CN102426153B - A kind of Wheat plant moisture monitoring method based on canopy high spectral index - Google Patents

Abstract

The present invention relates to a kind of Wheat plant moisture monitoring method based on canopy high spectral index, 2 years wheat ponds under 2 years 2 kinds, 4 different in moisture process are utilized to plant test figure, adopt the meticulous sampling method of decrement, analyze any hyperspectral index of band combination and the quantitative relationship of wheat plant water cut and leaf layer water cut between two of original spectrum and spectrum reciprocal in 350 ~ 2500nm wavelength band, found that based on original spectrum NDVI (R ₈₃₆, R ₇₉₃) and spectrum RVI (RC reciprocal ₈₃₇, RC ₇₉₃) wheat plant moisture can be monitored; Based on original spectrum NDVI (R ₁₁₀₀, R ₇₇₀) and RVI (R ₈₉₃, R ₈₀₅) wheat leaf layer moisture can be monitored.Research conclusion of the present invention provides new band combination and theoretical foundation for utilizing high-spectral data quick nondestructive to monitor wheat water content situation.

Description

A kind of Wheat plant moisture monitoring method based on canopy high spectral index

Technical field

The present invention relates to Wheat plant moisture monitoring, be specifically related to a kind of Wheat plant moisture monitoring method based on canopy high spectral index.

Background technology

Wheat is one of most important cereal crops in the world.The whole world has the population of 35% ~ 40% to take wheat as staple food.China's wheat annual production is about 100,000,000 tons, accounts for 22% of national total output of grain, accounts for 20% of world wheat total production.Moisture is the element factor of crop vital movement, and green plants water cut can reach more than 80% ~ 90%.Lack of water occurs the form of crop, physiology course all has an impact, and finally makes output reduce.Implementing precision irrigation according to crop water status is the important channel of improving efficiency of water application and water production efficiency.

In recent years, spectral remote sensing technology develops rapidly and opens new way for obtaining crop water status.Research shows, utilizes during Crop water deficits and causes the response characteristic of the changes such as blade interior physiological ecological and external morphological structure on EO-1 hyperion, can obtain crop water information quickly and accurately.Research finds, near 970mm, 1450mm and 1940mm wave band, the peak energy of the spectral reflectivity of the plants such as wheat, gerbera and soybean reflects the water regime of blade preferably, therefore, the vegetation index be made up of visible ray and near infrared region wave band can be used for the monitoring of plant moisture situation.The research such as Gregory thinks that the primary effect of moisture to spectrum is the direct radiation-absorbing of hydrone, and secondary influences is that moisture causes blade interior structure to change, and primary effect effect is much larger than secondary influences effect.Gao shows by analyzing the impact of Vegetation canopy scattering spectrum on moisture, and NDWI can comparatively monitor index canopy moisture.Tian Qing research of waiting so long finds, the characteristic absorption peak degree of depth near wheat leaf blade relative water content and 1450mm and area present good linear dependence.Gu Yanfang etc. and the achievement in research of Wang Jihua etc. confirm and utilize spectral reflectivity can the feasibility of Accurate Prediction water content in plant leaf.Field forever superfine research finds, ratio vegetation index R ₈₁₀/ R ₄₆₀the water percentage of different growth stage rice plant and blade can be monitored preferably; Find based on crop canopies spectral vegetation indexes RVI simultaneously _(610,560)/ NDVI _(810,610)wheat vegetation water regime can be predicted.Ah not all gas carries me and draws wood etc. to research and propose to can be used for the short-wave infrared vertical dehydration index SPSI monitoring large scale vegetation water content, and the precision of monitoring reaches 74%.

