CN114202702A - Based on D-fGRemote sensing estimation method for dynamic harvest index of winter wheat acquired by parameter remote sensing - Google Patents

Abstract

The invention discloses a method based on D-f_GThe invention discloses a method for estimating a dynamic harvest index of winter wheat by parameter remote sensing, which is based on field control experiments of winter wheat and provides dynamic parameters D-f of the ratio between aboveground biomass accumulated at different periods during the period from a flowering period to a mature period and aboveground biomass at corresponding periods_GOn the basis, narrow-band spectral indexes NDSI and D-f constructed by hyperspectral data of crop canopy_GScreening out D-f by the correlation relationship between_GThe estimated sensitive wave band center and the optimal wave band are combined, so that the high spectrum pair D-f of the canopy is utilized_GAccurate estimation of. Finally, based on D-f_GParameter remote sensing information and measured D-f_GQuantitative relationship with measured Dynamic crop Harvest Index (D-HI)The method realizes the optimal estimation of the harvest index of the winter wheat crops based on canopy hyperspectral remote sensing, so as to further improve the estimation precision and level of the harvest index of the winter wheat, and provide theoretical basis and technical reference for the high-precision acquisition of the harvest index of the crops by using a remote sensing satellite technology in a large range.

Description

Based on D-fGWinter wheat dynamic harvesting finger obtained by parameter remote sensingNumerical remote sensing estimation method

Technical Field

The invention relates to a hyperspectral sensitive band D-f based on a canopy_GA remote sensing estimation method for the dynamic harvest index of winter wheat obtained by parameter remote sensing.

Background

Harvest Index (HI), also called economic coefficient, refers to the ratio of the economic yield (grain, fruit) to the biological yield at crop Harvest, which reflects the proportion of assimilation products distributed between grain and vegetative organs. For food crops (e.g., wheat, corn, etc.), harvest index, i.e., the percentage of grain yield to the above-ground biological yield, where above-ground biological yield refers to the above-ground total dry matter mass (Donald, 1962; Donald and Hamblin, 1976; Pandawn, et al, 2007). As the crop harvest index can play an important role in aspects of crop yield simulation and estimation (Fan et al, 2017; Hu et al, 2019; Lorenz et al, 2010), crop variety breeding (Hay, 1995; river-Amado et al, 2019), crop growth and cultivation environment evaluation (Porter et al, 2020; Yang and Zhang, 2010), crop carbon fixation capacity evaluation (Chen et al, 2021; Unkovich et al, 2010) and agricultural response to climate change and the like, once the concept is provided, the crop harvest index becomes a research hotspot of scholars at home and abroad (Walter et al, 2018; Jixingjie et al, 2010).

Currently, crop harvest index estimation is extensively studied mainly from 2 cases on field scale and regional scale. Among field-scale crop HI estimates, some researchers have conducted intensive studies on the simulation of crop harvest index estimation and the influence of environmental stress factors (such as high temperature, water deficit, soil nutrient deficit or excess, etc.) on the formation of crop harvest index, mainly from agronomic and crop science perspectives (Fletcher and Jamieson, 2009; Kemanian et al, 2007; solani et al, 2005). For example, Fletcher and Jamieson (2009) carry out dynamic simulation of wheat harvest index change along with time and research on influence factors thereof, and research results show that the change rate of the wheat harvest index is closely related to the aboveground biomass of crops at the early stage of grain filling and the growth rate of the crops in the grain filling process,and the wheat harvest index shows curve change along with time, which has important guiding function for developing dynamic simulation and estimation of the winter wheat harvest index. Kemanian et al (2007) studied crops of wheat, barley and sorghum according to the ratio of the accumulated dry matter of HI and flowering season to the total dry matter of the whole growing season (f)_G) In a linear or curvilinear relationship, f is established in the field scale_GAnd the statistical model between HI realizes accurate simulation and estimation of the harvest index of the field-scale crops. Meanwhile, Li and the like (2011) are used for field control experiments of winter wheat in Yu cities in Shandong province in China based on different nitrogen levels, and the ratio of the accumulated dry matter amount of the flowering crops to the total dry matter amount of the whole growing season is utilized (f)_G) And the measured data is used for carrying out research on the winter wheat HI estimation method, so that a better crop harvest index simulation result is obtained. Results of the above study on the utilization of f_GThe estimation of the crop harvest index by parameters is of great reference significance, but since the above studies only consider the crop f at the mature stage_GParameters and maturity crop harvest index, both without taking into account f_GThe dynamic changes of the parameters and the harvest index influence the precision of estimation and simulation of the crop harvest index, thereby influencing the stability of the estimation result of the harvest index and further improving the estimation precision to a certain extent. In addition, some scholars have conducted wheat HI estimation studies based on the ratio of the transpiration from flowering to maturity of a crop to the total transpiration throughout the growth period (Li et al, 2011; Richards and Townley-Smith, 1987; Sadras and Connor, 1991), and the methods proposed in the studies have effectively estimated the HI of winter wheat under water deficit conditions, but under conditions of sufficient water and stress of other environmental factors (such as nitrogen stress), crop HI estimation still needs further intensive research.

In the estimation of the HI of the crops based on the regional scale, the traditional method adopts a point-to-surface method and a space interpolation method to obtain the harvest index of the regional crops (Nijian Qiang et al. 2010). Wherein, the point-by-point surface method is to take the mean value of the annual harvest indexes obtained by the fixed-point test as the regional harvest index; the spatial interpolation method is to carry out spatial interpolation on the harvesting indexes of the actual investigation multipoint crops to obtain the regional spatial distribution of the harvesting indexes in the current year. In recent years, the followingWith the rapid development of the remote sensing technology, the remote sensing technology provides a reliable technical means for accurately acquiring regional crop harvest index spatial information by virtue of the advantages of large coverage, rapid and accurate acquisition of surface crop parameter information (Campoy et al, 2020; Walter et al, 2018). The scholars at home and abroad develop a series of crop harvest index estimation researches (Li et al, 2011; Morion et al, 2007) on the basis of time sequence vegetation remote sensing information (such as normalized vegetation index, leaf area index and the like) which is acquired by remote sensing satellites and can reflect the growth conditions of crops. For example, Morion et al (2007) divides the whole growth period of winter wheat into two stages of germination-flowering and flowering-maturation, and constructs a model 1-NDVI according to the average value of NDVI in two time periods before and after flowering_post/NDVI_preThe spatial distribution of HI is estimated, the method can acquire the NDVI data of the winter wheat in the growing season by a remote sensing means, and the method has important reference significance for acquiring regional scale HI by utilizing remote sensing information. Meanwhile, the method is further applied by Chinese scholars, such as Du et al (2009) utilize MERIS NDVI time sequence data to carry out inversion and verification of the regional winter wheat harvest index in Limonitum in Shandong, and the regional winter wheat harvest index result is applied to the crop yield estimation research. And (2010) taking winter wheat in the plain region of Huang-Huai-Hai in China as a research object, representing the harvest index of the winter wheat by using the ratio of the NDVI (normalized difference of viscosity) accumulated value in the flowering period-milk stage of the wheat and the NDVI accumulated value before returning green and flowering, and well estimating the harvest index of the winter wheat on the regional scale by establishing a statistical model between the ratio and the actually-measured harvest index. The method is simple and easy to implement, the time sequence of the required remote sensing data is short and easy to obtain, and the method is beneficial to the practical application of the method, but the methods only aim at the estimation of the harvest index in the mature period, and the acquisition of index information in the dynamic change process of the harvest index change is not realized, so that further enhanced research is needed.

Accordingly, the prior art is deficient and needs improvement.

Disclosure of Invention

The invention aims to solve the technical problem of providing a canopy-based hyperspectral sensitive band D-f_GWinter wheat dynamic harvest index remote sensing acquired by parameter remote sensingAn estimation method.

