CN113052433B

CN113052433B - Crop unit yield estimation method based on key time phase and farmland landscape characteristic parameters

Info

Publication number: CN113052433B
Application number: CN202110198175.8A
Authority: CN
Inventors: 杜鑫; 李强子; 朱炯; 张源; 王红岩
Original assignee: Aerospace Information Research Institute of CAS
Current assignee: Aerospace Information Research Institute of CAS
Priority date: 2021-02-22
Filing date: 2021-02-22
Publication date: 2024-02-06
Anticipated expiration: 2041-02-22
Also published as: CN113052433A

Abstract

The invention provides a crop unit yield estimation method based on key time phases and farmland landscape characteristic parameters. On the basis, redundant parameters are removed through correlation analysis, the optimal characteristic parameter combination is selected, and an optimal model is constructed to perform crop unit yield estimation. Based on the key time phase and characteristic parameters of screening, the efficiency of crop unit yield estimation modeling is greatly improved, and stable estimation accuracy is ensured.

Description

Crop unit yield estimation method based on key time phase and farmland landscape characteristic parameters

Technical Field

The invention relates to the technical field of agricultural production, in particular to a crop unit yield estimation method based on key time phases and farmland landscape characteristic parameters.

Background

Traditional crop estimation methods are mainly based on agronomic and meteorological models. And obtaining a large number of samples through sampling investigation, and establishing a regression model between the yield and the main agronomic factors and meteorological factors. Although the method has high precision, the method requires a great deal of time and labor cost, is easy to be interfered by external factors, and is particularly obvious in estimating the yield of crops in a large area. In addition, various statistical indexes and mathematical models are used for statistical yield estimation, the accuracy of the method is higher, theoretical support is lacked, the estimation accuracy is poor in the year with larger yield fluctuation, a great amount of statistical data in the current year is needed to be input into the yield estimation model, and the hysteresis of the model estimation result is serious.

Early remote sensing estimation methods mainly built statistical regression models between original spectral features or vegetation indexes acquired in a single growth period and yield, and the empirical models have poor space-time generalization capability, so that the models are often required to be reconstructed when applied to different areas or years. In addition, studies have also been made on utilizing a light energy utilization efficiency (LUE) model to perform crop yield estimation based on time-series remote sensing data; or the crop unit yield estimation is carried out by utilizing a remote sensing and crop growth model assimilation method. Although the theory of estimating yield by using a crop growth model or a light energy utilization model is stronger, many difficulties are faced in the application of estimating yield on a large scale in county and city level, such as difficulty in localization of the crop growth model, too many input parameters, difficulty in determining the harvest index of the light energy utilization model, and the like, and in addition, the models are low in efficiency and difficult to verify in large-area estimation.

As can be seen, the prior art estimates yield per crop, either the estimation process is underutilized for critical growth periods/phases of the crop or involves processing large amounts of data, which is inefficient.

Disclosure of Invention

The invention provides a crop unit yield estimation method based on key time phases and farmland landscape characteristic parameters, which is used for solving the problems that in the prior art, the estimation process is insufficient for the key growth period/time phases of crops or the estimation process is low in efficiency due to the fact that a large amount of data are processed.

The invention provides a crop unit yield estimation method based on key time phases and farmland landscape characteristic parameters, which comprises the following steps:

screening key time phases of crops to be subjected to unit yield estimation; the key time phase is determined according to the influence of the growth characteristic parameter and the environment characteristic parameter of the crop in different alternative time phases on the unit yield estimation precision;

screening key characteristic parameters for unit yield estimation of the crops; the key characteristic parameters are determined according to the sensitivity degree of the characteristic parameters to the influence of the unit yield estimation precision of the crops;

determining an optimal characteristic parameter combination by eliminating data redundancy according to the key time phase and the key characteristic parameter, constructing a unit yield estimation model of the crop based on the optimal characteristic parameter combination, and performing unit yield estimation on the crop by the unit yield estimation model;

the key characteristic parameters comprise growth characteristic parameters, environment characteristic parameters and farmland landscape characteristic parameters of crops.

