Crop agronomic parameter and yield estimation method and system coupling crop growth model and remote sensing data

Abstract

The invention discloses a crop agronomic parameter and yield estimation method and system coupling a crop growth model and remote sensing data, and relates to the technical field of agronomic parameter and yield estimation.The method comprises the steps that crop growth data are collected, a CERES-Rice model is constructed according to the crop growth data, and rice agronomic parameters and yield data are output; a PROSAIL canopy reflection model is constructed; hyperspectral data are output through a PROSAIL canopy reflection model, and parameter inversion optimization is carried out in combination with remote sensing data; and outputting final rice agronomic parameters and yield data according to the model after inversion optimization. The remote sensing radiation transmission model and the crop growth model are adopted to monitor agricultural parameters and yield of crops, the adjustment of parameters directly influencing crop states shared between ecological physiology and the radiation transmission model is emphasized, the availability of a hyperspectral instrument is increased, narrow-band spectral reflectivity data is collected, and the measurement accuracy is improved. And the accuracy of agricultural parameter and yield monitoring is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of agronomic parameter and yield estimation, and more particularly to a crop agronomic parameter and yield estimation method and system by coupling a crop growth model and remote sensing data. Background Art

[0002] Currently, remote sensing sensors are commonly mounted on tractors, drones, or satellite platforms to monitor agricultural production dynamics, such as crop leaf area, biomass, nitrogen content, nitrogen nutrition indicators, and yield. The measurement principle of active remote sensing optical sensors is primarily based on detecting the amount of light reflected from the crop scene after the incident radiation interacts with the crop canopy and underlying soil background, i.e., canopy reflectance. How to effectively utilize the spectral information of crop canopies to characterize their phenotypic characteristics or physiological parameters, and apply these characteristics and parameters to agricultural production monitoring, precise nutrient management, rational water management, and seasonal yield prediction, is a major challenge currently facing scientists in the field of agricultural remote sensing. A common approach is to use remote sensing to observe crop canopy spectral reflectance to infer crop physiological and biochemical data or to drive eco-physiological crop growth models.

[0003] Mechanistic models, whether crop growth models or remote sensing radiative transfer models, typically convert system attribute data into observable variables. For example, the PROSAIL radiative transfer model uses plant canopy properties and solar geometry to simulate canopy spectra and bidirectional reflectance at a given time. Model inversion is a common method for inversely acquiring data using optimization algorithms, using sensor observations to infer system attribute data. Many researchers have used observed canopy spectral reflectance to invert the PROSAIL model and estimate crop biophysical properties such as LAI, leaf chlorophyll content, and leaf water content. Model inversion provides researchers with a reasonable method for estimating system attributes from remote sensing observations. This method is simple, practical, and its reliability has been proven by many researchers. However, due to the limited information ultimately obtained from remote sensing sensor spectral data, incomplete understanding of system processes, simplifications in model design, and incorrect input parameters, different model configurations may produce equally plausible results, ultimately leading to erroneous decision-making. This problem is common in the inversion of many mechanistic models, including PROSAIL. Therefore, it is necessary to set default values ​​for its basic parameters based on prior knowledge before simulating the model to facilitate subsequent inversion work.

[0004] The potential for coupling remote sensing models or data with crop growth models was discovered as early as the late 1970s. With the rapid development of computers, this research area has achieved significant breakthroughs in recent years and has become a hot topic. However, remote sensing and ecophysiological modeling techniques have been developed largely independently. For example, the Cropping System Model (CSM) provided in the Decision Support System for Agricultural Technology Transfer (DSSAT) is an ecophysiological model that simulates crop growth and development processes and the impact of soil moisture and nutrient status on crop yield. However, this model does not simulate any processes related to radiation interactions within the crop canopy. Similarly, while PROSAIL can simulate crop canopy spectra and bidirectional reflectance, it does not model any processes related to crop, water, or nutrient dynamics. Recently, radiation transfer models have been explicitly linked to ecophysiological models through common state variables, particularly LAI. This approach has resulted in more comprehensive models capable of simulating the temporal canopy spectral reflectance response and the underlying crop, water, and nutrient processes of the cropping system.

[0005] Establishing a link between radiative transfer and ecophysiological models offers several advantages for model inversion applications involving remote sensing observations. Radiative transfer models facilitate ecophysiological model parameterization and allow model inversion by providing a direct link to readily observable reflectance characteristics of the crop canopy. Simulation results from ecophysiological models can then be used to constrain radiative transfer model input parameters and allow estimation of crop yield and other biophysical variables that cannot be estimated from radiative transfer model inversions alone. Ecophysiological models also allow model inversions based on time series of remote sensing observations, rather than static point-in-time measurements.

[0006] When designing model inversion strategies for linked radiative transfer and ecophysiological models, a number of procedural issues arise, including which model parameters to adjust, which parameters to constrain, which spectral wavelengths to use, and which time-series remote sensing observations to incorporate. First, ecophysiological models are often complex, with many input parameters that can be adjusted. However, it is unclear whether the additional spectral information provides advantages for the model inversion problem compared to common broadband vegetation indices such as NDVI. Finally, the timing and availability of remote sensing observations will affect the performance of the model inversion procedure and the accuracy of estimated crop biophysical properties. Few studies have investigated these aspects of model inversion based on canopy spectral reflectance data.

[0007] Therefore, how to propose a crop agronomic parameter and yield estimation method and system that couples crop growth models and remote sensing data, use remote sensing radiation transfer models and crop growth models to carry out crop agronomic parameter and yield monitoring, and adjust the parameters that directly affect the crop status shared between the eco-physiology and radiation transfer models to improve the accuracy of agronomic parameter and yield monitoring is an urgent problem that technicians in this field need to solve. Summary of the Invention

[0008] In view of this, the present invention provides a method and system for estimating crop agronomic parameters and yield by coupling crop growth models and remote sensing data. This method uses a remote sensing radiation transfer model and a crop growth model to monitor crop agronomic parameters and yield, and adjusts parameters that directly affect the crop state shared between the ecophysiology and radiation transfer models to improve the accuracy of agronomic parameter and yield monitoring. To achieve the above objectives, the present invention adopts the following technical solutions:

[0009] A method for estimating crop agronomic parameters and yield by coupling crop growth models and remote sensing data, comprising:

[0010] Collect crop growth data, build the CERES-Rice model based on the crop growth data, and output rice agronomic parameters and yield data;

[0011] Preprocessing of rice agronomic parameters and yield data;

[0012] The PROSAIL canopy reflectance model was constructed based on the preprocessed rice agronomic parameters, yield data, and crop growth data;

[0013] Hyperspectral data is output through the PROSAIL canopy reflectance model and combined with multi-source remote sensing data to perform parameter inversion and optimization of the CERES-Rice model;

[0014] A bivariate collaborative assimilation framework of leaf area index and chlorophyll content was constructed, and the CERES-Rice model after parameter inversion and optimization was deeply integrated through multi-source remote sensing data.

