CN103868880A - Wheat leaf nitrogen content monitoring method based on spectrum double-peak index and method for establishing monitoring model - Google Patents

Abstract

The invention discloses a wheat leaf nitrogen content monitoring method based on spectrum double-peak indexes and a method for establishing a monitoring model, aiming at solving the technical problems that the conventional wheat leaf nitrogen content monitoring model is low in universality and accuracy. The method for establishing the wheat leaf nitrogen content monitoring model based on the spectrum double-peak indexes comprises the following steps: acquiring information, delimiting a red-edge wave band area, establishing red-edge double-peak indexes, confirming an optimal critical segmentation wave band, establishing the monitoring model and verifying the model. The wheat leaf nitrogen content monitoring model with high precision and good universality is established, the expansibility and the stability of the monitoring model are enhanced, and high precision and accuracy in monitoring of nitrogen content in wheat leaves are achieved. By adopting the monitoring model, the wheat leaf nitrogen content monitoring method based on the spectrum double-peak indexes, which is disclosed by the invention, is high in accuracy and simple and easy to operate when the content of nitrogen in wheat leaves in a land to be tested is estimated.

Description

Wheat leaf blade nitrogen content monitoring method based on the bimodal index of spectrum and the construction method of monitoring model thereof

Technical field

The present invention relates to agriculture vegetation growth information lossless monitoring art field, be specifically related to a kind of wheat leaf blade nitrogen content monitoring method based on the bimodal index of spectrum and the construction method of monitoring model thereof.

Background technology

Wheat is the important cereal crops of China, in wheat planting process, using of nitrogenous fertilizer is the effective way that improves its yield and quality, along with amount of nitrogen increases, though yield of wheat improves to some extent, but nitrogen utilization efficiency obviously reduces, environmental pollution effect increases the weight of, and more has plant lodging, remaining green when it is due to become yellow and ripe ripe, disease and pest late to increase the weight of and the of inferior quality danger of product qualitative change.According to statistics, the chemical nitrogen fertilizer in being manured into soil only has 35%～50% utilization absorbed by crops, and especially wheat is to the utilization factor of nitrogenous fertilizer only 32%, and all the other most of leaching losses are to lower soil, water body and evaporate in atmosphere.The excessive chemical fertilizer of using, not only causes the environmental problems such as soil, water body, atmosphere, also can pollute agricultural product, affects health, has increased production cost simultaneously.Therefore, greatly develop wheat crop growth real time monitoring, wheat plant Nitrogen Status is carried out in real time, detected fast and exactly, accurately management and regulation and control amount of application of nitrogen fertilizer, to guaranteeing crop yield and QT Quality Target, improve fertilizer utilization efficiency, reduce environmental pollution significant, this is also modern agriculture fertilizing management gordian technique in the urgent need to address.

Traditional crop plant nitrogen content detection method mainly depends on destructive sampling, in office analysis is measured, although result is comparatively reliable, time and effort consuming cost is high, and has certain hysteresis quality.In recent years, along with the fast development of high spectral technique, application of spectral lossless detection technology monitoring crop nitrogen nutritional status real-time, has become the current and important method of field crops growing way and nutrient diagnosis in the future, for crop nitrogenous fertilizer accurately management reliable technical support is provided.At present, numerous scholars propose sensitive wave length and the method for a lot of non-destructive monitoring green vegetation nitrogen contents, as direct employing spectral reflectivity estimation plant nitrogen content, utilize different sensitive bands to build ratio vegetation index monitoring nitrogen content, application of spectral characteristic variable (such as curve of spectrum reflection peak or absorption paddy feature and derivative spectrum characteristic quantity etc.) direct inversion N content of crop tissue; There are many researchers further to adopt polynary progressively linear regression technique to set up N content of crop tissue Forecast of Spectra equation.Above method is carried out based on multispectral data mostly, have spectral resolution low, the shortcoming that wavelength band is little, and analysis based on high spectrum is only used a few characteristic wave bands, cause the excavation application level of the high spectrum abundant information of vegetation inadequate, make the nitrogen spectrum evaluation method building have the limitation that extendability is not strong, stability lacks.In addition, in existing spectrum monitoring method research, single band model stability is poor, there is oversaturated defect in two waveband vegetation index model, especially in wheat increase yield process, colony is large, Dry-matter production is high, makes the accurate instruction susceptibility variation of two waveband vegetation index to plant nitrogen nutrition.Meanwhile, forefathers choose spectral signature variable and the multiband model of establishment and on structure and algorithm, go back slightly complicated, thereby cause temporal and spatial stability, universality and the monitoring accuracy of monitoring model poor, have slackened to a certain extent its production application.Therefore, under the rigid demand constantly refreshing at Unit Grain Production, try to explore new core sensitive band, build that algorithm is simple and the spectrum index of reliable results will be the Focal point and difficult point that non-destructive monitoring technology is applied in modern agricultural production.