Along with the develop rapidly of high spectrum resolution remote sensing technique, in recent years, the research report both at home and abroad about this respect is more and more.Ceccato etc. propose and utilize spherical vegetation humidity index (GVMI) inverting vegetation water content; Field superfine research forever finds that R (610,560)/ND (810,610) is the vegetation index predicting that wheat plant water regime is good; The research such as Kakani finds R1689/R1657 and outdoor Growth of Potted Cotton leaf water potential height correlation; Lucky Umihiko etc. measures Wheat Leaves reflectance spectrum within the scope of 1400 ~ 1600nm, establishes the model of moisture and reflectance spectrum by partial least square method.To sum up, existing research is always most to be carried out based on multispectral data, spectral resolution is lower, wavelength band is less, and based in the analysis of EO-1 hyperion, the institute's likely high-spectral data within the scope of less analysis 350 ~ 2500nm, and spectrum is not through pre-service, noise is comparatively large, and the sensitive band of some reflection vegetation moisture just may be caused not yet fully to be excavated for this or result has deviation.Therefore, be necessary to adopt and sample and analytical approach, the sensitive band that exploration discovery is new and hyperspectral index thereof with meticulous EO-1 hyperion more comprehensively.Simultaneously, because EO-1 hyperion magnanimity information extracts and the reason of data preprocessing method, Quantitative Monitoring model accuracy based on spectral technique structure is lower or can only be applied to specific breeding time, or model is slightly complicated in structure and algorithm, cause the universality of monitoring model poor, weaken its application to a certain extent.Explore new core bands, build simple and reliable spectrum index is the Focal point and difficult point that spectrum monitoring is applied in agricultural remote sensing field always.

In high-spectral data is analyzed, in order to reduce the impact of background noise, people adopt various technical finesse spectroscopic data, to improve the accuracy of spectral information.Deepening continuously about the sensitive band selection technique of wheat plant and canopy leaves water cut and method always.The research of Shibayama etc. shows, can diagnose double cropping of rice canopy water regime by the first order derivative of the spectral reflectivity of near infrared spectrum (1190 ~ 1320nm) or short infrared wave band (1600nm).Dobrowski etc. find that the canopy spectra at 690nm and 740nm place can reflect that plant is subject to the state of water stress, the research such as Zhang Jiahua finds 469nm, 645nm, 700nm and 710nm wave band of visible region, and 760nm, 815nm, 855nm, 930nm, 1075nm and 1100nm wave band of near infrared region and 1550nm, 1600nm, 1640nm, 1750nm and 2130nm wave band in short-wave infrared region are the sensitive bands of detection leaf water content change; Graeff etc. analyze the wheat leaf blade spectroscopic data under 6 kinds of different in moisture process, find 510 and 780nm, 540 and 780nm, 490 and 1300nm and 540 and 1300nm be the water physiological mechanism representing wheat the best instruction wave band.

Summary of the invention

The object of the invention is the deficiency existed for prior art, the Wheat plant moisture monitoring method based on canopy high spectral index is provided.

Utilize high-spectral data analytical approach and technology, the normalization of all combinations be made up of original spectrum reflectivity, first order derivative, inverse logarithm and inverse in multianalysis 350 ~ 2500nm wavelength band and Ratio index, explore the new sensitive band of instruction wheat plant moisture, and the wheat plant moisture content index monitoring model set up based on EO-1 hyperion parameter, its technical scheme is:

Based on a Wheat plant moisture monitoring method for Canop hyperspectrum, the real time measure soil moisture content, concrete steps are as follows:

1) hyperspectral information obtains

Utilize spectrometer measurement wheat canopy hyperspectral index data, significant wave segment limit is 350 ~ 2500nm, and wherein 350 ~ 1050nm spectrum sample is spaced apart 1.4nm, and spectral resolution is 3nm; 1050 ~ 2500nm spectrum sample is spaced apart 2nm, and spectral resolution is 10nm, and spectroscopic assay is selected to carry out when ceiling unlimited, calm or gentle breeze, and minute is 10:00-14:00.During measurement, sensor probe is vertically downward in canopy top, and spectrometer field angle is 25 °, highly about 1.0m, ground field range diameter is 0.44m, each cell measurement 10 sampling points, each sampling point duplicate measurements 5 times (visual field), the spectral reflectance value using its mean value as this community;

2) plant determination of moisture

With step 1) synchronous, at different growth stage, the strain of representative wheat 20 is got in each community, first press organ to be separated, again blade is separated by various position leaves, laboratory is taken back in the valve bag that rapid loading is weighed, its fresh weight is claimed with ten thousand/precision electronic balance, then weigh after drying to constant weight at 80 DEG C after putting into baking oven 105 DEG C of 30min that complete, obtain different parts organ dry weight, calculate plant water content (Plant water content respectively, PWC), leaf water content (Leaf water content, and leaf layer water cut (Canopy leaf water content LWC), CLWC), formula is as follows:

