wavelength points, judge m-th of wavelength points n-th of grader votes for Bnm, the ballot sum of m-th of wavelength points is Tm,
Middle m=1,2 ... ..., M.
T is selected according to actual demandmWhole wavelength points (k=1,2 ... ..., N) of=k are used as spectral signature wavelength.Tm's
Size determines the size of the wavelength points weighted value.If Spectral range is that (W can be full spectral limit or intersection method multi-categorizer to W
Spectral region after fusion), if ballot number is i, wijFor i ballot number Spectral range in j-th of wavelength, from more to less
Wavelength points are followed successively by mI..., mi..., m1,αiFor miWeight, 2≤i≤I≤N.m_newiFor m after weightingiNew wavelength collection
It closes.
m_newi={ wij∈ W }, mi-(αi*M)≤wij≤mi+αi*M
α1≤ ... ,≤αi≤ ... ,≤aI
Wherein, αiIt, can be according to actual demand setting value in the case where meeting conditions above.M_new_all is that all extractions are special
Levy wavelength.
M_new_all={ m_new1..., m_newi..., m_newI}
(5) the spectral signature wavelength filtered out and its soil nutrient content value are used into a series of chemometrics methods
Spectra inversion model is established, which is analyzed by model evaluation judgment of standard.Modeling method includes
Least square regression (PLS), least square method supporting vector machine (LS-SVM), artificial neural network (BPNN) scheduling algorithm；Evaluation mark
Standard includes absolute coefficient R2, predicted root mean square error RMSEP, the standards such as relation analysis error RPD.
The technical program is based on spectral technique, and Combining Multiple Classifiers, which are introduced spectral signature wavelength, extracts field.It is more
Multiple Classifier Fusion puts forward relative to single classifier method, and the classification information of multiple single classifiers can be carried out to comprehensive point
Analysis has certain complementarity for different graders.Regard each conventional characteristic wavelength algorithm as a grader, it can
To give full play to the advantage of each characteristic wavelength algorithm, learn from other's strong points to offset one's weaknesses, by its effective integration, be capable of providing optimal spectral wavelength,
Convenient for establishing more accurate soil nutrient content model.
Description of the drawings
Fig. 1：The method flow diagram of soil nutrient spectral signature wavelength is found based on multiple Classifiers Combination.
Specific implementation mode
Based on the method that multiple Classifiers Combination finds soil nutrient spectral signature wavelength, to predict soil nutrient total nitrogen (TN)
For, steps are as follows：
(1) pedotheque spectrum is obtained
178 pedotheques are obtained, using marine optics QE65000 spectrometers, are divided into 1nm between spectrum sample, when integral
Between 600ms, Spectral range 200-1100nm, obtain pedotheque visible-near-infrared spectrum, spectroscopic data using original spectrum into
Row subsequent processing.By 178 pedotheques using Kennard-Stone algorithms according to 2:120 parts of 1 ratio point modeling collection is examined
58 parts of collection.
(2) pedotheque nutrient content value is obtained；
Total nitrogen of soil (TN) content is measured using carbon blood urea/nitrogen analyzer.
(3) various features wavelength algorithm is used to calculate characteristic wavelength
Using genetic algorithm (GA), without three kinds of information variable method (UVE), successive projection algorithm (SPA) characteristic wavelength extractions
Algorithm calculates separately the characteristic wavelength of soil spectrum, and the spectral signature wavelength that each algorithm obtains see the table below.
(4) multiple Classifiers Combination
The four kinds of calculations of intersection method, union method, four kinds of ballot method, weighted voting algorithm fusion methods to above-mentioned calculating are respectively adopted
The characteristic wavelength that method obtains carries out multiple Classifiers Combination.Wherein, the wavelength points of two tickets are thrown in the selection of ballot method, and weighted voting algorithm uses
Intersection adjacent wavelengths, i.e. W can be the spectral region after intersection method multiple Classifiers Combination, and α is arranged1=0, α2=0.1, α3=0.2,
Four kinds of fusion method extraction Spectral characteristics of soil wavelength see the table below.
