CN105675821A - Image evaluation index establishing method for nondestructive diagnosis of crop nitrogen nutrition - Google Patents

Abstract

The invention discloses an image evaluation index establishing method for nondestructive diagnosis of crop nitrogen nutrition. The method comprises the steps that wheat canopy images and the nitrogen content of leaves are obtained; the wheat canopy images are subjected to partition extraction, average pixel values of R, G, B, H, S and I components of all non-zero pixel points of leaves in each wheat canopy image are calculated to serve as basic color characteristic parameters r, g, b, h, s and i; nine image characteristic parameters r, g, b, r-g-b, r-g, r-b, (r-g-b)/(r+g+b), (r-g)/(r+g+b), and (r-b)/(r+g+b) are calculated; CMI=(xr-yg-zb) is established, fitting with the nitrogen content of the leaves is carried out, weight coefficients x, y and z are determined, CMI is subjected to standardized processing, and it is determined that NCMI=(xr-yg-zb)/(r+g+b). The NCMI established in the method is a relatively-proper winter wheat nitrogen nutrition evaluation index, accuracy and stability are achieved, and the crop nitrogen content can be quantitatively evaluated.

Description

A kind of method for building up of the picture appraisal index of crop nitrogen nutrition Nondestructive

Technical field

The present invention relates to the method for building up of the picture appraisal index of a kind of crop nitrogen nutrition Nondestructive.

Background technology

Along with the high speed development of remote sensing and image processing techniques, extraction canopy image parameter carries out the lossless Nitrogen nutritional status of crop is just becoming current research focus^[1-5]. But the data analysis obtain canopy reflectance spectra apparatus expensive, relating to is complicated, limits remote sensing technology extensive use in agricultural production; And digital camera is easy and simple to handle, quick, in the lossless Nitrogen nutritional status of crop, there is certain application prospect^[2]. Crop leaf color changes with the change of internal nutrition, and the determination of leaf image characteristic parameter and acquisition are the keys of Accurate Diagnosis crop N Nutrition. Correlational study achievement shows, the base image parameter relevant to nutrient index, Different Crop is different. J.Vollmann^[3]And Shibayama^[4]Research show, the image parameter relevant to Semen sojae atricolor and rice leaf SPAD value of chlorophyll content respectively green component G and green color index (leafgreennessindex, LGI), Kyu-Jong etc.^[5]Find that G and Nitrogen in Rice cumulant are the significant correlation of 0.83; Wang Fangyong^[6]Analyze and find the b* parameter in L*a*b* space and the S component in HSI space and Cotton Gossypii SPAD value, Different Nitrogen Concentration good relationship. Contrast forefathers study discovery, canopy base image parameter is interacted process and combination, the dependency relation of the parameter obtained and chlorophyll content in leaf blades, N content of crop tissue becomes apparent from, Karcher, Wang Yuan and RobertL.Rorie experiment show simultaneously, the Nitrogen Nutrition on the bottle green index D GCI under HSI space and green meadow^[7], oryza sativa l. NC, SPAD^[8]And maize leaf nitrogen content^[9]All there is good dependency; Image combination parameter (R-B), (G-B) under rgb space are ideal with the expression of nitrogen content, plant nitrogen content to the Top-three Leaves SPAD value of each period of duration of winter wheat^[10]. In these parameters, DGCI=[(H-60)/60+ (1-S)+(1-I)]/3^[7]It is that the geometric ratio heavy phase after base image parameter H, S, I have been carried out normalized adds;HONGGUANG standardized value NRI=R/ (R+G+B)^[11]Being red component and the ratio of 3 monochromatic component summations, similar also has green glow standard value (NormalizedBluenessIntensity, NGI)^[12], blu-ray standard value (NormalizedRednessIntensity, NBI)^[13]. Visible, to the conversion of crop base image nonlinearity in parameters and standardization, improve its Precise Diagnosis to nitrogen nutrition to a certain extent. But crop growth conditions and nutrient content change along with the change of kind, period of duration, intensity of illumination, fertilising etc.^[14], and then cause the canopy image index difference to its nutriture ability to express. The research of contrast period of duration finds, the NBI and the NRI of pustulation period of 12 leaf phases is higher with spring maize nitrogen nutrition index degree of correlation^[13], NRI can better reflect the N Nutrition of period of seedling establishment and jointing stage Semen Tritici aestivi^[11,15], Li Hongjun etc.^[16]That then find Semen Tritici aestivi period of seedling establishment good relationship is green glow and the HONGGUANG ratio G/R on reflective blade face; Hu Hao^[17]The research display each monochromatic component of rgb space and (R+G+B) and Wheat Leaves nitrogen content in negative correlation in various degree, (R-G-B) is proportionate. Visible, wheat canopy leaf color is the concentrated expression of green 3 primary colors of bluish red, the change of any monochrome all can cause overall canopy leaf color to change in various degree, typical image features parameter NRI under aforementioned rgb space, NGI, NBI, G/R etc., when prominent monochromatic component, do not consider the residue monochromatic component entire effect to Nutrition monitoring.