Optionally, the screening the critical phases of the crops to be subjected to the unit yield estimation comprises:

determining a first model for estimating the unit yield of the crop according to the growth characteristic parameter and the environment characteristic parameter of the crop in any alternative time phase in at least one growing season and the actual unit yield corresponding to the crop, and acquiring the accuracy of the unit yield estimation of the crop by the first model as the accuracy corresponding to the any alternative time phase;

According to the relative precision of different alternatives of the crops, selecting three alternative time phases with highest precision as screened alternative single time phases;

determining a second model for estimating the unit yield of the crop according to the growth characteristic parameter and the environment characteristic parameter of the crop in the optional time phase combination in at least one growth season and the actual unit yield corresponding to the crop aiming at any time phase combination determined by the screened single optional time phase, and acquiring the accuracy of estimating the unit yield of the crop by the second model as the accuracy corresponding to the optional time phase combination;

selecting the alternative time phase combination with the minimum time phase number, which has the precision within a preset precision range, as a screened alternative time phase combination according to the precision corresponding to the time phase combination of different alternatives of the crops; the preset precision range is determined by precision of unit yield estimation of the crops when all the alternative list phases participate;

and if the precision corresponding to the selected alternative time phase combination is higher than that of any selected single alternative time phase, taking the selected alternative time phase combination as the key time phase, otherwise, taking the single time phase corresponding to the highest precision in the selected alternative single time phase as the key time phase.

Optionally, the screening the key characteristic parameters for unit yield estimation of the crops includes:

determining a third model for estimating the unit yield of the crop according to the characteristic parameters of the growth vigor of the crop in at least one growing season and the actual unit yield corresponding to the crop, and acquiring the accuracy of the unit yield estimation of the crop by the third model as the accuracy corresponding to the characteristic parameters of the growth vigor;

determining a fourth model for estimating the unit yield of the crop according to the growth characteristic parameter and the first environmental characteristic parameter of the crop in at least one growing season and the actual unit yield corresponding to the crop, and acquiring the accuracy of the unit yield estimation of the crop by the fourth model as the accuracy corresponding to the growth characteristic parameter and the first environmental characteristic parameter; wherein the first environmental characteristic parameter is an environmental characteristic parameter related to moisture;

determining a fifth model for estimating the unit yield of the crop according to the growth characteristic parameter and the second environment characteristic parameter of the crop in at least one growing season and the actual unit yield corresponding to the crop, and acquiring the accuracy of the unit yield estimation of the crop by the fifth model as the accuracy corresponding to the growth characteristic parameter and the second environment characteristic parameter; wherein the second environmental characteristic parameter is a temperature-related environmental characteristic parameter;

Determining a sixth model for estimating the unit yield of the crop according to the growth characteristic parameter, the first environmental characteristic parameter and the second environmental characteristic parameter of the crop in at least one growing season and the actual unit yield corresponding to the crop, and acquiring the accuracy of the unit yield estimation of the crop by the sixth model as the accuracy corresponding to the growth characteristic parameter and the environmental characteristic parameter;

and acquiring a combination with highest precision from the precision corresponding to the combination of the growth characteristic parameters and the different environment characteristic parameters, wherein if the precision corresponding to the alternative combination is higher than the first set improvement precision compared with the precision corresponding to the growth characteristic parameters, the characteristic parameters contained in the alternative combination are used as primarily determined key characteristic parameters, otherwise, the growth characteristic parameters are used as primarily determined key characteristic parameters.

Optionally, after obtaining the preliminary determined key feature parameter, the method further includes:

determining a seventh model for estimating the unit yield of the crop according to the key characteristic parameters and the farmland landscape characteristic parameters which are preliminarily determined for the crop in at least one growing season and the actual unit yield corresponding to the crop, and acquiring the accuracy of the unit yield estimation of the crop by the seventh model as the accuracy corresponding to the key characteristic parameters and the farmland landscape characteristic parameters which are preliminarily determined;

And if the precision of the precision improvement corresponding to the primarily determined key characteristic parameter and the farmland landscape characteristic parameter relative to the precision improvement corresponding to the growth characteristic parameter is larger than the second set precision improvement, taking the characteristic parameter contained in the primarily determined key characteristic parameter and the farmland landscape characteristic parameter as the determined key characteristic parameter.

Optionally, the determining an optimal characteristic parameter combination by eliminating data redundancy according to the key time phase and the key characteristic parameter, and constructing a unit yield estimation model of the crop based on the optimal characteristic parameter combination includes:

according to the correlation degree among the key characteristic parameters in the key time phase, carrying out correlation screening on the key characteristic parameters, so that only one key characteristic parameter with a correlation coefficient larger than a preset coefficient representing the correlation degree is reserved;

and taking the parameter combination obtained after the correlation screening in the key time phase as the optimal characteristic parameter combination, and determining a model for estimating the unit yield of the crop according to each characteristic parameter in the crop optimal characteristic parameter combination in at least one growing season and the actual unit yield corresponding to the crop.

Optionally, before screening the critical phases of the crop to be subjected to the unit yield estimation, the method further comprises:

constructing a growth curve of the crop through a time sequence remote sensing parameter NDVI, determining a first alternative time phase according to characteristic points representing the growth process of the crop in the growth curve, extracting at least one alternative time phase of the growth process before the first alternative time phase based on the first alternative time phase, and at least one alternative time phase of the growth process after the first alternative time phase.