PWC(％)＝(PFW-PDW)/PFW×100％

LWC(％)＝(LFW-LDW)/LFW×100％

CLWC(％)＝(∑LFW-∑LDW)/∑LFW×100％

Wherein, PFW (Plant Fresh Weight) is plant fresh weight, PDW (Plant Dry Weight) is plant weights, and LFW (Leaf FreshWeight) is fresh weight, and LDW (Leaf Dry Weight) is leaf dry weight;

3) vegetation index builds

Original spectrum NDVI=(R _{λ 1}-R _{λ 2})/(R _{λ 1}+ R _{λ 2}) RVI=R _{λ 1}/ R _{λ 2}

Spectrum NDVI=(RC reciprocal _{λ 1}-RC _{λ 2})/(RC _{λ 1}+ RC _{λ 2}) RVI=RC _{λ 1}/ RC _{λ 2}

Wherein R _{λ 1}and R _{λ 2}be respectively the reflectivity of any two wavelength, the scope of λ 1 and λ 2 is 350 ~ 2500nm, FD _{λ 1}and FD _{λ 2}for its corresponding first derivative spectrum, AL _{λ 1}and AL _{λ 2}for its corresponding inverse logarithm spectrum, RC _{λ 1}and RC _{λ 2}for its corresponding spectrum reciprocal, according to normalized differential vegetation index NDVI (Normalized Difference Vegetation Index) and ratio vegetation index RVI (Ratio Vegetation Index) the estimation ability to wheat water content content, filter out the result behaved oneself best.

4) data analysis and utilization

Utilize step 3) in data, comprehensive analysis plant water content and leaf water content and the direct quantitative relationship of canopy spectra reflectivity, adopt the meticulous sampling method of high-spectral data decrement, filter out the wavelength band to moisture-sensitive and spectrum parameter, and build moisture monitoring model;

With the correlationship of plant water content and leaf layer water cut between further analysis wheat canopy original spectrum and first derivative spectrum thereof, spectrum reciprocal, inverse logarithm spectrum;

Build the plant water content quantitative model based on wheat canopy hyperspectral index:

Utilize step 1) and step 2) testing data that obtains, calculate the coefficient of determination R of NDVI and RVI that the original spectrum of all any two band combinations in 350-2500nm wavelength band, first derivative spectrum, inverse logarithm spectrum and spectrum reciprocal forms and plant water content PWC ²;

Build the leaf layer water cut quantitative model based on wheat canopy hyperspectral index:

NDVI and RVI adopting the original spectrum of all any band combinations between two in same method calculating 350-2500nm wavelength band, first derivative spectrum, inverse logarithm spectrum and spectrum reciprocal to form and the coefficient of determination R of leaf layer water cut CLWC ²;

5) structure of monitoring model and inspection

Adopt relative root-mean-square deviation (RRMSE) to carry out assay, wherein RRMSE computing formula is as follows:

$\mathrm{RRMSE}=\sqrt{\frac{1}{n}\×\underset{i=1}{\overset{n}{\mathrm{\Σ}}}{(P_i-O_i)}^2\×\frac{100}{\stackrel{-}{O_i}}}$

O in above formula _ifor the moisture content value of observing in test, P _ifor the moisture content value of model estimation, n is model measurement test samples number;

The test and check of wheat plant water cut monitoring model:

In order to check reliability and the applicability of wheat water content monitoring model, utilize step 1) and step 2) the independent experiment data that obtains carries out test and check to above-mentioned model, utilizes the coefficient of determination R between predicted value and observed reading simultaneously ², these 3 indexs of relative error RE and relative root-mean-square deviation RRMSE carry out the performance of integrated survey model, and make 1: 1 graph of a relation of predicted value and observed reading, predictive ability of showing model directly perceived;

The test and check of wheat leaf layer water cut monitoring model:

Utilize step 1) and step 2) the monitoring equation of independent experiment data to above-mentioned different spectrum types modeling that obtain test respectively, utilizes the coefficient of determination R between predicted value and observed reading ², these 3 indexs of relative error RE and relative root-mean-square deviation RRMSE carry out integrated survey model, and make 1: 1 graph of a relation of predicted value and observed reading, predictive ability of showing model directly perceived.