[0015] The final rice agronomic parameters and yield data are output based on the deeply integrated model.

[0016] Optionally, the crop growth data includes: soil data, variety data, meteorological data, management data, leaf pigment content, leaf water content, canopy structure, soil background reflectivity, hotspot size, solar diffusivity and solar geometry.

[0017] Optionally, the leaf pigment content includes: chlorophyll a and b content, carotenoid content and brown pigment content; leaf water content is defined as equivalent water thickness; canopy structure includes leaf dry matter content, leaf structure coefficient, leaf area index and average leaf inclination; the characteristics of solar geometry include: solar zenith, observer zenith and solar azimuth.

[0018] Optionally, the preprocessing of the rice agronomic parameters and yield data includes: dividing the leaf dry matter mass by the leaf area and converting the units to obtain the leaf dry matter content, which is used as an input variable of the PROSAIL canopy reflectance model.

[0019] Optionally, the PROSAIL canopy reflectance model is used to output canopy-scale spectral reflectance at wavelengths of 400-2400 nm.

[0020] Optionally, the multi-source remote sensing data includes: the drone remote sensing camera used is a four-band camera, the spectral value of the camera is directly taken as the value of the central band, and the hyperspectral data output by the PROSAIL canopy reflectance model is converted into four-band reflectance data.

[0021] Optionally, the parameter inverse optimization employs the SCE-UA algorithm for minimum error parameter optimization. This algorithm performs a global search within the feasible parameter space. The SCE-UA algorithm performs best in high-dimensional nonlinear systems and exhibits excellent stability. The SCE-UA algorithm combines determinism and probabilism, and the concepts of complex point system evolution and competitive evolution. It has been widely used in parameter optimization problems in hydrological and crop models.

[0022] Optionally, the cost function for performing parameter inversion optimization includes:

[0023]

[0024] Among them, RES is the final value of the cost function, i is the current band, starting from 1, n is the total number of bands, K is the dimensionally normalized weight vector of each band, R obs is the reflectivity obtained by UAV remote sensing observation, R sim Refers to the reflectance of the corresponding band output by the PROSAIL canopy reflectance model.

[0025] Optionally, the method further includes: converting the nitrogen absorption amount of the leaves into chlorophyll content of the leaves.

[0026] Optionally, a crop agronomic parameter and yield estimation system that couples crop growth models and remote sensing data, including:

[0027] Collection module: used to collect crop growth data;

[0028] CERES-Rice model construction module: used to build the CERES-Rice model based on crop growth data and output rice agronomic parameters and yield data;

[0029] Preprocessing module: used to preprocess rice agronomic parameters and yield data;

[0030] PROSAIL canopy reflectance model construction module: used to construct the PROSAIL canopy reflectance model based on pre-processed rice agronomic parameters and yield data and crop growth data;

[0031] Parameter inversion and optimization module: used to output hyperspectral data through the PROSAIL canopy reflectance model and combine it with multi-source remote sensing data to perform CERES-Rice model parameter inversion and optimization;

[0032] Remote sensing data and crop model data assimilation module: This module is used to construct a two-variable collaborative assimilation framework for leaf area index and chlorophyll content, and deeply integrates the CERES-Rice model after parameter inversion and optimization using multi-source remote sensing data. By constructing a two-variable collaborative assimilation framework for leaf area index (LAI) and chlorophyll content (Cab), it achieves a deep integration of the CERES-Rice crop growth model and multi-source remote sensing data, resulting in more accurate crop simulation yield estimates.

[0033] Output module: used to output the final rice agronomic parameters and yield data based on the deeply integrated model.

[0034] As can be seen from the above technical solutions, compared with the prior art, the present invention discloses a method and system for estimating crop agronomic parameters and yield by coupling crop growth models and remote sensing data, which has the following beneficial effects:

[0035] This invention proposes a method for estimating crop agronomic parameters and yields by coupling crop growth models and remote sensing data. The method comprises the following steps: collecting crop growth data, constructing a CERES-Rice model based on the crop growth data, and outputting rice agronomic parameters and yield data; preprocessing the rice agronomic parameter and yield data; constructing a PROSAIL canopy reflectance model based on the preprocessed rice agronomic parameter and yield data and the crop growth data; outputting hyperspectral data from the PROSAIL canopy reflectance model, and inverting and optimizing CERES-Rice model parameters in combination with multi-source remote sensing data; constructing a bivariate collaborative assimilation framework for leaf area index and chlorophyll content, and deeply fusing the inverted and optimized CERES-Rice model using multi-source remote sensing data; and outputting final rice agronomic parameter and yield data based on the deeply fused model. The invention uses a model inversion program to estimate crop biophysical characteristics using linked ecophysiological and radiation transfer models. Focusing on estimating LAI, biomass, plant nitrogen uptake, and yield of cold-region rice crops in Jiansanjiang, the team investigated the availability of remote sensing data by implementing a genetic algorithm to identify a set of remote sensing observations that optimally estimated LAI using model inversion. Agronomic parameter and yield monitoring of crops was conducted using remote sensing radiation transfer models and crop growth models, with a focus on adjusting parameters that directly affect crop status shared between ecophysiological and radiation transfer models. Furthermore, the increasing availability of hyperspectral instrumentation allowed for the collection of narrowband spectral reflectance data, improving the accuracy of agronomic parameter and yield monitoring. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are merely embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on the provided drawings without paying any creative work.

[0037] Figure 1 This is a flow chart of a method for estimating crop agronomic parameters and yield by coupling a crop growth model and remote sensing data, provided by the present invention.

[0038] Figure 2 The present invention provides a linear regression relationship diagram between nitrogen uptake and chlorophyll content in cold-region rice leaves.

[0039] FIG3( a ) is a schematic diagram showing the simulation of the leaf area index simulated by PROSAIL+CERES-Rice coupling and the measured leaf area index at the jointing stage (SE) provided by the present invention.

[0040] FIG3( b ) is a schematic diagram showing the simulation of the leaf area index simulated by PROSAIL+CERES-Rice coupling and the measured leaf area index at the heading stage (HD) provided by the present invention.

