CN109387484A - A kind of ramee variety recognition methods of combination EO-1 hyperion and support vector cassification - Google Patents

A kind of ramee variety recognition methods of combination EO-1 hyperion and support vector cassification Download PDF

Info

Publication number
CN109387484A
CN109387484A CN201811242492.XA CN201811242492A CN109387484A CN 109387484 A CN109387484 A CN 109387484A CN 201811242492 A CN201811242492 A CN 201811242492A CN 109387484 A CN109387484 A CN 109387484A
Authority
CN
China
Prior art keywords
ramie
leaf
hyperspectral
sample
model
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Pending
Application number
CN201811242492.XA
Other languages
Chinese (zh)
Inventor
曹晓兰
崔国贤
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Hunan Agricultural University
Original Assignee
Hunan Agricultural University
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Hunan Agricultural University filed Critical Hunan Agricultural University
Priority to CN201811242492.XA priority Critical patent/CN109387484A/en
Publication of CN109387484A publication Critical patent/CN109387484A/en
Pending legal-status Critical Current

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
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F30/00—Computer-aided design [CAD]
    • G06F30/30—Circuit design
    • G06F30/36—Circuit design at the analogue level
    • G06F30/367—Design verification, e.g. using simulation, simulation program with integrated circuit emphasis [SPICE], direct methods or relaxation methods

Landscapes

  • Engineering & Computer Science (AREA)
  • Computer Hardware Design (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • Evolutionary Computation (AREA)
  • Chemical & Material Sciences (AREA)
  • Geometry (AREA)
  • Microelectronics & Electronic Packaging (AREA)
  • Spectroscopy & Molecular Physics (AREA)
  • Health & Medical Sciences (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • General Engineering & Computer Science (AREA)
  • Analytical Chemistry (AREA)
  • Biochemistry (AREA)
  • General Health & Medical Sciences (AREA)
  • Immunology (AREA)
  • Pathology (AREA)
  • Measuring Or Testing Involving Enzymes Or Micro-Organisms (AREA)

Abstract

The present invention proposes the ramee variety recognition methods of a kind of combination EO-1 hyperion and support vector cassification, comprising the following steps: sample collection and hyperspectral measurement, sample sets division, PCA feature extraction, SVC identification model is established and determining best model；The present invention determines the method for extracting the SVC ramie EO-1 hyperion variety ecotype model that feature establishes Linear, Polynomial, RBF and Sigmoid kernel function respectively by using grid data service, model recognition correct rate can achieve 95% or more, have the advantages that reliable and effective and quick, easy, the theoretical foundation and key technology for improving the ramee variety identification based on EO-1 hyperion, assistant breeding, the high yield and high quality to realize ramie and numb field Precision management support, ramee variety recognition cycle can be shortened, reduce manpower and material resources consumption.

