CN115081718A

CN115081718A - Method and device for predicting plant disease rate

Info

Publication number: CN115081718A
Application number: CN202210736132.5A
Authority: CN
Inventors: 肖颖欣; 董莹莹; 黄文江; 刘林毅
Original assignee: Aerospace Information Research Institute of CAS
Current assignee: Aerospace Information Research Institute of CAS
Priority date: 2022-06-27
Filing date: 2022-06-27
Publication date: 2022-09-20

Abstract

The application provides a method and a device for predicting plant disease rate, comprising the following steps: acquiring remote sensing data of target plants and vegetation coverage in the jointing stage, which correspond to a plurality of pixel areas respectively; determining initial prediction time corresponding to the pixel areas according to target plant remote sensing data corresponding to the pixel areas respectively; acquiring daily meteorological data and daily date related information corresponding to a plurality of pixel areas respectively; determining the total planting quantity, the initial infection quantity and the daily disease prevalence rate which correspond to the pixel areas respectively according to the target plant remote sensing data, the jointing vegetation coverage, the initial prediction time, the daily meteorological data and the daily date related information which correspond to the pixel areas respectively; and determining the daily plant disease rate of the target plants corresponding to the pixel areas according to the total planting number, the initial infection number, the initial prediction time and the daily disease prevalence rate corresponding to the pixel areas. The method and the device can realize the planar continuous and time continuous disease incidence prediction.

Description

Method and device for predicting plant disease rate

Technical Field

The application relates to the field of plant disease epidemiology, in particular to a method and a device for predicting plant disease rate of plants.

Background

The existing plant diseases have the characteristics of rapid onset of disease, prevention, controllability and irreversible performance, and the high-precision and high-frequency dynamic prediction of the plant diseases is favorable for accurate green prevention and control of the diseases.

Currently, domestic and foreign scholars establish an agronomic-meteorological model between meteorological factors and disease information based on years of observation data, and the model can analyze meteorological popular key factors of diseases, such as temperature, precipitation, rainy days, wind directions and the like in a key growth period, so that the time resolution of the model is improved, but the model still has the following defects: firstly, a space-time extensive scale early warning mode combining current point data driving and agricultural gas is limited by the spatial distribution condition of data, and the influence of host conditions on the occurrence and development of diseases is not considered; secondly, the current research of predicting diseases by combining meteorological data fails to analyze the epidemic process of the diseases.

Disclosure of Invention

In view of the above, the present application provides a method and an apparatus for predicting a plant disease rate, which are used to solve the above technical problems, and the technical solution is as follows:

a method for predicting the plant disease rate comprises the following steps:

acquiring remote sensing data of target plants and jointing-stage vegetation coverage corresponding to the pixel areas respectively, wherein the jointing-stage vegetation coverage refers to vegetation coverage under a remote sensing image when the target plants are in jointing stage;

determining the green turning periods of the target plants corresponding to the pixel areas respectively according to the target plant remote sensing data corresponding to the pixel areas respectively, and using the green turning periods as initial prediction time corresponding to the pixel areas respectively;

acquiring daily meteorological data and daily date related information which correspond to the multiple pixel areas respectively, wherein the daily meteorological data and the daily date related information refer to the daily meteorological data and the daily date related information in a time range formed by initial prediction time and future time to be predicted;

determining the total planting quantity, the initial infection quantity and the daily disease prevalence rate of target plants corresponding to the pixel areas according to target plant remote sensing data, jointing stage vegetation coverage, initial prediction time, daily meteorological data and daily date related information corresponding to the pixel areas, wherein the target plants are in any one of healthy, latent, infected and infection-removing states, the initial infection quantity refers to the quantity of the target plants in an infection state in initial prediction, the disease prevalence rate comprises an infection rate, a first conversion rate and a second conversion rate, the infection rate refers to the probability that the target plants in the healthy state are infected by the target plants in the infection state and become the target plants in the latent state, and the first conversion rate refers to the conversion rate when the target plants in the latent state are converted into the target plants in the infection state, the second conversion rate is the conversion rate when the target plant in the infection state is converted into the target plant in the infection removal state;

and determining the daily plant disease rate of the target plants corresponding to the pixel areas respectively according to the total planting quantity, the initial infection quantity, the initial prediction time and the daily disease prevalence rate of the target plants corresponding to the pixel areas respectively.

A plant disease rate prediction device comprising:

the first basic data acquisition module is used for acquiring target plant remote sensing data and jointing stage vegetation coverage which correspond to the pixel areas respectively, wherein the jointing stage vegetation coverage refers to vegetation coverage under a remote sensing image when the target plant is in jointing stage;

the initial prediction time determining module is used for determining the green turning periods of the target plants corresponding to the pixel areas according to the target plant remote sensing data corresponding to the pixel areas respectively, and the green turning periods are used as the initial prediction time corresponding to the pixel areas respectively;

the second basic data acquisition module is used for acquiring daily meteorological data and daily date related information which correspond to the multiple pixel areas respectively, wherein the daily meteorological data and the daily date related information refer to the daily meteorological data and the daily date related information in a time range formed by the initial prediction time and the future time to be predicted;