(5) total nitrogen of soil (TN) model and evaluation model precision are established
Spectral signature wavelength and TN content uses are partially minimum after modeling collection pedotheque is extracted by four kinds of fusion methods
Two multiply regression algorithm establishes inverting spectral model respectively, is used in combination inspection set pedotheque to test model, by being absolutely
Number R2, predicted root mean square error RMSEP, relation analysis error RPD evaluate soil TN content forecast result of model.Due to union method
Characteristic wavelength is not filtered out, so the following distal side effect that three kinds of fusion methods are only discussed, the TN contents that extraction wavelength is established
Model modeling collection and inspection set absolute coefficient R2, predicted root mean square error RMSEP, relation analysis error RPD see the table below.
| Fusion method | Modeling collection R2 | Inspection set R2 | RMSEP | RPD |
| Intersection method | 0.9047 | 0.9399 | 0.0133 | 3.7413 |
| Ballot method | 0.4378 | 0.5244 | 0.0325 | 1.4201 |
| Weighted voting algorithm | 0.8606 | 0.7624 | 0.0240 | 2.0299 |
The TN content model modeling collection and inspection set absolute coefficient R that traditional characteristic wavelength extracting method is established2, prediction it is equal
Square error RMSEP, relation analysis error RPD see the table below
| Characteristic wavelength algorithm | Modeling collection R2 | Inspection set R2 | RMSEP | RPD |
| Full spectrum | 0.8641 | 0.8126 | 0.0203 | 2.3274 |
| spa | 0.9185 | 0.8958 | 0.0158 | 2.9988 |
| uve | 0.871 | 0.8379 | 0.0189 | 2.4989 |
| ga1 | 0.9605 | 0.7307 | 0.0256 | 1.8427 |
It can be seen that by above two table, the modelling effect that three kinds of characteristic wavelength algorithms are merged by intersection method is significantly improved,
Modeling collection R2And inspection set R20.9 or more, and RMSEP is obviously reduced, RPD values have peak, 3.7413.
Embodiment only illustrates technical scheme of the present invention, rather than carries out any restrictions to it；Although with reference to the foregoing embodiments
Invention is explained in detail, for those of ordinary skill in the art, still can be to previous embodiment institute
The technical solution of record is modified or equivalent replacement of some of the technical features；And these modifications or substitutions, and
The essence of corresponding technical solution is not set to be detached from the spirit and scope of claimed technical solution of the invention.
Claims (9)
1. the method for finding soil nutrient spectral signature wavelength based on multiple Classifiers Combination, which is characterized in that include the following steps：
(1) soil spectrum is obtained；
(2) soil nutrient content value is obtained；
(3) two or more algorithms are used to calculate characteristic wavelength；
(4) multiple Classifiers Combination screens characteristic wavelength；
(5) model is established, prediction effect is analyzed.
2. the method according to claim 1 for finding soil nutrient spectral signature wavelength based on multiple Classifiers Combination, special
Sign is that the characteristic wavelength algorithm in step (3) is correlation coefficient process and/or successive projection algorithm and/or disappears without information variable
Except algorithm and/or genetic algorithm.
3. the method according to claim 1 for finding soil nutrient spectral signature wavelength based on multiple Classifiers Combination, special
Sign is that the screening technique of step (4) is：Regard each characteristic wavelength algorithm as a single classifier, by a variety of single classification
Device merges, and selects suitable fusion method to merge multi-categorizer according to actual demand, filters out characteristic wavelength.
4. the method according to claim 3 for finding soil nutrient spectral signature wavelength based on multiple Classifiers Combination, special
Sign is that fusion method is intersection method and/or union method and/or ballot method and/or weighted voting algorithm.
5. the method according to claim 4 for finding soil nutrient spectral signature wavelength based on multiple Classifiers Combination, special
Sign is that intersection method and union method calculating process are：Equipped with N number of grader, XnFor the calculating characteristic wavelength collection of n-th of grader
Close n=1,2 ..., N, fusion method be intersection method and union method characteristic wavelength fusion results Y1、Y2Respectively：
Y1=X1∪X2∪...∪Xn∪...∪XN
Y2=X1∩X2∩...∩Xn∩...∩XN
6. the method according to claim 4 for finding soil nutrient spectral signature wavelength based on multiple Classifiers Combination, special
Sign is that ballot method is crossed referred to as：Equipped with N number of grader, XnFor the calculating characteristic wavelength set of n-th of grader, wherein n=1,
2 ..., N；Full spectrum has M wavelength points, judge m-th of wavelength points n-th of grader votes for Bnm, m-th of wave
The ballot sum of long point is Tm, wherein m=1,2 ..., M.