The technical scheme of the invention is as follows:

based on D-f_GThe remote sensing estimation method for the dynamic harvest index of winter wheat acquired by parameter remote sensing comprises the following steps:

a1, according to the ground actual measurement dynamic biomass data, constructing a ratio dynamic parameter D-f between the overground biomass accumulated at different periods during the flowering period-mature period of the crops and the overground biomass at the corresponding period_G；

D-f_GThe calculation method is as follows:

in the formula, sigma W_postIs the overground biomass (kg/hm) accumulated at different periods during the flowering period and the mature period of the winter wheat²)；

∑W_wholeIs total aboveground biomass (kg/hm) corresponding to the sampling period²) (ii) a t is the sampling time, W_tWeight of dry matter (kg/hm) at time t sample²)，W_aWeight in terms of dry matter (kg/hm) at flowering stage²)，D-f_G,tA ratio parameter representing the t sample time;

a2, constructing any two canopy high-spectrum narrow-band spectral indexes NDSI based on ground crop canopy high-spectrum data, and establishing NDSI and winter wheat D-f_GA linear model in between;

a3, then drawing and analyzing NDSI and winter wheat D-f_GFitting accuracy R between²A two-dimensional map;

a4, by determining R²Maximum value area and center of gravity of the maximum value area, thereby obtaining winter wheat D-f_GA sensitive band center;

a5, determination of D-f_GEstimating an optimal band combination;

a6 based on NDSI and D-f_GD-f of a relation_GA remote sensing estimation model;

A7、D-f_Gremote sensing estimation of (2);

a8, obtaining based on D-f_GAnd a dynamic harvest index estimation model of the D-HI relationship;

remote sensing estimation of A9, D-HI.

The method, said step A4, according to R²Obtaining the center of the sensitive wave band by the gravity center method of the maximum area so as to determine D-f_GEstimated sensitive bands, i.e. narrow band spectral indices NDSI and D-f corresponding to each band of hyperspectral in the canopy_GOn the basis of inter-correlation calculation, a maximum area is determined according to a threshold value of a correlation coefficient meeting the statistical significance requirement, and on the basis, the gravity center of the maximum area of the correlation coefficient is calculated, so that NDSI and D-f are obtained_GThe combination of the center and the band of the spectral band with larger parameter correlation; the specific process is as follows:

firstly, drawing NDSI and winter wheat D-f_GFitting between R²Determining NDSI and winter wheat D-f on the basis of two-dimensional map_GA band region in which the inter-correlation is high; secondly, find R in this region²Maximum value point, traversing all points meeting significance condition in the neighborhood of the point 8, and marking the set of the points as R²A maximum region Ω; finally, calculate R²The center of gravity of the local maximum value region is defined as each R²A sensitivity band center of a maximum region; the calculation formula (7) of the center of gravity is as follows:

wherein f (u, v) is R with band coordinates (u, v)²The value, omega, is the area of maximum,

center coordinates of the sensitive band.

The method, the step A4, for winter wheat D-f_GThe 6 sensitive band centers of sensitivity include λ (443nm, 506nm), λ (442nm, 635nm), λ (732nm, 834nm), λ (787nm, 804nm), λ (810nm, 877nm) and λ (861nm, 985 nm).

Said method, said step A6In NDSI and D-f_GD-f of a relation_GThe remote sensing estimation model is as follows:

D-f_G,t＝m×NDSI_i，j，t+n (5)

wherein i and j are hyperspectral wave bands of 350-1000nm respectively, t is different sampling time, and m and n are fitting parameters in a linear equation obtained after fitting; according to the formula, calculating and solving the ratio parameter D-f between the accumulated aboveground biomass of the crops from the flowering period to the t period and the aboveground biomass of the t growth period_G,t。

The method, the step A8, obtains the D-f-based_GThe dynamic harvest index estimation model for the D-HI relationship is:

D-HI_t＝HI₀+s×D-f_G,t (6)

wherein, HI₀Is the intercept, i.e. the value of the dynamic harvest index without a change in biomass after the flowering phase of the crop, i.e. when D-f_G,tA value for D-HI harvest index at 0; s is D-HI or D-f_GSlope constant in linear relationship.

The method comprises the step A8 of calculating the D-f of 128 sample points in a winter wheat cell according to dynamic aboveground biomass data of the winter wheat at different acquisition times during the flowering phase and the mature phase and dynamic data of grain yield in the grouting process_GAnd a dynamic harvest index D-HI, on the basis of which D-f is normalized by the formula (6)_GAnd fitting the correlation between the dynamic harvest index D-HI to obtain D-f_GAnd a dynamic harvest index D-HI inter-estimation model, which is specifically as follows:

D-HI_t＝0.1018+0.8093*D-f_G,t

a winter wheat yield estimation method based on a winter wheat dynamic harvest index is disclosed, wherein the winter wheat dynamic harvest index is obtained by adopting any one of the methods.

The method of the invention has the following beneficial effects:

(1) from a static state f_GThe parameter is developed into a dynamic f_GParameter(s)

Predecessor studies were based on f_GThe crop harvest index estimation study ofThe ratio of the aboveground biomass accumulated in the flowering period and the mature period of the crops to the aboveground biomass in the mature period is not considered, and the parameter belongs to a static parameter because the dynamic change of the yield of grains in the growth process of the crops is not considered. Taking into account that there is already a static-based f_G(S-f_G) In the crop harvest index estimation research, because the time years of modeling data and verification data are short, the harvest index estimation model possibly has the problem of low stability, and aiming at the situation, the general static parameters S-f are further used by the invention_GDynamic parameters D-f developed to take into account the ratio between aboveground biomass accumulated during different periods of growth during the flowering-maturation phase and aboveground biomass at the corresponding periods_G(Dynamic f_G). Due to the consideration of the dynamic D-f of the flowering period-mature period of the crops_GThe crop growth dynamic change process information can be considered to a certain extent, the number of modeling samples per year can be increased, and therefore a crop harvest index estimation model and an estimation result with high stability and accuracy can be obtained by utilizing modeling data in a short period.

(2) Proposes to obtain f based on remote sensing technology_GTechnical method of parameter information

General acquisition f_GThe method is to obtain the biomass on the crop ground by artificially observing the biomass on the ground on the field scale, and the invention adopts the static S-f_GDevelopment to dynamic D-f_GOn the basis, NDSI remote sensing information is obtained through canopy hyperspectrum, and then a gravity center model is used for D-f_GThe remote sensing estimation sensitive wave band is screened, and D-f is realized_GAnd remote sensing estimation of the parameters. The invention provides a method for constructing NDSI D-f based on the center of a canopy spectrum sensitive waveband_GThe remote sensing estimation technical method can be used for unmanned aerial vehicle remote sensing and D-f in a large range by using a remote sensing satellite technology_GThe method lays a technical foundation for obtaining parameter remote sensing and is also based on actual measurement S-f for field scale_GThe up-scaling application of the crop harvest index estimation method provides a new idea and a new technical method.

(3) Propose based on dynamic f_GDynamic harvest index remote sensing estimation technical method obtained by parameter remote sensing

To is directed atThere has been a conventional f-based_GThe parameter crop harvest index estimation is realized by completely adopting ground measured data, the scale-up area application of the method cannot be realized by utilizing remote sensing information, the research of estimating the crop harvest index by utilizing the remote sensing information generally only estimates the crop mature period harvest index, but the dynamic change process of the crop harvest index is not considered enough, and the further improvement of the estimation precision of the crop harvest index is influenced to a certain extent, so that the dynamic f is obtained by fully utilizing the remote sensing data_GOn the basis of the parameter remote sensing information, the invention provides a dynamic crop harvest index remote sensing estimation method considering crop biomass change and grain formation process in different periods of a grouting period and a mature period, improves the stability and the precision of a crop harvest index remote sensing estimation model to a certain extent, and breaks through the traditional f-based remote sensing estimation model_GThe parameter crop harvest index estimation method can not utilize remote sensing information to carry out the bottleneck of upscaling application, and realizes the dynamic f-based_GAnd performing remote sensing high-precision estimation on the dynamic harvest index obtained by parameter remote sensing.