a process data determination module, configured to determine, according to target plant remote sensing data, jointing vegetation coverage, initial prediction time, daily meteorological data and daily date related information corresponding to a plurality of pixel areas, a total planting number, an initial infection number, and a daily disease prevalence rate of target plants corresponding to the plurality of pixel areas, where the target plants are in any one of healthy, latent, infected, and infection-removing states, the initial infection number is the number of target plants in an infection state at the time of initial prediction, the disease prevalence rate includes an infection rate, a first conversion rate, and a second conversion rate, the infection rate is a probability that a target plant in a healthy state is infected by a target plant in an infection state and becomes a target plant in a latent state, and the first conversion rate is a conversion rate when a target plant in a latent state is converted into a target plant in an infection state, the second conversion rate is the conversion rate when the target plant in the infection state is converted into the target plant in the infection removal state;

and the disease rate prediction module is used for determining the daily disease rate of the target plants corresponding to the pixel areas according to the total planting quantity, the initial infection quantity, the initial prediction time and the daily disease prevalence rate of the target plants corresponding to the pixel areas.

According to the technical scheme, the method for predicting the plant disease rate comprises the steps of firstly obtaining target plant remote sensing data and jointing stage vegetation coverage degree corresponding to a plurality of pixel areas respectively, then determining the green turning stages of target plants corresponding to the plurality of pixel areas respectively according to the target plant remote sensing data corresponding to the plurality of pixel areas respectively as initial prediction time, then obtaining daily meteorological data and daily date related information corresponding to the plurality of pixel areas respectively, then determining the total planting number, the initial infection number and the daily disease prevalence rate of the target plants corresponding to the plurality of pixel areas respectively according to the target plant remote sensing data, jointing stage vegetation coverage degree, initial prediction time, daily meteorological data and daily date related information corresponding to the plurality of pixel areas respectively, and finally determining the total planting number, the initial infection number, the daily disease prevalence rate of the target plants corresponding to the plurality of pixel areas respectively according to the total planting number of the target plants corresponding to the plurality of pixel areas respectively, And determining the daily disease rate of the target plants corresponding to the pixel areas respectively by the initial infection number, the initial prediction time and the daily disease prevalence rate. According to the method and the device, the plant disease rate of the target plants corresponding to the pixel areas is predicted by combining meteorological data and remote sensing data, planar continuous plant disease rate accurate prediction is achieved, continuous plant disease rate prediction in a time range formed by initial prediction time and future time to be predicted can be achieved, the epidemic process of diseases can be analyzed, and the prediction accuracy of the plant disease rate is improved.

Drawings

In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings needed to be used in the description of the embodiments or the prior art will be briefly introduced below, it is obvious that the drawings in the following description are only embodiments of the present application, and for those skilled in the art, other drawings can be obtained according to the provided drawings without creative efforts.

Fig. 1 is a schematic flow chart of a method for predicting a plant disease rate provided in an embodiment of the present application;

FIG. 2 is a schematic diagram of a normal distribution curve for determining jointing vegetation coverage according to an embodiment of the present disclosure;

FIG. 3a is a _RH 、β _RH And 1-gamma _RH A schematic diagram of the response curves to humidity, respectively;

FIG. 3b is a _T 、β _T And 1-gamma _T A schematic diagram of the response curves to temperature, respectively;

FIG. 3c is a _age 、β _age And 1-gamma _age A schematic of the response curves to the respective growth periods;

FIG. 4 is a graph showing the first order sensitivity and the total effect sensitivity corresponding to 25 parameters, respectively;

FIG. 5 is a schematic diagram of an optimization process for three sets of sensitivity parameter sets;

FIG. 6 is a graph of Init versus 11 vegetation indices _I Carrying out initialization training on the schematic diagram of the average RMSE obtained by the model;

FIG. 7 is a set of training set and test set basedTo RMSE and R ² A schematic diagram of (a);

FIG. 8 is a diagram showing the number of target plants in four states obtained by continuous prediction using a disease mechanism model;

fig. 9 is a schematic structural diagram of a plant disease rate prediction apparatus provided in the embodiment of the present application.

Detailed Description

The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application, and it is obvious that the described embodiments are only a part of the embodiments of the present application, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present application.

The present application provides a method for predicting a plant disease rate, which is described in detail by the following examples. Referring to fig. 1, a schematic flow chart of a plant disease rate prediction method provided in an embodiment of the present application is shown, where the plant disease rate prediction method may include:

s101, obtaining target plant remote sensing data and jointing vegetation coverage corresponding to the pixel areas respectively.

The vegetation coverage in the jointing stage refers to the vegetation coverage of the target plant in the remote sensing image in the jointing stage.

The remote sensing image of the target plant planted in the area to be predicted can be obtained, and data contained in the remote sensing image is target plant remote sensing data. It can be understood that the remote sensing image comprises a plurality of pixels, so that the region to be predicted can be divided into a plurality of pixel regions, and each pixel region corresponds to the target plant remote sensing data. For example, if the resolution of the remote sensing image is 20, the pixel area in this step may be 20 × 20m ² The area of (a).

It should be noted that the target plant remote sensing data obtained in this step is continuous remote sensing data of the target plant in the growth period range, and the target plant remote sensing data obtained in this step is the target plant remote sensing data collected last time in history. Here, the growth period of the target plant includes the following phenological periods: wintering period, green turning period, jointing period, heading period and flowering period.