Summary of the invention

For above-mentioned the deficiencies in the prior art, the technical problem to be solved in the present invention is to provide a kind of universality is good, accuracy is high and mechanism the is strong wheat leaf blade nitrogen content monitoring method based on the bimodal index of spectrum and the construction method of monitoring model thereof.

The construction method that the present invention is based on the wheat leaf blade nitrogen content monitoring model of the bimodal index of spectrum, comprises the following steps:

(1) information acquisition: first gather respectively according to a conventional method the hyperspectral information of the wheat canopy blade of the different growing stage in zones of different, then adopt five conventional step smoothing methods to carry out smoothing processing to the curve of spectrum; Synchronize with spectral information collection, corresponding collection wheat leaf blade sample also determines the wheat leaf blade actual measurement nitrogen content of the different growing stage in zones of different with conventional method of analysis;

(2) delimit red side wave section region: select wavelength 680nm as red limit starting point, 755nm as red limit terminal, within the scope of this red limit, there is the feature of obvious double-peak feature according to the first derivative spectra of wheat canopy blade, determine that this critical cutting wave band of bimodal is k;

(3) build the bimodal index in red limit: calculate the bimodal index raNVI in red limit by following formula:

$\mathrm{raNVI}=\frac{(R_{755}+R_{680}-2\×R_k)}{(R_{755}-R_{680})},$

Wherein, k is the critical cutting wave band of wheat canopy blade the first derivative spectra within the scope of red limit, R ₇₅₅, R ₆₈₀, R _kbe respectively the spectral reflectivity of wheat canopy blade corresponding to 755nm, 680nm, k nm wave band;

(4) determine optimum critical cutting wave band: bimodal red limit index raNVI is carried out to correlation analysis with the corresponding wheat leaf blade actual measurement nitrogen content of step (1) gained, adopt the meticulous sampling method of high-spectral data decrement, get the k value of correlativity when the highest, be defined as optimum critical cutting wave band;

(5) set up monitoring model: according to wheat leaf blade actual measurement nitrogen content and the direct quantitative relationship of the bimodal index raNVI in red limit of different growing stage, construct the wheat leaf blade nitrogen content monitoring model of following correspondence: Y=a × raNVI+b, wherein, Y is wheat leaf blade nitrogen content, raNVI is the bimodal index in red limit, a is coefficient, and b is constant term, and the value of a, b is tried to achieve via linear regression mode.

Also comprise afterwards in step (5): step (6) testing model:

Use accuracy and the universality of the described wheat leaf blade nitrogen content of independent time wheat test figure Test And Checkout step (5) monitoring model, adopt coefficient of multiple correlation R ², root-mean-square deviation RMSE and relative error RE carry out comprehensive evaluation to it, wherein the computing formula of RMSE and RE is as follows:

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

$\mathrm{RE}=\sqrt{\frac{1}{n}\×\underset{i=1}{\overset{n}{\mathrm{\Σ}}}{\left(\frac{P_i-O_i}{O_i}\right)}^2}\×100\%$

Wherein O _ifor wheat leaf blade actual measurement nitrogen content value in test, P _ifor the wheat leaf blade nitrogen content value of monitoring model estimation, n is model measurement test samples number.

In described step (1), the hyperspectral information of wheat canopy leaves is measured and is adopted the high spectral radiometer in field, and wave band value is 350～1100nm, and spectrum sample is spaced apart 1.4nm, and spectral resolution is 3nm.

The method of measuring wheat leaf blade actual measurement nitrogen content in described step (1) is: synchronize with spectral information collection, wheat 20 strains are got in every community, pressing organ separates, each organ is dried, is weighed at 80 DEG C, after pulverizing, adopt its total nitrogen content of Kjeldahl nitrogen determination, go out wheat leaf blade actual measurement nitrogen content by various position leaves leaf dry weight weighted calculation.

In described step (4), when critical cutting wave band k is between red border area in 680nm～755nm under the step change of interval 1nm, calculate the value of its corresponding bimodal index raNVI in red limit, and carry out correlation analysis with the corresponding wheat leaf blade actual measurement nitrogen content of described step (1) gained, draw in the time of k=705nm, the correlativity of the bimodal index raNVI in red limit and wheat leaf blade actual measurement nitrogen content is the highest, and then obtains:

$\mathrm{raNVI}=\frac{(R_{755}+R_{680}-2\×R_{705})}{(R_{755}-R_{680})}.$

The wheat that is applicable to the Yellow River and Huai He River region from the jointing stage to the wheat leaf blade nitrogen content monitoring model in grouting later stage is: Y=3.7032 × raNVI+0.6922.