Drawings

FIG. 1 shows the location of a research area and the layout of test cells;

FIG. 2 is a high spectral curve of winter wheat canopy at different growth periods in the test plot;

FIG. 3 is a technical roadmap;

FIG. 4 is a graph of crop D-f estimation based on NDSI_GDetermining a schematic diagram of the center of the remote sensing sensitive wave band;

FIG. 5 is S-f_GAnd maturity harvest index G-HI;

FIG. 6 is a graph based on actual measurements of S-f_GVerifying the estimation precision of the crop harvest index in the mature period;

FIG. 7 shows NDSI two-dimensional distribution (2020) of winter wheat in normal horizontal fertilization and irrigation treatment (N2W 2); a. flowering period (5 months and 10 days), b, early grouting period (5 months and 18 days), c, middle grouting period (5 months and 24 days), d, later grouting period (6 months and 3 days), and e, mature period (6 months and 19 days);

FIG. 8 shows NDSI and winter wheat D-f_GFitting between R²A two-dimensional map;

FIG. 9 shows NDSI and winter wheat D-f_GFitting between R²A two-dimensional contour map;

FIG. 10 shows NDSI and D-f constructed based on the center of the sensitive band_GConstructing an inter model;

FIG. 11 shows D-f based on the center of the sensitive band_GVerifying an estimation result;

FIG. 12 is a diagram based on D-f_GEstablishing a D-HI estimation model of the remote sensing parameters;

FIG. 13 shows a spectrum-based sensitive band D-f_GD-HI estimation result verification of the parameters;

FIG. 14 shows a spectrum-based sensitive band D-f_GVerifying the estimation result of the parameter maturity D-HI;

sensitive band centers corresponding to the respective graphs in fig. 10, 11, 13, 14: a (443nm, 506nm), b (442nm, 635nm), c (732nm, 834nm), d (787nm, 804nm), e (810nm, 877nm), f (861nm, 985 nm);

Detailed Description

The present invention will be described in detail with reference to specific examples.

1 materials and methods

1.1 study area and design of experiments

The field test of the invention is arranged in a comprehensive test demonstration base (116 degrees 92 '-116 degrees 94' E, 40 degrees 05 '-40 degrees 06' N) of the agricultural environment of the cisterm district in Beijing city. The research area belongs to a warm-temperate zone semi-humid continental monsoon climate, the annual average air temperature is about 11.5 ℃, the annual average precipitation is about 625mm, the annual sunshine is about 2750 hours, and the frost-free period is about 195 days. The main planting system in the research area is winter wheat-summer corn which is cooked twice a year. The study area location map is shown in figure 1.

In order to obtain winter wheat observation crops with different growth vigor and growth states, considering that the growth vigor and the yield of winter wheat in a research area are mainly influenced by fertilization and irrigation at present, and particularly, nitrogen and soil moisture have an important control effect on the growth vigor and the yield of the winter wheat, the design test of the invention mainly considers two factors of nitrogen fertilizer and moisture. Winter wheat was tested between 10 months in 2019 and 6 months in 2020. The wheat variety selected for the test is round selection 987, and the base fertilizer is calcium superphosphate and potassium sulfate (P)₂O₅Is 135kghm²、K₂O is 90kg/hm²). The experiment set up 4 nitrogen supply levels: n0 (no nitrogen fertilizer application), N1 (nitrogen application amount 160 kg/hm)²) N2 (nitrogen application of 240 kg/hm)²) N3 (nitrogen application of 320 kg/hm)²). The nitrogen fertilizer is applied in two times, namely, the base fertilizer and the top dressing in the green turning period are respectively applied by half of the fertilizer amount during sowing, and the top dressing time is the green turning-jointing period of the winter wheat. The treatment of the irrigation water quantity in the whole growth period is divided into four levels: w0(0mm), W1(100mm), W2(150mm), W3(200 mm). Wherein, W2 is the total amount of normal irrigation of local winter wheat in the whole growth period. The winter wheat is irrigated for 4 times in total, the irrigation time is respectively before winter, the green turning period, the jointing period and the booting period, and each cell is irrigated with water of 0mm, 25mm, 37.5mm and 50mm on average. Other cultivation management of the winter wheat test is consistent with the local traditional winter wheat management measures. Setting 4 different nitrogen applying amounts and 4 different water treatments in the whole growth period of winter wheat, wherein the total number of the treatments is 16, each treatment is set to be 3 times, the total number of the treatments is 48 test cells, and the area of each cell is 30m²(5 m.times.6 m). In the invention, the seeding time of the control experiment is 10 and 9 months in 2019, the heading-blooming period of the winter wheat is 5 and above months in 2020, the filling-milk stage (5 and 6 middle months in 2020), and the maturity stage (6 and 6 middle months in 2020).

1.2 data acquisition and processing

The ground data acquisition mainly comprises the observation of indexes such as the overground fresh biomass of the winter wheat, the overground dry biomass and the hyperspectral data of the canopy of the winter wheat in each cell, and in the data acquisition process, the ground data acquisition work is carried out by mainly selecting the date with clear and cloudy weather and stable sunlight intensity. Finally, the invention carries out 5 times of ground data acquisition work in 5 months and 10 days (flowering period), 5 months and 18 days (early stage of grouting), 5 months and 24 days (middle stage of grouting), 6 months and 3 days (late stage of grouting) and 6 months and 19 days (mature period) in 2020. The test is carried out in 48 cells in total, 2 sampling points are arranged in each cell in total, 5 times of sample collection is carried out in total, 10 wheat plant areas (about 1 row and 20cm) with basically consistent growth vigor are selected in each cell in the flowering period in order to ensure the data collection quality, and a label is tied to be used as a mark for carrying out the on-ground fresh biomass and canopy hyperspectral data collection of winter wheat at each sampling point in the flowering-maturation period of wheat. In order to accurately obtain the aboveground fresh biomass, aboveground dry biomass data and canopy spectrum data of each cell, the data obtained by 2 sampling points in each cell are respectively subjected to average processing, so that the quality of ground observation data participating in modeling and model verification is improved.

(1) Aboveground biomass harvesting

During ground observation each time, the overground part of the winter wheat with the row length of 20cm is taken as a sample at the mark point in each cell, and then the collected winter wheat sample is sealed and stored by a self-sealing bag and transported back to a laboratory for analysis. In a laboratory, firstly weighing and recording the total fresh weight of wheat at each sampling point; secondly, separating the stem leaves and the ears of each sampling point, respectively putting the stem leaves and the ears into paper bags, and weighing the fresh weight of the stem leaves and the ears; thirdly, the separated wheat stem leaves and ears are placed in an oven for fixation treatment at 105 ℃ for 30min, then the sample is dried to constant weight at 85 ℃, and the dry weight weighing results of the stem leaves and ears of each cell observation sample point are recorded. And finally, obtaining the dry biomass weight of each cell sample point.

(2) Acquisition of dynamic harvest index

On the basis of obtaining the dry weight of the winter wheat stem leaves and ears with the row length of 20cm at each cell sampling point, threshing the wheat ears at each sampling point respectively, and then weighing and recording the seed weight of each sampling point. And finally, calculating the winter wheat harvest index of each ground observation time in the winter wheat grain filling process. As the crop Harvest Index is gradually formed and dynamically changed along with the change of the Harvest Index along with time in the grain filling process, the winter wheat Harvest Index of each ground observation time in the winter wheat grain filling process is called as a Dynamic Harvest Index (D-HI).

Wherein t is the ground sampling time during the period from grouting to mature period, W_z，t、W_J，t、W_Y，t、W_S，tThe grain, stem, leaf, and the like of winter wheat are respectively sampled at the time t from the filling to the mature period,Dry weight of ear, D-HI_tIs the dynamic harvest index at time t sample during the grout to maturity phase.

(3) Canopy hyperspectral data measurement

The canopy hyperspectral measurement mainly utilizes an American ASD Field Spec 4 spectral radiometer to carry out ground spectrum collection on a uniform growth area in 48 cells, and the measurement range of the spectrometer is 350-2500 nm. Wherein the sampling interval in the wavelength of 350-1000nm is 1.4nm, the sampling interval in the wavelength of 1000-2500 nm is 2nm, and the data interval after resampling is 1 nm. Before each measurement, the standard white board is used for correction, the probe is vertically downward during measurement, the visual angle of the probe of the spectrum device is 25 degrees, and the height of the probe from the top of the crop canopy is about 0.5 m. Each cell takes 2 sampling points, 5 spectral data are read at the optimal time interval of each sampling point, and the average value is taken as the spectral reflectance value of the cell so as to reduce noise interference and randomness. In the invention, the collection of the crop canopy spectrum is carried out at the local time of 10: 00-14: 00 under the conditions of good weather conditions and sufficient sunlight irradiation.

The preprocessing of the canopy hyperspectral data mainly comprises spectrum averaging and spectrum smoothing. Wherein, the spectral data mean value processing utilizes ViewSpecPro software, and the mean value is used as the reflection spectrum value of the corresponding sampling point. The spectral smoothing process mainly utilizes a 9-point weighted moving average method in ENVI software to carry out smoothing and denoising on spectral reflectivity data. And finally, obtaining the high spectral reflectance data of the winter wheat canopy at each sampling point of the observation cell. The spectra of the 48 test cells are averaged and smoothed to obtain different canopy hyperspectral curves for winter wheat at different growth periods as shown in fig. 2.