List of references:

[1] Ding Yongjun, Li Minzan, Sun Hong, etc. based on tomato nutrient element diagnostic cast [J] of multispectral image technology. Transactions of the Chinese Society of Agricultural Engineering, 2012,28 (8): 175 180.

[2] Liu Ying, Li Zhihong. utilize Digital image technology to carry out the research [J] of Semen Maydis Nitrogen Nutrition Diagnosis. Maize Sciences, 2010,18 (4): 147 149. [3] J.VollmannJ, WalterH, SatoT, etal.Digitalimageanalysisandchlorophyllmeteringforphenot ypingtheeffectsofnodulationinsoybean. [J] .ComputersandElectronicsinAgriculture, 2011 (1): 190 195.

[4]ShibayamaM,SakamotoT,TakadaE,etal.Estimatingriceleafgreenness(SPAD)usingFixed‐Pointcontinuousobservationsofvisibleredandnearinfrarednarrow‐banddigitalimages[J].PlantProductionScience,2012,15(4):293‐309.

[5]KyuJL.ByuuWLEstimationofricegrowthandnitrogennutritionstatususingcolordigitalcameraimageanalysis[J].EuropeanjorunalofAgronomy.2013.(6):57‐68.

[6] Wang Fangyong, Wang Keru, Li Shaokun, etc. utilize digital camera and imaging spectrometer estimation cotton leaf chlorophyll and nitrogen content [J]. Acta Agronomica Sinica, 2010,36 (11): 1,981 1989.

[7] KarcherDE, RichardsonMD.Quantifyingturfgrasscolorusingdigitalimagea nalysis [J] .CropSci, 2003,43:943 951.

[8] Wang Yuan, Wang Dejian, Zhang Gang, etc. the rice canopy image based on digital camera is split and Nitrogen Nutrition Diagnosis [J]. Transactions of the Chinese Society of Agricultural Engineering, and 2012,28 (17): 131 136.

[9] RorieRL, PurcellLC, KarcherDE, etal.Theassessmentofleafnitrogenincornfromdigitalimages [J] .CropSci, 2011 (51): 2,174 2180.

[10] Wu Funing. image processing technique application [D] in winter wheat Nitrogen nutritional status. China Agricultural University, 2004.

[11] Xiao Yanbo, Jia Liangliang, Chen Xinping, etc. Applied Digital image analysis technology carries out winter wheat jointing stage Nitrogen nutritional status [J]. China's agronomy circular, and 2008,24 (8): 448 453.

[12] Li Lantao, Zhang Meng, Ren Tao, Li little Kun, Cong Huan, Wu Lishu, Lu Jianwei. Applied Digital image technique carries out Nitrogen Nutrition of Paddy Rice Plant diagnosis [J]. plant nutrient and fertilizer journal, 2015,01:259 268.

[13] platinum is suitable, Cao Weidong, Xiong Jing, etc. application digital camera carries out green manure and turns over rear spring maize Nitrogen Nutrition Diagnosis and production forecast [J]. spectroscopy and spectrum analysis, and 2013, (12): 3,334 3338.

[14] Wang Renhong, Song Xiaoyu, Li Zhenhai, etc. the winter wheat nitrogen nutrition index based on EO-1 hyperion estimates [J]. Transactions of the Chinese Society of Agricultural Engineering, and 2014,30 (19): 191 198

[15] Zhang Lizhou, Hou Xiaoyu, Zhang Yuming, etc. the application [J] in winter wheat Nitrogen Nutrition Diagnosis of the digital picture diagnostic techniques. Chinese Ecological Agriculture journal, 2011,19 (5): 1,168 1174.

[16] Li Hongjun, Zhang Lizhou, Chen Ximing, etc. Applied Digital image carries out the research [J] of image analysis method in wheat nitrogen nutrient diagnosis. Chinese Ecological Agriculture journal, and 2011,19 (1): 155 159.