The beneficial effect that the present invention is based on the construction method of the wheat leaf blade nitrogen content monitoring model of the bimodal index of spectrum is:

1. based on the bimodal index of spectrum, build the wheat leaf blade nitrogen content monitoring model that precision is high, universality is good, significantly strengthened the mechanism that adopts monitoring model estimation wheat leaf blade nitrogen content;

2. the method makes full use of in characteristic wave bands the spectral information that enriches that multiple wavelength provide, and has simplified comprehensive computing and the structure optimization of wheat canopy blade multiband light spectrum information, and extendability and the stability of the monitoring model that the method builds are enhanced;

3. overcome existing vegetation index and in wheat increase yield process, tended to defect saturated and that the sensitivity of Leaf N content is declined, especially wheat leaf blade nitrogen content monitoring under different location, different cultivars, different Nitrogen Level has been had to higher precision and accuracy;

4. realized the Real-time Obtaining of wheat nitrogen information, for the hardware product exploitation of the harmless tacheometer of portable nitrogen provides crucial band selection, promote the accurately widespread use of management and Efficient Techniques of Regulating of crop nitrogenous fertilizer based on spectral analysis technique, for technical foundation is established in the direct application of remote sensing technology in accurate agricultural.

The wheat leaf blade nitrogen content monitoring method that the present invention is based on the bimodal index of spectrum comprises the following steps:

(a) information acquisition: first gather according to a conventional method the hyperspectral information of the wheat canopy blade in field to be measured, then with five step smoothing methods of routine, the curve of spectrum is carried out to smoothing processing;

(b) calculate the bimodal index raNVI in red limit: according to the hyperspectral information of step (a) gained, calculate the bimodal index raNVI in red limit of the wheat canopy blade in field to be measured according to described step (3), (4) in the construction method of above-mentioned wheat leaf blade nitrogen content monitoring model;

(c) calculate wheat leaf blade nitrogen content: calculate the wheat leaf blade nitrogen content Y:Y=a × raNVI+b in field to be measured according to above-mentioned wheat leaf blade nitrogen content monitoring model, wherein, Y is wheat leaf blade nitrogen content, and raNVI is the bimodal index in red limit, a is coefficient, and b is constant term.

In described step (a), the hyperspectral information of wheat canopy leaves is measured and is adopted the high spectral radiometer in field, and wave band value is 350～1100nm, and spectrum sample is spaced apart 1.4nm, and spectral resolution is 3nm.

In described step (b), k=705nm, $\mathrm{raNVI}=\frac{(R_{755}+R_{680}-2\×R_{705})}{(R_{755}-R_{680})}.$

The wheat leaf blade nitrogen content monitoring equation of the wheat that is applicable to the Yellow River and Huai He River region from the jointing stage to the grouting later stage is: Y=3.7032 × raNVI+0.6922.

The beneficial effect that the present invention is based on the wheat leaf blade nitrogen content monitoring method of the bimodal index of spectrum is:

1. use the aforementioned wheat leaf blade nitrogen content monitoring model based on the bimodal index of spectrum, estimate wheat Leaf N content in field to be measured, method is simple, easy operating;

2. comprehensive computing and the structure optimization of wheat canopy blade multiband light spectrum information have been simplified, overcome existing vegetation index and in wheat increase yield process, tended to defect saturated and that the sensitivity of Leaf N content is declined, wheat leaf blade nitrogen content monitoring under different location, different cultivars, different Nitrogen Level has been had to higher precision and accuracy;

3. realized the Real-time Obtaining of wheat nitrogen information, for the hardware product exploitation of the harmless tacheometer of portable nitrogen provides crucial band selection, promote the accurately widespread use of management and Efficient Techniques of Regulating of crop nitrogenous fertilizer based on spectral analysis technique, for technical foundation is established in the direct application of remote sensing technology in accurate agricultural.

For concrete operation step of the present invention and flow process, further detailed explanation will be made below by reference to the accompanying drawings.

Brief description of the drawings

Fig. 1 is the process flow diagram that the present invention is based on the construction method of the wheat leaf blade nitrogen content monitoring model of the bimodal index of spectrum.

Fig. 2 is the correlativity schematic diagram of raNVI and wheat leaf blade nitrogen content under different critical cutting wave band of the present invention (k nm).

Fig. 3 is the modelling effect figure of raNVI of the present invention and wheat leaf blade nitrogen content monitoring model.

Fig. 4 is the test effect figure of raNVI of the present invention and wheat leaf blade nitrogen content monitoring model.

Embodiment

The present invention is based on the wheat leaf blade nitrogen content monitoring model of the bimodal index of spectrum construction method process flow diagram as shown in Figure 1, describe the concrete implementation step of the method below in detail.

(1) information acquisition: utilize the high spectral radiometer in field to adopt observation technology vertically downward to measure the hyperspectral information data that wheat canopy comprises 350-1100nm wavelength band, then adopt five conventional step smoothing methods to carry out smoothing processing to the curve of spectrum; Synchronize with spectral information collection (simultaneously, same to place), corresponding collection wheat plant sample, divides sample, oven dry, pulverizes and weighs through plant, adopts Kjeldahl nitrogen determination wheat leaf blade actual measurement nitrogen content (%).