1.3 methods of investigation

1.3.1 basic concept

(1) Dynamic harvest index

The general Harvest Index (Harvest Index, HI) considers only the percentage of the total dry matter mass above the yield of mature crop kernels (Donald and Hamblin 1976), which is the maximum value of the crop Harvest Index, the final Harvest Index. In order to improve the precision of the estimation of the crop Harvest Index and the stability of a crop Harvest Index model, the method considers the Harvest Index in the mature period and the Dynamic information of the Harvest Index of the crops, which gradually forms and changes along with time, and is called as the Dynamic Harvest Index (D-HI). To distinguish the proposed D-HI indices, the general Harvest index is defined herein as G-HI (general Harvest index). For food crops (such as wheat, corn, etc.), the dynamic harvest index refers to the dynamic change process of gradually increasing the yield of crop seeds in percentage of the overground dry matter of the crops from the formation of the crop seeds to the gradual filling to the maturation process, and the index is increased along with the gradual increase of the growth and development time of the crops after the seeds are formed until the maximum harvest index is reached.

(2) Dynamic f_GParameter(s)

The ratio parameter f between the biomass accumulated on the ground in the common crop flowering period-mature period and the biomass on the ground in the mature period_GIs a Static parameter (Static f)_G，S-f_G) And only applied during the maturation period f_GCalculated, but lack of between flowering-maturity period f_GDynamic course study of parameters, which resulted in f obtained with a shorter year trial_GThe phenomenon of unstable relation possibly exists in the correlation between the static parameters and the crop harvest indexes, so that the estimation precision of the crop harvest indexes is reduced to a certain degree. In order to improve the stability and the estimation precision of the crop harvest index estimation model, the method is in the original static state f_GOn the basis of the parameters, a dynamic f is provided_GIndex, namely dynamic parameter D-f considering ratio between aboveground biomass accumulated in different periods during flowering period-mature period and aboveground biomass in corresponding period_G(Dynamic f_G). The index D-f_GThe calculation method is as follows:

in the formula, sigma W_postIs the overground biomass (kg/hm) accumulated at different periods during the flowering period and the mature period of the winter wheat²)；∑W_wholeIs total aboveground biomass (kg/hm) corresponding to the sampling period²) (ii) a t is the time of the sampling,W_tweight of dry matter (kg/hm) at time t sample²)，W_aWeight in terms of dry matter (kg/hm) at flowering stage²)，D-f_G,tA ratio parameter representing the t sample time.

(3) Canopy high spectral narrow band spectral index NDSI

A large number of remote sensing monitoring researches of crops show that strong correlation (strong construction and the like, 2015; von Meichen and the like, 2010) exists between the normalized vegetation index (NDVI) and the aboveground biomass of the crops and the yield of crop grains, and meanwhile, the index is also applied to certain extent in remote sensing estimation of the crop harvest index in recent years and obtains a better research result, so that the most commonly used normalized vegetation index is also used for carrying out remote sensing estimation research on the dynamic harvest index. The formula for calculating the NDVI is as follows:

NDVI＝(ρ_nir-ρ_red)/(ρ_nir+ρ_red) (3)

where nir and red represent the near infrared and red bands, respectively, ρ_nirAnd ρ_redRespectively representing the spectral reflectivity of a near infrared band and the spectral reflectivity of a red light band. When the near-infrared band reflectivity and the red-light band reflectivity are not limited to the near-infrared region and the red-light region of the electromagnetic spectrum, but are combined for any two bands of the hyperspectral, the near-infrared band reflectivity and the red-light band reflectivity can be expressed by a hyperspectral narrow-band spectral index (NDSI), which is specifically as follows:

NDSI＝(ρ_i-ρ_j)/(ρ_i+ρ_j) (4)

wherein i and j are respectively the wavelength and rho corresponding to the hyperspectral band_iAnd ρ_jThe spectral reflectivities corresponding to the i and j wavelengths, respectively. Wherein the range of NDSI value is [ -1,1 [)]. In order to facilitate the research on the hyperspectral narrow-band spectral index and f of the crop canopy_GConsidering that the spectrum of the crop canopy is greatly influenced by the atmosphere and the water vapor in 1350-1415 nm and 1800-1950 nm, the invention mainly aims at the range of visible light and near infrared wave band, therefore, the invention enters the range of 350-1000nm wave band (containing 650 wave bands)Line D-f_GAnd (4) screening the estimated remote sensing sensitive wave band and estimating the dynamic crop harvest index by remote sensing.

1.3.2 general technical route

Firstly, according to the actually measured biomass data on the ground, the ratio parameter of the biomass accumulated on the ground in the flowering period-mature period of the crops to the biomass on the ground in the mature period is constructed as a static parameter (S-f)_G) And dynamic parameters (D-f) of the ratio of aboveground biomass accumulated during different periods during the flowering-maturation periods to aboveground biomass during the corresponding periods_G). Then, constructing hyperspectral narrow-band spectral indexes (NDSI) of any two canopy layers based on hyperspectral data of the canopy layers of the ground crop, and establishing the NDSI and the D-f of the winter wheat_GA linear model in between; then drawing and analyzing the NDSI and the winter wheat D-f_GFitting accuracy R between²A two-dimensional map; on the basis of the above, by determining R²Maximum area and center of gravity of maximum area, thereby obtaining winter wheat f_GA sensitive band center; finally, a dynamic harvest index D-HI and a parameter D-f are constructed_GThe best model of (1). Secondly, constructing a structure based on S-f according to the actually measured biomass data on the ground_GThe harvest index estimation model of (1). And finally, verifying by using a reserved harvest index verification data set, and carrying out precision comparison analysis on the crop harvest index estimation models constructed by the two methods. The specific technical route is shown in figure 3.

1.3.3 high spectral sensitivity based waveband D-f_GParameter-derived dynamic crop harvest index estimation

1.3.3.1 dynamic crop harvest index estimation model construction

The invention provides a method based on D-f_GA dynamic harvest index D-HI remote sensing estimation method of remote sensing information is provided. Dynamic parameters D-f_GAs intermediate variables, first NDSI and D-f are determined_GModel of winter wheat f is determined according to the model_GA sensitive band center; then determining D-f_GAnd a model of statistical relationship between dynamic harvest indexes D-HI; finally, according to the screened sensitive wave band center, the corresponding D-f can be determined_GAnd further estimating the crop harvest index and carrying out precision verification. Methods for calculating dynamic harvest indices of crops such asThe following:

D-f_G,t＝m×NDSI_i，j，t+n (5)

D-HI_t＝HI₀+s×D-f_G,t (6)

equation (5) is used mainly for NDSI and D-f_GThe model of (2) is constructed. Wherein i and j are hyperspectral wave bands of 350-1000nm respectively, t is different sampling time, and m and n are fitting parameters in a linear equation obtained after fitting. According to the formula, calculating and solving the ratio parameter D-f between the accumulated aboveground biomass of the crops from the flowering period to the t period and the aboveground biomass of the t growth period_G,t。

Equation (6) applies primarily to D-f_GAnd constructing a model between the D-HI and the D-HI. Wherein, HI₀Is the intercept, i.e. the value of the dynamic harvest index without a change in biomass after the flowering phase of the crop, i.e. when D-f_G,tA value for D-HI harvest index at 0; s is D-HI or D-f_GSlope constant in linear relationship. According to the formula, the dynamic harvest index at the t growth stage can be calculated.

1.3.3.2D-f_GCanopy hyperspectral sensitive band center determination by remote sensing estimation

The invention utilizes remote sensing technology to obtain the dynamic parameter D-f of the ratio between the overground biomass accumulated in different periods during the flowering period-mature period of crops and the overground biomass in corresponding periods_GInformation to thereby utilize the measured D-f_GThe correlation between the index and the crop dynamic harvest index is realized based on D-f_GAnd accurately estimating the crop harvest index of the remote sensing information. In the research, the hyperspectrum of the canopy of the winter wheat is obtained and the D-f of the winter wheat is actually measured_GOn the basis of parameters, developing the winter wheat D-f based on the high-spectrum remote sensing narrow-band spectral index NDSI_GAnd (5) remote sensing estimation research. Due to numerous wave bands of the hyperspectral data and high correlation among the wave bands, the redundancy of spectral information is increased, and the D-f is improved_GThe accuracy of the parameter remote sensing estimation model needs to utilize the narrow-band spectral index NDSI to D-f_GAnd (4) screening the center of the wave band sensitive to the parameters and the spectral wave band.