[17] Hu Hao. based on winter wheat Nitrogen Nutrition Diagnosis and the growth monitoring [D] of visible ray near infrared spectrum. the Chinese Academy of Agricultural Sciences, 2009.

Summary of the invention

It is an object of the invention to for the deficiencies in the prior art, the method for building up of the picture appraisal index of a kind of crop nitrogen nutrition Nondestructive is provided, inventor is by analyzing each base image parameter sign ability to nitrogen under different Plant plane (different planting density levels, different nitrogen amount applied), find representative monochromatic component and strengthen its proportion, the combining weights of mutual tuning residual components, and standardization further, search out more suitable picture appraisal index.First 9 image features are determined, i.e. 3 monochromatic component r, g, b, 3 linear combination parameter r-g-b, r-g, r-b, and 3 standardized feature parameter (r-g-b)/(r+g+b), (r-g)/(r+g+b), (r-b)/(r+g+b), explore more representative base image parameter and weight thereof, and carry out best fit with nitrogen nutrition index, determine color combination standard index (NCMI) proposed under rgb space, with 3 typical image parameter (DGCI, NRI, G/R) different cultivars, the contrast of different sowing schemes finds, it is more stable with the dependency of Leaf nitrogen concentration (LNC) and error of fitting, for relatively suitable winter wheat nitrogen nutrition picture appraisal index.

It is an object of the invention to be achieved through the following technical solutions:

A kind of method for building up of the picture appraisal index of crop nitrogen nutrition Nondestructive, it comprises the following steps:

(1), sampling: obtain the Leaf nitrogen concentration of wheat canopy image, wheat plant;

(2), the K means clustering method segmented extraction wheat canopy image based on H component is utilized, calculate Color characteristics parameters based on the average pixel value of every all non-zero pixel R of width wheat canopy image Leaf, G, B, H, S, I component, be respectively defined as r, g, b, h, s, i; Calculate 9 image features, i.e. 3 monochromatic component r, g, b, 3 linear combination parameter r-g-b, r-g, r-b and 3 normalizing parameter (r-g-b)/(r+g+b), (r-g)/(r+g+b), (r-b)/(r+g+b);

(3) color combinatorial index CMI (ColorMixIndex)=(xr-yg-zb) treating tuning, is built, then by the Leaf nitrogen concentration matching of CMI Yu identical field, determine the weight coefficient x, y, z of 3 components, finally to CMI standardization, it is determined that NCMI:NCMI=(xr-yg-zb)/(r+g+b).

In step (1), the acquisition methods of wheat canopy image: at wheat during jointing stage, adopt camera from canopy 1.0m height and the shooting sampling of 90 ° of ground. Preferably in obtaining wheat canopy image under wheat during jointing stage, fair weather, the jointing stage is the period of monitoring important in wheat growth and nitrogen fertilizer application; Under fair weather, the picture of shooting is clear, and segmentation effect is good. During concrete operations, it is possible to adopt slr camera Olympus E-620, fair weather from canopy 1.0m height with 90 ° of ground shooting sampling, it is thus achieved that wheat canopy image should collect whole plant.

The acquisition methods of the Leaf nitrogen concentration of wheat plant: wheatland top destructiveness was sampled the same day by wheat canopy image taking, selects the plant that 15～25 strain growing ways are close, the Leaf nitrogen concentration of Kjeldahl nitrogen determination wheat plant, averages.

Sample point picks up from different cultivars, different growing, different nitrogen amount applied, different planting density and different year.

In step (3), fit procedure is as follows:

A, set x ∈ [1,2], y ∈ [-1,1], z ∈ [-1,1], the contribution degree of absolute value more big then corresponding color component is more big, and 0 namely without contribution degree;

B, in above-mentioned scope, constantly adjust x, y, z value by step-length (being set to 0.05), carry out regression analysis with Leaf nitrogen concentration, calculate x, y, z value and coefficient of determination R², build four-dimensional array [x, y, z, R²];

C, drafting four-dimensional array [x, y, z, R²] three-dimensional distribution map (see Fig. 1); Shade represents R²Just, the red area that CMI is the highest with Leaf nitrogen concentration dependency, it is mainly distributed on x ∈ [1.5,2], z ∈ [-1 ,-0.5] or y ∈ [-0.5,0], z ∈ [-1 ,-0.5];

d、R²When being 0.83353 to the maximum, corresponding x, y, z are the optimal value of r, g, b linear fit, x=1.6, y=-0.95, z=-0.8.

E, to CMI standardization, it is determined that the concrete formula of NCMI is as follows:

NCMI=(1.6r-0.95g-0.8b)/(r+g+b).