Further preferably, Germany is used to produce Portable soil moisture tacheometer TRIME-EZ the real time measure soil moisture content.

Further preferably, step 1) the middle FieldSpec Pro FR2500 type back hanging type field EO-1 hyperion radiation gauge measured wheat canopy hyperspectral index data acquisition and produce with U.S. Analytical SpectralDevice (ASD) company.

Further preferably, step 4) in based on original spectrum NDVI (R ₈₃₆, R ₇₉₃) and spectrum RVI (RC reciprocal ₈₃₇, RC ₇₉₃) the Wheat plant moisture monitoring model that builds, the coefficient of determination (R ²) be respectively 0.851 and 0.852; Based on original spectrum NDVI (R ₁₁₀₀, R ₇₇₀) and RVI (R ₈₉₃, R ₈₀₅) the wheat leaf layer moisture monitoring model that builds, coefficient of determination R ²be respectively 0.730 and 0.812.

Further preferably, step 4) in data analysis and arranging all carry out at Exce12007 and MATLAB7.0 (The Math Works, 2000).

Further preferably, step 5) in the coefficient of determination (R of model measurement ²) being all greater than 0.783, relative root-mean-square deviation RRMSE is all less than 0.205.

Beneficial effect of the present invention:

The wheat pond that the present invention is based under different year, different cultivars, different in moisture gradient and different growth stage plants test, wheat canopy original spectrum reflectivity and the inverse normalized differential vegetation index of band combination, ratio vegetation index and the quantitative relationship with wheat plant water cut, canopy leaves water cut thereof between two arbitrarily within the scope of multianalysis 350 ~ 2500nm, and to establish respectively based on original spectrum NDVI (R ₈₃₆, R ₇₉₃) and spectrum RVI (RC reciprocal ₈₃₇, RC ₇₉₃) plant water content monitoring model; Based on original spectrum NDVI (R ₁₁₀₀, R ₇₇₀) and RVI (R ₈₉₃, R ₈₀₅) leaf layer moisture monitoring model.Utilize independent experiment data to show the assay of model, the coefficient of determination of all models is all greater than 0.783, RRMSE circle in 0.148-0.205, and Stability and veracity is better.These results are that the quick Precise Diagnosis of wheat plant and canopy leaves moisture provides reliable estimation models, exploitation for portable moisture spectromonitor provides critical bands and selects, and utilizes high-spectral data monitoring crop water regime to provide reference for relevant from now on.

Accompanying drawing explanation

Fig. 1 is the Wheat plant moisture monitoring method schematic diagram that the present invention is based on canopy high spectral index;

Fig. 2 wheat plant water cut and original spectrum NDVI (R ₈₃₆, R ₇₉₃) between graph of a relation (n=141);

Fig. 3 wheat plant water cut and spectrum RVI (RC reciprocal ₈₃₇, RC ₇₉₃) between graph of a relation (n=141);

Fig. 4 wheat leaf layer water cut and original spectrum NDVI (R ₁₁₀₀, R ₇₇₀) between quantitative relationship figure (n=141);

Fig. 5 wheat leaf layer water cut and original spectrum RVI (R ₈₉₃, R ₈₀₅) between quantitative relationship figure (n=141);

Fig. 6 is based on original spectrum NDVI (R ₈₃₆, R ₇₉₃) and spectrum RVI (RC reciprocal ₈₃₇, RC ₇₉₃) wheat plant water cut predicted value and 1: 1 graph of a relation (n=112) of observed reading;

Fig. 7 is based on original spectrum NDVI (R ₁₁₀₀, R ₇₇₀) and original spectrum RVI (R ₈₉₃, R ₈₀₅) wheat leaf layer water cut predicted value and 1: 1 graph of a relation (n=112) of observed reading.

Embodiment

Below in conjunction with concrete accompanying drawing and embodiment, method of the present invention is described in more detail.

The present invention has carried out 2 field tests altogether, and relate to different year, different types of varieties and different soils moisture solution, concrete test design is described below.