For example, if the growth period of the target plant is 10/month 1/day to 6/month 1/day of the next year, and the current time is 2022/year 4/month 1/day, the remote sensing data of the target plant obtained in this step is the remote sensing data of 2021/month 10/day to 2022/year 4/month 1/day, and the remote sensing data of 2021/year 4/month 2/day to 2021/year 6/month 1.

The vegetation coverage in the jointing stage is the vegetation coverage of the target plant in the remote sensing image in the jointing stage. It should be understood by those skilled in the art that the vegetation coverage can be calculated based on the remote sensing data of the target plant, and since the vegetation coverage calculated based on the remote sensing data of the target plant exists in the prior art, the vegetation coverage in the jointing stage can be directly obtained according to the time of the jointing stage in the historical data without calculation in the step.

Optionally, the target plant can be wheat, and the step can predict the daily occurrence of the gibberellic disease of wheat (namely the disease strain rate). Of course, the target plants in this embodiment may be other plants, and this is not limited in this application.

And S102, determining the green turning periods of the target plants corresponding to the pixel areas respectively according to the target plant remote sensing data corresponding to the pixel areas respectively, and using the green turning periods as initial prediction time corresponding to the pixel areas respectively.

The step can determine the green turning period of the target plant corresponding to each pixel area according to the target value plant remote sensing data corresponding to each pixel area, and the green turning period is used as the initial prediction time corresponding to the pixel area.

In an optional embodiment, for each of the plurality of pixel areas, the process of determining the green return period of the target plant corresponding to the pixel area may include:

step a1, determining a vegetation index time sequence curve of the target plant corresponding to the pixel area according to the remote sensing data of the target plant corresponding to the pixel area.

Optionally, the Vegetation Index in this step may be an NDVI (Normalized Difference Vegetation Index), that is, the NDVI timing curve of the target plant corresponding to the pixel region may be determined, and the NDVI timing curve represents a change condition of the daily NDVI value with time.

In this step, a formula can be adopted

To determine the daily NDVI value.

Where NDVI represents the daily NDVI value, ρ _nir Representing the daily near infrared band reflectivity, p _red Indicating the daily red band reflectivity. Where ρ is _nir And ρ _red The target plant remote sensing data corresponding to the pixel area is obtained.

Step a2, smoothing the vegetation index time sequence curve of the target plant corresponding to the pixel area to obtain a smoothed vegetation index time sequence curve corresponding to the pixel area.

Optionally, in this step, a Whittaker smoothing algorithm may be used to smooth a daily NDVI timing curve of a MODIS (mode Resolution Imaging spectrometer) of the target plant corresponding to the pixel region.

And a3, fitting the smoothed vegetation index time sequence curve corresponding to the pixel area through a logistic function to obtain a fitting curve corresponding to the pixel area.

The Logistic function refers to a Logistic function, and in the step, a smoothed vegetation index time sequence curve corresponding to the pixel area can be fitted through the Logistic function to obtain a fitting curve.

Step a4, calculating the second-order curvature of the fitting curve corresponding to the pixel area, determining the maximum value point of the second-order curvature, and taking the time corresponding to the maximum value point as the green turning period of the target plant corresponding to the pixel area.

Specifically, the second-order curvature of the fitting curve can be calculated, and the maximum point of the second-order curvature is extracted and corresponds to the green turning period of the target plant.

In the step, the green turning period can be used as initial prediction time, namely, the step starts to predict from the green turning period no matter whether the current time is the green turning period of the target plant.

The green-turning period of the target plant corresponding to each pixel area can be obtained through the steps a 1-a 4, so that the green-turning periods of the target plant corresponding to a plurality of pixel areas can be obtained, and the initial prediction time corresponding to the pixel areas is obtained.

And S103, acquiring daily meteorological data and daily date related information corresponding to the multiple pixel areas respectively.

The weather data and the date-of-day related information refer to weather data and date-of-day related information in a time range formed by the initial prediction time and the future time to be predicted.

For example, the green turning period of the target plant corresponding to the pixel region is 2/1/2022/1/2022, the current time is 4/1/2022, and the present embodiment needs to predict the plant disease rate of the target plant every day from 4/2/2022/6/1/2022, then this step can obtain the weather data and the information related to the date every day from 2/1/2022/6/1/2022, wherein for the weather data every day from 2/1/2022/4/1/2022, the actual weather data can be used, and for the weather data every day from 2/4/2/2022/6/1/2022, the predicted weather data can be used.

Alternatively, the daily date-related information may be DOY (day of year) information, and the DOY information may indicate a position of the specified date in one year, for example, DOY information corresponding to 1 month and 1 day of the year is 1, DOY information corresponding to 2 months and 2 days of the year is 2, and so on.

And S104, determining the total planting quantity, the initial infection quantity and the daily disease prevalence rate of the target plants respectively corresponding to the pixel areas according to the target plant remote sensing data, the jointing vegetation coverage, the initial prediction time, the daily meteorological data and the daily date related information respectively corresponding to the pixel areas.