A kind of method of wheat nitrogen nutrition Nondestructive, comprises the following steps:

(1) obtain the wheat canopy image of field to be measured, with the K means clustering method segmented extraction wheat canopy image based on H component, calculate 3 monochromatic component r, g, b;

(2) x, y, z in color combination standard index NCMI is determined according to the method described in the present invention, wherein, NCMI=(xr-yg-zb)/(r+g+b);

(3) value of 3 the monochromatic component r obtained in step (1), g, b being substituted in the NCMI computing formula determined in claim 2, calculated color combination standard index NCMI is the Semen Tritici aestivi nitrogen content of the field to be measured of prediction.

In method of the present invention, the definition of all parameters and computational methods are with the method for building up of above-mentioned a kind of picture appraisal index for wheat nitrogen nutrition Nondestructive.

Beneficial effects of the present invention:

1, the inventive method is by under different bearing stage difference Plant plane, the forecast error of wheat leaf blade nitrogen content is contrasted and correlation analysis by NCMI and 3 typical image features parameter, it was shown that NCMI possesses suitability, Stability and veracity as winter wheat nitrogen nutrition evaluation index;

2, sampling digital camera carries out the data sampling of image, and method is easy, and equipment (camera) expense is low, the application under the environment of convenient actual land for growing field crops;

3, crop nitrogen content is carried out quantitatively evaluating, peasant can be instructed to carry out the rational application of fertilizer, namely ensure the nitrogenous fertilizer needed for plant, excessive fertilization can be avoided again, improve yield, it is to avoid excessive fertilization contaminated environment;

4, the construction method of evaluating goes for other crops, but needs with reference to the method, the sampled data of Different Crop to be carried out the matching of coefficient x, y.z.

Accompanying drawing explanation

Fig. 1 is 3 monochromatic component r, g, b and relative coefficient distribution graphics.

Detailed description of the invention

1 experiment and data

1.1 experimental designs

2012-2014 tills the land in national information agricultural centre Rugao (north latitude 32 °～32 ° 30 ', east longitude 120 ° 20 '～120 ° 50 ') Experimental Base rice wheat wheel, 2 kinds: life is selected No. 6 (V1), raised wheat No. 18 (V2); 3 Nitrogen Levels: purity nitrogen 0kg/hm²(N0), purity nitrogen 150kg/hm²(N1), purity nitrogen 300kg/hm²(N2); 2 planting densities: D1 line-spacing 40cm (100,000 Seedlings/mu), D2 line-spacing 20cm (200,000 Seedlings/mu). Adopt random split block design, 12 process, repeat for 3 times, Gong36Ge community. Plot area 35m²(7m × 5m), drilling, the gross area is about 1080m². Executing nitrogen to carry out at twice: during sowing, N0, N1, N2, the jointing stage carries out when second time (mid-March), nitrogen amount applied and sowing the same.

1.2 data acquisitions

1.2.1 canopy image acquisition

Jointing stage is the nitrogen diagnosis critical period, adopts slr camera Olympus E-620 (being called for short Olympus), and the fair weather in 3-4 month is from canopy 1.0m height and 90 ° of ground shooting sampling (table 1). The every sub-sampling in every district 1 time, 36 communities sample 36 wheat canopy images altogether.

Table 1 Semen Tritici aestivi is taken pictures, the Date of Sampling

1.2.2 Leaf nitrogen concentration is measured

Wheat leaf blade nitrogen content chemical assay and image acquisition carry out the same period, and overground part destructiveness was sampled the same day by shooting, and every community selects the 20 close plant of strain growing way, the Leaf nitrogen concentration (LNC) of Kjeldahl nitrogen determination wheat plant, averages.

1.2.3 model evaluation standard

Following index is selected to contrast the degree of correlation of each image features and wheat leaf blade nitrogen content, the fitting effect of evaluation and foreca value (namely according to the calculated NCMI of method provided by the invention) and measured value (Leaf nitrogen concentration that the same period measures).

1) coefficient of determination R²: the square value of correlation coefficient r, it was shown that predictive value explains the degree that actual value is deteriorated, computing formula:

$R^2={\left(\frac{{\mathrm{\Σ}}_{i-=1}^n(x_i-\stackrel{\‾}{x})(y_i-\stackrel{\‾}{y})}{\sqrt{{\mathrm{\Σ}}_{i=1}^n-{(x_i-\stackrel{\‾}{x})}^2{\mathrm{\Σ}}_{i=1}^n-{(y_i-\stackrel{\‾}{y})}^2}}\right)}^2$

In formula, x_iFor the measured value of sample i,For x_iMeansigma methods; y_iFor the measured value of sample i,For y_iMeansigma methods; N is sample number.