(2) delimit red side wave section region: select wavelength 680nm as red limit starting point, 755nm as red limit terminal, within the scope of this red limit, there is the feature of obvious double-peak feature according to the first derivative spectra of wheat canopy blade, determine that this critical cutting wave band of bimodal is k.

Red limit (rededge) is because the reflectivity that vegetation forms at the Multiple Scattering of blade interior at the strong absorption of red spectral band chlorophyll and near-infrared band light sharply rises, and forms steep and is bordering on the hypotenuse of straight line.This specific form in reflectance spectrum curve is peculiar by plant, and Chinese scholars gives the concern of height always to this, and is called " red limit ", and its wavelength location is generally in 660～770nm.Red limit is that wheat leaf blade reflectivity between 680nm～760nm increases the fastest region, belong to the zone of transition of ruddiness region and near infrared region, also be the flex point of the first derivative spectra in this interval, comprise ground object target quantity of information very abundant, the description on red limit comprises the slope on the He Hong limit, position on red limit.The various physical and chemical parameters of red limit and vegetation are closely related, and are the important instruction wave bands of describing plant pigment state and health status, and therefore, red limit is the ideal tools of remote sensing investigation vegetation state.

Under different nitrogenous fertilizer gradient growing ways, in the red edge regions of wheat canopy blade, (680～760nm) the first derivative spectra has obvious response feature, near 680nm, single order differentiated data is stable, can be red limit starting point, and absorb by force paddy for oxygen and water near 760nm, its the first derivative spectra is unstable, can be using 755nm as red limit terminal.Plant nitrogen nutrition is more sufficient, and leaf area index is larger, and red limit slope is also just larger, and corresponding vegetation growth state is better, and position, red limit there will be " moving to right ", and it is large that red limit slope becomes; Otherwise position, red limit can corresponding " blue shift ", red limit slope diminishes.Based on this, whole red limit (680～755nm) is done as a whole quadraturing and is calculated to obtain red limit area (SDr), and it is all regular with nutrition status of the plant, growing way feature and growthdevelopmental stage and changes.Although it is larger that red limit area SDr is affected by crop growing state and nitrogen nutrition, its sensitive to wheat leaf blade nitrogen content is not high enough, is difficult to directly utilize this formal parameter directly to indicate wheat plant nitrogen nutritional status.There is obvious double-peak feature in the canopy leaves the first derivative spectra of different growing way wheats, determine that a boundary site or line are as critical cutting wave band k, such two unimodal area (A at bimodal within the scope of red limit _680-k, A _k-755) relative size different with nitrogen level performance with wheat growth period.

(3) build the bimodal index in red limit: calculate the bimodal index raNVI in red limit by following formula:

$\mathrm{raNVI}=\frac{(R_{755}+R_{680}-2\×R_k)}{(R_{755}-R_{680})},$

Wherein, k is the critical cutting wave band of wheat canopy blade the first derivative spectra within the scope of red limit, R ₇₅₅, R ₆₈₀, R _kbe respectively the spectral reflectivity of wheat canopy blade corresponding to 755nm, 680nm, k nm wave band;

Standardization vegetation index (Normalized difference vegetation index is called for short NVI) also claims green color index, and it is the best indicator of vegetation growth state and coverage, closely related with vegetation nitrogen nutritional status.It is the reflected light of being estimated by earth exploration satellite, and its non-linear conversion red and near-infrared band, is obtained divided by the gap of red and near infrared channels by total amount.It is that one is used the most extensively, also good a kind of vegetation index form of effect in sensor information extraction at present.Current NVI vegetation index based on ruddiness and near infrared reflectivity exist easily saturated, proofread and correct the shortcomings such as limited, noise is more, be difficult to adapt to the actual conditions that current China land for growing field crops cereal crops resource highly dense type is produced, it is indicative not accurate enough to wheat growing way and nitrogen nutritional status.

The present invention adopts the bimodal index raNVI in red limit to build wheat leaf blade nitrogen content monitoring model, both can give full play to the advantage of NVI index framework, can utilize again red marginal ray spectrum information to enrich and the cumulative advantage of sensitive information forward, realize precise monitoring and the Nitrogen Status diagnosis of resource-intensive High-yield Wheat Fields wheat growing way.