In the step, the target plant is in any state of healthy state, latent state, infectious state and infection removal state in the whole growth period, wherein the healthy state is marked as S state; the latent state is marked as an E state, and when the target plant is in the latent state, the target plant is infected by pathogenic bacteria, but is in the latent stage and has no infectivity; the infection state is recorded as the I state, and the target plant in the healthy state can be infected when the target plant is in the infection state; the infection removal state is denoted as an R state, and the target plant is no longer infectious when the infection removal state is achieved, for example, the target plant dies or pathogenic bacteria are artificially removed.

The initial infection number is the number of target plants in an infection state at the initial prediction time, that is, the number of target plants in an infection state at the initial prediction time, and for example, the initial infection number is the number of target plants in an infection state at the turning stage when the target plants are predicted from the turning stage in this step.

The disease prevalence rate includes an infection rate, which is a probability that a target plant in a healthy state is infected with a target plant in an infected state and becomes a target plant in a latent state, a first transition rate, which is a transition rate at which the target plant in the latent state is transformed into the target plant in the infected state, and a second transition rate, which is a transition rate at which the target plant in the infected state is transformed into the target plant in a state where the infection is removed.

And S105, determining the daily plant diseases and plant diseases of the target plants corresponding to the pixel areas according to the total planting quantity, the initial infection quantity, the initial prediction time and the daily disease prevalence rate of the target plants corresponding to the pixel areas.

It is understood that when pathogenic bacteria exist in target plants, a large amount of disease onset usually starts after a certain period (prediction is more practical), and the initial infection number and the total planting number in the initial prediction and the daily disease prevalence rate all have an influence on the disease onset of the target plants, so that the daily disease prevalence rate of the target plants corresponding to the pixel areas can be accurately determined based on the total planting number, the initial infection number, the initial prediction time and the daily disease prevalence rate of the target plants corresponding to the pixel areas.

The method for predicting the plant disease and plant rate comprises the steps of firstly obtaining target plant remote sensing data and jointing stage vegetation coverage corresponding to a plurality of pixel areas respectively, then determining the green turning periods of target plants corresponding to the pixel areas respectively according to the target plant remote sensing data corresponding to the pixel areas respectively, using the green turning periods as initial prediction time, then obtaining weather data and date-of-day related information corresponding to the pixel areas respectively, then determining the total planting number, initial infection number and disease prevalence rate of the target plants corresponding to the pixel areas respectively according to the target plant remote sensing data, jointing stage vegetation coverage, initial prediction time, weather data and date-of-day related information corresponding to the pixel areas respectively, and finally determining the total planting number, initial infection number and disease prevalence rate of the target plants corresponding to the pixel areas respectively according to the total planting number, and date-of the target plants corresponding to the pixel areas respectively, And determining the daily disease rate of the target plants corresponding to the pixel areas respectively by the initial infection number, the initial prediction time and the daily disease prevalence rate. According to the method and the device, the plant disease rate of the target plants corresponding to the pixel areas is predicted by combining meteorological data and remote sensing data, planar continuous plant disease rate accurate prediction is achieved, continuous plant disease rate prediction in a time range formed by initial prediction time and future time to be predicted can be achieved, the epidemic process of diseases can be analyzed, and the prediction accuracy of the plant disease rate is improved.

In an embodiment of the present application, a description is given of the aforementioned "step S104 of determining the total planting number, the initial infection number, and the disease prevalence rate per day" of the target plant corresponding to each of the plurality of pixel areas according to the remote sensing data, the jointing vegetation coverage, the initial prediction time, the weather data per day, and the date-per-day related information of the target plant corresponding to each of the plurality of pixel areas.

Optionally, the daily weather data includes daily temperature data and daily humidity data, and based on this, the process of step S104 may include:

and b1, determining the total planting quantity of the target plants corresponding to the pixel areas according to the jointing vegetation coverage corresponding to the pixel areas.

Specifically, the process of this step may include the following steps b11 to b 12:

and b11, determining linear regression coefficients according to the jointing vegetation coverage corresponding to the pixel areas respectively.

For convenience of description, the vegetation coverage at jointing stage is represented by FVC _jointing And (4) showing.

In this embodiment, the target plant has no ineffective tillering at the jointing stage, and no ridge is sealed in the field, and the target plant passes through the FVC corresponding to each of the plurality of pixel regions _jointing And carrying out statistics, and displaying that the statistical result shows that the normal distribution shows that the normal distribution shows that the normal distribution. Suppose 95% FVC _jointing Corresponding to the government-directed range of plant number per target plant species (e.g., 300- _jointing Two points in the corresponding normal distribution curve (see A and B in FIG. 2, and FVC with 95% hatching in FIG. 2) _jointing ) Determining the linear regression coefficient N _par1 And N _par2 。

And b12, determining the total planting quantity of the target plants corresponding to the pixel areas respectively according to the linear regression coefficient and the jointing vegetation coverage degree corresponding to the pixel areas respectively.

Specifically, in this embodiment, it is assumed that in each pixel region, the total planting number of the target plants is always a fixed value N, that is, no new target plants are planted after the prediction is started, and the number of target plants removed by the whole plant can be ignored.