[0041] FIG4( a ) is a schematic diagram showing the simulation of the aboveground biomass at the jointing stage (SE) using PROSAIL+CERES-Rice coupling simulation and the measured aboveground biomass provided by the present invention.

[0042] FIG4( b ) is a schematic diagram of the simulation of the aboveground biomass at the heading stage (HD) using PROSAIL+CERES-Rice coupling simulation and the measured aboveground biomass provided by the present invention.

[0043] FIG5( a ) is a schematic diagram showing the simulation of the aboveground nitrogen uptake at the jointing stage (SE) using PROSAIL+CERES-Rice coupling simulation and the measured aboveground nitrogen uptake provided by the present invention.

[0044] FIG5( b ) is a schematic diagram of the simulation of the aboveground nitrogen uptake at the heading stage (HD) using PROSAIL+CERES-Rice coupling simulation and the measured aboveground nitrogen uptake provided by the present invention.

[0045] FIG6( a ) is a schematic diagram showing the simulation of yield simulated by PROSAIL+CERES-Rice coupling and the measured aboveground yield at the jointing stage (SE) provided by the present invention.

[0046] FIG6( b ) is a schematic diagram of the simulation of yield simulated by PROSAIL+CERES-Rice coupling and the measured aboveground yield at the heading stage (HD) provided by the present invention. DETAILED DESCRIPTION

[0047] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0048] The embodiment of the present invention discloses a method for estimating crop agronomic parameters and yield by coupling crop growth models and remote sensing data, such as Figure 1 Shown, including:

[0049] Collect crop growth data, build the CERES-Rice model based on the crop growth data, and output rice agronomic parameters and yield data;

[0050] Preprocessing of rice agronomic parameters and yield data;

[0051] The PROSAIL canopy reflectance model was constructed based on the preprocessed rice agronomic parameters, yield data, and crop growth data;

[0052] Hyperspectral data is output through the PROSAIL canopy reflectance model and combined with multi-source remote sensing data to perform parameter inversion and optimization of the CERES-Rice model;

[0053] A framework for the collaborative assimilation of leaf area index and chlorophyll content, using multi-source remote sensing data, was established. The CERES-Rice model, after parameter inversion and optimization, was deeply integrated with this model. The advantages of multi-source remote sensing data and multiple assimilation variables were utilized in crop yield estimation based on data assimilation. In terms of remote sensing data selection, integrating multi-source remote sensing observations helped overcome the inaccuracy or lack of single-source remote sensing observations. In terms of assimilation variables, chlorophyll, given its importance to crop growth, was used alongside leaf area index as an assimilation variable for yield estimation, improving the accuracy of yield estimates.

[0054] The final rice agronomic parameters and yield data are output based on the deeply integrated model.

[0055] Furthermore, the crop growth data includes: soil data, variety data, meteorological data, management data, leaf pigment content, leaf water content, canopy structure, soil background reflectivity, hotspot size, solar diffusivity and solar geometry.

[0056] Furthermore, the leaf pigment content includes: chlorophyll a and b content, carotenoid content and brown pigment content; leaf water content is defined as equivalent water thickness; canopy structure includes leaf dry matter content, leaf structure coefficient, leaf area index and average leaf inclination angle; the characteristics of solar geometry include: solar zenith, observer zenith and solar azimuth.

[0057] Furthermore, the wide dynamic range vegetation index WDRVI includes: WDRVI effectively expands the dynamic range of the vegetation index through parameter adjustment, and is suitable for various vegetation monitoring scenarios from sparse to dense, and is a practical tool for precision agriculture and ecological research. In practical applications, parameter calibration needs to be combined with field data to maximize its effectiveness. WDRVI remains sensitive in medium and high coverage vegetation areas (LAI>3), avoiding the saturation failure problem of NDVI, and is particularly suitable for monitoring densely planted crops such as rice and corn. Multi-phase WDRVI divides the planting area, and the random forest model is used to divide the planting area.

[0058] Furthermore, the preprocessing of the rice agronomic parameters and yield data includes: dividing the leaf dry matter mass by the leaf area and converting the units to obtain the leaf dry matter content, which is used as an input variable of the PROSAIL canopy reflectance model.

[0059] Furthermore, the PROSAIL canopy reflectance model is used to output canopy-scale spectral reflectance at wavelengths of 400-2400 nm.

[0060] Furthermore, the multi-source remote sensing data includes: the drone remote sensing camera used is a four-band camera, the spectral value of the camera is directly taken as the value of the central band, and the hyperspectral data output by the PROSAIL canopy reflectance model is converted into four-band reflectance data.

[0061] Furthermore, the parameter inversion optimization adopts PSO to optimize the parameters with the minimum error.

[0062] Furthermore, the cost function for performing parameter inversion optimization includes:

[0063]

[0064] Among them, RES is the final value of the cost function, i is the current band, starting from 1, n is the total number of bands, K is the dimensionally normalized weight vector of each band, R obs is the reflectivity obtained by UAV remote sensing observation, R sim Refers to the reflectance of the corresponding band output by the PROSAIL canopy reflectance model.

[0065] Furthermore, the method further includes: converting the nitrogen absorption amount of the leaves into chlorophyll content of the leaves.

[0066] In a specific embodiment, a crop agronomic parameter and yield estimation system that couples crop growth models and remote sensing data includes:

[0067] Collection module: used to collect crop growth data;

[0068] CERES-Rice model construction module: used to build the CERES-Rice model based on crop growth data and output rice agronomic parameters and yield data;

[0069] Preprocessing module: used to preprocess rice agronomic parameters and yield data;

[0070] PROSAIL canopy reflectance model construction module: used to construct the PROSAIL canopy reflectance model based on pre-processed rice agronomic parameters and yield data and crop growth data;

[0071] Parameter inversion and optimization module: used to output hyperspectral data through the PROSAIL canopy reflectance model and combine it with multi-source remote sensing data to perform CERES-Rice model parameter inversion and optimization;

[0072] Remote sensing data and crop model data assimilation module: used to build a bivariate collaborative assimilation framework of leaf area index and chlorophyll content, and deeply integrate the CERES-Rice model after parameter inversion optimization through multi-source remote sensing data;

[0073] Output module: used to output the final rice agronomic parameters and yield data based on the deeply integrated model.