Due to NDSI and D-f_GR of (5)²In the two-dimensional graph, R²The areas of maxima are not uniformly distributed, R²Maximum point and R²The centers of gravity of the areas of maxima do not necessarily coincide completely, resulting in R²The corresponding wave band of the maximum point does not necessarily coincide with the center of the optimal wave band. Thus, to ensure that a NDSI constructed using the center of a selected band can achieve high precision D-f_GThe estimation result further ensures that the estimation result of the crop harvest index is more stable, and the method is based on the R²Obtaining the center of the sensitive wave band by the gravity center method of the maximum area so as to determine D-f_GEstimated sensitive bands, i.e. narrow band spectral indices NDSI and D-f corresponding to each band of hyperspectral in the canopy_GOn the basis of inter-correlation calculation, a maximum area is determined according to a threshold value of a correlation coefficient meeting the statistical significance requirement, and on the basis, the gravity center of the maximum area of the correlation coefficient is calculated, so that NDSI and D-f are obtained_GThe spectral band center and band combination with large parameter correlation. The specific process is as follows:

firstly, drawing NDSI and winter wheat D-f_GFitting between R²Determining NDSI and winter wheat D-f on the basis of two-dimensional map_GA band region in which the inter-correlation is high; secondly, find R in this region²Maximum value point, traversing all points meeting significance condition in the neighborhood of the point 8, and marking the set of the points as R²A maximum region Ω; finally, calculate R²The center of gravity of the local maximum value region is defined as each R²The sensitive band center of the maxima region. Winter wheat D-f_GThe sensitive band center determination diagram is shown in fig. 4, and the calculation formula (7) of the center of gravity is as follows:

wherein f (u, v) is R with band coordinates (u, v)²The value, omega, is the area of maximum,

center coordinates of the sensitive band.

1.3.4 based on measured S-f_GIs estimated by the maturity crop harvest index (G-HI)

In order to compare with the estimation accuracy of crop harvest index before improvement, the invention utilizes the proportionality coefficient (S-f) of the accumulated dry matter amount of the flowering period to the mature period of the crop to the total accumulated amount of the whole growth period, which is proposed by Kemanian et al (2007)_G) Based on measured S-f with maturity harvest index (G-HI) linear relation model_GThe maturity crop harvest index G-HI estimate of (a) is in the form of a linear model as follows:

G-HI＝HI₀+k×S-f_G (8)

wherein G-HI is the crop harvest index at harvest, HI₀Is the intercept in a linear relationship; k is the slope in the linear relationship; s-f_GThe ratio of the accumulation amount of the dry matters in the flowering-maturation period of the crops to the total accumulation amount in the whole growth period.

1.3.5 evaluation of precision of model for estimating crop harvest index

To evaluate the crop harvest index estimation_G(including S-f)_GAnd D-f_G) Estimation accuracy and harvest index (including G-HI and D-HI) estimation accuracy, the present invention selects a coefficient of determination (R)²) The precision of the harvest index estimation model is checked for Root Mean Square Error (RMSE), Normalized Root Mean Square Error (NRMSE) and Mean Relative Error (MRE). Wherein R is²The fitting degree of the analog value and the measured value is represented, and the value range is 0-1. R²The higher and closer to 1, the better the fitting effect of the model, and conversely, R²The lower and closer to 0, the worse the fitting effect of the model. The RMSE represents the deviation degree of the analog value and the measured value, the larger the root mean square error value is, the larger the deviation of the analog value and the measured value is, namely the worse the analog effect is; conversely, the smaller the root mean square error value, the better the simulation. NRMSE refers to the ratio of the root mean square error to the average of the measured values. MRE is an average value of the sum of absolute values of the respective relative errors, and indicates an average degree of deviation between the analog value and the actually measured value. When NRMSE and MRE<When the simulation result is 10%, the precision of the simulation result is judged to be excellent; when the content is 10 percent<NRMSE and MRE<When the simulation result is 20%, the precision of the simulation result is judged to be good; when the content is 20 percent<NRMSE and MRE<When the simulation result is 30%, judging that the precision of the simulation result is normal; when NRMSE and MRE>When the simulation result is 30%, judging that the precision of the simulation result is poor, and giving priority to the size of the NRMSE value according to the judgment standard, wherein the specific formula is as follows:

in the formula, x_iIs f_G(e.g., S-f)_GAnd D-f_G) Or measured value of HI (e.g., G-HI and D-HI); y is_iIs winter wheat f_G(e.g., S-f)_GAnd D-f_G) Or an estimated value of HI (e.g., G-HI and D-HI);

are respectively x_i，y_iN is the number of samples.

2 results and analysis

2.1 based on measured S-f_GIs estimated by the maturity crop harvest index (G-HI)

In the research, the average actually measured harvest index of winter wheat in the maturation stage of each cell is calculated and obtained according to the overground biomass data of the maturation stage of winter wheat in 19 days in 6 months and the unit area yield of wheat grains; then, the overground biomass of the winter wheat at the flowering period collected in 5, 10 and 2020 months is used as a reference, and the flowering-maturation period of the winter wheat is calculated and obtainedThe ratio S-f of the cumulative amount of dry matter to the total cumulative amount over the entire growth period_GAnd (4) parameters. On the basis, the ratio S-f of the accumulation amount of the dry matters in the flowering-maturation period of the winter wheat to the total accumulation amount in the whole growth period is constructed according to the formula (8)_GAnd constructing a model with the harvest index of the mature period. In the research, according to the principle that the ratio of a modeling data set to a verification data set is 2:1, the actually measured S-f of one winter wheat field control experiment repeated treatment group is randomly determined_GAnd G-HI as validation data set (16 data samples total), S-f measured in the remaining 2 experimental replicate treatment groups_GAnd G-HI as a modeling dataset (32 data samples total). Wherein, as can be seen from FIG. 5, S-f_GAnd harvest index HI the linear model constructed with coefficients of determination of 0.5058. Verification with the reserved verification dataset resulted in an RMSE of 0.0603, an NRMSE and an MRE of 11.78%, 11.31%, respectively, as shown in fig. 6.

2.2 high spectral sensitivity based band D-f_GParameter derived remote sensing estimation of crop harvest index

2.2.1 Hyperspectral NDSI calculation results of winter wheat canopy treated by different control experiments

On the basis of crop canopy hyperspectral data preprocessing collected by different experimental processing cells of each ground observation, calculating and drawing a narrow-band spectral index (NDSI) of any two-band combination of each experimental processing cell of each ground observation according to a formula (4). Wherein, the total number of the combination and related NDSI values between any two bands in a hyperspectral range of 350-1000nm is 650 multiplied by 650. The invention only shows the canopy hyperspectral NDSI result obtained by calculating 5 times of ground observation data of normal horizontal fertilization and irrigation treatment (N2W2) in a winter wheat control experiment. The 5 NDSI distribution diagrams shown in FIG. 7 are the NDSI calculation results of different growth periods, such as normal horizontal fertilization and irrigation treatment (N2W2) in the winter wheat control experiment, including the flowering period, the early stage of filling, the later stage of filling and the mature period of the winter wheat. Wherein, the abscissa (λ)₁) Ordinate (λ)₂) The high spectrum wavelength of the crop canopy is 350-1000nm, and the point corresponding to the two-dimensional space formed by the horizontal axis and the vertical axis is any two wave bands lambda₁、λ₂The corresponding calculated NDSI value for the reflectivity.

2.2.2 NDSI-based D-f_GEstimation of canopy hyperspectral sensitive band center determination

(1) Based on NDSI and D-f_GR of correlation of (1)²Two-dimensional distribution map

The invention firstly calculates and obtains NDSI and D-f of each cell by four times of ground observation of 5 months and 18 days, 5 months and 24 days, 6 months and 3 days and 6 months and 19 days_GThe data indexes are equal, wherein, D-f_GThe calculation of (A) takes 5 months and 10 days as the standard of overground biomass of winter wheat in the flowering period. On the basis, NDSI and D-f are constructed_GStatistical relationship model between them. Wherein, 48 cells carry out four times of ground observation, 192 cell ground observation data samples are obtained through accumulation, and the data indexes comprise NDSI and D-f_G. According to the principle that the ratio of the modeling data set to the verification data set is 2:1, one of the field control experiment repeated treatment groups of the winter wheat is randomly calculated to obtain NDSI and D-f_GAs a validation data set (64 cell ground observation data samples in total), NDSI and D-f were obtained by calculation of the remaining 2 experimental replicate-treatment groups_GAs a modeled data set (128 cell ground data samples total).