Select Tm=k wholes wavelength points (k=1,2 ..., N) it is used as spectral signature wavelength, as fusion results.
7. the method according to claim 4 for finding soil nutrient spectral signature wavelength based on multiple Classifiers Combination, special
Sign is that weighted voting algorithm is crossed referred to as：Equipped with N number of grader, XnFor the calculating characteristic wavelength set of n-th of grader, wherein n
=1,2 ..., N；Full spectrum has M wavelength points, judge m-th of wavelength points n-th of grader votes for Bnm, m
The ballot sum of a wavelength points is Tm, wherein m=1,2 ..., M,
Select Tm=k whole wavelength points (k=1,2 ..., N) be used as spectral signature wavelength, TmSize determine the wavelength
The size of point weighted value, if Spectral range is W, W is the spectral region after full spectral limit or intersection method multiple Classifiers Combination, if throwing
Ticket number is i, wijFor j-th of wavelength in i ballot number Spectral range, wavelength points from more to less are followed successively by mI...,
mi..., mi, αiFor miWeight, 2≤i≤I≤N, m_newiFor m after weightingiNew wavelength set,
m_newi={ wij∈ W }, mi-(αi*M)≤wij≤mi+αi*M
α1≤ ... ,≤α i≤... ,≤aI
Wherein, αiCan be all extraction characteristic waves according to actual demand setting value, m_new_all in the case where meeting conditions above
It is long.
M_new_all={ m_new1..., m_newi..., m_newI}
8. the method according to claim 1 for finding soil nutrient spectral signature wavelength based on multiple Classifiers Combination, special
Sign is that the method for establishing model includes least square regression and/or least square method supporting vector machine and/or artificial neural network
Network.
9. the method that soil nutrient spectral signature wavelength is found based on multiple Classifiers Combination according to claim 1 or 8,
It is characterized in that, evaluation criterion includes absolute coefficient R2And/or predicted root mean square error RMSEP and/or relation analysis error RPD.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810705297.XA CN108663334B (en) | 2018-07-02 | 2018-07-02 | Method for searching spectral characteristic wavelength of soil nutrient based on multi-classifier fusion |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810705297.XA CN108663334B (en) | 2018-07-02 | 2018-07-02 | Method for searching spectral characteristic wavelength of soil nutrient based on multi-classifier fusion |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN108663334A true CN108663334A (en) | 2018-10-16 |
| CN108663334B CN108663334B (en) | 2021-03-30 |
Family
ID=63772390
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810705297.XA Active CN108663334B (en) | 2018-07-02 | 2018-07-02 | Method for searching spectral characteristic wavelength of soil nutrient based on multi-classifier fusion |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108663334B (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111389741A (en) * | 2020-04-16 | 2020-07-10 | 长春光华学院 | Automatic sorting system for detecting surface defects of automobile brake pads based on machine vision |
| CN111504981A (en) * | 2020-04-26 | 2020-08-07 | 上海交通大学 | Method for determining chemical components and moisture content in powder material |
| CN111540155A (en) * | 2020-03-27 | 2020-08-14 | 北京联合大学 | Intelligent household fire detector |
Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106644978A (en) * | 2016-11-16 | 2017-05-10 | 山东省科学院海洋仪器仪表研究所 | Detection method capable of judging precision and used for analyzing content of soil nutrients based on spectral characteristic wavelength |
-
2018
- 2018-07-02 CN CN201810705297.XA patent/CN108663334B/en active Active
Patent Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106644978A (en) * | 2016-11-16 | 2017-05-10 | 山东省科学院海洋仪器仪表研究所 | Detection method capable of judging precision and used for analyzing content of soil nutrients based on spectral characteristic wavelength |
Non-Patent Citations (5)
| Title |
|---|
| 徐兆阳: ""基于多层次控制的多分类器融合遥感影像分类"", 《中国优秀硕士学位论文全文数据库 基础科学辑》 * |
| 李凯 等: ""近红外光谱和多分类器融合的葡萄酒品种判别研究"", 《光谱学与光谱分析》 * |