[0074] In a specific embodiment, because the nitrogen concentration-related indicator input to the PROSAIL model is chlorophyll, rather than the nitrogen content or nitrogen uptake output by the CERES-Rice model, a conversion model between the two is required to achieve conversion between the two data. The chlorophyll measurement test was conducted using field trials conducted at the Jiansanjiang Experimental Station of China Agricultural University in 2013. The test varieties included Kongyu 131, Longjing 21, 31, 36, and 40, a total of five different rice varieties. At the critical growth stages of tillering, panicle differentiation, and jointing, 220 fully expanded leaves were selected based on the differences in rice leaf color at different nitrogen levels to ensure the stability of the chlorophyll estimation model. The results were used to measure indicators such as chlorophyll, leaf nitrogen content, leaf area, and nitrogen uptake. Chlorophyll measurement was based on the rice chlorophyll determination method published by Feng Shuanghua in 1997. A 5mm punch was used to obtain 10 samples from the tip, middle, and lower parts of the leaf. Ten holes were punched evenly throughout each leaf, resulting in 10 samples. The 10 samples were placed in a 15ml centrifuge tube. A 95% ethanol extract was poured into the tube and extracted for 24 hours until the leaves turned white. The volume was then adjusted to 14ml. Chlorophyll concentration was determined using a UV-Vis spectrophotometer (UV757CRT) at a wavelength of 652nm, using 95% ethanol as a blank control. Chlorophyll concentration in rice was calculated using the formula.

[0075]

[0076] Among them, D 652 is the absorbance of the extract, D0 is the absorbance of the blank control, V is the volume of the extract (ml), A is the leaf area (cm 2 ), chlorophyll concentration unit is ug / cm -2 .

[0077] In a specific embodiment, the steps of setting the PROSAIL model parameters specifically include:

[0078] S1: The PROSAIL canopy reflectance model was developed by linking the PROSPECT leaf optical property model with the SAIL canopy bidirectional reflectance model. PROSAIL uses 14 input parameters to define leaf pigment content, leaf water content, canopy structure, soil background reflectance, hotspot size, solar diffusivity, and solar geometry. Leaf pigment content is determined by chlorophyll a and b content (Ca; μg / cm -2 ), carotenoid content (Ccr; μg / cm -2 ) and brown pigment content (Cbp). Leaf water content was defined as equivalent water thickness (Cw; cm). Canopy structure was defined using four parameters, including leaf dry matter content (Cm; g / cm -2), leaf structure coefficient (N), leaf area index (LAI), and mean leaf inclination (θl; degrees). The solar geometry is characterized by the solar zenith, observer zenith, and solar azimuth. Based on these inputs, the model calculates canopy bidirectional reflectance from 400 to 2500 nm in 1 nm increments.

[0079] S2: For the independent PROSAIL inversion and the linkage with the ecophysiological model, the Cab and LAI parameters were of primary interest. Other parameters were either kept constant or confirmed in a specified manner based on measured values, as shown in Table 1. The Ccr and Cbp parameters were fixed at 20.0 μg / cm, respectively. -2 and 0.0, the equivalent water thickness was fixed at 0.02 cm, and the leaf dry matter content was fixed at 0.006 g / cm based on biomass measurement. -2 , structural parameters, N, and the mean blade pitch were manually adjusted to improve the PROSAIL inversion of LAI. The resulting parameter values, N of 1.35 and θl of 59°, were used for all subsequent PROSAIL simulations. The soil background reflectance parameter was determined from bare soil reflectance observations for each measurement date. During the field study, the solar diffusivity parameter was fixed at 15%, based on observations of shaded and sunlit Spectralon panels. The hotspot size parameter was fixed at 1.0. The solar zenith angle was calculated based on the timestamp of each radiometric observation at the site. Both the observer zenith angle and the solar azimuth were fixed at 0°.

[0080] Table 1 List of PROSAIL model input parameters

[0081]

[0082]

[0083] S3: In order to better utilize the initial parameters that are difficult to obtain accurately at the regional scale by inverting the growth model, a simple parameter sensitivity method was used to conduct sensitivity analysis on the CERES-Rice model. Finally, the cultivation and management parameters (sowing date, sowing rate and nitrogen fertilizer) that vary significantly within the regional scope were optimized as model input parameters, and the remaining management parameters were based on the empirical values ​​obtained from local surveys. The Wheat Grow model was used to simulate LNA (g / m -2 ), leaf dry matter (g / m -2 ) and LAI. Cab(μg / cm -2 ) values ​​were obtained from LNA and LAI using an algorithm, and Cm is the divisor of leaf dry matter and LAI (g / cm -2). Therefore, these values ​​(simulated by CERES-Rice) serve as input parameters for the PROSAIL model. A coupled model is then developed by combining CERES-Rice with PROSAIL.

[0084] In a specific embodiment, Figure 1 As shown in the figure, the agronomic parameters output by the CERES-Rice model are exactly the input parameters of the PROSAIL model, and the spectral curve of the PROSAIL model can be combined with the multispectral reflectance output by UAV remote sensing. Through the intermediate output variables, the two mechanism models of CERES-Rice and PROSAIL can be dynamically coupled with UAV remote sensing.

[0085] The CERES-Rice model takes as input four types of data: soil, variety, meteorological, and management data. The model outputs leaf area, leaf, stem, and seed biomass, and leaf, stem, and seed nitrogen content, along with crop yield data. Intermediate variables such as leaf area, leaf dry matter mass, and chlorophyll content are converted to chlorophyll, leaf dry matter content, and leaf area. Leaf dry matter content is obtained by dividing leaf dry matter mass by leaf area and converting units. This is then input into the PROSAIL model. The solar zenith angle, observation zenith angle, and relative azimuth angle correspond to those in the remote sensing data header file. Other parameter settings are based on existing research literature. High-resolution spectral data covering a wavelength range of 400–2500 nm are generated.

[0086] There is a significant correlation between leaf nitrogen uptake and chlorophyll content. Approximately 75% of leaf nitrogen is directly involved in chloroplast composition, and its content directly affects chlorophyll synthesis capacity. A statistical relationship was established based on measured data. Leaf area, leaf dry matter content, and chlorophyll content, as output by the CERES-Rice model, were input into the PROSAIL model. Other parameters were set to the default values ​​shown in Table 1. The solar zenith angle, observation zenith angle, and relative azimuth angle were consistent with those in the remote sensing data header file. Other parameter settings were based on existing research references. High-resolution spectral data covering the wavelength range of 400–2500 nm were generated.

[0087] This drives the PROSAIL model to output canopy-scale spectral reflectance at wavelengths of 400-2400nm. The drone remote sensing camera used is a four-band Sequoia camera (green light 550nm, red light 670, red edge 730nm, near-infrared 780nm). Since this camera has a narrow band, the camera's spectral value can be directly taken as the value of the center band. The hyperspectral data output by the PROSAIL model is converted into four-band reflectance data. The cost function is shown below:

[0088]

[0089] Among them, RES is the final value of the cost function, i is the current band, starting from 1, n is the total number of bands, K is the dimensionally normalized weight vector of each band, R obs is the reflectivity obtained by UAV remote sensing observation, R sim Refers to the reflectivity of the corresponding band output by the PROSAIL model.