Description

A kind of ramee variety recognition methods of combination EO-1 hyperion and support vector cassification
Technical field
The present invention relates to the technical field of ramee variety identification more particularly to a kind of combination EO-1 hyperion and support vector machines point The ramee variety recognition methods of class.
Background technique
The perennial root type herbaceous plant of Section of Genus Boehmeria Urticaceae Section of Genus Boehmeria, is the specialty in China, is known as Chinese grass.I State is one of main Chan Ma state in the world, possesses the most abundant ramee variety resource, ramie planting area and raw material output account for generation 95% or more of boundary, in national economy, ramie has always higher economic status, currently, the type of China's ramee variety, Mainly there are the ecotype divided by planting area, the morphotype divided by botanic conformation, by yield and quality division Economic ecology type and the ripe phase type etc. divided by breeding time.
What the identification of traditional ramee variety was mainly divided according to the ecotype of planting area division, by botanic conformation Morphotype, the economic ecology type divided by yield and quality and the standards such as ripe phase type divided by breeding time, then rely on Artificial experience is identified that time-consuming for these recognition methods, at high cost, subjectivity is strong, and accuracy rate is low, is unsuitable for a large amount of Ramee variety screening identification, although some research achievements for carrying out crop identification using EO-1 hyperion are existing many, for not The bloom spectral property of homogenic type ramie establishes the model of ramee variety identification still with (selection) effective spectral signature is extracted So more lack, although the research that some methods using support vector cassification SVC carry out qualitative classification identification is also very much, The selection of support vector cassification Kernel Function is extremely crucial, kernel function and and its criterion that does not uniquely determine of parameter setting, need It wants experience and the determination that makes repeated attempts, takes time and effort larger.Therefore, the present invention proposes a kind of combination EO-1 hyperion and support vector machines point The ramee variety recognition methods of class, to solve shortcoming in the prior art.
Summary of the invention
In view of the above-mentioned problems, the present invention propose by using grid data service determine extract feature establish respectively Linear, The method of the SVC ramie EO-1 hyperion variety ecotype model of Polynomial, RBF and Sigmoid kernel function, model recognition correct rate It can achieve 95% or more, have the advantages that reliable and effective and quick, easy, improve ramee variety based on EO-1 hyperion and know Not, assistant breeding, the theoretical foundation of the high yield and high quality to realize ramie and numb field Precision management and key technology support, can contract Short ramee variety recognition cycle reduces manpower and material resources consumption.
The present invention proposes the ramee variety recognition methods of a kind of combination EO-1 hyperion and support vector cassification, including following step It is rapid:
Step 1: sample collection and hyperspectral measurement
The ramie for collecting different cultivars, as ramie sample, then by ramie sample using portable field spectroradiometer and portable The matched hand-held leaf clip leaf spectra detector of formula field spectroradiometer selected in ramie leaf samples 4 sampled points measure into Row high-spectral data collection, sample point data do the blade EO-1 hyperion number being averaged after breakpoint correction again as the ramie sample According to；
Step 2: sample set divides
In the above step 1 after the blade high-spectral data collection of the ramie sample of different cultivars, by the ramie sample of different cultivars This blade high-spectral data is randomly assigned according to the ratio of 2:1, successively labeled as at modeling collection and forecast set；
Step 3: PCA feature extraction
The leaf characteristic that ramie sample is extracted using PCA selects ingredient of all characteristic values greater than 1 as PCA main gene or side Poor contribution rate of accumulative total reaches the preceding n main gene of 85%-95% as PCA main gene；
Step 4: SVC identification model is established
Utilize SVC algorithm by the PCA main cause subcharacter after extracting in above-mentioned steps three in stoichiometry software Unscrambler Variable since variance accumulate n-th of main gene of contribution rate >=85%, successively increase PCA main gene number, then respectively with difference SVC kernel function and the optimal parameter of kernel function be combined, the moulds of multiple identification ramee varieties are established after then being analyzed Type obtains best features variable number, SVC kernel function and parameter combination；
Step 5: best model is determined
The prediction data of forecast set in step 2 is substituted into the various combined models that step 4 is established, evaluates and identify pre- It surveys as a result, obtaining the model of optimal parameter combination.