Finally, the Matlab software is used to obtain the NDSI and the D-f_GFitting accuracy R²Two-dimensional plot (as shown in FIG. 8) with abscissa (. lamda.)₁) Ordinate (λ)₂) The hyperspectral wavelength of the crop canopy is 1nm in the wavelength range of 350-1000nm, and the horizontal axis and the vertical axis form R²The total number of points corresponding to the two-dimensional space is 650 × 650, and each R²The two-dimensional space points correspond to two bands (lambda)₁，λ₂) Combining the constructed NDSI value with the measured D-f_GFitting accuracy R between². Wherein each R is²The two-dimensional space point is NDSI constructed by two wave bands with certain wavelength in 128 cells and corresponding cells D-f_GAnd constructing the decision coefficient of the linear model. As can be seen from FIG. 8, the fitting accuracy R²The connection lines between the two points (350 ) and (1000 ) are in axial symmetry distribution, so that the NDSI pair D-f can be obtained_GThe region with large correlation and the related band information. On the upper side of the symmetry axis between the two points (350 ) and (1000 ), within the dotted line box in FIG. 8The dark red part shows that NDSI and winter wheat D-f_GThe two areas mainly have high correlation, and the areas comprise two-dimensional areas with a transverse axis of 370-550 nm, a longitudinal axis of 480-720 nm, a transverse axis of 750-970 nm and a longitudinal axis of 770-1000 nm, wherein R is²Up to 0.75 or more, which is D-f_GAnd a foundation is laid for estimating the center of the hyperspectral sensitive band of the canopy.

(2)D-f_GEstimation of canopy hyperspectral sensitive band center determination

In the study, by looking up the correlation coefficient significance check table, when the number of samples n is 128, at the 0.05 significance level, when R is²>NDSI and D-f at 0.030-_GPresenting a significant correlation; at the 0.01 significance level, when R²>NDSI and D-f at 0.051_GA very significant correlation is present. In order to ensure the accuracy and reliability of the determination of the sensitive wave band center, the invention selects the coincidence R²>R of 0.051²Two-dimensional region study was conducted to find R in FIG. 8²>Maximum point of 0.051, traverse all R in 8 neighborhoods of this point²>0.051 points, marking the set of the points as a maximum area and using R²Gradient of 0.1 indicates R²For a more intuitive display of the distribution range of the sensitive band, here for R²Results of ≧ 0.6 are displayed as shown in FIG. 9.

To improve the center estimate D-f of the selected hyperspectral sensitive band_GTo improve the accuracy of estimating the harvest index, the invention selects R²>R of 0.8²And determining and researching the sensitive waveband center in the two-dimensional area. Omega_A～Ω_FTo satisfy R²R > 0.8²The maximum value area has the following specific results: omega_AThe range is 400-500 nm on the horizontal axis and 480-540 nm on the vertical axis; omega_BThe range is 370-550 nm on the horizontal axis and 550-720 nm on the vertical axis; omega_CThe range is 720-750 nm on the horizontal axis and 720-980 nm on the vertical axis; omega_DThe range is 770-810 nm on the horizontal axis and 790-820 nm on the vertical axis; omega_EThe range is 770-870 nm on the horizontal axis and 830-930 nm on the vertical axis; omega_FIn the range ofThe horizontal axis is 750-970 nm and the vertical axis is 950-1000 nm. In determining NDSI and D-f_GAfter the sensitive region(s), each R is calculated according to the formula (7)²Center of maximum area (lambda)₁，λ₂) Wherein λ is₁Is the central abscissa, λ, of the area of maximum value₂Maximum area center ordinate. Finally, omega is obtained_A～Ω_FThe maxima have centroids A (443nm, 506nm), B (442nm, 635nm), C (732nm, 834nm), D (787nm, 804nm), E (810nm, 877nm), and F (861nm, 985nm), respectively.

2.2.3 construction of D-f of NDSI based on canopy spectral sensitivity band center_GRemote sensing estimation

D-f selected according to 2.2.2_GAnd estimating the center result of the hyperspectral sensitive waveband of the canopy to respectively calculate the hyperspectral narrow waveband spectral index NDSI of the canopy. The center results of the 6 sensitive bands include lambda (443nm, 506nm), lambda (442nm, 635nm), lambda (732nm, 834nm), lambda (787nm, 804nm), lambda (810nm, 877nm) and lambda (861nm, 985 nm). Then, NDSI data of 128 cell ground observation data samples and winter wheat D-f are utilized according to formula (5)_GAnd constructing a linear model, and performing precision verification by using the rest 64 cell ground observation data samples as verification data sets. NDSI and D-f constructed based on sensitive band center_GInter-statistical relationship and D-f_GThe specific results of the estimation accuracy are shown in table 1, fig. 10, and fig. 11. As can be seen from Table 1, 6D-f were selected_GNDSI fitting D-f constructed by estimating center of hyperspectral sensitive band of canopy_GAt P<All reach a very significant level at the level of 0.01, and the model determines a coefficient R²Between 0.8138 and 0.8613. NDSI and D-f established by a reservation validation dataset pair_GThe 6 screened D-f are verified by a statistical model_GNDSI and D-f constructed by estimating center of hyperspectral sensitive waveband of canopy_GThe inter-statistical models have better D-f_GEstimation of Effect, estimation of D-f_GAll achieve high precision level. Wherein the RMSE is between 0.0267 and 0.0411, the NRMSE is between 9.27 and 14.27 percent, and the MRE is between 8.81 and 14.26 percent. Among them, NDSI estimation constructed by screened sensitive band center (732nm, 834nm)Winter wheat D-f_GWith the highest precision, determining the coefficient R²0.9522 is achieved, the sum of NRMSE and MRE is respectively 9.27 percent and 8.81 percent; secondly, the NDSI constructed by the screened sensitive band center (443nm, 506nm) estimates the D-f of the winter wheat_GThe precision reaches a higher level, and the coefficient R is determined²0.9347 for NRMSE and MRE and 10.94% and 9.86% respectively; NDSI estimation of winter wheat D-f constructed by screened sensitive band center (861nm, 985nm)_GRelatively low precision, and its determination coefficient R²0.8981 for NRMSE and MRE, 14.27% for NRMSE and 14.26% for MRE respectively.

TABLE 1 NDSI and D-f constructed based on the center of the sensitive band_GInter-statistical relationship and D-f_GEstimation accuracy

Note: x in the fitting equation is the wave band lambda₁，λ₂The constructed NDSI, y is fitted winter wheat D-f_G. And N is the number of samples.

Indicates a very significant correlation at p < 0.01 levels.

2.2.4 high spectral sensitivity based waveband D-f_GParameter-acquired D-HI remote sensing estimation model establishment and verification

Based on NDSI and D-f_GThe invention utilizes the high spectrum of the canopy to screen out D-f_GThe center of a sensitive wave band is estimated through parameters; then, the accurate D-f is realized by using the NDSI index constructed by the sensitive waveband center_GAnd (6) remote sensing estimation. Based on the measured D-f_GAnd D-HI using D-f_GAnd remote sensing parameter information realizes remote sensing estimation of the dynamic harvest index D-HI. And finally, verifying the D-HI remote sensing estimation model by using the reserved verification data.

2.2.4.1 based on D-f_GD-HI estimation model establishment of remote sensing parameters

Calculating the small winter wheat according to the dynamic aboveground biomass data of the winter wheat at different acquisition times during the flowering period and the mature period and the dynamic data of the grain yield in the grouting processD-f of 128 sample points_GAnd a dynamic harvest index D-HI, on the basis of which D-f is normalized by the formula (6)_GAnd fitting the correlation between the dynamic harvest index D-HI to obtain D-f_GAnd a dynamic harvest index D-HI inter-estimation model, which is specifically as follows:

D-HI_t＝0.1018+0.8093*D-f_G,t

research shows that the present invention has practical measured winter wheat D-f_GThe method has a remarkable linear relation with a crop dynamic harvest index, wherein the dynamic D-f_GAnd the coefficient of determination of the linear model constructed from the dynamic harvest index D-HI reached 0.9679 (FIG. 12), which was developed based on the dynamic D-f_GThe dynamic harvest index estimation lays a good foundation.