2) root-mean-square error RMSE: inspection predictive value and measured value meet degree of accuracy, computing formula:

$RMSE=\sqrt{\frac{1}{n-1}\underset{i=1}{\overset{n}{\Σ}}{(y_i-y_i^{\′})}^2}$

In formula, y_i' for the actual value that sample i reference instrument method measures.

2 wheat canopy picture appraisal index study

For the over-segmentation phenomenon that avoid uneven illumination under the environment of land for growing field crops, background complicated and shadow occlusion etc. is caused as far as possible, utilize the K means clustering method (Huang Fen based on H component, Yu Qi, Yao Xia, Deng. K mean cluster segmentation [J] of wheat canopy image H component. computer engineering and application, 2014, (3): 129 134.) segmented extraction wheat canopy image pattern, calculate Color characteristics parameters based on the average pixel value of all non-zero pixel R of every width blade, G, B, H, S, I component, be respectively defined as r, g, b, h, s, i; Calculate pseudoreduced 9 image features accordingly, i.e. 3 monochromatic component r, g, b, 3 linear combination parameter r g b, r g, r b and 3 normalizing parameter (r g b)/(r+g+b), (r g)/(r+g+b), (r b)/(r+g+b), correlation analysis and best fit is carried out, it is determined that wheat canopy picture appraisal index with Leaf nitrogen concentration.

The correlation analysis of 2.1 image features and nitrogen nutrition index

Select experimental data in 2013, contrast the dependency relation (R of lower 9 image features of different growing, kind, density and nitrogen amount applied and Leaf nitrogen concentration (LNC) respectively²)。

Table 2 shows, the different bearing stage, and monochromatic component r is apparently higher than g and b; 3 linear combination parameter r-g-b, r-g, r-b are then substantially higher than the monochromatic component same period, and r-b is slightly higher, March 14 and April r-b Yu LNC on the 1st R²Respectively-0.74 and-0.77; After these 3 linear combination parameter standardization, degree of correlation improves further, and wherein, March 14 (r-g-b)/(r+g+b) and LNC is pole significant correlation (R²=-0.85).

The coefficient of determination R of wheat canopy characteristic parameter and LNC under the different Plant plane of table 2²(different cultivars, level of density, execute nitrogen scheme)

Note: V1 makes a living and selects No. 6, and V2 is for raising wheat No. 18; D1 is line-spacing 40cm (100,000 Seedlings/mu), and D2 is line-spacing 20cm (200,000 Seedlings/mu); N0 is purity nitrogen 0kg/hm², N1 is purity nitrogen 75kg/hm², N2 is purity nitrogen 150kg/hm², lower same.

For 2 kinds (V1, V2), the dependency of monochromatic component r and g is better than b; Linear combination parameter r-g-b, r-b R²Then being relatively higher than r and g, wherein, the linear combination parameter r-b of V1 and V2 all reaches-0.66; After standardization, contrast itself and 3 combination parameters (r-g-b, r-b, r-g), V1 kind dependency is promoted to-0.73 ,-0.78 ,-0.59, V2 kind dependency respectively and is promoted to-0.53 ,-0.67 ,-0.37 respectively from-0.45 ,-0.66 ,-0.31 from-0.48 ,-0.77 ,-0.55.

2 lower 9 image features of planting density are similar with kind to the relevant level of Leaf nitrogen concentration, and the dependency of monochromatic component r and g is higher than b;The degree of correlation of linear combination parameter r-b is better than 3 single amounts and other 2 combination parameters, the R of r-b and LNC under D2 density²Reach-0.61; Relative to 3 monochromatic component and 3 linear combination parameter, the dependency of the normalizing parameter of D2 density promotes to some extent.

Dependency under 3 nitrogen amount applied is similar to planting density, and monochromatic component is more weak; 3 linear combination parameter then relatively strong, compared with r-g and r-g-b, r-b is higher and stable; Meanwhile, the dependency of 3 normalizing parameters further enhances, the R of r-b under different Nitrogen applications²Respectively-0.62 ,-0.73 and-0.4, and the R of (r-b)/(r+g+b)²Then it is respectively increased-0.72 ,-0.74 and-0.41; The dependency that N1 executes nitrogen scheme is comparatively notable relative to N0 and N2.