Further improvement lies in that: sample collection is using local varieties, Cultivars, different root and stem of certain plants type, no in the step 1 With the sample in maturity period and different output, sample collection selects 9 kinds of ramee varieties, and in the prosperous long-term acquisition ramie of ramie sample The blade high-spectral data of sample, each kind acquire 162 high light datas of blade, and 9 kind symbiosis are at 1458 blade samples This high-spectral data.
Further improvement lies in that: 9 kinds of ramee varieties are respectively Jinsha Chinese holly skin fiber crops, Bijie circle is numb, Xiang Tan chicken bone is white, river flowing from Guizhou Province through Hunan into Dongting Lake River Huang shell morning, Pingtang machete fiber crops, middle ramie 1, Shaoyang 4, bimodal great Ye fiber crops and Suining piemarker.
Further improvement lies in that: in the step 1 when high-spectral data collection, the main lobe arteries and veins of ramie leaf samples is avoided, The blade clamp holder clamping of hand-held leaf clip leaf spectra detector is first tested to the leaf of ramie sample when high-spectral data collection The surveyed position of piece, then the blade EO-1 hyperion of the probe measurement ramie sample with hand-held leaf clip leaf spectra detector, ramie sample For this blade Samples selecting on blade main lobe arteries and veins both sides, blade main lobe arteries and veins both sides respectively select 2 sampled points, select 4 altogether Sampled point, sample point data do the blade high-spectral data being averaged after breakpoint correction again as the ramie sample.
Further improvement lies in that: in order to which the blade high-spectral data for eliminating ramie sample is first in acquisition in the step 1 The noise that end is generated with end, the spectrum number when blade high-spectral data collection of ramie sample between selection 420nm-2450nm According to being analyzed, and portable field spectroradiometer per half an hour does an OPT optimization and blank reference.
Further improvement lies in that: the blade high-spectral data modeling collection of the ramie sample of different cultivars is used in the step 2 In establishing variety ecotype model, the blade high-spectral data forecast set of the ramie sample of different cultivars is not involved in modeling, is only used for Evaluate and test the accuracy rate of model.
Further improvement lies in that: use PCA to accumulate when extracting the leaf characteristic of ramie sample with variance in the step 3 It is appropriate to increase main gene number based on contribution rate judging result, eventually by the accuracy of choosing comprehensively model prediction collection, drop Dynamics and variance accumulation contribution rate of dimension etc. is because usually determining main gene number.
Further improvement lies in that: the kernel function and parameter setting in the step 4 in order to avoid SVC algorithm are to model Precision has an impact with complexity, while in order to find best kernel function and parameter, needing using different SVC kernel function and core Function establish ramee variety identification mould, specially four kinds different SVC kernel functions, including Linear, Polynomial, RBF and Sigmoid kernel function.
Further improvement lies in that: Linear, Polynomial, RBF and Sigmoid kernel function establishes ramee variety knowledge When other model, use grid data service and cross validation accuracy for selection criteria come determine optimal Linear, The penalty factor and kernel functional parameter γ value of Polynomial, RBF and Sigmoid kernel function.
Further improvement lies in that: first forecast set data in step 2 are substituted into above-mentioned steps four in the step 5 and are established Various combined models in predicted, SVC kernel function and its parameter, most are then determined as standard using forecast set accuracy The model of good main gene number, it is optimal selection that accuracy highest kernel function is selected in the case where identical main gene number, Forecast set accuracy of the best kernel function model in different main gene numbers is analyzed again, then adds main gene number, Select forecast set accuracy be no longer significantly increased or reach expected requirement when main gene number be used as best main cause Subnumber.
The invention has the benefit that the present invention passes through the blade to multiple and different kinds, different genotype ramie sample The leaf characteristic of a variety of ramie samples is studied and compared to bloom spectral property, and principal component analysis PCA effect can be improved just True rate, and the hair method of contribution rate is accumulated to determine best main gene by tradeoff model accuracy, the dynamics of dimensionality reduction and variance Number, and determine that extract feature establishes Linear, Polynomial, RBF and Sigmoid core respectively by using grid data service The method of the SVC ramie EO-1 hyperion variety ecotype model of function, then using forecast set accuracy as standard, available optimal ramie Numb variety ecotype model, in the reasonable situation of selection parameter, model recognition correct rate can achieve 95% or more, the method for the present invention Have the advantages that reliable and effective and quick, easy applied to ramee variety identification, improves the ramee variety based on EO-1 hyperion Identification, assistant breeding, the high yield and high quality to realize ramie and numb field Precision management theoretical foundation and key technology support, can be with Shorten ramee variety recognition cycle, reduces manpower and material resources consumption, while cost can be shortened, be suitable for large batch of ramee variety Identification.