The invention also provides a crop unit yield estimation device based on the key time phase and the farmland landscape characteristic parameters, which comprises:

the first screening module is used for screening key time phases of crops to be subjected to unit yield estimation; the key time phase is determined according to the influence of the growth characteristic parameter and the environment characteristic parameter of the crop in different alternative time phases on the unit yield estimation precision;

the second screening module is used for screening key characteristic parameters for unit yield estimation of the crops; the key characteristic parameters are determined according to the sensitivity degree of the characteristic parameters to the influence of the unit yield estimation precision of the crops;

the generation module is used for determining an optimal characteristic parameter combination through eliminating data redundancy according to the key time phase and the key characteristic parameters, constructing a unit yield estimation model of the crops based on the optimal characteristic parameter combination, and carrying out unit yield estimation on the crops through the unit yield estimation model;

The invention also provides an electronic device comprising a memory, a processor and a computer program stored on the memory and capable of running on the processor, wherein the processor realizes the steps of the crop unit yield estimation method based on the key time phase and the farmland landscape characteristic parameters when executing the program.

The present invention also provides a non-transitory readable storage medium having stored thereon a computer program which, when executed by a processor, performs the steps of any of the above-described crop yield per unit estimation methods based on key time phases and farm landscape characteristic parameters.

According to the crop unit yield estimation method based on the key time phase and the farmland landscape characteristic parameters, which is provided by the invention, key time phases which are important for unit yield formation are screened out from the growth process of crops according to the influence on the accuracy of unit yield estimation, and key characteristic indexes for unit yield formation are screened out from the crop growth condition characteristics, the environment influence characteristics and the farmland landscape characteristics. On the basis, redundant parameters are removed through correlation analysis, the optimal characteristic parameter combination is selected, and an optimal model is constructed to perform crop unit yield estimation. Based on the key time phase and characteristic parameters of screening, the efficiency of crop unit yield estimation modeling is greatly improved, and stable estimation accuracy is ensured.

Drawings

In order to more clearly illustrate the embodiments of the present invention or the technical solutions of the prior art, the following description will briefly explain the drawings used in the embodiments or the description of the prior art, and it is obvious that the drawings in the following description are some embodiments of the present invention, and other drawings can be obtained according to these drawings without inventive effort for a person skilled in the art.

FIG. 1 is a flow diagram of a crop yield per unit estimation method based on key time phases and farmland landscape characteristic parameters;

FIG. 2 is a block diagram of a crop yield per unit estimation device based on key time phases and farmland landscape characteristic parameters according to another embodiment of the present invention;

fig. 3 is a schematic physical structure of an electronic device according to another embodiment of the present invention.

Detailed Description

For the purpose of making the objects, technical solutions and advantages of the embodiments of the present invention more apparent, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention, and it is apparent that the described embodiments are some embodiments of the present invention, but not all embodiments of the present invention. All other embodiments, which can be made by those skilled in the art based on the embodiments of the invention without making any inventive effort, are intended to be within the scope of the invention.

While the prior art is more theoretical in estimating yield using crop growth models or light energy utilization models, many difficulties are faced in performing extensive yield estimation applications (e.g., county-level yield estimation applications), such as difficulty in localization of crop growth models, excessive input parameters, difficulty in determining harvest index of light energy utilization models, etc., and in addition, these models often act on pixel or plot scale, and are inefficient and difficult to verify in large area estimation. For the county and city level unit production estimation, compared with a complex mechanism model, sampling statistical investigation and an empirical model based on multi-source data such as remote sensing weather and the like are better in reliability and operability, and still are mainstream schemes in the current or future in a certain time. The yield is the result of the combined action of the gene characteristics of crops, the growth environment and the farmland management, quantitative observation of all factors is difficult in a large area, and the information which can be obtained by the prior technical means mainly comprises light temperature (such as sunlight hours, radiant quantity and the like), moisture (such as precipitation, soil moisture and the like), the distribution of crops, the growth state and the like (such as canopy height, leaf area index, vegetation index and the like). Many researches have performed regression analysis by using parameters such as statistical unit yield and spectral reflectivity, vegetation index or temperature, precipitation and the like, good effects are obtained, but no systematic research has been performed in the aspects of multi-source feature screening and analysis, and regression modeling is often performed by using only a single NDVI or meteorological feature. In addition, studies have shown that the degree of field breakage and the degree of agricultural concentration during crop growth can also have an effect on yield formation. At the county scale, crop landscape pattern indexes can be calculated to represent the information, and the existing research lacks the utilization of the indexes. Meanwhile, the influence of time sequence characteristics on crop estimation is comprehensively and systematically considered from the production forming process in the current research, and on the county scale, the research mostly adopts a single time period to perform model estimation, and although the mode is simple and easy to implement, other lost time characteristics may reduce the accuracy of estimation. Furthermore, since there may be differences in sowing times for different area or year crops, it is not reasonable to construct a unit yield model using only the same DOY (Day Of Year) period.