Embodiment 1: carry out in decorated archway testing station of Agricultural University Of Nanjing (118 ° of 51 ' E, 32 ° of 1 ' N) rainwater-proof mud sump in 2008-2009.Experimental cultivar is Huaihe River wheat 25 and Yang Mai 18.Experimental field soil types is yellowish soil, and soil organic matter content is 1.63%, and total nitrogen content is 0.09%, and available phosphorus contents is 58.3mgkg ^-1, quick-acting potassium content is 91.64mgkg ^-1, PH is 7.20.If four moisture solution: middle drought (W1), light drought (W2), contrast (W3) and heavy irrigation (W4) (being about 40% ~ 45% of field capacity, 60% ~ 65%, 75% ~ 80% and 100% respectively).Plot area is 6m ², random district group arranges, and repeats 3 times.Adopt seed drilling, line-spacing is 25cm, and Basic Seedling is 1,800,000 strain hm ^-2.Nitrogen fertilizer amount is 210kghm ^-2, Dressing ratios is 1: 1, respectively processes to join and executes phosphate fertilizer P ₂o ₅110kghm ^-2with potash fertilizer K ₂o 135kghm ^-2all as base manure.Other management is with conventional field management.This test figure is used for the inspection of model.

Embodiment 2: carry out in the rainwater-proof mud sump of decorated archway testing station of Agricultural University Of Nanjing (118 ° of 51 ' E, 32 ° of 1 ' N) in 2009-2010.Experimental cultivar is Huaihe River wheat 25 and Yang Mai 18.Experimental field soil types is yellowish soil, and soil organic matter content is 1.78%, and total nitrogen content is 0.10%, and available phosphorus contents is 52.7mgkg ^-1, quick-acting potassium content is 95.82mgkg ^-1, PH 7.25.If four moisture solution: middle drought (W1), light drought (W2), contrast (W3) and heavy irrigation (W4) (being about 40% ~ 45% of field capacity, 60% ~ 65%, 75% ~ 80% and 100% respectively).Plot area is 10m ², random district group arranges, and repeats 3 times.Adopt seed drilling, line-spacing is 25cm, and Basic Seedling is 1,800,000 strain hm ^-2.Nitrogen fertilizer amount is 210kghm ^-2, Dressing ratios is 1: 1, respectively processes to join and executes phosphate fertilizer P ₂o ₅110kghm ^-2with potash fertilizer K ₂o 135kghm ^-2all as base manure.Other management is with conventional field management.This test figure is used for the structure of model.

Following five steps is the present invention includes with reference to Fig. 1:

1) hyperspectral information obtains

The FieldSpec Pro FR2500 type back hanging type field EO-1 hyperion radiation gauge that wheat canopy hyperspectral index DATA REASONING adopts Analytical Spectral Device (ASD) company of the U.S. to produce, significant wave segment limit is 350 ~ 2500nm, wherein 350 ~ 1050nm spectrum sample is spaced apart 1.4nm, and spectral resolution is 3nm; 1050 ~ 2500nm spectrum sample is spaced apart 2nm, and spectral resolution is 10nm, and spectroscopic assay is selected to carry out when ceiling unlimited, calm or gentle breeze, and minute is 10:00-14:00.During measurement, sensor probe is vertically downward in canopy top, and spectrometer field angle is 25 °, highly about 1.0m, ground field range diameter is 0.44m, each cell measurement 10 sampling points, each sampling point duplicate measurements 5 times (visual field), the spectral reflectance value using its mean value as this community.

2) plant determination of moisture

With step 1) synchronous, at different growth stage, the strain of representative wheat 20 is got in each community, first press organ to be separated, again blade is separated by various position leaves, laboratory is taken back in the valve bag that rapid loading is weighed, its fresh weight is claimed with ten thousand/precision electronic balance, then weigh after drying to constant weight at 80 DEG C after putting into baking oven 105 DEG C of 30min that complete, obtain different parts organ dry weight, calculate plant water content (Plant water content respectively, PWC), leaf water content (Leaf water content, and leaf layer water cut (Canopy leaf water content LWC), CLWC), formula is as follows:

Plant water content PWC (%)=(PFW-PDW)/PFW × 100%

Leaf water content LWC (%)=(LFW-LDW)/LFW × 100%

Leaf layer water cut CLWC (%)=(∑ LFW-∑ LDW)/∑ LFW × 100%

Wherein, PFW is plant fresh weight, and PDW is plant weights, and LFW is fresh weight, and LDW is leaf dry weight.