Two unimodal areas of red edge regions carry out computing based on single order differential numerical value, and owing to relating to, wave band number is many, and computation process is loaded down with trivial details, adopt integral algorithm to simplify the computation process of vegetation index, obtain:

$A_{680-k}={\∫}_{680}^k\frac{\mathrm{dR\λ}}{\mathrm{d\λ}}\mathrm{d\λ}=R_k-R_{680}$

$A_{k-755}={\∫}_k^{755}\frac{\mathrm{dR\λ}}{\mathrm{d\λ}}\mathrm{d\λ}=R_{755}-R_k$

$\mathrm{raNVI}=\frac{(A_{k-755}-A_{680-k})}{(A_{k-755}+A_{680-k})}=\frac{(R_{755}+R_{680}-2\×R_k)}{(R_{755}-R_{680})}.$

(4) determine optimum critical cutting wave band: bimodal red limit index raNVI is carried out to correlation analysis with the corresponding wheat leaf blade actual measurement nitrogen content of step (1) gained, adopt the meticulous sampling method of high-spectral data decrement, get the k value of correlativity when the highest, be defined as optimum critical cutting wave band.

(5) set up monitoring model: according to wheat leaf blade actual measurement nitrogen content and the direct quantitative relationship of the bimodal index raNVI in red limit of different growing stage, construct the wheat leaf blade nitrogen content monitoring model of following correspondence:

Y＝a×raNVI+b；

Wherein, Y is wheat leaf blade nitrogen content, and raNVI is the bimodal index in red limit, and a is coefficient, and b is constant term, and the value of a, b is tried to achieve via linear regression mode.

(6) testing model: use accuracy and the universality of the described wheat leaf blade nitrogen content of independent time wheat test figure Test And Checkout step (5) monitoring model, adopt coefficient of multiple correlation R ², root-mean-square deviation RMSE and relative error RE carry out comprehensive evaluation to it, wherein the computing formula of RMSE and RE is as follows:

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

$\mathrm{RE}=\sqrt{\frac{1}{n}\×\underset{i=1}{\overset{n}{\mathrm{\Σ}}}{\left(\frac{P_i-O_i}{O_i}\right)}^2}\×100\%$

Wherein O _ifor wheat leaf blade actual measurement nitrogen content value in test, P _ifor the wheat leaf blade nitrogen content value of monitoring model estimation, n is model measurement test samples number.

Describe the present invention in detail with embodiment below, but following examples also limit the scope of the invention never in any form.

Embodiment 1: carry out altogether 3 field tests, relate to different year, different quality types and different nitrogenous fertilizer processing, concrete test design is described below.

Test 1:2009-2010 carries out at Agricultural University Of He'nan's science and education Demonstration Garden (Zhengzhou).Experimental cultivar is Henan wheat 49-198 and Henan wheat 34.The soil texture is moisture soil, organic matter 1.06%, full nitrogen 0.091%, rapid available phosphorus 25.6mgkg ^-1, available potassium 124.5mgkg ^-1.Front stubble is corn.Test is established 5 and is executed nitrogen level, is respectively 0,75,150,225 and 300kghm ^-2purity nitrogen.Respectively process to join and execute 150kghm ^-2p ₂o ₅and 120kghm ^-2k ₂o, phosphorus potash fertilizer is all as base manure, respectively processes nitrogenous fertilizer 50% as base manure, and 50% topdressed for the jointing stage.Test random alignment, repeats community area 7 × 2.9=20.3m for 3 times ², Basic Seedling 2.25 × 10 ⁶hm ^-2, line-spacing 20cm, sowing on October 14.Other cultivation management measures are with general High-yield Wheat Fields.This testing data is for the structure of monitoring model.

Test 2:2010-2011 carries out in auspicious cloud town, Jiaozuo City Wen County.Experimental cultivar is Henan wheat 49-198 and Zheng wheat 366.The soil texture is cinnamon soil, organic matter 1.57%, full nitrogen 0.114%, rapid available phosphorus 28.3mgkg ^-1, available potassium 121.6mgkg ^.1.Front stubble is corn.Test is established 4 and is executed nitrogen level, is respectively 0,90,180 and 270kghm ^-2purity nitrogen.Respectively process to join and execute 120kghm ^-2p ₂o ₅and 150kghm ^-2k ₂o, phosphorus potash fertilizer is all as base manure, respectively processes nitrogenous fertilizer 50% as base manure, and 50% topdressed for the jointing stage.Test random alignment, repeats community area 6 × 4.2=25.2m for 4 times ², Basic Seedling 1.8 × 10 ⁶hm ^-2, line-spacing 20cm, sowing on October 12.Other cultivation management measures are with general High-yield Wheat Fields.This testing data is for the structure of monitoring model.

Test 3:2011-2012 carries out in Kaifeng County, Kaifeng.Experimental cultivar is short anti-58, Henan wheat 49-198 and Zheng wheat 366.The soil texture is moisture soil, organic matter 1.18%, full nitrogen 0.104%, rapid available phosphorus 45.9mgkg ^-1, available potassium 90.3mgkg ^-1.Front stubble is corn.Test is established 5 and is executed nitrogen level, is respectively 0,75,150,225 and 300kghm ^-2purity nitrogen.Respectively process to join and execute 120kghm ^-2p ₂o ₅and 120kghm ^-2k ₂o, phosphorus potash fertilizer is all as base manure, respectively processes nitrogenous fertilizer 50% as base manure, and 50% topdressed for the jointing stage.Test random alignment, repeats community area 7 × 2.9=20.3m for 3 times ², Basic Seedling 2.7 × 10 ⁶hm ^-2, line-spacing 20cm, sowing on October 15.Other cultivation management measures are with general High-yield Wheat Fields.This testing data is for the inspection of monitoring model.