2.2.4.2 based on D-f_GD-HI estimation model verification of remote sensing parameters

Based on D-f_GOn the basis of establishing a D-HI estimation model of the remote sensing parameters, the corresponding NDSI of each remote sensing sensitive waveband center is calculated by utilizing the spectrum information in the reserved 64 groups of verification data. Wherein 64 NDSI substitution data are available for each band center. On the basis, the NDSI is substituted into the corresponding NDSI and D-f in the table 1_GIn the inter-statistical model, 64D-f corresponding to the center of each sensitive waveband are obtained_GAnd (6) remote sensing parameters. Then, the above D-f is mixed_GRemote sensing parameter substitution based on D-f in FIG. 12_GAnd (3) obtaining 64D-HI estimation results of the center of each sensitive wave band by using a D-HI remote sensing estimation model of the parameters, and carrying out D-HI remote sensing estimation result precision verification. And respectively carrying out precision evaluation on the remote sensing estimation results of the D-HI under different sensitive wave band conditions by using the reserved 64D-HI data.

(1) Winter wheat dynamic harvest index overall precision verification

The NDSI estimation D-f is constructed in the centers of 6 remote sensing sensitive wave bands of lambda (443nm, 506nm), lambda (442nm, 635nm), lambda (732nm, 834nm), lambda (787nm, 804nm), lambda (810nm, 877nm), lambda (861nm, 985nm) and the like_GUnder the condition, respectively carrying out D-f treatment on the center of the hyperspectral sensitive wave band of each canopy_GDynamic crop harvest index result carrying accuracy estimated by parameter remote sensingAnd (6) verifying. Finally, the results of the overall verification of the dynamic harvest index of the winter wheat at different sampling periods in the flowering-mature period are shown in fig. 13 and table 2. As can be seen from Table 2, NDSI estimation D-f constructed by 6 remote sensing sensitive waveband centers_GUnder the condition, the D-f obtained by hyperspectral utilization of the canopy_GThe parameter remote sensing information can realize accurate estimation of the dynamic crop harvest index. According to the result of the overall accuracy evaluation index estimated from the D-HI, under the condition that the centers of the hyperspectral sensitive wave bands of the 6 canopies are screened out, the hyperspectral sensitive wave band D-f is based on_GThe D-HI estimation results of the parameters are verified to reach a high-precision level, and the fitting precision R of the parameters is²Between 0.9169 and 0.9584, RMSE is between 0.0380 and 0.0507, NRMSE is between 10.83 and 14.45 percent, and MRE is between 9.62 and 13.99 percent. Wherein the D-f is estimated based on the center lambda (732nm, 834nm) of the hyperspectral sensitive band_GThe D-HI estimation result of the parameter has the highest precision, and the NRMSE and the MRE are respectively 10.83 percent and 9.62 percent; secondly, estimating D-f based on the center lambda (443nm, 506nm) of the hyperspectral sensitive waveband_GThe D-HI estimation result of the parameter has higher precision, and the NRMSE and the MRE are respectively 11.60 percent and 10.24 percent; estimation of D-f based on center lambda (861nm, 985nm) of hyperspectral sensitive band_GThe D-HI estimation result of the parameter has relatively low precision, and NRMSE and MRE are respectively 14.45% and 13.99%.

TABLE 2 base on D-f_GD-HI estimation model overall accuracy verification of remote sensing parameters

(2) D-HI estimation model precision verification of winter wheat in different growth periods from filling to maturity

In the invention, the total precision is estimated by remote sensing by analyzing the dynamic harvest index of winter wheat in different sampling periods of the flowering period and the mature period by using 64 groups of reserved verification data, and the precision evaluation is respectively carried out on the D-HI estimation model results of the early stage of grouting, the middle stage of grouting and the late stage of grouting corresponding to different sampling dates such as 18 days in 5 months, 24 days in 5 months, 3 days in 6 months, 19 days in 6 months and the like, and the specific results are shown in tables 3 to 6. Among the reserved 64 groups of data, 16 groups of verification data corresponding to each sampling date are provided.

A. Precision verification of D-HI estimation model of winter wheat in different grouting stages

In the early stage of winter wheat filling (Table 3), the fitting precision R between the harvest index and the actually measured crop harvest index is estimated by remote sensing²Between 0.3128 and 0.5819, RMSE is between 0.0273 and 0.0393, NRMSE is between 13.21 and 19.04 percent, and MRE is between 11.50 and 17.94 percent. Wherein the D-f is estimated based on the center lambda (732nm, 834nm) of the hyperspectral sensitive band_GThe D-HI estimation result of the parameter in the pre-grouting period has the highest accuracy, and the NRMSE and the MRE are respectively 13.21 percent and 11.50 percent; secondly, estimating D-f based on the center lambda (443nm, 506nm) of the hyperspectral sensitive waveband_GThe D-HI estimation result of the parameter has higher precision, and the NRMSE and the MRE are respectively 13.82 percent and 11.92 percent; estimation of D-f based on center lambda (861nm, 985nm) of hyperspectral sensitive band_GThe D-HI estimation result of the parameters has relatively low precision, and NRMSE and MRE are respectively 19.04% and 17.94%.

In the middle stage of winter wheat filling (Table 4), the fitting precision R between the harvest index and the actually measured crop harvest index is estimated by remote sensing²Between 0.3597 and 0.4391, the RMSE is between 0.0321 and 0.0437, the NRMSE is between 11.27 and 15.36 percent, and the MRE is between 9.05 and 13.18 percent. Wherein the D-f is estimated based on the center lambda (732nm, 834nm) of the hyperspectral sensitive band_GThe D-HI estimation result of the parameter in the early stage of grouting has the highest precision, and NRMSE and MRE are respectively 11.27% and 9.05%; secondly, estimating D-f based on the center lambda (443nm, 506nm) of the hyperspectral sensitive waveband_GThe D-HI estimation result of the parameter has higher precision, and NRMSE and MRE are respectively 11.69 percent and 9.17 percent; estimation of D-f based on center lambda (861nm, 985nm) of hyperspectral sensitive band_GThe D-HI estimation result of the parameters has relatively low precision, and NRMSE and MRE are respectively 15.36% and 13.18%.

In the late stage of winter wheat filling (Table 5), the fitting precision R between the harvest index and the actually measured crop harvest index is estimated by remote sensing²Between 0.4928 and 0.5964, the RMSE is between 0.0396 and 0.0565, the NRMSE is between 9.92 and 14.16 percent, and the MRE is between 8.66 and 13.40 percent. Wherein the D-f is estimated based on the center lambda (732nm, 834nm) of the hyperspectral sensitive band_GThe D-HI estimation result of the parameter in the pre-grouting period has the highest precision, and NRMSE and MRE are respectively 9.92% and 8.66%; secondly, estimating D-f based on the center lambda (443nm, 506nm) of the hyperspectral sensitive waveband_GThe D-HI estimation result of the parameter has higher precision, and NRMSE and MRE are respectively 10.92 percent and 9.95 percent; estimation of D-f based on center lambda (861nm, 985nm) of hyperspectral sensitive band_GThe D-HI estimation result of the parameters has relatively low precision, and NRMSE and MRE are respectively 14.16% and 13.40%.

B. Compared with the traditional method, the D-HI remote sensing estimation model precision verification of the winter wheat in the mature period

In the winter wheat mature period (Table 6 and FIG. 14), the fitting precision R between the harvest index and the actually measured crop harvest index is estimated by remote sensing²Between 0.2724 and 0.6762, the RMSE is between 0.0492 and 0.0601, the NRMSE is between 9.62 and 11.74 percent, and the MRE is between 9.27 and 11.43 percent. Wherein the D-f is estimated based on the center lambda (732nm, 834nm) of the hyperspectral sensitive band_GThe D-HI estimation result of the parameter in the maturation period has the highest precision, and NRMSE and MRE are respectively 9.62 percent and 9.27 percent; secondly, estimating D-f based on the center lambda (443nm, 506nm) of the hyperspectral sensitive waveband_GThe D-HI estimation result of the parameter has higher precision, and NRMSE and MRE are respectively 10.33 percent and 9.92 percent; estimation of D-f based on center lambda (861nm, 985nm) of hyperspectral sensitive band_GThe D-HI estimation of the parameters has relatively low accuracy, with NRMSE and MRE being 11.74% and 11.43%, respectively. By convention based on measured S-f as in section 2.1_GIn comparison of the estimation results of the maturity period G-HI, the method provided by the invention estimates the D-f based on the central lambda (732nm and 834nm) of the hyperspectral sensitive waveband_GThe precision of the harvest index of the parameter estimation mature period is more accurate than the traditional actual measurement S-f_GThe G-HI estimation accuracy RMSE is improved by 0.0111, the NRMSE is improved by 2.16 percent, and the MRE is improved by 2.04 percent, which shows that the invention utilizes the remote sensing technology to improve the traditional actually measured S-f_GThe effectiveness of the G-HI estimation method of (1).