In order to solve the problems in the prior art, the scheme provided by the application divides the existing agricultural condition features easy to obtain into three types of crop growth features, environment features and farmland landscape features, comprehensively evaluates the influences of various features and yield formation key time on estimation of the yield of the winter wheat, and aims to create a crop unit yield estimation model based on remote sensing and meteorological data to realize high-precision estimation of the yield of the winter wheat in county scale. It mainly surrounds the following three problems: (a) analyzing a critical period of crop estimation; (b) assessing the impact of environmental characteristics on crop estimation; (c) And analyzing the promotion effect of farmland landscape characteristics on crop estimation. And finally, carrying out variable screening analysis on the basis of the optimal time phase and the characteristics, and constructing an optimal model to realize the estimation of the county-scale winter wheat unit yield.

Fig. 1 is a flow chart of a crop unit yield estimation method based on key time phase and farmland landscape characteristic parameters provided in this embodiment, referring to fig. 1, the crop unit yield estimation method based on key time phase characteristic parameters and farmland landscape characteristic parameters includes:

step 101: screening key time phases of crops to be subjected to unit yield estimation; the key time phase is determined according to the influence of the growth characteristic parameter and the environment characteristic parameter of the crop in different alternative time phases on the unit yield estimation precision;

Step 102: screening key characteristic parameters for unit yield estimation of the crops; the key characteristic parameters are determined according to the sensitivity degree of the characteristic parameters to the influence of the unit yield estimation precision of the crops;

step 103: determining an optimal characteristic parameter combination by eliminating data redundancy according to the key time phase and the key characteristic parameter, constructing a unit yield estimation model of the crop based on the optimal characteristic parameter combination, and performing unit yield estimation on the crop by the unit yield estimation model;

The crop unit yield estimation method based on the key time phase characteristic parameter and the farmland landscape characteristic parameter provided by the embodiment is executed by a computer or a server, and crops in the embodiment can be any crop, for example, wheat. The time phase is obtained by dividing the crop growth period through remote sensing data.

In this embodiment, on the one hand, the time phase most representing the estimation of the unit yield of the crop is selected, and on the other hand, the characteristic parameter most reflecting the formation process of the unit yield of the crop in the growth period of the crop is selected, and a model for estimating the unit yield of the crop is determined based on the selected time phase and the data obtained by selecting the characteristic parameter. The data for establishing the model is simplified from the time phase and the characteristic parameters, so that the model for single-product estimation can be established based on the minimum data quantity, and the efficiency of model establishment is improved.

The embodiment provides a crop unit yield estimation method based on key time phases and farmland landscape characteristic parameters, wherein key time phases which are important for unit yield formation are screened out from the growth process of crops according to the influence on the accuracy of unit yield estimation, and key characteristic indexes for unit yield formation are screened out from crop growth condition characteristics, environment influence characteristics and farmland landscape characteristics. On the basis, redundant parameters are removed through correlation analysis, the optimal characteristic parameter combination is selected, and an optimal model is constructed to perform crop unit yield estimation. Based on the key time phase and characteristic parameters of screening, the efficiency of crop unit yield estimation modeling is greatly improved, and stable estimation accuracy is ensured.

Further, on the basis of the above embodiment, the screening the critical phases of the crops to be subjected to the unit yield estimation includes:

Specifically, when the key time phase is screened, three alternative single time phases can be screened according to the precision of each alternative time phase, then the alternative single time phases are combined in pairs to obtain alternative double time phase combinations, and at least one alternative double time phase combination is selected according to the precision when the crop unit yield estimation is carried out based on each alternative double time phase combination. The corresponding precision of the selected alternative double-time phase combination is higher than that of the Yu Renyi alternative single-time phase, and the corresponding precision of the selected alternative double-time phase combination is higher (usually obviously higher) than that of the unselected alternative double-time phase combination.

And adding an alternative time phase which is not contained in the selected alternative double-time-phase combination to obtain an alternative three-time-phase combination, and if the corresponding precision of the alternative three-time-phase combination is obviously higher than that of the alternative double-time-phase combination, taking the alternative double-time-phase combination as a key time phase. And so on, until a certain time phase combination is obtained, after an alternative time phase which is not contained in the current time phase combination is added into the certain time phase combination, the corresponding precision of forming a new time phase combination is not improved, and the certain time phase combination is taken as a key time phase.