| 王鹏: ""基于差异性度量的多分类器融合研究"", 《中国优秀硕士学位论文全文数据库 信息科技辑》 * |
| 祝志慧 等: ""基于光谱技术和多分类器融合的异物蛋检测"", 《农业工程学报》 * |
| 郭红玲: ""多分类器选择关键技术的研究"", 《中国优秀硕士学位论文全文数据库 信息科技辑》 * |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111540155A (en) * | 2020-03-27 | 2020-08-14 | 北京联合大学 | Intelligent household fire detector |
| CN111540155B (en) * | 2020-03-27 | 2022-05-24 | 北京联合大学 | Intelligent household fire detector |
| CN111389741A (en) * | 2020-04-16 | 2020-07-10 | 长春光华学院 | Automatic sorting system for detecting surface defects of automobile brake pads based on machine vision |
| CN111504981A (en) * | 2020-04-26 | 2020-08-07 | 上海交通大学 | Method for determining chemical components and moisture content in powder material |
| CN111504981B (en) * | 2020-04-26 | 2021-10-22 | 上海交通大学 | Method for determining chemical components and moisture content in powder material |
Also Published As
| Publication number | Publication date |
|---|---|
| CN108663334B (en) | 2021-03-30 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Tan et al. | Estimating the distribution trend of soil heavy metals in mining area from HyMap airborne hyperspectral imagery based on ensemble learning | |
| Ball et al. | Morphometric analysis of phytoliths: recommendations towards standardization from the International Committee for Phytolith Morphometrics | |
| CN105630743B (en) | A kind of system of selection of spectrum wave number | |
| Fuentes et al. | Automated grapevine cultivar classification based on machine learning using leaf morpho-colorimetry, fractal dimension and near-infrared spectroscopy parameters | |
| Mishra et al. | A Deep Learning-Based Novel Approach for Weed Growth Estimation. | |
| CN101299237B (en) | High spectroscopic data supervision classifying method based on information quantity dimensionality sequence | |
| CN108416378A (en) | A kind of large scene SAR target identification methods based on deep neural network | |
| CN104008551B (en) | A kind of Citrus Huanglongbing pathogen detection method based on visible images | |
| CN110232419A (en) | A kind of method of side slope rock category automatic identification | |
| CN110991064B (en) | Soil heavy metal content inversion model generation method, system and inversion method | |
| CN109253985B (en) | Method for identifying wood grade for koto panel by near infrared spectrum based on neural network | |
| CN109544537A (en) | The fast automatic analysis method of hip joint x-ray image | |
| CN107679569A (en) | Raman spectrum substance automatic identifying method based on adaptive hypergraph algorithm | |
| CN108663334A (en) | The method for finding soil nutrient spectral signature wavelength based on multiple Classifiers Combination | |
| CN104867150A (en) | Wave band correction change detection method of remote sensing image fuzzy clustering and system thereof | |
| CN101894270A (en) | Method for full-automatic sample selection oriented to classification of remote-sensing images | |
| CN111161362A (en) | Tea tree growth state spectral image identification method | |
| May et al. | Automated ripeness assessment of oil palm fruit using RGB and fuzzy logic technique | |
| Mousavirad et al. | Design of an expert system for rice kernel identification using optimal morphological features and back propagation neural network | |
| CN104680185B (en) | Hyperspectral image classification method based on boundary point reclassification | |
| CN110443139A (en) | A kind of target in hyperspectral remotely sensed image noise wave band detection method of Classification Oriented | |
| CN107132190A (en) | A kind of soil organism spectra inversion model calibration samples collection construction method | |
| CN103955711B (en) | A kind of mode identification method in imaging spectral target identification analysis | |
| CN112200042A (en) | Method for analyzing ecological change trend by using space-time ecological environment remote sensing fractal dimension | |
| CN108827909B (en) | Rapid soil classification method based on visible near infrared spectrum and multi-target fusion |
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 |