Detailed description of the invention
Fig. 1 is that the present invention is based on EO-1 hyperions and SVC ramee variety to identify modeling procedure figure.
Fig. 2 is that the present invention is different principal component numbers, different kernel function SVC Model checking result schematic diagrams.
Fig. 3 is the accumulation contribution rate schematic diagram of 20 principal components before the present invention.
Specific embodiment
In order to deepen the understanding of the present invention, the present invention is further described below in conjunction with embodiment, the present embodiment For explaining only the invention, it is not intended to limit the scope of the present invention..
According to Fig. 1,2 and 3, the present embodiment proposes the ramie product of a kind of combination EO-1 hyperion and support vector cassification Kind recognition methods, comprising the following steps:
Step 1: sample collection and hyperspectral measurement
Sample collection using local varieties, Cultivars, different root and stem of certain plants type, different ripening stages and different output sample, sample receives Collection selection Jinsha Chinese holly skin fiber crops, Bijie circle fiber crops, Xiang Tan chicken bone is white, the yellow shell of Yuanjiang is early, Pingtang machete fiber crops, middle ramie 1, Shaoyang 4, double Peak great Ye fiber crops and Suining piemarker totally 9 kinds of ramee varieties, as ramie sample, and in the prosperous long-term acquisition ramie sample of ramie sample Blade high-spectral data, each kind acquires 162 high light datas of blade, and 9 kind symbiosis are high at 1458 blade samples Then ramie sample is used portable field spectroradiometer and the matched hand-held leaf clip of portable field spectroradiometer by spectroscopic data Leaf spectra detector carries out high-spectral data collection in ramie leaf samples, when high-spectral data collection, avoids ramie sample The main lobe arteries and veins of this blade, first by the blade clamp holder clamping quilt of hand-held leaf clip leaf spectra detector when high-spectral data collection Survey the surveyed position of blade of ramie sample, then the blade of the probe measurement ramie sample with hand-held leaf clip leaf spectra detector EO-1 hyperion, for the blade Samples selecting of ramie sample on blade main lobe arteries and veins both sides, blade main lobe arteries and veins both sides respectively select 2 samplings Point, selects altogether 4 sampled points, and sample point data does the blade bloom being averaged after breakpoint correction again as the ramie sample Modal data；
Step 2: sample set divides
In the above step 1 after the blade high-spectral data collection of the ramie sample of different cultivars, by the ramie sample of different cultivars This blade high-spectral data is randomly assigned according to the ratio of 2:1, is successively labeled as into modeling collection and forecast set, different cultivars For the blade high-spectral data modeling collection of ramie sample for establishing variety ecotype model, the blade of the ramie sample of different cultivars is high Spectroscopic data forecast set is not involved in modeling, is only used for the accuracy rate of evaluation and test model；
Step 3: PCA feature extraction
The leaf characteristic that ramie sample is extracted using PCA selects ingredient of all characteristic values greater than 1 as PCA main gene or side Poor contribution rate of accumulative total reaches 90% preceding 20 main genes as PCA main gene, and the leaf characteristic of ramie sample is extracted using PCA When contribution rate judging result is accumulated by variance based on, it is appropriate to increase main gene number, eventually by choosing comprehensively model prediction The accuracy of collection, the dynamics of dimensionality reduction and variance accumulation contribution rate etc. are because usually determining main gene number；
Step 4: SVC identification model is established
Utilize SVC algorithm by the PCA main cause subcharacter after extracting in above-mentioned steps three in stoichiometry software Unscrambler Variable successively increases PCA main gene number since variance accumulates contribution rate >=85% 20th main gene, then respectively and not The optimal parameter of same SVC kernel function and kernel function is combined, and multiple identification ramee varieties are established after then being analyzed Model obtains best features variable number, SVC kernel function and parameter combination, in order to avoid the kernel function and parameter of SVC algorithm are set It sets and the precision of model is had an impact with complexity, while in order to find best kernel function and parameter, needing using different SVC Kernel function and kernel function establish ramee variety identification mould, specially four kinds different SVC kernel functions, including Linear, Polynomial, RBF and Sigmoid kernel function, Linear, Polynomial, RBF and Sigmoid kernel function establish ramie product When kind of identification model, use grid data service and cross validation accuracy for selection criteria come determine optimal Linear, The penalty factor and kernel functional parameter γ value of Polynomial, RBF and Sigmoid kernel function；