For each of the plurality of pixel regions, this step may use the formula N ═ N (N) _par1 ×FVC _jointing ×N _par2 )×s ² Determining the total planting number of the target plants corresponding to the pixel region, wherein FVC _jointing Refers to the pixel regionField-corresponding jointing stage vegetation coverage, s ² Representing the area of the picture element region, e.g. s ² Can be 20 ² 。

And b2, determining the initial infection quantity of the target plant corresponding to each of the pixel areas according to the target plant remote sensing data and the daily temperature data corresponding to each of the pixel areas and the initial prediction time corresponding to each of the pixel areas.

Optionally, for each of the plurality of pixel areas, the process of determining the initial infection number of the target plant corresponding to the pixel area in this step may include the following steps b21 to b 24:

and b21, determining the green-turning vegetation index corresponding to the pixel area according to the remote sensing data of the target plants corresponding to the pixel area and the initial prediction time.

Optionally, the vegetation index of the seedling stage in this step may be an optimized soil-adjusted vegetation index (OSAVI).

Here, the calculation formula of OSAVI is:

therefore, according to the initial prediction time (namely the green turning period of the target plant), the near-infrared band reflectivity rho of the green turning period can be determined from the remote sensing data of the target plant corresponding to the pixel area _nir And red band reflectivity ρ _red Therefore, the vegetation index of the seedling stage corresponding to the pixel area can be calculated by adopting the calculation formula of the OSAVI.

Of course, the vegetation index in the striking stage in this step may be other types of vegetation indexes, which is not limited in this application.

And b22, acquiring the vegetation coverage of the green-turning stage corresponding to the pixel area according to the initial prediction time corresponding to the pixel area.

The vegetation coverage in the green turning stage refers to the vegetation coverage under the remote sensing image when the target plant is in the green turning stage.

After determining the green-turning period of the target plant corresponding to the pixel area in the previous step, the step can obtain the vegetation coverage of the green-turning period corresponding to the pixel area from the existing vegetation coverage data.

Step b23, calculating average temperature data of the target plants in the range from the wintering period to the green turning period according to the daily temperature data corresponding to the pixel area, and taking the average temperature data as the average temperature data corresponding to the pixel area.

Considering that the infection rate of pathogenic bacteria on the target plant is related to the early growth environment and host of the target plant, especially the overwintering period temperature and the growth vigor of the vegetative growth node influence the concentration of pathogenic bacteria, based on this, the step can also calculate the average temperature data of the target plant in the range from the overwintering period to the green-turning period to determine the initial infection number based on the average temperature data.

And b24, determining the initial infection quantity of the target plants respectively corresponding to the pixel areas according to the average temperature data, the green-turning vegetation index and the green-turning vegetation coverage respectively corresponding to the pixel areas.

Specifically, for each of the plurality of pixel regions, this step may adopt a formula

To determine the initial infection number of the target plant corresponding to the image element area.

In the formula, Init _I Representing the initial infection number, VI, corresponding to the pixel region _green Expressing the vegetation index, FVC, of the corresponding green-turning period of the pixel region _green Expressing the vegetation coverage, T, of the corresponding green turning period of the pixel region _overwinter And representing average temperature data corresponding to the pixel area, wherein n, m and k are constants.

And b3, determining the flowering periods of the target plants corresponding to the pixel areas respectively according to the target plant remote sensing data and the daily temperature data corresponding to the pixel areas respectively.

Optionally, for each of the multiple pixel areas, the process of determining the flowering period of the target plant corresponding to the pixel area in this step may include the following steps b31 to b 34:

and b31, determining a vegetation index time sequence curve of the target plant corresponding to the pixel area according to the remote sensing data of the target plant corresponding to the pixel area.

The process of determining the vegetation index time-series curve in this step is the same as that of the step a1, and reference may be made to the description of the step a, which is not repeated herein.

And b32, smoothing the vegetation index time sequence curve of the target plant corresponding to the pixel area to obtain a smoothed vegetation index time sequence curve corresponding to the pixel area.

The process of determining the vegetation index time-series curve in this step is the same as that of the step a2, and reference may be made to the description of the step a, which is not repeated herein.

And b33, determining a maximum value point from the smoothed vegetation index time sequence curve corresponding to the pixel area, wherein the time corresponding to the maximum value point is used as the heading date of the target plant corresponding to the pixel area.

And b34, determining the flowering period of the target plant corresponding to the image element area according to the heading period and the daily temperature data of the target plant corresponding to the image element area.

Specifically, in this step, temperature accumulation may be performed from the heading stage temperature of the target plant until a preset temperature threshold is reached, so as to obtain the flowering stage (also referred to as flowering stage) of the target plant.

The temperature threshold may be set according to actual conditions, for example, in the embodiment, the temperature threshold is 105 ℃.

And b4, determining the daily disease prevalence rate of the target plant corresponding to each of the plurality of pixel areas according to the daily temperature data, the daily humidity data and the daily date related information corresponding to each of the plurality of pixel areas and the flowering period corresponding to each of the plurality of pixel areas.

Optionally, the process of this step may include the following steps b41 to b 46:

step b41, according to the daily humidity data corresponding to the multiple pixel areas respectively and the pre-established response curves of the infection rate, the first conversion rate and the second conversion rate respectively to humidity, determining the response values of the infection rate, the first conversion rate and the second conversion rate respectively corresponding to the multiple pixel areas respectively to humidity as a daily first response value, a daily second response value and a daily third response value respectively.