[0090] Furthermore, the optimized CERES-Rice model incorporates target LAI and Cab parameters obtained from multiple sources (such as drones and satellites). The EnKF data assimilation algorithm is used to assimilate remotely sensed LAI and Cab data, optimizing the model's dynamic adjustment mechanism for leaf weight, leaf area, and leaf nitrogen content. By constructing a bivariate collaborative assimilation framework for leaf area index and chlorophyll content, the optimized CERES-Rice model is deeply integrated with multi-source remote sensing data. This deeply integrated model then outputs final rice agronomic parameters and yield data.

[0091] In a specific embodiment, the deep integration of the CERES-Rice model after parameter inversion optimization using multi-source remote sensing data includes:

[0092] For the determined sensitive parameter set results, the ENKF algorithm was used to assimilate the leaf area index and chlorophyll content retrieved by drone multispectral remote sensing in 2022 with the winter wheat leaf area index and chlorophyll content simulated by the DSSAT model. The specific steps of the model assimilation process are as follows:

[0093] (1) Based on the determined sensitive parameter set, Monte Carlo sampling was performed on the sensitive parameters according to the simlab input parameter text format to generate 4000 sampling sets. Each sampling set was input into the CERES-Wheat model to generate LAI and Cab, i.e., the forecast state variables. When obtaining each set of forecast state variables, if LAI and Cab were observed by remote sensing, a Gaussian perturbation with a mean of 0 and a standard deviation of 0.15 was set for the remote sensing observation data.

[0094] (2) The ENKF algorithm is used to assimilate the forecast state variables and the remote sensing observation data after Gaussian perturbation to obtain the updated LAI and Cab model forecast state variables. At the same time, the updated state variables are used as the next forecast state variables.

[0095] (3) Repeat steps (1) and (2) until all 4000 sampling sets have been run.

[0096] (4) At the end of each model run, a set of leaf area index, chlorophyll content and yield will be generated. The measured phenological period and yield will be used to filter out a new set from this set, and the final assimilation result will be the average value of this set.

[0097] In a specific embodiment, an experimental analysis of a method for estimating crop agronomic parameters and yield by coupling a crop growth model with remote sensing data is conducted, specifically including:

[0098] Example (1) Correlation between nitrogen uptake and chlorophyll content in rice leaves in cold regions

[0099] Since the CERES-Rice model outputs the amount of nitrogen absorbed by leaves, rather than the chlorophyll content, it cannot be directly coupled with the PROSAIL model. According to previous research experience, there is a good correlation between the amount of nitrogen absorbed by leaves and the chlorophyll content of leaves. In this example, a simple linear regression was established to convert the amount of nitrogen absorbed by leaves into the chlorophyll content of leaves. The results are as follows: Figure 2 As shown, the leaf content ranges from 10 to 80 kg ha -1 , and the chlorophyll content ranges from 30 to 70 ug / cm 2 There is a good linear relationship between leaf nitrogen uptake and chlorophyll content, and the R 2 The regression equation between the two is Y = 0.4593X + 30.672, which shows good consistency. Most of the data are distributed on both sides of the regression line. In addition, the unit of leaf nitrogen uptake is kg ha -1 , chlorophyll content unit is ug / cm 2 Therefore, in this embodiment, for cold-region rice, the nitrogen absorption amount of the aboveground part can be converted into chlorophyll content.

[0100] Example (2) Estimating LAI Effect Using the PROSAIL+CERES-Rice Model

[0101] This example uses the PROSAIL + CERES-Rice model coupling to establish a leaf area estimation method. To explore which period has the best effect in estimating leaf area, remote sensing data acquired by multispectral cameras during the jointing stage (SE) and heading stage (HD) are used as intermediate variables for coupling. The results are shown in Figures 3(a) and 3(b). During the jointing stage, the leaf area ranges from 1 to 5, and the simulated leaf area index has a good correlation with the measured leaf area index, R 2 =0.8, the two sets of data are distributed around the 1:1 line, and the convergence effect is good, RMSE = 0.53, and RE = 18.81%, the relative error performance is average, close to 20%. For the heading stage, the leaf area index is distributed between 2 and 7, which is generally greater than that during the jointing stage, and the variation between different treatments is large. The R between the simulated leaf area index and the measured leaf area index is 2 =0.71, which is lower than that in the jointing stage, and RMSE =0.65, which is higher than that in the jointing stage, but RE =15.44%, which is better than that in the jointing stage. RE in both the jointing and heading stages are less than 20%, indicating that the established models can be used. In summary, PROSAIL+CERES-Rice can well estimate LAI in the jointing and heading stages.

[0102] Example (3) Estimating biomass effects using the PROSAIL+CERES-Rice model

[0103] In order to better optimize nitrogen management, aboveground biomass is also one of the key factors. This example explored the effect of PROSAIL+CERES-Rice in simulating aboveground biomass at the jointing and heading stages. The results are shown in Figures 4(a) and 4(b). During the jointing stage, the AGB range is 2 to 8 t ha. -1 The simulated AGB has a good correlation with the measured AGB, R 2 =0.74, but the regression line between the two deviates from the 1:1 regression line, RMSE = 0.98t ha -1 , RE=26.95%, RE exceeded 20%, and the relative error was large. At the heading stage, the range of AGB was 4~11t ha -1 , the variation between different treatments is large, and the measured and simulated AGB are distributed on both sides of the 1:1 regression line. 2 =0.74, indicating a good correlation between the two, RMSE = 1.12 kg ha -1, RE = 17.8%, and RE is less than 20%, which is an acceptable level. Although the measured and simulated AGBs are well correlated across different periods, the RE is higher at the jointing stage, exceeding 20%, while it is less than 20% at the heading stage. Overall, the AGB data simulated by the PROSAIL+CERES-Rice model perform inconsistently across different periods, performing better at the heading stage and worse at the jointing stage.