In general, the hyperspectral sensitive waveband D-f based on the invention_GPrecision results with higher levels of different grouting stage (early grouting stage, middle grouting stage and late grouting stage) dynamic harvest indexes and maturity stage harvest indexes obtained by the parameter-obtained D-HI remote sensing estimation methodAnd the highest precision of the dynamic harvest indexes obtained in different growth periods is ranked as the early stage of grouting<Middle stage of grouting<Late stage of grouting<In the mature period, the accuracy evaluation result fully illustrates the feasibility of the D-HI remote sensing estimation method provided by the invention, and the method has important significance for accurately acquiring the dynamic harvest index information and the mature period harvest index by considering the crop dynamic growth information. In addition, compared with the common remote sensing estimation result of the harvest index of the crops in the mature period (Morinondo et al, 2007; strong construction, 2010;), the harvest index estimation result of the mature period obtained by the method provided by the invention has consistency with the result of the common remote sensing estimation method, such as D-f estimation by using the center of the hyperspectral sensitive band_GAs for the D-HI estimation result precision result of the parameters, the sensitive wave band center lambda (732nm and 834nm) with the highest D-HI estimation precision is mainly positioned in red light and near infrared wave bands, and the NDSI formed by the center is basically the same as the NDVI wave band combination adopted by the traditional estimation crop harvest index, which proves that the method is based on the hyperspectral sensitive wave band D-f_GThe D-HI remote sensing estimation method for parameter acquisition has certain consistency with the traditional method for estimating the crop harvest index by using NDVI.

TABLE 3 base on D-f_GPrecision verification of grouting early-stage harvest index of D-HI estimation model of remote sensing parameters (5 month and 18 days)

Note: n represents the number of samples of D-HI actually measured in the verification data set during the pre-grouting period. Indicates a very significant correlation at p < 0.01 levels and indicates a significant correlation at p < 0.05 levels.

TABLE 4 base on D-f_GD-HI estimation model of remote sensing parameters grouting medium harvest index precision verification (24 days in 5 months)

Note: n represents the number of samples of D-HI actually measured in the middle stage of grouting in the verification data set. Indicates a very significant correlation at p < 0.01 levels and indicates a significant correlation at p < 0.05 levels.

TABLE 5 bases on D-f_GD-HI estimation model of remote sensing parameters grouting later harvest index precision verification (6 month and 3 days)

Note: and N represents the number of samples of D-HI actually measured in the later stage of grouting in the verification data set. Indicates a very significant correlation at p < 0.01 levels and indicates a significant correlation at p < 0.05 levels.

TABLE 6 bases on D-f_GD-HI estimation model maturation period harvest index precision verification of remote sensing parameters (6 month and 19 days)

Note: n represents the number of samples of observed D-HI at maturity in the validation dataset. Indicates a very significant correlation at p < 0.01 levels and indicates a significant correlation at p < 0.05 levels.

It will be understood that modifications and variations can be made by persons skilled in the art in light of the above teachings and all such modifications and variations are intended to be included within the scope of the invention as defined in the appended claims.

Claims (7)

1. Based on D-f_GThe remote sensing estimation method for the dynamic harvest index of winter wheat obtained by parameter remote sensing is characterized by comprising the following steps:

a1, according to the ground actual measurement dynamic biomass data, constructing a ratio dynamic parameter D-f between the overground biomass accumulated at different periods during the flowering period-mature period of the crops and the overground biomass at the corresponding period_G；

D-f_GCalculation methodThe method comprises the following steps:

in the formula, sigma W_postIs the overground biomass (kg/hm) accumulated at different periods during the flowering period and the mature period of the winter wheat²)；∑W_wholeIs total aboveground biomass (kg/hm) corresponding to the sampling period²) (ii) a t is the sampling time, W_tWeight of dry matter (kg/hm) at time t sample²)，W_aWeight in terms of dry matter (kg/hm) at flowering stage²)，D-f_G，tA ratio parameter representing the t sample time;

a2, constructing any two canopy high-spectrum narrow-band spectral indexes NDSI based on ground crop canopy high-spectrum data, and establishing NDSI and winter wheat D-f_GA linear model in between;

a3, drawing and analyzing NDSI and winter wheat D-f_GFitting accuracy R between²A two-dimensional map;

a4, by determining R²Maximum value area and center of gravity of the maximum value area, thereby obtaining winter wheat D-f_GA sensitive band center;

a5, determination of D-f_GEstimating an optimal band combination;

a6 based on NDSI and D-f_GD-f of a relation_GA remote sensing estimation model;

A7、D-f_Gremote sensing estimation of (2);

a8, obtaining based on D-f_GAnd a dynamic harvest index estimation model of the D-HI relationship;

remote sensing estimation of A9, D-HI.

2. The method according to claim 1, wherein said step A4 is based on R²Obtaining the center of the sensitive wave band by the gravity center method of the maximum area so as to determine D-f_GEstimated sensitive bands, i.e. narrow band spectral indices NDSI and D-f corresponding to each band of hyperspectral in the canopy_GBased on the calculation of the inter-correlation, the correlation coefficient satisfiesDetermining a maximum region by a threshold value of statistical significance requirement, and calculating the gravity center of the maximum region of the correlation coefficient on the basis of the maximum region, thereby obtaining NDSI and D-f_GThe combination of the center and the band of the spectral band with larger parameter correlation; the specific process is as follows:

firstly, drawing NDSI and winter wheat D-f_GFitting between R²Determining NDSI and winter wheat D-f on the basis of two-dimensional map_GA band region in which the inter-correlation is high; secondly, find R in this region₂Maximum value point, traversing all points meeting significance condition in the neighborhood of the point 8, and marking the set of the points as R²A maximum region Ω; finally, calculate R²The center of gravity of the local maximum value region is defined as each R²A sensitivity band center of a maximum region; the calculation formula (7) of the center of gravity is as follows:

wherein f (u, v) is R with band coordinates (u, v)²The value, omega, is the area of maximum,

center coordinates of the sensitive band.

3. The method as claimed in claim 1, wherein the 6 sensitive band centers sensitive to winter wheat D-fG of step a4 include λ (443nm, 506nm), λ (442nm, 635nm), λ (732nm, 834nm), λ (787nm, 804nm), λ (810nm, 877nm) and λ (861nm, 985 nm).

4. The method of claim 1, wherein step A6 is based on NDSI and D-f_GD-f of a relation_GThe remote sensing estimation model is as follows:

D-f_G，t＝m×NDSI_i，j，t+n (5)

wherein i and j are respectively high spectral bands between 350nm and 1000nm, and t is different sampling timeM and n are fitting parameters in a linear equation after fitting; according to the formula, calculating and solving the ratio parameter D-f between the accumulated aboveground biomass of the crops from the flowering period to the t period and the aboveground biomass of the t growth period_G，t。

5. The method according to claim 1, wherein the step A8, obtaining is based on D-f_GThe dynamic harvest index estimation model for the D-HI relationship is:

D-HI_t＝HI₀+s×D-f_G，t (6)

wherein, HI₀Is the intercept, i.e. the value of the dynamic harvest index without a change in biomass after the flowering phase of the crop, i.e. when D-f_G，tA value for D-HI harvest index at 0; s is D-HI or D-f_GSlope constant in linear relationship.

6. The method as claimed in claim 5, wherein the step A8 is used for calculating D-f of 128 sample points in the winter wheat cell according to the dynamic aboveground winter wheat biomass data at different collection times during the flowering stage-mature stage and the dynamic grain yield data during grain filling process_GAnd a dynamic harvest index D-HI, on the basis of which D-f is normalized by the formula (6)_GAnd fitting the correlation between the dynamic harvest index D-HI to obtain D-f_GAnd a dynamic harvest index D-HI inter-estimation model, which is specifically as follows:

D-HI_t＝0.1018+0.8093*D-f_G，t。

7. a method for estimating the yield of winter wheat based on the dynamic harvest index of winter wheat, wherein the dynamic harvest index of winter wheat is obtained by the method according to any one of claims 1 to 6.

Images