Alternatively, the Logistic curve can be used to simulate the response curves of the infection rate, the first conversion rate and the second conversion rate, respectively, to humidity.

In this example, the infection rate α is to ensure a positive driving effect of humidity on disease prevalence _RH And a first conversion ratio beta _RH Are all in positive correlation with relative humidity RH, and the second conversion rate gamma _RH Inversely related to relative humidity RH, i.e. 1-gamma _RH Is positively correlated with the relative humidity RH.

In this step, the formula of the humidity response curve is: f. of _RH (x)＝1/e ^(-x+δ)/τ 。

Wherein x represents humidity data for infection rate α _RH ，f _RH (x) Representing the response curve of infection rate to humidity for a first conversion rate beta _RH ，f _RH (x) Representing the response curve of the first conversion rate to humidity for 1-gamma _RH ，f _RH (x) A response curve representing the second conversion rate versus humidity; delta and tau represent two parameters in the Logistic curve, for alpha _RH And beta _RH ，δ＝δ _αβ ，τ＝τ _αβ (ii) a For 1-gamma _RH ，δ＝δ _γ ，τ＝τ _γ 。

Referring to FIG. 3a, a is shown _RH 、β _RH And 1-gamma _RH Schematic representation of the response curves to humidity, respectively. For given delta and tau, the step can input the daily humidity data corresponding to each pixel area into the humidity response curve corresponding to the pixel area to obtain the daily infection rate-humidity response value corresponding to the pixel area as the daily first response value; similarly, a daily response value of the first conversion rate to humidity corresponding to the pixel area can be obtained as a daily second response value, and the pixel area can be obtainedAnd the response value of the second daily conversion rate to the humidity corresponding to the image element area is used as a third daily response value.

Step b42, according to the daily temperature data corresponding to the multiple pixel areas respectively and the pre-established response curves of the infection rate, the first conversion rate and the second conversion rate to the temperature respectively, determining the response values of the infection rate, the first conversion rate and the second conversion rate to the temperature respectively corresponding to the multiple pixel areas respectively as the fourth response value, the fifth response value and the sixth response value to the temperature respectively.

Optionally, a gaussian curve may be used to simulate response curves of the infection rate, the first conversion rate, and the second conversion rate to the temperature, and it is noted that in this embodiment, different variances are used on two sides of a mean value of the gaussian curve.

In this example, the infection rate α _T A first conversion rate beta _T Respectively, with respect to the relative temperature T, and a second conversion rate gamma _T The relationship with the relative temperature T is in an opposite trend, i.e. the infection rate alpha _T A first conversion rate beta _T 、1-γ _T The same trend is observed with respect to the relative temperature T.

In this step, the formula of the temperature response curve is:

wherein x represents temperature data for infection rate α _T ，f _T (x) Representing the response curve of infection rate to temperature for a first conversion rate beta _T ，f _T (x) Showing the response curve of the first conversion rate to temperature for 1-gamma _T ，f _T (x) A response curve representing the second conversion rate versus temperature; mu, sigma _T1 And σ _T2 Representing three parameters in a Gaussian curve, for alpha _T And beta _T ，μ＝μ _αβ ，σ _T1 ＝σ _{T1_αβ} ，σ _T2 ＝σ _{T2_αβ} (ii) a For 1-gamma _T ，μ＝μ _γ ，σ _T1 ＝σ _{T1_γ} ，σ _T2 ＝σ _{T2_γ} 。

See the figure3b, shows α _T 、β _T And 1-gamma _T Schematic representation of the response curves to temperature, respectively. For a given μ, σ _T1 And σ _T2 In the step, the daily temperature data corresponding to each pixel area can be input into the temperature response curve corresponding to the pixel area, and the response value of the daily infection rate to the temperature corresponding to the pixel area is obtained and used as the daily fourth response value; similarly, a response value of a first conversion rate of each day to temperature corresponding to the image element area is obtained as a fifth response value of each day, and a response value of a second conversion rate of each day to temperature corresponding to the image element area is obtained as a sixth response value of each day.

Step b43, according to the information related to the daily date and the flowering period corresponding to the plurality of pixel areas respectively, and the pre-established response curves of the infection rate, the first conversion rate and the second conversion rate to the growth period respectively, determining the response values of the infection rate, the first conversion rate and the second conversion rate to the growth period respectively corresponding to the plurality of pixel areas respectively as a seventh response value, an eighth response value and a ninth response value of each day.

Optionally, a gaussian curve may be used to simulate the response curves of the infection rate, the first conversion rate, and the second conversion rate to the growth period, and it is noted that in this embodiment, different variances are used on two sides of the mean value of the gaussian curve.

In this example, the infection rate α _age A first conversion rate beta _age Respectively, the growth period age, and a second conversion rate gamma _age The relation with the growth period age is in an opposite trend, namely the infection rate alpha _age A first conversion rate beta _age 、1-γ _age The relationship with the growth period age shows the same trend.