3) vegetation index builds

Original spectrum NDVI=(R _{λ 1}-R _{λ 2})/(R _{λ 1}+ R _{λ 2}) RVI=R _{λ 1}/ R _{λ 2}

Spectrum NDVI=(RC reciprocal _{λ 1}-RC _{λ 2})/(RC _{λ 1}+ RC _{λ 2}) RVI=RC _{λ 1}/ RC _{λ 2}

Wherein R _{λ 1}and R _{λ 2}be respectively the reflectivity of any two wavelength, the scope of λ 1 and λ 2 is 350 ~ 2500nm, FD _{λ 1}and FD _{λ 2}for its corresponding first derivative spectrum, AL _{λ 1}and AL _{λ 2}for its corresponding inverse logarithm spectrum, RC _{λ 1}and RC _{λ 2}for its corresponding spectrum reciprocal, according to the estimation ability of normalized differential vegetation index NDVI and ratio vegetation index RVI to wheat water content content, filter out the result behaved oneself best.

4) data analysis and utilization

Utilize step 3) in data, comprehensive analysis plant water content and leaf water content and the direct quantitative relationship of canopy spectra reflectivity, adopt the meticulous sampling method of decrement, filter out the wavelength band to moisture-sensitive and spectrum parameter, and build moisture monitoring model, data analysis and arrangement are all carried out at Exce12007 and MATLAB7.0 (The Math Works, 2000).

5) structure of monitoring model and inspection

Adopt relative root-mean-square deviation (RRMSE) to carry out assay, and draw 1: 1 graph of a relation between experimental observation value and model predication value, wherein RRMSE computing formula is as follows:

$\mathrm{RRMSE}=\sqrt{\frac{1}{n}\×\underset{i=1}{\overset{n}{\mathrm{\Σ}}}{(P_i-O_i)}^2\×\frac{100}{\stackrel{-}{O_i}}}$

O in above formula _ifor the moisture content value of observing in test, P _ifor the moisture content value of model estimation, n is model measurement test samples number.

With the correlationship of plant water content and leaf layer water cut between further analysis wheat canopy original spectrum and first derivative spectrum thereof, spectrum reciprocal, inverse logarithm spectrum.

Build the plant water content quantitative model based on wheat canopy hyperspectral index

Utilize the testing data of 2009-2010, calculate NDVI and RVI of the original spectrum of all any two band combinations in 350-2500nm wavelength band, first derivative spectrum, inverse logarithm spectrum and spectrum reciprocal formation and the coefficient of determination (R of plant water content PWC ²).Show see Fig. 2, Fig. 3 result, based on original spectrum NDVI (R ₈₃₆, R ₇₉₃) and based on spectrum RVI (RC reciprocal ₈₃₇, RC ₇₉₃) the model coefficient of determination R that builds ²performance better, is respectively 0.851 and 0.852.

Build the leaf layer water cut quantitative model based on wheat canopy hyperspectral index

The NDVI adopting the original spectrum of all any band combinations between two in same method calculating 350-2500nm wavelength band, first derivative spectrum, inverse logarithm spectrum and spectrum reciprocal to form and the coefficient of determination (R of RVI and CLWC ²).Show see Fig. 4, Fig. 5 result, based on original spectrum NDVI (R ₁₁₀₀, R ₇₇₀) and RVI (R ₈₉₃, R ₈₀₅) show all better, coefficient of determination R ²be respectively 0.730 and 0.812.

The test and check of wheat plant water cut monitoring model

In order to check reliability and the applicability of wheat water content monitoring model, utilizing 2008-2009 year independent experiment data to carry out test and check to above-mentioned model, utilizing the coefficient of determination (R between predicted value and observed reading simultaneously ²), relative error (RE) and relative root-mean-square deviation (RRMSE) 3 indexs carry out the performance (specifically in table 1) of integrated survey model, and select 1: 1 graph of a relation (Fig. 6) that result makes predicted value and observed reading preferably, predictive ability of showing model directly perceived.Fig. 6 shows, based on original spectrum NDVI (R ₈₃₆, R ₇₉₃) and spectrum RVI (RC reciprocal ₈₃₇, RC ₇₉₃) the plant moisture model test results performance that builds better, the matching coefficient of determination (R ²) being respectively 0.830 and 0.834, relative error (RE) is respectively 0.136 and 0.138, and relative root-mean-square deviation (RRMSE) is respectively 0.151 and 0.152, have passed the test and check of model.