Step 5: best model is determined
First forecast set data in step 2 are substituted into the various combined models established in above-mentioned steps four and predicted, then The model for determining SVC kernel function and its parameter, best main gene number as standard using forecast set accuracy, in identical main gene It is optimal selection that the highest kernel function of accuracy is selected in the case where number, then analyzes best kernel function model in different main genes Forecast set accuracy in the case of number, then add main gene number, select forecast set accuracy be no longer significantly increased or Person reach expected requirement when main gene number be used as best number of main factor.
Forecast set data are substituted into above-mentioned steps four after being predicted in the various combined models established, it is different it is main at Number, different kernel function SVC models is divided to differentiate that accuracy is as shown in table 1 below in detail in modeling collection and forecast set:
Different principal component numbers, different kernel function SVC Model checking accuracy (%)
Table 1
It follows that the SVC ramee variety of Linear, Polynomial, RBF and Sigmoid kernel function is high according to upper table 1 In spectrum discrimination model, RBF kernel function model effect is best, overall to be higher than other three kinds.
Selecting the forecast set effect of 20 principal components is best, modeling collection and pre- in similar kernel function SVC model Survey collects overall and each kind and determines that result is as shown in table 2 below:
20 principal component parameter SVC differentiate accuracy (%)
Table 2
It can be concluded that, choosing comprehensively accuracy and calculation amount, 20 principal components and RBF method are best according to upper table 2 Selection, forecast set accuracy are 96.91%.
PCA is carried out to modeling collection sample to analyze, and takes preceding 20 principal components PC, the detailed contributions rate such as following table of each ingredient Shown in 3:
The accumulation contribution rate of preceding 20 principal components
Table 3
It can be concluded that, it is that contribution rate is maximum in all PC that the 1st PC contribution rate, which is 75.78%, as shown in upper table 3；Preceding 2 PC It accumulates contribution rate to increase rapidly, the 2nd PC accumulation contribution rate is 86.68%, and each PC accumulation contribution rate is slowly increased later；Preceding 20 A principal component accumulation contribution rate is to 99.98%, and only surplus 0.02% spectral information fails to express.
9 different genotype ramee variety sample sets are divided, dividing condition is as shown in table 4 below:
9 different genotype ramee variety sample set dividing conditions
Table 4
The present invention is studied and is compared by the blade bloom spectral property to multiple and different kinds, different genotype ramie sample The accuracy of principal component analysis PCA effect can be improved in the leaf characteristic of a variety of ramie samples, and correct by tradeoff model Rate, the dynamics of dimensionality reduction and variance accumulate the hair method of contribution rate to determine best main gene number, and by using grid search Method, which determines, extracts the SVC ramie EO-1 hyperion kind that feature establishes Linear, Polynomial, RBF and Sigmoid kernel function respectively The method of identification model, then using forecast set accuracy as standard, available optimal ramee variety identification model, selection parameter In reasonable situation, model recognition correct rate can achieve 95% or more, and the method for the present invention has applied to ramee variety identification can By effective and quick, easy advantage, improves the ramee variety identification based on EO-1 hyperion, assistant breeding, is to realize ramie High yield and high quality and numb field Precision management theoretical foundation and key technology support, ramee variety recognition cycle can be shortened, subtracted Few manpower and material resources consumption, while cost can be shortened, it is suitable for large batch of ramee variety and identifies.
The basic principles, main features and advantages of the invention have been shown and described above.The technical staff of the industry should Understand, the present invention is not limited to the above embodiments, and the above embodiments and description only describe originals of the invention Reason, without departing from the spirit and scope of the present invention, various changes and improvements may be made to the invention, these changes and improvements It all fall within the protetion scope of the claimed invention.The claimed scope of the invention is by appended claims and its equivalent circle It is fixed.