For example, 5 single alternative phases, P1, P2, P3, P4 and P5, respectively, are determined for the crop. The accuracy of the first model (which can be obtained by comparing the unit yield predicted from the first model based on the growth characteristics and environmental characteristics in 2018 in the single alternative time phase with the actual unit yield of 18 years) is verified by software (e.g., matlab) fitting the first model based on the growth characteristics and environmental characteristics of each single alternative time phase, and the actual unit yield (e.g., based on the growth characteristics and environmental characteristics in 2014-2017 in the single alternative time phase and the actual unit yield in each year), respectively. Specifically, modeling and precision evaluation are sequentially performed by using all characteristic parameters in the time phases P1, P2, P3, P4 and P5, and the precision influence of each single alternative relative unit production estimation is judged according to precision evaluation indexes (R2, RMSE and MRE). The results show that the three alternative phases P2, P3 and P4 are far more accurate than the two alternative phases P1 and P5.

Then, double-alternative time phase combination analysis is carried out on the basis of P2, P3 and P4 alternative time phase combinations, modeling is carried out by sequentially utilizing all characteristic parameters in the P23, P24 and P34 alternative time phase combinations, and accuracy evaluation is carried out, and the result shows that the P24 alternative time phase combination is lifted maximally;

Adding P3 based on P24 double alternative phases, namely P234, analyzing the effect of adding P3, adding P1 based on P234, and adding P5 based on P12345, namely all alternative phases, and analyzing the effect of adding P1 and P5. The results showed no significant differences between P24, P234, P12345.

Thus, two alternative phases P2 and P4 are determined to be critical phases.

In this embodiment, the accuracy of unit yield estimation based on different single alternative phases and combinations of different alternative phases is analyzed through model comparison established based on different single alternative phases and combinations of alternative phases, so that a key phase is determined, and the amount of data for modeling can be reduced from a time dimension based on the key phase, so that the efficiency of unit yield estimation is improved.

Further, on the basis of the foregoing embodiments, the screening the key characteristic parameters for estimating the yield of the crop includes:

The growth characteristics may include NDVI, EVI, LAI, FPAR, GPP, SR (band 1-band 7) and the like, the environmental characteristics may include a moisture-related environmental characteristic (i.e., a first environmental characteristic) and a temperature-related environmental characteristic (i.e., a second environmental characteristic), wherein the moisture-related environmental characteristic may include ET, PET, precipitation, average water vapor pressure, and the like, and the temperature-related environmental characteristic may include lst_day, lst_night, solar hours, average air temperature, maximum air temperature, minimum air temperature, and the like.

In this embodiment, if the environmental characteristic parameter is added on the basis of the growth characteristic parameter, the determined model for performing unit yield estimation can improve the accuracy of unit yield estimation, the environmental characteristic parameter and the growth characteristic parameter can be used as key characteristic parameters together, otherwise, the growth characteristic parameter is only used as key characteristic parameter, so that the characteristic parameter is simplified, the data volume for modeling is reduced, and the efficiency of unit yield estimation is improved.

Further, on the basis of the above embodiments, after obtaining the primarily determined key characteristic parameters, the method further includes:

The farmland landscape characteristic parameters comprise ED, AREA, PROP and the like.

And adding the farmland landscape characteristic parameters on the basis of the preliminarily determined key characteristic parameters obtained by screening, and if the accuracy of the model established after the farmland landscape characteristic parameters are added in the single-product estimation is improved, adding the farmland landscape characteristic parameters into the preliminarily determined key characteristic parameters to obtain the finally determined key characteristic parameters.

Specifically, the characteristic parameters can be screened by the following process:

firstly modeling by utilizing growth characteristic parameters of all alternative time phases and obtaining estimated production precision;

adding all environment characteristic parameters in the alternative time phases on the basis of the growth characteristic parameters, analyzing whether the addition of the environment characteristic parameters is helpful to estimating the yield, and indicating that the improvement is not obvious;

and adding farmland landscape characteristic parameters on the basis of growth characteristic parameters, analyzing whether the addition of the farmland landscape characteristic parameters is helpful to the estimated yield precision, and the result shows that the method has a great effect on improving the estimated yield precision.

In the embodiment, the farmland landscape characteristic parameter which is beneficial to improving the unit yield estimation precision is added from the other dimension, the farmland landscape characteristic parameter does not increase a large operation data amount, and the efficiency and the accuracy of unit yield estimation can be improved.

Further, on the basis of the foregoing embodiments, the determining, by removing data redundancy, an optimal characteristic parameter combination according to the key time phase and the key characteristic parameter, and constructing a unit yield estimation model of the crop based on the optimal characteristic parameter combination includes:

The degree of correlation between the key characteristic parameters is usually determined according to experience or actual physical meaning of the key characteristic parameters, and when the degree of correlation between a plurality of key characteristic parameters is higher, only one characteristic parameter can be reserved, so that the key characteristic parameters are further simplified.