[0104] Example (4) Estimating PNU Effect Using the PROSAIL+CERES-Rice Model

[0105] PNU is the product of nitrogen content and biomass. Nitrogen content is difficult to estimate and the model stability is poor. However, PNU is an indicator that can replace nitrogen content. Therefore, this example uses the PROSAIL+CERES-Rice model to carry out PNU estimation research. The results are shown in Figures 5(a) and 5(b). During the jointing stage, the range of PNU is 15 kg ha -1 ~100kg ha -1 The average value is 62 kg ha -1 , there is a correlation between the simulated PNU and the measured PNU, R 2 =0.71, RMSE = 13.34 kgha -1 , RE = 21.32%, and the relative error exceeds 20%. At the heading stage, the measured PNU range is 30-160 kg ha -1 The average value is 97.2 kg ha -1 , the correlation between the measured and simulated values ​​is good, R 2 =0.81, the data are distributed on both sides of the 1:1 line, with good consistency, RMSE = 16.35 kg ha -1 , RE=16.81%, which is lower than 20%. In terms of simulating PNU, the final result of heading stage is better than jointing stage.

[0106] Example (5) Using the PROSAIL+CERES-Rice model to estimate yield effects

[0107] Since remote sensing data or the PROSAIL model cannot directly output yield, the CERES-Rice model is a crop mechanism model that can directly output yield data. Therefore, this example couples the PROSAIL and CERES-Rice models to estimate cold-region rice yield at the jointing and heading stages. The results are shown in Figures 6(a) and 6(b). The yield range obtained in this study was 4.7 to 11.1 t ha. -1At the jointing stage and heading stage, there is a certain correlation between the simulated yield and the measured yield. The R 2 were 0.73 and 0.79 respectively, and the R 2 The data distribution of the heading stage is more evenly distributed on both sides of the 1:1 line, and the RMSE distribution of the two periods is 0.94 and 0.84 t ha -1 , the effect of estimation at the heading stage is better than that at the jointing stage. For RE, the RE of both is less than 20%, which are 11.18% and 9.94% respectively, which shows that both can estimate the yield of cold-region rice well at different periods, and the effect of estimation at the heading stage will be better.

[0108] The coupling of remote sensing technology and crop growth models has become a hot topic. Many technologies use empirical models to compare and optimize agronomic parameters obtained through remote sensing inversion with those output by crop growth models. This method assumes that agronomic parameters obtained through remote sensing inversion are more accurate than those simulated by crop growth models. However, in actual research and production, remote sensing inversion also suffers from the problems of different spectra for the same object and different objects for the same spectrum, as well as uncertainty and error. Remote sensing inversion is also known as pathological inversion. Therefore, this embodiment uses the vegetation index output by remote sensing as the key point of coupling, which can effectively avoid the errors caused by such problems. This embodiment constructs a method for estimating agronomic parameters and yield by coupling the remote sensing optical model PROSAIL with the growth model CERES-Rice. The effectiveness of estimation at different time periods is studied, ultimately demonstrating that this method can effectively estimate the LAI, AGB, PNU, and yield of cold-region rice.

[0109] When simulating wheat parameters in different years and regions, the best results were achieved using a three-band approach consisting of red, green, and near-infrared, or SAVI. This example used GSAVI as the intermediate parameter, demonstrating the importance of coupling appropriate vegetation indices. The optimal timing for wheat is the flowering and heading stage, which in this example was the heading stage. Because rice transitions from vegetative growth to reproductive growth during this stage, the number of tillers is essentially set. Yield potential is essentially established after the heading stage, and unless there are significant disasters, yield variation is minimal.

[0110] Since the output and input parameters of the CERES-Rice model and the PROSAIL model are inconsistent, this example establishes the correlation between leaf nitrogen uptake and chlorophyll. The regression model R established in this example 2=0.85 is superior to existing research. LAI is a parameter directly related to the PROSAIL model. According to research on PROSAIL's global sensitivity, LAI is one of the main variables affecting the entire spectral curve, especially in the near-infrared band. LAI is the main influencing factor, but reflectance saturation occurs in areas with high LAI. The PROSAIL model input parameter leaf dry matter content is obtained by dividing leaf dry weight by leaf area. This indicator has a good correlation with leaf mass. The CERES-Rice model outputs leaf mass and also outputs aboveground plant biomass. Due to the significant correlation between leaf area and biomass, biomass estimation can also be carried out through a coupled method in this embodiment.

[0111] Leaf nitrogen content, or chlorophyll content, is another hot topic in inversion research, ranking second only to LAI. This example obtained nitrogen uptake from the CERES-Rice model. Chlorophyll content is generally more sensitive to green light. This example successfully inverted cold-region rice yield, achieving an RE of 9.94% at the heading stage. This is consistent with the conclusions of other researchers. Because CERES-Rice is a process-based model, it fully accounts for differences in weather, soil, management, and variety. Furthermore, through remote sensing calibration, it can fully account for seasonal conditions.

[0112] This example uses the crop growth model CERES-Rice coupled with UAV remote sensing and radiation transfer models, and establishes a method for estimating LAI, AGB, PNU, and yield through vegetation indices. A conversion method for nitrogen uptake and chlorophyll content of cold-region rice plants was established, and the regression equation was Y = 0.4593X + 30.672, with R 2 The coupled method was effective in estimating LAI, with an RMSE of 0.53 and 0.65 at the jointing and heading stages, respectively. The coupled method also performed well in estimating biomass, with RMSEs of 0.98 and 1.21 t ha at the jointing and heading stages, respectively. -1 The model can also effectively estimate PNU, with RMSE of 13.34 and 16.35 kg ha at the jointing and heading stages, respectively. -1 The model and remote sensing coupled approach, incorporating meteorological, soil, management, and variety data, and dynamically calibrated, enabled yield estimation with RMSEs of 0.94 and 0.84 t ha, respectively. -1 .

[0113] In a specific embodiment, a method for estimating crop agronomic parameters and yields by coupling crop growth models and remote sensing data is described, taking the Jiangsanjiang Experimental Station of China Agricultural University located in the Sanjiang Plain of Heilongjiang Province in Northeast China as an example. The Sanjiang Plain is a typical temperate semi-humid continental monsoon climate zone. The main soil type is white pulp black soil, and japonica rice is the main crop in this cold region. The average annual sunshine time is about 2300h-2600h, and the frost-free period is only about 110-135 days per year. During the growing season, the annual average temperature is about 2°C, and the daily average temperature is 19.9°C. The average annual rainfall is 500-600 mm, of which about 72% occurs from June to September. The specific experiment is as follows:

[0114] (1) Field experiment setup

[0115] The study was conducted in 10 plots in 2017 and 2018, involving two japonica rice varieties, Longjing 31 (11 leaves) and Longjing 21 (12 leaves), with five nitrogen fertilizer rates of 0, 40, 80, 120, and 160 kg ha -1 , two different planting densities (27 and 33 plants m -2 All experiments were conducted using a completely randomized block design with three replicates. The size of each plot was 7 m × 9 m and did not change during the study period. In the nitrogen level experiment, nitrogen fertilizer was applied in three parts: 40% as a basal application before transplanting, 30% at the tillering stage, and 30% at jointing. Phosphate and potassium fertilizers were applied at 50 kg ha each throughout the field experiment. -1 P2O5 and 105K2O kg ha -1 Phosphorus fertilizer should be applied once before transplanting, and potassium fertilizer should be applied in equal parts before transplanting and during the jointing period.