2.2. the matching of image evaluation of nutrition index with determine

Correlation analysis finds, the sign power of Leaf nitrogen concentration is differed by 3 monochromatic component of rgb space, but the related advantages of 3 linear combination parameter is better than 3 monochromatic component, and the dependency of r-b is higher than r-g and r-g-b and stablizes. Visible, although g, b judge that the degree of accuracy of Leaf nitrogen concentration is relatively low and built on the sand, but there is also the contribution degree can not ignore; After the further standardization of linear combination parameter, the lifting of degree of correlation becomes apparent from. Therefore, the structure of wheat nitrogen nutrition picture appraisal index, 3 foundation characteristics component r, g, b and weight thereof should be weighed, be optimized combination and standardization.

First color combinatorial index CMI (ColorMixIndex)=(xr-yg-zb) treating tuning is built, then with the LNC of identical field through multi-scheme matching repeatedly, determine the weight coefficient x, y, z of 3 components, finally to CMI standardization, it is determined that NCMI.

If NCMI=(xr-yg-zb)/(r+g+b), fit procedure is as follows:

(1) setting x ∈ [1,2], y ∈ [-1,1], z ∈ [-1,1], the contribution degree of absolute value more big then corresponding color component is more big, and 0 namely without contribution degree;

(2) in above-mentioned scope, constantly adjust x, y, z value by step-length (being set to 0.05), carry out regression analysis with Leaf nitrogen concentration, calculate x, y, z value and coefficient of determination R², build four-dimensional array [x, y, z, R²]。

(3) four-dimensional array [x, y, z, R are drawn²] three-dimensional distribution map (see Fig. 1). Distribution of color represents R²Just, the red area that CMI is the highest with nitrogen nutrition dependency, it is mainly distributed on x ∈ [1.5,2], z ∈ [-1 ,-0.5] or y ∈ [-0.5,0], z ∈ [-1 ,-0.5].

(4)R²When being 0.83353 to the maximum, corresponding x, y, z are the optimal value of r, g, b linear fit, x=1.6, y=-0.95, z=-0.8.

(5) to CMI standardization, it is determined that the concrete formula of NCMI is as follows:

NCMI=(1.6r-0.95g-0.8b)/(r+g+b) (1)

R, g, b data of field to be measured are substituted into above-mentioned formula, so that it may calculating the color combination standard index NCMI obtaining corresponding field, this index is nitrogen content predictive value.

3 interpretations of result

Divide the experimental data in independent time (2014) by different growing and different Plant plane, adopt the coefficient of determination (R²) and the dependencys of forecast error (RMSE) quantitative analysis NCMI and comparatively ripe image parameter HONGGUANG standardized value NRI, the bottle green index D GCI of 3 research and green glow and HONGGUANG ratio G/R with Leaf nitrogen concentration, the feasibility of research NCMI sign Wheat Leaves nitrogen content.

$NRI=\frac{r}{r+g+b}---\left(2\right)$

$DGCI=\frac{(h-60)/60+(1-s)+(1-i)}{3}---\left(3\right)$

G/R=g/r (4)

The analysis contrast of 3.1 Wheat Cultivars

Result shows (table 3), the R of 4 image features of 2 kinds², March 8 is relatively low, and April 15 is the highest, and error is minimum.Think, March 8, Semen Tritici aestivi was in period of seedling establishment and jointing stage separation, the wheat nitrogen of this growing stage absorbs, transhipment is more active, the Nitrogen Accumulation of self stem and leaf is in dynamic changes phase, to a certain degree have impact on the accuracy measuring Leaf nitrogen concentration LNC, and the nutriture of Semen Tritici aestivi instability reflects canopy leaves surface, it is also possible to cause that the image features extracted exists deviation, reduce further dependency between the two; Enter into late growth stage (April 15), Semen Tritici aestivi enters the second stage on nitrogen accumulation peak, nitrogen conveying operating and other Related Physiological Characteristics feed back to surface and the structure of canopy leaves all more completely, the image now obtained and nutrition parameters are all comparatively firm, and therefore degree of fitting and estimation precision are higher.