Specifically, two key time phases P2 and P4, characteristic parameters of growth vigor and characteristic parameters of farmland landscapes can be obtained through the analysis result, the estimated yield is maximally helped, the correlation degree analysis of the characteristic parameters is carried out on the basis, and one of two variables with correlation coefficients larger than 0.8 is removed;

and then, carrying out accumulated modeling analysis on the rest characteristic parameters from high to low by using random forest importance ranking, wherein the result shows that the accuracy of the estimated model tends to be stable and is not improved after the fifth variable is added. These 5 key features were finally screened (PROP, NDVI_P2, B2_P2, ED, B1_P4).

In the embodiment, the reduction of the influence parameters is further realized through the correlation degree between the key characteristic parameters, so that the efficiency and the accuracy of unit yield estimation are improved.

Further, on the basis of the above embodiments, before screening the critical phases of the crops to be subjected to the unit yield estimation, the method further includes:

For example, for wheat, the NDVI curve can be determined by using the MODIS 16-day synthetic product, and since the MODIS 16-day synthetic product generates a sampling point with a period of 16 days, the candidate phase corresponding to the peak value of the NDVI curve can be taken as P3, the phase corresponding to the peak value of the NDVI curve can be sequentially determined to the left to determine P2 and P1, and the phase corresponding to the peak value of the NDVI curve can be sequentially determined to the right to the left to determine P4 and P5.

In the embodiment, the determination of the alternative time phase of the crops is realized based on the NDVI curve, and the foundation is laid for the selection of the key time phase by dividing the alternative time phase.

As a specific embodiment, the present embodiment is classified into three types of crop growth characteristic parameters, environmental characteristic parameters (including moisture characteristics and light temperature characteristics), and farmland landscape characteristic parameters based on remote sensing data, meteorological data, and statistical data. And 5 alternative time phases (P1-P5) are divided by using the NDVI curve, then single-product estimation modeling is carried out based on different characteristic parameters or different characteristic parameter combinations under different alternative time phases, and the influence of the alternative time phases and the characteristic parameters on county level estimation is discussed by comprehensively analyzing the accuracy of the single-product estimation model on a prediction result. Finally, the optimal characteristic parameter combination is selected, characteristic parameter optimization is carried out on the basis, redundant variables are removed, and an optimal model is built for county winter wheat unit yield estimation.

The method specifically comprises the following steps:

step one: to explore the impact of phase characteristics on crop estimates, it is necessary to obtain accurate growth periods of winter wheat in different years from county to county. In the case that the observation data of the growing period in a large range are difficult to obtain accurately, the characteristic of the NDVI curve is used by many scholars to divide the growing period of winter wheat due to the fact that the characteristic can well reflect the growing condition of winter wheat. 5 key alternative phases were extracted from the curved profile of NDVI synthesized by the winter wheat growing season MODIS for 16 days, noted as P1-P5. These 5 key growth periods cover the growth process of winter wheat from green turning to maturation, and are also the most critical periods for yield impact. Wherein P3 is the vigorous growth phase of winter wheat and is represented as a peak point on the NDVI curve. P1 and P2 are respectively two points of peak point to left, and P4 and P5 are respectively two points of peak point to right.

Can be represented by the formula pk=i+16 x (k-3), wherein: k=1, 2,3,4,5.i is the DOY for the county NDVI curve for the maximum NDVI between day 65 and day 145.

Step two: the characteristic parameter set is constructed as shown in the following table:

step three: and screening and analyzing key time phase combinations of winter wheat unit yield estimation based on crop growth characteristic parameters and environmental characteristic parameters. Firstly, the estimated production precision of a single time phase is analyzed, and on the basis, precision analysis is carried out on double-time phase combination, three-time phase combination and all time phase combinations to investigate the optimal time phase combination.

Step four: on the basis of the growth characteristic parameters, gradually adding water (M), light temperature (T) and all environment characteristic parameters (ENV), and researching the influence degree of different environment characteristic parameters on the estimated production precision.

Step five: further adding farmland landscape characteristic parameters such as Edge Density (ED), winter wheat sowing AREA (AREA), winter wheat occupied AREA ratio (PROP) and the like, and deeply analyzing the influence of the addition of different farmland landscape characteristic parameters on the estimation accuracy.

Step six: and selecting the optimal characteristic parameter combination, carrying out characteristic optimization on the basis, removing redundant variables, and constructing an optimal model to carry out county-level winter wheat unit yield estimation.

It can be seen that, in this embodiment, 5 key time phases (denoted as P1-5) in the winter wheat yield formation process are extracted based on the NDVI process curve, each alternative time phase is comprehensively analyzed, and the influence of the alternative time phase combination on the unit yield estimation is determined, so that the optimal alternative time phase combination of the unit yield estimation is determined; by dividing the existing agricultural condition data into crop growth characteristic parameters, environment characteristic parameters and farmland landscape characteristic parameters, the influence of various characteristics on winter wheat county level unit yield estimation is explored by the system, and the addition of landscape indexes is found to greatly improve estimation accuracy; based on the combination analysis of different time phases and characteristic types, the variable correlation and importance analysis is further carried out, redundant variables are removed, an optimal model of winter wheat county estimated products is obtained, and data support can be provided for county estimated products rapidly and effectively.