Concrete modeling method is as follows, and wherein, related instrument and equipment if no special instructions, is routine instrument device; The related raw material of industry if no special instructions, is the commercially available conventional raw material of industry.

(1) information acquisition: wheat canopy blade hyperspectral information is measured the high spectral radiometer in FieldSpec Pro FR2500 type back hanging type field that adopts U.S. Analytical Spectral Device (ASD) company to produce.Wave band value is 350～1100nm, and spectrum sample is spaced apart 1.4nm, and spectral resolution is 3nm.When information measurement is chosen in fine, calm or wind speed is very little, carry out, time range is 10:00-14:00.When measurement, vertically downward, instrument field angle is 25 ° to sensor probe, and apart from the about 1.0m of canopy top vertical height, ground field range diameter is 0.44m.Taking 10 spectrum as a sampling spectrum, 10 sampling spectrum of each observation station record, the spectral reflectance value using its mean value as this observation station.Synchronize with spectral information collection, wheat 20 strains are got in every community, separate by organ, and each organ is dried, weighed at 80 DEG C, after pulverizing, adopt its total nitrogen content of Kjeldahl nitrogen determination, go out wheat leaf blade actual measurement nitrogen content by various position leaves leaf dry weight weighted calculation.

(2) delimit red side wave section region: select wavelength 680nm as red limit starting point, 755nm as red limit terminal, within the scope of this red limit, there is the feature of obvious double-peak feature according to the first derivative spectra of wheat canopy blade, determine that this critical cutting wave band of bimodal is k; The correlativity of the lower raNVI of different critical cutting wave band (k nm) and wheat leaf blade nitrogen content as shown in Figure 2;

(3) build the bimodal index in red limit: calculate the bimodal index raNVI in red limit by following formula:

$\mathrm{raNVI}=\frac{(R_{755}+R_{680}-2\×R_k)}{(R_{755}-R_{680})},$

Wherein, k is the critical cutting wave band of wheat canopy blade the first derivative spectra within the scope of red limit, R ₇₅₅, R ₆₈₀, R _kbe respectively the spectral reflectivity of wheat canopy blade corresponding to 755nm, 680nm, k nm wave band.

(4) determine optimum critical cutting wave band: bimodal red limit index raNVI is carried out to correlation analysis with the corresponding wheat leaf blade actual measurement nitrogen content of step (1) gained, adopt the meticulous sampling method of high-spectral data decrement, get the k value of correlativity when the highest, be defined as optimum critical cutting wave band; When critical cutting wave band k is between red border area in 680nm～755nm under the step change of interval 1nm, calculate the value of its corresponding bimodal index raNVI in red limit, and carry out correlation analysis with the corresponding wheat leaf blade actual measurement nitrogen content of described step (1) gained, draw in the time of k=705nm, the correlativity of the bimodal index raNVI in red limit and wheat leaf blade actual measurement nitrogen content is the highest, r=0.914***, n=190, and then obtain:

$\mathrm{raNVI}=\frac{(R_{755}+R_{680}-2\×R_{705})}{(R_{755}-R_{680})}.$

(5) set up monitoring model: according to wheat leaf blade actual measurement nitrogen content and the direct quantitative relationship of the bimodal index raNVI in red limit of different growing stage, construct the wheat leaf blade nitrogen content monitoring model of following correspondence:

Y＝a×raNVI+b；

Wherein, Y is wheat leaf blade nitrogen content, and raNVI is the bimodal index in red limit, and a is coefficient, and b is constant term, and the value of a, b is tried to achieve via linear regression mode.

In red side wave section (680nm～755nm), choose arbitrarily k nm wave band position, random combine is set up the linear regression model (LRM) of raNVI and wheat leaf blade nitrogen content, gets higher multiple correlation coefficient (R ²) corresponding K-band as optimize R _kthereby, utilize optimum raNVI to set up wheat leaf blade nitrogen content monitoring model.Based on the raNVI of above-mentioned new structure, utilize the test figure in 2009-2012 Nian Ge field experiment district, build and be applicable to the Leaf N content monitoring model of the Yellow River and Huai He River wheat during jointing stage to the later stage of being in the milk.Multiple raNVI of test statistics are set up to monitoring model Y=a × raNVI+b, and adopt multiple correlation coefficient (R ²) and standard error (SE) goodness of fit of monitoring model is evaluated.Fig. 3 has shown the simulate effect of raNVI of the present invention and wheat leaf blade nitrogen content monitoring model, and wherein ordinate represents wheat leaf blade nitrogen content, and horizontal ordinate represents the corresponding bimodal index raNVI in red limit.Result shows, the nitrogen content monitoring equation of the Yellow River and Huai He River wheat from the jointing stage to the grouting later stage is: Y=3.7032 × raNVI+0.6922, coefficient of multiple correlation R ²be 0.847, standard error SE is 0.390, sample number n=464.