Table 1

The test and check of wheat leaf layer water cut monitoring model

Utilize the monitoring equation of 2008-2009 year independent experiment data to above-mentioned different spectrum types modeling to test respectively, utilize the coefficient of determination (R between predicted value and observed reading ²), relative error (RE) and relative root-mean-square deviation (RRMSE) 3 indexs carry out integrated survey model specifically in table 2, and 1: 1 graph of a relation making predicted value and observed reading intuitively shows the predictive ability (Fig. 7) showing good model.Fig. 7 shows, with original spectrum NDVI (R ₁₁₀₀, R ₇₇₀) and RVI (R ₈₉₃, R ₈₀₅) for variable set up leaf layer water cut monitoring model behave oneself best, coefficient of determination R ²be respectively 0.783 and 0.812, relative error RE is respectively 0.242 and 0.182, and relative root-mean-square deviation RRMSE is respectively 0.205 and 0.148.

Table 2

Claims (6)

1. the Wheat plant moisture monitoring method based on canopy high spectral index, it is characterized in that, moisture solution period is that the jointing stage is to the maturity stage, the real time measure soil moisture content, soil moisture content according to actual measurement is volumetric(al) moisture content, by artificial recharge, the soil moisture of each moisture solution is adjusted to preset value, control each community soil moisture content, concrete steps are as follows:

1) hyperspectral information obtains

Measure wheat canopy hyperspectral index data, significant wave segment limit is 350 ~ 2500nm, and wherein 350 ~ 1050nm spectrum sample is spaced apart 1.4nm, and spectral resolution is 3nm; 1050 ~ 2500nm spectrum sample is spaced apart 2nm, and spectral resolution is 10nm, and spectroscopic assay is selected to carry out when ceiling unlimited, calm or gentle breeze, and minute is 10:00-14:00; During measurement, sensor probe is vertically downward in canopy top, and spectrometer field angle is 25 °, highly about 1.0m, ground field range diameter is 0.44m, each cell measurement 10 sampling points, each sampling point duplicate measurements 5 visual fields, the spectral reflectance value using its mean value as this community;

2) plant determination of moisture

With step 1) synchronous, at different growth stage, the strain of representative wheat 20 is got in each community, first presses organ and is separated, again blade is separated by various position leaves, take back laboratory in the valve bag that rapid loading is weighed, claim its fresh weight with ten thousand/precision electronic balance, weigh after drying to constant weight at 80 DEG C after then putting into baking oven 105 DEG C of 30min that complete, obtain different parts organ dry weight, calculate plant water content PWC respectively, leaf water content LWC and leaf layer water cut CLWC, formula is as follows:

PWC(％)＝(PFW-PDW)/PFW×100％

LWC(％)＝(LFW-LDW)/LFW×100％

CLWC(％)＝(ΣLFW-ΣLDW)/ΣLFW×100％

Wherein, PFW is plant fresh weight, and PDW is plant weights, and LFW is fresh weight, and LDW is leaf dry weight;

3) vegetation index builds

Original spectrum NDVI=(R _{λ 1}-R _{λ 2})/(R _{λ 1}+ R _{λ 2}) RVI=R _{λ 1}/ R _{λ 2}

Spectrum NDVI=(RC reciprocal _{λ 1}-RC _{λ 2})/(RC _{λ 1}+ RC _{λ 2}) RVI=RC _{λ 1}/ RC _{λ 2}

Wherein R _{λ 1}and R _{λ 2}be respectively the reflectivity of any two wavelength, the scope of λ 1 and λ 2 is 350 ~ 2500nm, FD _{λ 1}and FD _{λ 2}for its corresponding first derivative spectrum, AL _{λ 1}and AL _{λ 2}for its corresponding inverse logarithm spectrum, RC _{λ 1}and RC _{λ 2}for its corresponding spectrum reciprocal, according to the estimation ability of normalized differential vegetation index NDVI and ratio vegetation index RVI to wheat water content content, filter out the result behaved oneself best;