Specifically, for the estimation of the county scale winter wheat unit yield, the estimation precision of the phase based on P2, P3 and P4 is obviously better than that of the phase combined by P1 and P5 for a single phase combined by the county scale winter wheat unit yield, wherein the phase combined by P2, P3 and P4 is not obviously different from the phase combined by the phases combined by P2, P3 and P4; the accuracy of the combination of multiple alternative time phases in the estimated yield is obviously higher than that of any single time phase, wherein the effect of the P2 and P4 combination of the alternative time phases is optimal, and the addition of other time phases has no obvious effect on improving the estimated yield precision.

The growth characteristic parameters have the greatest influence on the estimation precision of the unit yield of the county-scale winter wheat, the addition of the environmental characteristic parameters such as moisture, light temperature and the like hardly contributes to the improvement of the estimation precision, and the influence of the farmland landscape characteristic parameters on the estimation precision is not neglected, wherein the winter wheat occupies a ploughed area ratio (PROP) and the Edge Density (ED) are particularly required to be emphasized.

The most important 5 key characteristic parameters for county-level scale winter wheat unit yield estimation modeling are screened and determined, and the key characteristic parameters are respectively as follows: PROP, NDVI_P2, B2_P2, ED, B1_P4. On the basis, the model is built to realize the estimation of the unit yield of winter wheat in certain province and county, the average relative error is less than 3%, R2 reaches more than 0.8, and the root mean square error is about 250kg/ha.

Fig. 2 is a block diagram of a crop yield estimation device based on key time phase and farmland landscape characteristic parameters according to the present embodiment, and referring to fig. 2, the crop yield estimation device based on key time phase and farmland landscape characteristic parameters includes a first screening module 201, a second screening module 202 and a generating module 203, wherein,

a first screening module 201, configured to screen a critical time phase of a crop to be subjected to unit yield estimation; the key time phase is determined according to the influence of the growth characteristic parameter and the environment characteristic parameter of the crop in different alternative time phases on the unit yield estimation precision;

A second screening module 202, configured to screen key feature parameters for unit yield estimation of the crop; the key characteristic parameters are determined according to the sensitivity degree of the characteristic parameters to the influence of the unit yield estimation precision of the crops;

the generating module 203 is configured to determine an optimal characteristic parameter combination by removing data redundancy according to the key time phase and the key characteristic parameter, construct a unit yield estimation model of the crop based on the optimal characteristic parameter combination, and perform unit yield estimation on the crop through the unit yield estimation model;

The crop unit yield estimation device based on the key time phase and the farmland landscape characteristic parameter provided by the embodiment is applicable to the crop unit yield estimation method based on the key time phase and the farmland landscape characteristic parameter provided by the above embodiments, and is not described herein.

The embodiment provides a crop unit yield estimation device based on key time phases and farmland landscape characteristic parameters, wherein key time phases which are important for unit yield formation are screened out from the growth process of crops according to the influence on the accuracy of unit yield estimation, and key characteristic indexes for unit yield formation are screened out from crop growth condition characteristics, environment influence characteristics and farmland landscape characteristics. On the basis, redundant parameters are removed through correlation analysis, the optimal characteristic parameter combination is selected, and an optimal model is constructed to perform crop unit yield estimation. Based on the key time phase and characteristic parameters of screening, the efficiency of crop unit yield estimation modeling is greatly improved, and stable estimation accuracy is ensured.

determining a first model for estimating the unit yield of the crop according to the growth characteristic parameter and the environment characteristic parameter of the crop in any alternative time phase in at least one growing season and the corresponding actual unit yield of the crop, and acquiring the accuracy of the unit yield estimation of the crop by the first model as the corresponding accuracy in any time;

according to the different corresponding precision of the crops, selecting three single time phases with highest precision as screened single time phases;

determining a second model for estimating the unit yield of the crop according to the growth characteristic parameter and the environment characteristic parameter of the crop in the combination of any one time in at least one growing season and the actual unit yield corresponding to the crop aiming at the combination of any one time determined by the screened single time phase, and acquiring the accuracy of estimating the unit yield of the crop by the second model as the accuracy corresponding to the combination of any one time;

selecting the time phase combination with the minimum time phase number, which is within a preset precision range, as the screened time phase combination according to the precision corresponding to the different time phase combinations of the crops; the preset precision range is determined by precision of unit yield estimation of the crops when all the alternative list phases participate;

And if the precision corresponding to the time phase combination of the screening is higher than that of any screened single time phase, taking the time phase combination of the screening as the key time phase, otherwise, taking the single time phase corresponding to the highest precision in the screened single time phases as the key time phase.