[0116] In addition to the plot experiment, three farmers were selected to conduct a validation experiment in Qixing Farm in 2017 and 2018. The soil organic matter (OM) content of the three farmers was 30.2 g kg -1 , 37.5gkg -1 and 43.2g kg -1 . Treatments in each trial included (1) farmer practice (FP); (2) regional optimal management (ROM); (3) precision rice management 1 (PRM1) based on remote sensing-based nitrogen recommendation of rice at the jointing stage; (4) PRM2; and (5) PRM3. PRM2 and PRM3 used two different rates of controlled-release fertilizer as basal fertilizer. The plot size of each treatment varied from 20 m × 8 m to 30 m × 10 m, depending on the farmer's field conditions. The rice variety was Longjing 31, and each treatment was replicated three times. Detailed information on planting density and nitrogen fertilizer application rate is given in Table 2.

[0117] Table 2 Detailed information on planting density and nitrogen fertilizer application rate

[0118]

[0119] Table Notes: FP stands for farmer management, ROM stands for optimized regional management, PRM1 stands for precision rice management with nitrogen recommendation based on remote sensing, PRM2 and PRM3 respectively indicate that PRM1 and PRM2 use two controlled-release fertilizers with different rates as basal fertilizers.

[0120] (2) Drone data collection

[0121] This experiment used a Swiss-made eBeeSQ fixed-wing unmanned aerial vehicle (SenseFly, Cheseaux-sur-Lausanne, Switzerland) equipped with a multispectral redwood camera. This camera includes a quad-band multispectral camera (1.2 MP, 1280 × 960 pixels) for the green band (550 ± 20 nm), the red band (660 ± 20 nm), the red-edge band (735 ± 5 nm), and the near-infrared band (790 ± 20 nm), as well as a red, green, and blue (RGB) camera (16 MP, 4608 × 3456 pixels). The drone was also equipped with an upward-facing solar irradiance sensor that automatically controls the sensor integration time. The camera system can use a white radiometric calibration plate, and spectral reflectance was calibrated using a radiometric panel (Labsphere, Inc., North Sutton, NH, USA) before each flight. The drone survey missions were conducted between 10:00 AM and 2:00 PM during daytime hours under clear sky and calm wind conditions. A total of four multispectral reflectance orthophotos were acquired during the jointing and heading stages of rice growth in 2017 and 2018. Field map boundaries were vectorized and used as regions of interest. The average value of the image spectrum within the region was selected as the reflectance for the field, which was then used to calculate and align them with the ground data.

[0122] (3) Experimental data collection

[0123] During the critical growth periods of cold-region rice, mainly the jointing stage (SE) and the heading stage (HD), drone remote sensing aerial surveys were carried out and samples were taken immediately. During each sampling period, three holes were selected from a representative rice area according to the average number of tillers in the plot. After washing and removing the roots, the plants were placed in an oven at a high temperature of 105°C to fix the green leaves. The water was then dried in an oven at a constant temperature of 85°C and finally weighed. The aboveground biomass of the plot was obtained after conversion. The dried samples were then crushed, digested with sulfuric acid and hydrogen peroxide, and the standard Kjeldahl method was used to determine the nitrogen concentration of the plants. In order to ensure data quality, a certain number of blind samples were set in the middle to ensure data accuracy. The nitrogen absorption of the aboveground part of the plant is the aboveground biomass multiplied by the nitrogen concentration of the entire aboveground plant (leaves + stems).

[0124] (4) Data Analysis

[0125] In this embodiment, the reflectance data of the four band spectra acquired by the UAV Sequoia camera are used to calculate various VIs (as shown in Table 3), and the original reflectance data of the three bands and VI are used in the analysis. 2 The calculated VIs were ranked in relation to AGB, PNU and NNI, and the indicators with the highest performance were further studied. The data collected in 2017 and 2018 were pooled together and then randomly divided into a training dataset (70%) and a test dataset (30%). A total of 381 observations were obtained in the experiments in 2017 and 2018, of which 266 were used as training datasets and 115 were used as test datasets. The data range of all training datasets covered the test dataset range, ensuring that the test data did not exceed the range of the trained model. The training dataset was used to establish simple regression models using linear, quadratic, power, exponential and logarithmic functions between a single band and 72 VIs and AGB, PNU and NNI. The established model was evaluated using the test dataset. The coefficient of determination (R 2 ), root mean square error (RMSE) and relative error (RE) were used to evaluate the model. 2 The higher is and the lower RMSE and RE are, the higher the precision and accuracy of the model for predicting N state indicators.

[0126] Table 3 Common vegetation index

[0127]

[0128]

[0129] As shown in Table 4, the monitoring effects of the models constructed using different methods for key rice indicators in different periods are compared. For AGB in the SE period, the monitoring effects of the models constructed using different methods for the present embodiment, the single-band modeling (Band), and the commonly used vegetation index method (VI) R 2The results are 0.74, 0.45, and 0.68 respectively. This embodiment performs the best. However, for RMSE and RE, this embodiment has no significant difference from the other two methods. For AGB in HD period, R 2 The RMSE and RE results are basically consistent with those of the VI method, which are 0.74, 1.21 t / ha, and 17.8% respectively, but all of them significantly optimize the Band method. A unified analysis was also conducted for LAI, which is directly related to the near-infrared band. The results are basically similar to those of AGB. In the SE and HD periods, both this embodiment and the VI method perform better than the Band method. In addition, this embodiment performs better than AGB to a certain extent. 2 In the SE and HD periods, the values ​​were 0.81 and 0.71, the RMSEs were 0.53 and 0.65, and the REs were 18.8% and 15.4%, respectively.

[0130] The effect of this embodiment on estimating nitrogen absorption index is obviously better than that of band and VI methods. In the SE period, R 2 The estimated accuracy of the Band and VI methods has been greatly improved, but it is still not as significant as the effect of this embodiment. The three methods have R 2 were 0.81, 0.43, and 0.64, respectively; the RMSE were 16.3, 28.1, and 22.5 kg / ha, respectively; and the RE were 16.8%, 29.3%, and 23.6%, respectively.