4) data analysis and utilization

Utilize step 3) in data, comprehensive analysis plant water content and leaf water content and the direct quantitative relationship of canopy spectra reflectivity, adopt the meticulous sampling method of decrement, filter out the wavelength band to moisture-sensitive and spectrum parameter, and build moisture monitoring model;

With the correlationship of plant water content and leaf layer water cut between further analysis wheat canopy original spectrum and first derivative spectrum thereof, spectrum reciprocal, inverse logarithm spectrum;

Build the plant water content quantitative model based on wheat canopy hyperspectral index:

Utilize step 1) and step 2) testing data that obtains, calculate the coefficient of determination R of NDVI and RVI that the original spectrum of all any two band combinations in 350-2500nm wavelength band, first derivative spectrum, inverse logarithm spectrum and spectrum reciprocal forms and plant water content PWC ²;

Build the leaf layer water cut quantitative model based on wheat canopy hyperspectral index:

NDVI and RVI adopting the original spectrum of all any band combinations between two in same method calculating 350-2500nm wavelength band, first derivative spectrum, inverse logarithm spectrum and spectrum reciprocal to form and the coefficient of determination R of leaf layer water cut CLWC ²;

5) structure of monitoring model and inspection

Adopt relative root-mean-square deviation RRMSE to carry out assay, wherein RRMSE computing formula is as follows:

$\mathrm{RRMSE}=\sqrt{\frac{1}{n}\×\underset{i=1}{\overset{n}{\mathrm{\Σ}}}{(P_i-O_i)}^2}\×\frac{100}{{\stackrel{\‾}{O}}_i}$

O in above formula _ifor the moisture content value of observing in test, P _ifor the moisture content value of model estimation, n is model measurement test samples number;

The test and check of wheat plant water cut monitoring model:

In order to check reliability and the applicability of wheat water content monitoring model, utilize step 1) and step 2) the independent experiment data that obtains carries out test and check to above-mentioned model, utilizes the coefficient of determination R between predicted value and observed reading simultaneously ², these 3 indexs of relative error RE and relative root-mean-square deviation RRMSE carry out the performance of integrated survey model, and make 1: 1 graph of a relation of predicted value and observed reading, predictive ability of showing model directly perceived;

The test and check of wheat leaf layer water cut monitoring model:

Utilize step 1) and step 2) the monitoring equation of independent experiment data to above-mentioned different spectrum types modeling that obtain test respectively, utilizes the coefficient of determination R between predicted value and observed reading ², these 3 indexs of relative error RE and relative root-mean-square deviation RRMSE carry out integrated survey model, and make 1: 1 graph of a relation of predicted value and observed reading, predictive ability of showing model directly perceived.

2. the Wheat plant moisture monitoring method based on canopy high spectral index according to claim 1, is characterized in that, uses Germany to produce Portable soil moisture tacheometer TRIME-EZ the real time measure soil moisture content.

3. the Wheat plant moisture monitoring method based on canopy high spectral index according to claim 1, it is characterized in that, step 1) the middle FicldSpec Pro FR2500 type back hanging type field EO-1 hyperion radiation gauge measured Analytical Spectral Device (ASD) company of the wheat canopy hyperspectral index data acquisition U.S. and produce.

4. the Wheat plant moisture monitoring method based on canopy high spectral index according to claim 1, is characterized in that, step 4) in based on original spectrum NDVI (R ₈₃₆, R ₇₉₃) and based on spectrum RVI (RC reciprocal ₈₃₇, RC ₇₉₃) the Wheat plant moisture monitoring model that builds, coefficient of determination R ²be respectively 0.851 and 0.852; Based on original spectrum NDVI (R ₁₁₀₀, R ₇₇₀) and RVI (R ₈₉₃, R ₈₀₅) the wheat leaf layer moisture monitoring model that builds, coefficient of determination R ²be respectively 0.730 and 0.812.

5. the Wheat plant moisture monitoring method based on canopy high spectral index according to claim 1, is characterized in that, step 4) in data analysis and arrange all carry out at Excel2007 and MATLAB7.0 (The Math Works, 2000).

6. the Wheat plant moisture monitoring method based on canopy high spectral index according to claim 1, is characterized in that, step 5) in the coefficient of determination R of model measurement ²all be greater than 0.783, relative root-mean-square deviation RRMSE is all less than 0.205.