Because with plant height, the different plant type of wheat breed causes that the reflectivity curve of visible region canopy spectra is different, 2 kind R²Slightly distinguish with RMSE (see table 3), but compared to other 3 characteristic parameters, NCMI maintains more stable fitting degree and estimation precision, the R of V1 and 2 nutritive index²It is maintained at-0.76～-0.95, RMSE between 1.833～3.893; The R of V2²Floating-0.69～-0.98, RMSE is in 1.249～4.36 scopes. Wherein, V1 is higher at the dependency of 3 period of duration, error of fitting is relatively low, April 15, although the R of NRI and NCMI and LNC²It is-0.95, but the RMSE of NCMI is lower than NRI; The R in V2 kind only March 8²Slightly weak with RMSE, March 31 and April NCMI and NRI and LNC on the 15th R²Equal, respectively-0.84 ,-0.94, NCMI obtain the RMSE value lower than NRI equally. Thinking, the acquisition of image parameter and wheat canopy reflectance are closely related, and research shows from the jointing stage to florescence, it is seen that light regional reflex rises after taking the lead in dropping, and late growth stage, blade starts to turn yellow, it is seen that the canopy spectra reflectance of light part rises. This Changing Pattern of different growing wheat canopy EO-1 hyperion and multispectral reflectance, make later stage NCMI value because 3 monochromatic component are with the lifting of reflectance, because the proportion that optimizes arranged obtains enhancing in various degree further, comparing NRI and the G/R only considering single color component, degree of correlation and forecast error all increase.

The coefficient of determination R of table 3 different cultivars wheat canopy characteristic parameter and LNC²With root-mean-square error RMSE

From table 4, fertility early stage, except individual data, the degree of correlation of lower 4 image features of D2 density and LNC and estimation precision are all in various degree higher than D1; But with the passage of period of duration, the dependency gap of D1 and D2 tapers into, and R²Value rises, at peaking on April 15. From optics and imaging angle analysis, under the environment of land for growing field crops, the Semen Tritici aestivi image of shooting is by ambient interferences such as soil, chad, weeds, may result in wheat canopy spectrum, in red edge regions, " red shift " phenomenon occurs, reduce dependency stronger long-wave band (HONGGUANG) reflectance, and the short-wave band that dependency is more weak (blue light and green glow) reflectance increases. March 8, wheat plant was less, under low-density D1, the ambient interferences such as soil, weeds is more prominent relative to D2, later stage wheat plant is grown up, canopy is dense, under D1 and D2 density, the difference by ambient interferences reduces, thus image parameter related levels and estimation precision step up, to the sign ability of nitrogen nutrition closer to. But under D1 density, DGCI under March 8 and March 31 HSI space is slightly better than 3 image parameters under rgb space, thinking, component H and S relevant to color under HSI color space separates with brightness I, thus reduces the soil environment impact on reflectance spectrum.

Visible, the reflection of rgb space lower 3 image features wheat nitrogen nutriture to possessing certain Canopy cover degrees is more accurate, the NCMI proposed shows more prominent under D2 density, it is maintained at-0.85～-0.97 with the coefficient of determination of 2 nutritive indexs, estimation error is maintained at 1.299～3.505, wherein, the R of March 8 and LNC²With RMSE respectively-0.91 and 3.297, the R of April 15 and LNC²-0.97 and 1.299 it is divided into, with the R of SPAD with RMSE²-0.97 and 1.630 it is divided into RMSE.

The coefficient of determination R of the horizontal wheat canopy characteristic parameter of table 4 different densities and LNC²With root-mean-square error RMSE

For reducing the interference of level of density, only analyze the dependency of different Nitrogen Levels under D2 density. Table 5 shows, N0, N1 and N2 process lower R²Differing greatly with RMSE, N1 is relatively strong. Thinking, under different Nitrogen applications, wheat canopy visible light wave range reflectance there are differences, and along with nitrogen amount applied increases, chlorophyll content increases, and the absorption of major part solar visible radiation strengthens to some extent, and canopy reflection coefficient decreases; But excessive nitrogen (N2) of executing is likely to suppress the Semen Tritici aestivi absorption to phosphorus, potassium and other trace element, causes that canopy leaves is off color, and physiological status difference, the image parameter of acquisition and nutritive index all exist deviation, and dependency weakens to some extent. Simultaneously, during Semen Tritici aestivi nitrogen stress (N0), blade LNC is relatively low, but owing to the nitrogen in Lao Ye shifts to young leaves, shows as the first chlorisis yellow of plant lower blade, and gradually to top leaf expansion, this easy service performance makes the red component r under N0 rise, and b component declines, and g component is relatively stable, mistake causes NCMI value to go up not down (see formula (1)), reduces the dependency with LNC.

Table 5 shows simultaneously, NCMI shows, under different nitrogen amount applied, the degree of correlation and law characteristic that match with above-mentioned analysis, under same Nitrogen Level, there is the fitting degree higher than other parameters and preferably estimation precision, dependency under N1 is the most notable, error precision is minimum, with the dependency of LNC and error respectively-0.78 and 1.960.