[0131] The yield estimation effect of this embodiment is also significantly better than the traditional band and VI methods. Since this embodiment uses the crop model and remote sensing coupling to form a yield estimation method based on the crop growth mechanism, the error of remote sensing estimation can be adjusted by using parameters such as meteorology, soil, and crop field management. This significantly improves the problem of poor remote sensing yield estimation effect in the early stage of crop production. In the SE period, the yield estimation R of this embodiment is 2The VIR reached 0.73, significantly exceeding Band's 0.32 and VI's 0.46. The RMSE was 0.94 t / ha, surpassing Band's 1.34 t / ha and VI's 1.12 t / ha. Simultaneously, the RE was low at 11.2%, within a narrow relative error range. This method can achieve relatively accurate yield estimates in the early stages of crop growth and has broad application prospects. During the HD period, since the crop population is essentially complete, VI can reflect the basic characteristics of the population and characterize the composition of future yields. Therefore, the VI yield estimation method is significantly more effective than during the SE period, but still slightly inferior to the present embodiment. The VIR2 of this embodiment and VIR2 were 0.79 and 0.74, significantly higher than Band's 0.34. The RMSEs were 0.84 and 0.78, respectively, also surpassing Band's 10.9. Similarly, both methods had low REs of 9.94% and 9.74%, respectively, within a narrow range of variation.

[0132] Table 4 Comparison of the effects of constructing rice biomass, leaf area, nitrogen uptake, and yield monitoring models at the jointing and heading stages using this embodiment, the single-band modeling method, and the commonly used vegetation index method

[0133]

[0134] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on the differences from other embodiments. Reference can be made to the common and similar parts between the various embodiments. For the devices disclosed in the embodiments, since they correspond to the methods disclosed in the embodiments, the description is relatively simple, and the relevant parts can be referred to the method description.

[0135] The above description of the disclosed embodiments is intended to enable one skilled in the art to implement or use the present invention. Various modifications to these embodiments will be readily apparent to one skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention is not limited to the embodiments shown herein but is intended to conform to the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for estimating crop agronomic parameters and yield by coupling crop growth models and remote sensing data, characterized in that: include: Collect crop growth data, build the CERES-Rice model based on the crop growth data, and output rice agronomic parameters and yield data; Preprocessing of rice agronomic parameters and yield data; The PROSAIL canopy reflectance model was constructed based on the preprocessed rice agronomic parameters, yield data, and crop growth data; Hyperspectral data is output through the PROSAIL canopy reflectance model and combined with multi-source remote sensing data to perform parameter inversion and optimization of the CERES-Rice model; A bivariate collaborative assimilation framework of leaf area index and chlorophyll content was constructed, and the CERES-Rice model after parameter inversion and optimization was deeply integrated through multi-source remote sensing data. The final rice agronomic parameters and yield data are output based on the deeply integrated model.

2. The method for estimating crop agronomic parameters and yield by coupling crop growth models and remote sensing data according to claim 1, characterized in that: The crop growth data includes: soil data, variety data, meteorological data, management data, leaf pigment content, leaf water content, canopy structure, soil background reflectivity, hotspot size, solar diffusivity and solar geometry.

3. The method for estimating crop agronomic parameters and yield by coupling crop growth models and remote sensing data according to claim 2, characterized in that: The leaf pigment content includes: chlorophyll a and b content, carotenoid content and brown pigment content; leaf water content is defined as equivalent water thickness; canopy structure includes leaf dry matter content, leaf structural coefficient, leaf area index and average leaf inclination angle; the characteristics of solar geometry include: solar zenith, observer zenith and solar azimuth.

4. The method for estimating crop agronomic parameters and yield by coupling crop growth models and remote sensing data according to claim 1, wherein: The preprocessing of the rice agronomic parameters and yield data includes: dividing the leaf dry matter mass by the leaf area and converting the units to obtain the leaf dry matter content, which is used as an input variable of the PROSAIL canopy reflectance model.

5. The method for estimating crop agronomic parameters and yield by coupling crop growth models and remote sensing data according to claim 1, characterized in that: The PROSAIL canopy reflectance model is used to output canopy-scale spectral reflectance at wavelengths of 400-2400 nm.

6. The method for estimating crop agronomic parameters and yield by coupling crop growth models and remote sensing data according to claim 1, characterized in that: The multi-source remote sensing data includes: the drone remote sensing camera used is a four-band camera, the spectral value of the camera is directly taken as the value of the central band, and the hyperspectral data output by the PROSAIL canopy reflectance model is converted into four-band reflectance data.

7. The method for estimating crop agronomic parameters and yield by coupling crop growth models and remote sensing data according to claim 1, characterized in that: The parameter inversion optimization adopts the SCE-UA algorithm to perform parameter optimization with minimum error.

8. The method for estimating crop agronomic parameters and yield by coupling crop growth models and remote sensing data according to claim 7, characterized in that: The cost function of the parameter inversion optimization includes: Among them, RES is the final value of the cost function, i is the current band, starting from 1, n is the total number of bands, K is the dimensionally normalized weight vector of each band, R obs is the reflectivity obtained by UAV remote sensing observation, R sim Refers to the reflectance of the corresponding band output by the PROSAIL canopy reflectance model.

9. The method for estimating crop agronomic parameters and yield by coupling crop growth models and remote sensing data according to claim 1, wherein: Also includes: The data of leaf nitrogen absorption was converted into the chlorophyll content of the leaves.

10. A crop agronomic parameter and yield estimation system that couples crop growth models and remote sensing data, characterized in that: include: Collection module: used to collect crop growth data; CERES-Rice model construction module: used to build the CERES-Rice model based on crop growth data and output rice agronomic parameters and yield data; Preprocessing module: used to preprocess rice agronomic parameters and yield data; PROSAIL canopy reflectance model construction module: used to construct the PROSAIL canopy reflectance model based on pre-processed rice agronomic parameters and yield data and crop growth data; Parameter inversion and optimization module: used to output hyperspectral data through the PROSAIL canopy reflectance model and combine it with multi-source remote sensing data to perform CERES-Rice model parameter inversion and optimization; Remote sensing data and crop model data assimilation module: used to build a bivariate collaborative assimilation framework of leaf area index and chlorophyll content, and deeply integrate the CERES-Rice model after parameter inversion optimization through multi-source remote sensing data; Output module: used to output the final rice agronomic parameters and yield data based on the deeply integrated model.

Images