By the result of above-mentioned testing data and existing spectrum vegetation index, the performance in the estimation of wheat leaf blade nitrogen content contrasts, these conventional spectrum vegetation indexs comprise: VOG, CTR, GM2, SR705, ND705, mND705, RI-1dB, REP, CAI, NAOC, TVI and DCNI etc., and as shown in table 1 below.

The wheat leaf blade nitrogen content monitoring model parameter of the various common vegetation indexs of table 1

Continued 1

Comparing result shows, the accuracy of the wheat leaf blade nitrogen content monitoring model based on the bimodal index raNVI in red limit provided by the present invention is higher, is better than existing vegetation index.

(6) testing model: in order to check the dependable with function of wheat leaf blade nitrogen content monitoring model, utilize 2011-2012 independent experiment data to carry out test and check to above-mentioned model, utilize the multiple correlation coefficient (R between predicted value and observed reading simultaneously ²), root-mean-square deviation (RMSE) and average relative error (RE) carry out Comprehensive Assessment, and draw 1: 1 graph of a relation between observed value and predicted value, the predictive ability of intuitively showing model.

The assay of the wheat leaf blade nitrogen content monitoring model based on the bimodal index raNVI in red limit that the present invention is established and conventional optimization vegetation index (REPle and mND705) assay compare, the R of REPle model ², RMSE and RE be respectively 0.78,0.414 and 13.7%; The R of mND705 model ², RMSE and RE be respectively 0.80,0.391 and 14.1%; And the precision of prediction of raNVI model of the present invention is the highest, estimation error is minimum: as shown in Figure 4, the testing model of the wheat leaf blade nitrogen content based on the bimodal index in red limit from the jointing stage to the grouting later stage: y=0.9023x+0.5326, coefficient of multiple correlation R ²be 0.848, root-mean-square deviation RMSE is 0.338, and average relative error RE is 11.9%, sample number n=190.

In sum, the present invention is based on the construction method of the wheat leaf blade nitrogen content monitoring model of the bimodal index in red limit, built the wheat leaf blade nitrogen content monitoring model that precision is high, universality is good; Extendability and the stability of this monitoring model are enhanced, and wheat leaf blade nitrogen content monitoring under different location, different cultivars, different Nitrogen Level is had to higher precision and accuracy.The present invention is based on the wheat leaf blade nitrogen content monitoring method of the bimodal index in red limit, use aforementioned monitoring model, estimate wheat Leaf N content in field to be measured, accuracy is high, method is simple, easy operating, promote the accurately widespread use of management and Efficient Techniques of Regulating of crop nitrogenous fertilizer based on spectral analysis technique, for technical foundation is established in the direct application of remote sensing technology in accurate agricultural.

In conjunction with the accompanying drawings and embodiments the present invention is described in detail above, but, person of ordinary skill in the field can understand, do not departing under the prerequisite of aim of the present invention, can also change each design parameter in above-described embodiment, form multiple specific embodiments, be common variation range of the present invention, describe in detail no longer one by one at this.

Claims (10)

1. a construction method for the wheat leaf blade nitrogen content monitoring model based on the bimodal index of spectrum, is characterized in that, comprises the following steps:

(1) information acquisition: first gather respectively according to a conventional method the hyperspectral information of the wheat canopy blade of the different growing stage in zones of different, then adopt five conventional step smoothing methods to carry out smoothing processing to the curve of spectrum; Synchronize with spectral information collection, corresponding collection wheat leaf blade sample also determines the wheat leaf blade actual measurement nitrogen content of the different growing stage in zones of different with conventional method of analysis;

(2) delimit red side wave section region: select wavelength 680nm as red limit starting point, 755nm as red limit terminal, within the scope of this red limit, there is the feature of obvious double-peak feature according to the first derivative spectra of wheat canopy blade, determine that this critical cutting wave band of bimodal is k;

(3) build the bimodal index in red limit: calculate the bimodal index in red limit by following formula ranVI:

?，

Wherein, k is the critical cutting wave band of wheat canopy blade the first derivative spectra within the scope of red limit, r ₇₅₅, r ₆₈₀, r _kbe respectively the spectral reflectivity of wheat canopy blade corresponding to 755nm, 680nm, k nm wave band;

(4) determine optimum critical cutting wave band: by bimodal red limit index rathe corresponding wheat leaf blade actual measurement nitrogen content of NVI and step (1) gained carries out correlation analysis, adopts the meticulous sampling method of high-spectral data decrement, gets the k value of correlativity when the highest, is defined as the critical cutting wave band of optimum;

(5) set up monitoring model: according to wheat leaf blade actual measurement nitrogen content and the bimodal index in red limit of different growing stage rathe direct quantitative relationship of NVI, constructs the wheat leaf blade nitrogen content monitoring model of following correspondence:

Y=a× raNVI+b；

Wherein, Y is wheat leaf blade nitrogen content, ranVI is the bimodal index in red limit, and a is coefficient, and b is constant term, and the value of a, b is tried to achieve via linear regression mode.