Table 5 difference executes the coefficient of determination R of nitrogen scheme wheat canopy characteristic parameter and LNC²With root-mean-square error RMSE

Above-mentioned correlation analysis and precision estimation show, relative to other 3 typical image evaluatings, linear fit parameter NCMI by each base color component proportion of tuning standardization, under 3 period of duration, 2 wheat breeds, 2 level of density and 3 nitrogen amount applied, its dependency with Leaf nitrogen concentration LNC and fitting precision all maintain good stability.

Claims (6)

1. the method for building up for the picture appraisal index of wheat nitrogen nutrition Nondestructive, it is characterised in that it comprises the following steps:

(1), sampling: obtain the Leaf nitrogen concentration of wheat canopy image, wheat plant;

(2), the K means clustering method segmented extraction wheat canopy image based on H component is utilized, calculate Color characteristics parameters based on the average pixel value of every all non-zero pixel R of width wheat canopy image Leaf, G, B, H, S, I component, be respectively defined as r, g, b, h, s, i; Calculate 9 image features, i.e. 3 monochromatic component r, g, b, 3 linear combination parameter r-g-b, r-g, r-b and 3 normalizing parameter (r-g-b)/(r+g+b), (r-g)/(r+g+b), (r-b)/(r+g+b);

(3) color combinatorial index CMI=(xr-yg-zb) treating tuning, is built, then with the Leaf nitrogen concentration matching of identical field, determine the weight coefficient x, y, z of 3 components, finally to CMI standardization, it is determined that color combination standard index NCMI:NCMI=(xr-yg-zb)/(r+g+b).

2. the method for building up of the picture appraisal index of wheat nitrogen nutrition Nondestructive according to claim 1, it is characterized in that in step (1), the acquisition methods of wheat canopy image: at wheat during jointing stage, adopts camera from canopy 1.0m height and the shooting sampling of 90 ° of ground.

3. the method for building up of the picture appraisal index of wheat nitrogen nutrition Nondestructive according to claim 1, it is characterized in that in step (1), the acquisition methods of the Leaf nitrogen concentration of wheat plant: wheatland top destructiveness was sampled the same day by wheat canopy image taking, select 15～25 strain plant, the Leaf nitrogen concentration of Kjeldahl nitrogen determination wheat plant, averages.

4. the method for building up of the picture appraisal index of wheat nitrogen nutrition Nondestructive according to claim 1, it is characterized in that in step (1), sample point picks up from different cultivars, different growing, different nitrogen amount applied, different planting density and different year.

5. the method for building up of the picture appraisal index of wheat nitrogen nutrition Nondestructive according to claim 1, it is characterised in that in step (3), fit procedure is as follows:

A, set x ∈ [1,2], y ∈ [-1,1], z ∈ [-1,1], the contribution degree of absolute value more big then corresponding color component is more big, and 0 namely without contribution degree;

B, in above-mentioned scope, constantly adjust x, y, z value by step-length (being set to 0.05), carry out regression analysis with Leaf nitrogen concentration, calculate x, y, z value and coefficient of determination R², build four-dimensional array [x, y, z, R²];

C, drafting four-dimensional array [x, y, z, R²] three-dimensional distribution map; Distribution of color represents R²Just, the region that CMI is the highest with Leaf nitrogen concentration dependency, it is distributed in x ∈ [1.5,2], z ∈ [-1 ,-0.5] or y ∈ [-0.5,0], z ∈ [-1 ,-0.5];

D, for wheat crop, R²When being 0.83353 to the maximum, corresponding x, y, z are the optimal value of r, g, b linear fit, x=1.6, y=-0.95, z=-0.8.

E, to CMI standardization, it is determined that the concrete formula of NCMI is as follows:

NCMI=(1.6r-0.95g-0.8b)/(r+g+b).

6. the method for a wheat nitrogen nutrition Nondestructive, it is characterised in that comprise the following steps:

(1) obtain the wheat canopy image of field to be measured, with the K means clustering method segmented extraction wheat canopy image based on H component, calculate 3 monochromatic component r, g, b;

(2) method according to any one of Claims 1 to 5 determines the x, y, z in color combination standard index NCMI, wherein, and NCMI=(xr-yg-zb)/(r+g+b);

(3) value of 3 the monochromatic component r obtained in step (1), g, b being substituted in the NCMI computing formula determined in claim 2, calculated color combination standard index NCMI is the Semen Tritici aestivi nitrogen content of the field to be measured of prediction.