Claims (10)

1.一种结合高光谱和支持向量机分类的苎麻品种识别方法，其特征在于：包括以下步骤：1. a ramie species identification method in conjunction with hyperspectral and support vector machine classification, is characterized in that: comprise the following steps: 步骤一：样本收集及高光谱测量Step 1: Sample collection and hyperspectral measurement 收集不同品种的苎麻，作为苎麻样本，然后将苎麻样本采用便携式地物光谱仪和便携式地物光谱仪配套的手持叶夹式叶片光谱探测器在苎麻样本叶片上选择4个采样点测量进行高光谱数据采集，采样点数据做断点校正后再取平均值作为该苎麻样本的叶片高光谱数据；Collect different varieties of ramie as ramie samples, and then select 4 sampling points on the ramie sample leaves to measure the ramie samples for hyperspectral data collection by using a portable ground object spectrometer and a hand-held leaf clip leaf spectral detector matched with the portable ground object spectrometer. , the sampling point data is corrected by breakpoints, and then the average value is taken as the leaf hyperspectral data of the ramie sample; 步骤二：样本集划分Step 2: Divide the sample set 在上述步骤一中不同品种的苎麻样本的叶片高光谱数据采集后，将不同品种的苎麻样本的叶片高光谱数据按照2:1的比例随机分配，依次标记为成建模集和预测集；After the leaf hyperspectral data of different varieties of ramie samples are collected in the above step 1, the leaf hyperspectral data of different varieties of ramie samples are randomly allocated according to the ratio of 2:1, and marked as a modeling set and a prediction set in turn; 步骤三：PCA特征提取Step 3: PCA Feature Extraction 采用PCA提取苎麻样本的叶片特征，选择所有特征值大于1的成分作为PCA主因子或方差累计贡献率达到85%-95%的前n个主因子作为PCA主因子；The leaf characteristics of ramie samples were extracted by PCA, and all components with eigenvalues greater than 1 were selected as PCA main factors or the top n main factors with cumulative variance contribution rate of 85%-95% were selected as PCA main factors; 步骤四：SVC识别模型建立Step 4: Establish SVC recognition model 在化学计量软件Unscrambler中利用SVC算法将上述步骤三中提取后的PCA主因子特征变量从方差累积贡献率≥85%第n个主因子开始，依次增加PCA主因子个数，然后分别与不同的SVC核函数及核函数的最佳参数进行组合，然后进行分析后建立多个识别苎麻品种的模型，获得最佳特征变量个数、SVC核函数和参数组合；Using the SVC algorithm in the chemometric software Unscrambler, the characteristic variables of the PCA main factors extracted in the above step 3 start from the nth main factor whose variance cumulative contribution rate is greater than or equal to 85%, and increase the number of PCA main factors in turn. The SVC kernel function and the best parameters of the kernel function are combined, and after analysis, several models for identifying ramie varieties are established, and the optimal number of characteristic variables, SVC kernel function and parameter combination are obtained; 步骤五：确定最佳模型Step 5: Determine the best model 将步骤二中的预测集的预测数据代入步骤四建立的各种组合的模型中，评价并识别预测结果，获取最佳参数组合的模型。Substitute the prediction data of the prediction set in step 2 into the models of various combinations established in step 4, evaluate and identify the prediction results, and obtain the model with the best parameter combination. 2.根据权利要求1所述的一种结合高光谱和支持向量机分类的苎麻品种识别方法，其特征在于：所述步骤一中样本收集采用地方品种、选育品种、不同蔸型、不同成熟期和不同产量的样本，样本收集选择9种苎麻品种，并且在苎麻样本旺长期采集苎麻样本的叶片高光谱数据，每个品种采集162个叶片高光数据，9个品种共生成1458个叶片样本高光谱数据。2. a kind of ramie variety identification method combining hyperspectral and support vector machine classification according to claim 1, it is characterized in that: in the described step 1, sample collection adopts local varieties, breeding varieties, different types, different mature 9 ramie varieties were selected for sample collection, and the leaf hyperspectral data of ramie samples were collected in the long-term of ramie samples. 162 leaf highlight data were collected for each variety, and a total of 1458 leaf samples were generated for 9 varieties. Spectral data. 3.根据权利要求2所述的一种结合高光谱和支持向量机分类的苎麻品种识别方法，其特征在于：9种所述苎麻品种分别为金沙枸皮麻、毕节圆麻、湘潭鸡骨白、沅江黄壳早、平塘大刀麻、中苎1号、邵阳4号、双峰大叶麻和绥宁青麻。3. a kind of ramie variety identification method in conjunction with hyperspectral and support vector machine classification according to claim 2, is characterized in that: 9 kinds of described ramie varieties are respectively Jinsha wolf skin hemp, Bijie round hemp, Xiangtan chicken bone white , Yuanjiang Huanghuzao, Pingtang Da Dao Ma, Zhong Ramie No. 1, Shaoyang No. 4, Shuangfeng Da Ye Ma and Suining Qing Ma. 4.根据权利要求1所述的一种结合高光谱和支持向量机分类的苎麻品种识别方法，其特征在于：所述步骤一中高光谱数据采集时，避开苎麻样本叶片的主叶脉，高光谱数据采集时先将手持叶夹式叶片光谱探测器的叶片夹持器夹紧被测苎麻样本的叶片所测部位，再用手持叶夹式叶片光谱探测器的探头测定苎麻样本的叶片高光谱，苎麻样本的叶片采样点选择在叶片主叶脉两边，叶片主叶脉两边各选择2个采样点，一共选择4个采样点，采样点数据做断点校正后再取平均值作为该苎麻样本的叶片高光谱数据。4. a kind of ramie variety identification method combining hyperspectral and support vector machine classification according to claim 1, is characterized in that: during the hyperspectral data collection in described step 1, avoid the main vein of ramie sample leaf, hyperspectral During data collection, the blade holder of the hand-held leaf-clamp type leaf spectral detector was first clamped to the measured part of the leaf of the ramie sample to be tested, and then the probe of the hand-held leaf-clamp type blade spectral detector was used to measure the leaf hyperspectrum of the ramie sample. The leaf sampling points of the ramie sample are selected on both sides of the main leaf vein, and 2 sampling points are selected on each side of the main leaf vein, and a total of 4 sampling points are selected. Spectral data. 5.