In this step, the formula of the growth period response curve is:

wherein, t ₀ ＝DOY _anthesis +k，DOY _anthesis Denotes flowering time, k denotes time offset, t ₀ Is shown as rawMean value under the response curve of the fertile phase; x represents date-related information (e.g., DOY) for an infection rate α _age ，f _age (x) Representing the response curve of infection rate to fertility for a first conversion rate beta _age ，f _age (x) Shows the response curve of the first conversion rate to the growth period for 1-gamma _age ，f _age (x) A response curve representing the second conversion rate to the growth period; k. sigma _A1 And σ _A2 Representing three parameters in a Gaussian curve, for alpha _age ，k＝k _α ，σ _A1 ＝σ _{A1_α} ，σ _A2 ＝σ _{A2_α} (ii) a For beta _age ，k＝k _β ，σ _A1 ＝σ _{A1_β} ，σ _A2 ＝σ _{A2_β} (ii) a For 1-gamma _age ，k＝k _γ ，σ _A1 ＝σ _{A1_γ} ，σ _A2 ＝σ _{A2_γ} 。

Referring to FIG. 3c, a is shown _age 、β _age And 1-gamma _age Graph of the response curves to the growth phase, respectively. For a given k, σ _A1 And σ _A2 In the step, the data of the growth period per day and the DOY of the flowering period corresponding to each pixel area can be obtained _anthesis Inputting the response value of the daily infection rate to the growth period corresponding to the pixel region into a growth period response curve corresponding to the pixel region, and taking the response value as a seventh daily response value; similarly, a response value of the first conversion rate of each day to the growth period corresponding to the image element area can be obtained as an eighth response value of each day, and a response value of the second conversion rate of each day to the growth period corresponding to the image element area can be obtained as a ninth response value of each day.

And b44, determining the daily infection rate of the target plants corresponding to the pixel areas respectively according to the daily first response value, the daily fourth response value and the daily seventh response value.

Optionally, for each of the plurality of pixel areas, the step may use the formula α (t) ═ α ₀ ×α _T (t)×α _RH (t)×α _age (t) determining the infection rate of the target plant corresponding to the image element area on the t day.

Wherein α (t) represents the infection rate on day t, α _RH (t) denotes the first response value, α, on day t _T (t) fourth response value, α, on day t _age (t) represents a seventh response value, α, on day t ₀ The constant represents the infection rate regulation value (optimal infection rate) in the optimal state.

And b45, determining a daily first conversion rate of the target plant corresponding to each of the plurality of image element areas according to the daily second response value, the daily fifth response value and the daily eighth response value.

Optionally, for each of the plurality of pixel areas, this step may employ a formula

And determining the first conversion rate of the t day of the target plant corresponding to the image element area.

Wherein β (t) represents the first conversion rate on day t, β _RH (t) represents a second response value, β, on day t _T (t) represents the fifth response value, β, on day t _age (t) represents an eighth response value, β, on day t ₀ Is constant and represents the first slew rate adjustment value (optimal E-I slew rate) in the optimal state.

And b46, determining a second daily conversion rate of the target plant corresponding to each of the plurality of image element areas according to the third daily response value, the sixth daily response value and the ninth daily response value.

And determining a second conversion rate of the t day of the target plant corresponding to the image element area.

Wherein γ (t) represents the second conversion rate on day t, γ _RH (t) represents a third response value, γ, on day t _T (t) represents the sixth response value on day t, γ _age (t) represents a ninth response value, γ, on day t ₀ Is constant and represents the second slew rate adjustment value in the optimum state (optimum I-R slew rate).

In summary, through the steps b41 to b46, the daily infection rate, the first conversion rate and the second conversion rate of the target plant corresponding to each of the plurality of pixel areas can be obtained, that is, the daily disease prevalence rate of the target plant corresponding to each of the plurality of pixel areas can be obtained.

In another embodiment of the present application, a process of determining the daily plant disease rate of the target plant corresponding to each of the plurality of pixel areas according to the total planting number, the initial infection number, the initial prediction time, and the daily disease prevalence rate of the target plant corresponding to each of the plurality of pixel areas in step S105 is described.

Taking any one of the multiple pixel areas as an example, a process of determining the daily plant disease rate of the target plant corresponding to the pixel area according to the total planting number, the initial infection number, the initial prediction time and the daily disease prevalence rate of the target plant corresponding to the pixel area is given below.

Optionally, the process of determining the daily plant disease rate of the target plant corresponding to the image element area may include:

and c1, determining the number of target plants in healthy, latent, infectious and infection-removing states in the pixel area every day according to the total planting number, the initial infectious number, the initial prediction time and the disease prevalence rate of the target plants corresponding to the pixel area every day.

Optionally, in this step, the number of target plants in the pixel area in healthy, latent, infectious, and infection-removing states each day may be determined based on the total planting number, initial infectious number, initial prediction time, and disease prevalence rate of the target plants corresponding to the pixel area, in combination with the following formulas (1) to (4).

In the above-mentioned formula,

indicating the change of the number of the target plants in a healthy state on the t day, which indicates the difference of the number of the target plants in a healthy state on the t day and the t +1 day;

indicating the change of the number of target plants in the latent state on the t th day, which indicates the difference of the number of target plants in the latent state on the t th day and the t +1 th day;

indicating the change of the number of the target plants in the infection state on the t th day, which indicates the difference of the number of the target plants in the infection state on the t th day and the t +1 th day;

indicating the change in the number of target plants in the infection removal state on day t, which indicates the difference between the number of target plants in the infection removal state on day t and day t + 1.