2. the construction method of the wheat leaf blade nitrogen content monitoring model based on the bimodal index of spectrum according to claim 1, is characterized in that, also comprises afterwards in step (5):

Step (6) testing model: use accuracy and the universality of the described wheat leaf blade nitrogen content of independent time wheat test figure Test And Checkout step (5) monitoring model, adopt multiple correlation coefficient r ² , root-mean-square deviation RMSE and relative error RE carry out comprehensive evaluation to it, wherein the computing formula of RMSE and RE is as follows:

Wherein o _i for wheat leaf blade actual measurement nitrogen content value in test, p _i for the wheat leaf blade nitrogen content value of monitoring model estimation, n is model measurement test samples number.

3. the construction method of the wheat leaf blade nitrogen content monitoring model based on the bimodal index of spectrum according to claim 1, it is characterized in that, in described step (1), the hyperspectral information of wheat canopy leaves is measured and is adopted the high spectral radiometer in field, wave band value is 350 ~ 1100 nm, spectrum sample is spaced apart 1.4 nm, and spectral resolution is 3 nm.

4. the construction method of the wheat leaf blade nitrogen content monitoring model based on the bimodal index of spectrum according to claim 1, it is characterized in that, the method of measuring wheat leaf blade actual measurement nitrogen content in described step (1) is: synchronize with spectral information collection, wheat 20 strains are got in every community, pressing organ separates, each organ is dried, is weighed at 80 DEG C, after pulverizing, adopts its total nitrogen content of Kjeldahl nitrogen determination, goes out wheat leaf blade actual measurement nitrogen content by various position leaves leaf dry weight weighted calculation.

5. the construction method of the wheat leaf blade nitrogen content monitoring model based on the bimodal index of spectrum according to claim 1, it is characterized in that, in described step (4), when critical cutting wave band k is between red border area in 680nm～755nm under the step change of interval 1nm, calculate the value of its corresponding bimodal index raNVI in red limit, and carry out correlation analysis with the corresponding wheat leaf blade actual measurement nitrogen content of described step (1) gained, draw in the time of k=705nm, the correlativity of the bimodal index raNVI in red limit and wheat leaf blade actual measurement nitrogen content is the highest, and then obtains:

?。

6. the construction method of the wheat leaf blade nitrogen content monitoring model based on the bimodal index of spectrum according to claim 1, it is characterized in that, the wheat that is applicable to the Yellow River and Huai He River region from the jointing stage to the wheat leaf blade nitrogen content monitoring model in grouting later stage is: Y=3.7032 × ranVI+0.6922.

7. the wheat leaf blade nitrogen content monitoring method based on the bimodal index of spectrum, is characterized in that, comprises the following steps:

(a) information acquisition: first gather according to a conventional method the hyperspectral information of the wheat canopy blade in field to be measured, then with five step smoothing methods of routine, the curve of spectrum is carried out to smoothing processing;

(b) calculate the bimodal index raNVI in red limit: according to the hyperspectral information of step (a) gained, calculate the bimodal index raNVI in red limit of the wheat canopy blade in field to be measured according to described step (3), (4) in claim 1;

(c) calculate wheat leaf blade nitrogen content: the wheat leaf blade nitrogen content monitoring model of setting up according to claim 1 calculates the wheat leaf blade nitrogen content Y in field to be measured:

Y=a× raNVI+b；

Wherein, Y is wheat leaf blade nitrogen content, ranVI is the bimodal index in red limit, and a is coefficient, and b is constant term.

8. the wheat leaf blade nitrogen content monitoring method based on the bimodal index of spectrum according to claim 6, it is characterized in that, in described step (a), the hyperspectral information of wheat canopy leaves is measured and is adopted the high spectral radiometer in field, wave band value is 350 ~ 1100 nm, spectrum sample is spaced apart 1.4 nm, and spectral resolution is 3 nm.

9. a kind of wheat leaf blade nitrogen content monitoring method based on the bimodal index of spectrum according to claim 6, is characterized in that, in described step (b), and k=705nm,

.

10. the wheat leaf blade nitrogen content monitoring method based on the bimodal index of spectrum according to claim 6, is characterized in that, the monitoring equation of the wheat leaf blade nitrogen content of the wheat that is applicable to the Yellow River and Huai He River region from the jointing stage to the grouting later stage is:

Y=?3.7032× raNVI+0.6922。

Images