根据权利要求1所述的一种结合高光谱和支持向量机分类的苎麻品种识别方法，其特征在于：所述步骤一中为了消除苎麻样本的叶片高光谱数据在采集时首端与末端产生的噪音，苎麻样本的叶片高光谱数据采集时选择420nm-2450nm之间的光谱数据进行分析，并且便携式地物光谱仪每半小时做一次OPT优化和白板参比。5. a kind of ramie variety identification method combining hyperspectral and support vector machine classification according to claim 1, is characterized in that: in the described step 1, in order to eliminate the leaf hyperspectral data of ramie sample when collecting head end and end For the noise generated, the spectral data between 420nm-2450nm was selected for analysis when collecting the hyperspectral data of the leaves of ramie samples, and the portable ground object spectrometer did OPT optimization and whiteboard reference every half hour. 6.根据权利要求1所述的一种结合高光谱和支持向量机分类的苎麻品种识别方法，其特征在于：所述步骤二中不同品种的苎麻样本的叶片高光谱数据建模集用于建立品种识别模型，不同品种的苎麻样本的叶片高光谱数据预测集不参与建模，仅用于评测模型的准确率。6. A method for identifying ramie varieties combining hyperspectral and support vector machine classification according to claim 1, characterized in that: in the step 2, the leaf hyperspectral data modeling set of the ramie samples of different varieties is used to establish Variety identification model, the leaf hyperspectral data prediction set of ramie samples of different varieties does not participate in the modeling, and is only used to evaluate the accuracy of the model. 7.根据权利要求1所述的一种结合高光谱和支持向量机分类的苎麻品种识别方法，其特征在于：所述步骤三中采用PCA提取苎麻样本的叶片特征时以方差累积贡献率判断结果为基础，适当增加主因子个数，最终通过综合权衡模型预测集的正确率、降维的力度和方差累积贡献率等因素来确定主因子个数。7. a kind of ramie species identification method combining hyperspectral and support vector machine classification according to claim 1, is characterized in that: when adopting PCA to extract the leaf feature of ramie sample in described step 3, judge result with variance cumulative contribution rate Based on this, the number of main factors is appropriately increased, and finally the number of main factors is determined by comprehensively weighing factors such as the correct rate of the model prediction set, the strength of dimensionality reduction, and the cumulative contribution rate of variance. 8.根据权利要求1所述的一种结合高光谱和支持向量机分类的苎麻品种识别方法，其特征在于：所述步骤四中为了避免SVC算法的核函数和参数设置对模型的精度与复杂性产生影响，同时为了找到最佳核函数和参数，需要采用不同的SVC核函数及核函数建立苎麻品种识别模，具体为四种不同的SVC核函数，包括Linear、Polynomial、RBF和Sigmoid核函数。8. a kind of ramie species identification method in conjunction with hyperspectral and support vector machine classification according to claim 1, is characterized in that: in the described step 4, in order to avoid the kernel function of SVC algorithm and parameter setting to the precision and complexity of model At the same time, in order to find the best kernel function and parameters, it is necessary to use different SVC kernel functions and kernel functions to establish the ramie variety identification model, specifically four different SVC kernel functions, including Linear, Polynomial, RBF and Sigmoid kernel functions . 9.根据权利要求8所述的一种结合高光谱和支持向量机分类的苎麻品种识别方法，其特征在于：所述Linear、Polynomial、RBF和Sigmoid核函数建立苎麻品种识别模型时，采用网格搜索法和交叉验证正确率为选择标准来确定最佳的Linear、Polynomial、RBF和Sigmoid核函数的惩罚因子C和核函数参数γ值。9. a kind of ramie species identification method in conjunction with hyperspectral and support vector machine classification according to claim 8, is characterized in that: when described Linear, Polynomial, RBF and Sigmoid kernel function set up ramie species identification model, adopt grid The search method and cross-validation accuracy rate are the selection criteria to determine the optimal penalty factor C and kernel function parameter γ value of Linear, Polynomial, RBF and Sigmoid kernel functions. 10.根据权利要求1所述的一种结合高光谱和支持向量机分类的苎麻品种识别方法，其特征在于：所述步骤五中先将步骤二中预测集数据代入上述步骤四中建立的各种组合的模型中进行预测，然后以预测集正确率为标准来确定SVC核函数及其参数、最佳主因子个数的模型，在相同主因子个数的情况下选择正确率最高的核函数为最佳选择，再分析最佳核函数模型在不同主因子个数情况下的预测集正确率，然后添加主因子个数，选择预测集正确率不再有显著提高或者达到预期要求的时的主因子个数作为为最佳主因子数。10. The method for identifying ramie varieties combined with hyperspectral and support vector machine classification according to claim 1, wherein in the step 5, the prediction set data in the step 2 is first substituted into each of the established in the above-mentioned step 4. Then, the prediction set accuracy rate is used to determine the SVC kernel function and its parameters, the model with the best number of main factors, and the kernel function with the highest accuracy rate is selected under the same number of main factors. For the best choice, re-analyze the prediction set accuracy rate of the optimal kernel function model under different number of main factors, then add the number of main factors, and select the one when the prediction set accuracy rate is no longer significantly improved or meets the expected requirements. The number of main factors is taken as the optimal number of main factors.
CN201811242492.XA 2018-10-24 2018-10-24 A kind of ramee variety recognition methods of combination EO-1 hyperion and support vector cassification Pending CN109387484A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201811242492.XA CN109387484A (en) 2018-10-24 2018-10-24 A kind of ramee variety recognition methods of combination EO-1 hyperion and support vector cassification