Fig. 3 illustrates a physical schematic diagram of an electronic device, as shown in fig. 3, where the electronic device may include: processor 301, communication interface (Communications Interface) 302, memory (memory) 303 and communication bus 304, wherein processor 301, communication interface 302, memory 303 accomplish the communication between each other through communication bus 304. The processor 301 may call logic instructions in the memory 303 to perform the following method:

Further, the logic instructions in the memory 303 may be implemented in the form of software functional units and stored in a computer readable storage medium when sold or used as a stand alone product. Based on this understanding, the technical solution of the present invention may be embodied essentially or in a part contributing to the prior art or in a part of the technical solution, in the form of a software product stored in a storage medium, comprising several instructions for causing a computer device (which may be a personal computer, a server, a network device, etc.) to perform all or part of the steps of the method according to the embodiments of the present invention. And the aforementioned storage medium includes: a U-disk, a removable hard disk, a Read-Only Memory (ROM), a random access Memory (RAM, random Access Memory), a magnetic disk, or an optical disk, or other various media capable of storing program codes.

Further, embodiments of the present invention disclose a computer program product comprising a computer program stored on a non-transitory readable storage medium, the computer program comprising program instructions which, when executed by a computer, are capable of performing the methods provided by the above-described method embodiments, for example comprising:

In another aspect, an embodiment of the present invention further provides a non-transitory readable storage medium having stored thereon a computer program that is implemented when executed by a processor to perform the transmission method provided in the above embodiments, for example, including:

The apparatus embodiments described above are merely illustrative, wherein the elements illustrated as separate elements may or may not be physically separate, and the elements shown as elements may or may not be physical elements, may be located in one place, or may be distributed over a plurality of network elements. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution of this embodiment. Those of ordinary skill in the art will understand and implement the present invention without undue burden.

From the above description of the embodiments, it will be apparent to those skilled in the art that the embodiments may be implemented by means of software plus necessary general hardware platforms, or of course may be implemented by means of hardware. Based on this understanding, the foregoing technical solution may be embodied essentially or in a part contributing to the prior art in the form of a software product, which may be stored in a readable storage medium, such as a ROM/RAM, a magnetic disk, an optical disk, etc., including several instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to execute the method described in the respective embodiments or some parts of the embodiments.

Finally, it should be noted that: the above embodiments are only for illustrating the technical solution of the present invention, and are not limiting; although the invention has been described in detail with reference to the foregoing embodiments, it will be understood by those of ordinary skill in the art that: the technical scheme described in the foregoing embodiments can be modified or some technical features thereof can be replaced by equivalents; such modifications and substitutions do not depart from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. The crop unit yield estimation method based on the key time phase and the farmland landscape characteristic parameters is characterized by comprising the following steps of:

the key characteristic parameters comprise growth characteristic parameters, environment characteristic parameters and farmland landscape characteristic parameters of crops;

the screening of key phases of crops to be subjected to unit yield estimation comprises the following steps:

if the precision corresponding to the selected alternative time phase combination is higher than the precision corresponding to any selected single alternative time phase, the selected alternative time phase combination is used as the key time phase, otherwise, the single time phase corresponding to the highest precision in the selected alternative single time phase is used as the key time phase;

The determining an optimal characteristic parameter combination by eliminating data redundancy according to the key time phase and the key characteristic parameter, and constructing a unit yield estimation model of the crop based on the optimal characteristic parameter combination comprises the following steps:

2. The method for estimating unit yield of crops based on key time phase and farmland landscape characteristic parameters according to claim 1, wherein said screening key characteristic parameters for estimating unit yield of crops comprises:

3. The crop yield estimation method based on key time phase and farmland landscape characteristic parameters according to claim 2, further comprising, after obtaining the preliminary determined key characteristic parameters:

4. The method for estimating unit yield of crops based on key time phases and characteristic parameters of farmland landscapes according to claim 1, further comprising, before screening the key time phases of crops to be subjected to unit yield estimation:

5. Crop unit yield estimation device based on key time phase and farmland landscape characteristic parameters, and is characterized by comprising:

the first screening module is specifically configured to:

the generating module is specifically configured to:

6. An electronic device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, wherein the processor performs the steps of the crop yield estimation method based on key time phases and farm landscape characteristic parameters as claimed in any of claims 1 to 4.

7. A non-transitory readable storage medium having stored thereon a computer program, which when executed by a processor, implements the steps of the crop yield estimation method based on key time phases and farmland landscape characteristic parameters according to any of claims 1 to 4.