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201811242492.XA CN109387484A (en) 2018-10-24 2018-10-24 A kind of ramee variety recognition methods of combination EO-1 hyperion and support vector cassification

Publications (1)

Publication Number Publication Date
CN109387484A true CN109387484A (en) 2019-02-26

Family

ID=65427710

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201811242492.XA Pending CN109387484A (en) 2018-10-24 2018-10-24 A kind of ramee variety recognition methods of combination EO-1 hyperion and support vector cassification

Country Status (1)

Country Link
CN (1) CN109387484A (en)

Cited By (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN111272668A (en) * 2020-01-22 2020-06-12 中国农业科学院农产品加工研究所 Construction method of wheat variety identification model
CN111476320A (en) * 2020-05-09 2020-07-31 河南科技大学 A method for establishing a discrimination model for fertilized eggs in early hatching based on visible and near-infrared spectroscopy
CN114994150A (en) * 2022-05-31 2022-09-02 中国标准化研究院 Electronic tongue rapid classification method for spicy degree of zanthoxylum schinifolium

Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2013052824A1 (en) * 2011-10-05 2013-04-11 Cireca Theranostics, Llc Method and system for analyzing biological specimens by spectral imaging
CN104794368A (en) * 2015-05-15 2015-07-22 哈尔滨理工大学 Rolling bearing fault classifying method based on FOA-MKSVM (fruit fly optimization algorithm-multiple kernel support vector machine)
WO2018112441A1 (en) * 2016-12-15 2018-06-21 Progenity Inc. Ingestible device and associated methods

Patent Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2013052824A1 (en) * 2011-10-05 2013-04-11 Cireca Theranostics, Llc Method and system for analyzing biological specimens by spectral imaging
CN104794368A (en) * 2015-05-15 2015-07-22 哈尔滨理工大学 Rolling bearing fault classifying method based on FOA-MKSVM (fruit fly optimization algorithm-multiple kernel support vector machine)
WO2018112441A1 (en) * 2016-12-15 2018-06-21 Progenity Inc. Ingestible device and associated methods

Non-Patent Citations (3)

* Cited by examiner, † Cited by third party
Title
全东平 等: "基于流形学习算法柑橘叶片高光谱快速分类的研究", 《万方会议论文》 *
曹晓兰等: "高光谱参数和逐步判别的苎麻品种识别", 《光谱学与光谱分析》 *
王化: "高光谱技术在茶叶品种检测中的应用", 《科技创新导报》 *

Cited By (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN111272668A (en) * 2020-01-22 2020-06-12 中国农业科学院农产品加工研究所 Construction method of wheat variety identification model
CN111476320A (en) * 2020-05-09 2020-07-31 河南科技大学 A method for establishing a discrimination model for fertilized eggs in early hatching based on visible and near-infrared spectroscopy
CN111476320B (en) * 2020-05-09 2023-03-24 河南科技大学 Visible near infrared spectrum-based method for establishing early incubation fertilized egg discrimination model
CN114994150A (en) * 2022-05-31 2022-09-02 中国标准化研究院 Electronic tongue rapid classification method for spicy degree of zanthoxylum schinifolium
CN114994150B (en) * 2022-05-31 2023-10-27 中国标准化研究院 An electronic tongue rapid classification method for red pepper numbness

Similar Documents

Publication Publication Date Title
CN105630743B (en) A kind of system of selection of spectrum wave number
CN109387484A (en) A kind of ramee variety recognition methods of combination EO-1 hyperion and support vector cassification
CN108181263B (en) Tobacco leaf position feature extraction and discrimination method based on near infrared spectrum
CN104502283A (en) Two-band hyperspectral index and prediction model for estimating yield and shoot dry weight of soybean
CN112539785A (en) Tobacco grade identification system and method based on multi-dimensional characteristic information
CN108844917A (en) A kind of Near Infrared Spectroscopy Data Analysis based on significance tests and Partial Least Squares
CN114092839B (en) Unmanned aerial vehicle remote sensing-based soybean harvest period maturity judging method
CN102998438A (en) Maturity analysis and judgment method for field tobacco leaves
CN111562235A (en) Method for rapidly identifying black-leaf outbreak disease and infection degree of tobacco leaves based on near infrared spectrum
CN114694023A (en) Cotton aphid severity grading method based on spectral index data reconstruction
CN104255118B (en) Rapid lossless testing method based on near infrared spectroscopy technology for paddy rice seed germination percentage
CN119470828B (en) Traditional chinese medicine medicinal material true and false identification platform
CN112414967B (en) A near-infrared quality control method for rapid real-time detection of Puhuang charcoal processing
CN109685099B (en) Apple variety distinguishing method based on spectrum band optimization fuzzy clustering
CN104568771A (en) Economical, portable and efficient grading method for maturity of fresh tobacco leaves of flue-cured tobaccos
CN115436318A (en) A method for identifying species of Cinnamon based on near-infrared spectroscopy
CN109146270B (en) A method for establishing the robustness index of vegetable seedlings
CN114324195A (en) Multichannel agricultural product nondestructive testing method
CN111795932B (en) Hyperspectrum-based nondestructive testing method for sugar acidity of waxberry fruits
CN113919690A (en) Tobacco leaf field maturity state evaluation method and application
CN109374574A (en) A method of identifying the sense of cured tobacco leaf wax using near infrared light spectrum information
CN113267458A (en) Method for establishing quantitative prediction model of soluble protein content of sweet potatoes
CN106404699A (en) Nondestructive measurement method for contents of nitrogen elements of pear tree leaf blades
CN104897608B (en) A method for quality identification of oolong tea based on near-infrared spectroscopy
CN117990646A (en) A tobacco leaf quality evaluation and comparison system and method

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