In the above formula, s (t) represents the number of target plants in a healthy state on the t-th day, e (t) represents the number of target plants in a latent state on the t-th day, i (t) represents the number of target plants in an infectious state on the t-th day, α (t) represents the infection rate on the t-th day, β (t) represents the first conversion rate on the t-th day, γ (t) represents the second conversion rate on the t-th day, and N represents the total planting number of target plants corresponding to the pixel region.

In this embodiment, after obtaining the number variation of the target plants in the healthy, latent, infectious, and infectious removal states on the t-th day, the number variation is added to the number of the target plants in the healthy, latent, infectious, and infectious removal states on the t-th day, so as to obtain the number of the target plants in the healthy, latent, infectious, and infectious removal states on the t + 1-th day.

That is, the prediction process of this step is performed in the order of days, for example, assuming that the target plants in each pixel region are in a healthy state or an infectious state at the time of initial prediction, the initial infection number Init is determined _I And obtaining the number of target plants in a healthy state corresponding to each pixel region, namely Init, after the total planting number N _S ＝N-Init _I . Assuming that the initial prediction time is 2022/1/2022, to predict the number of target plants in healthy, latent, infectious, and infection-removing states on 5/1/2022, Init is first introduced _S 、Init _I N, and the infection rate, the first conversion rate and the second conversion rate corresponding to the initial prediction time (2/1/2022) are substituted into the formulas (1) to (4), thereby obtaining the variation of the number of target plants in healthy, latent, infectious, and infection-removing states on 2/1/2022, and obtaining the number of target plants in healthy, latent, infectious, and infection-removing states on 2/2022; then substituting the number of target plants in healthy, latent, infectious and infectious removal states on 2/2022, and the infection rate, the first conversion rate and the second conversion rate corresponding to 2/2022 into the formulas (1) to (4) to obtain the number change of target plants in healthy, latent, infectious and infectious removal states on 2/2022, and thus the number of target plants in healthy, latent, infectious and infectious removal states on 2/3/2022; by analogy, the number of target plants in healthy, latent, infectious and infection-removing states respectively on 5/1/2022 is finally obtained.

And c2, determining the daily plant disease rate of the target plant corresponding to the image element area according to the number of the target plants in the health, latent, infectious and infection removal states of the image element area every day.

Optionally, in this step, the daily plant disease rate of the target plant corresponding to the pixel region may be determined by using formula (5).

In the formula, DI (t) represents the disease rate of the target plant on the t-th day.

The daily plant disease rate of the target plant corresponding to each of the plurality of pixel areas can be obtained through the steps c1 to c 2.

In another embodiment of the present application, a disease mechanism model may be constructed, so as to implement, based on the disease mechanism model, the above-mentioned step b24, "determining the initial infection number of the target plant corresponding to each of the plurality of pixel areas according to the average temperature data, the vegetation index in the green-turning period, and the vegetation coverage in the green-turning period corresponding to each of the plurality of pixel areas", "step b4," determining the daily disease prevalence rate of the target plant corresponding to each of the plurality of pixel areas according to the daily temperature data, the daily humidity data, and the daily date-related information corresponding to each of the plurality of pixel areas, in combination with the flowering period corresponding to each of the plurality of pixel areas ", and a step S105 of determining the daily plant disease rate of the target plants corresponding to the pixel areas according to the total planting number, the initial infection number, the initial prediction time and the daily disease prevalence rate of the target plants corresponding to the pixel areas.

Here, the disease mechanism model is a sesr (confidential-Exposed-infected-Removed) model, which is obtained by training a training sample including training daily temperature data, training daily humidity data, training daily date related information, training average temperature data, a training green stage vegetation index, training green stage vegetation coverage, a training flowering stage, a total planting number of training target plants, and training initial prediction time corresponding to a plurality of pixel areas, and training a training label including a disease rate corresponding to the labeled training sample.

Therefore, the process of "determining the initial infection number of the target plant corresponding to each of the plurality of pixel areas according to the average temperature data, the vegetation index at the turning stage and the vegetation coverage at the turning stage corresponding to each of the plurality of pixel areas", "step b4, determining the disease prevalence rate of the target plant corresponding to each of the plurality of pixel areas per day according to the temperature data, the humidity data and the date-per-day related information of the target plant corresponding to each of the plurality of pixel areas in combination with the flowering stage corresponding to each of the plurality of pixel areas" in step b24 may include: and processing daily temperature data, daily humidity data, daily date related information, average temperature data, green turning vegetation index, green turning vegetation coverage, flowering phase, total planting quantity of target plants and initial prediction time which correspond to the pixel areas respectively by using a disease mechanism model obtained in advance to obtain the daily plant disease rate of the target plants contained in the pixel areas output by the disease mechanism model.

In an optional embodiment, the process of obtaining the disease mechanism model (SEIR model) in this embodiment may include the following steps d1 to d 4:

and d1, acquiring task processing models corresponding to the pre-constructed vegetation indexes respectively.

Alternatively, referring to table 1 below, a plurality of vegetation indexes in this step, and their types and calculation formulas are shown.

TABLE 1 multiple vegetation indices, types and calculation formulas

In Table 1 above,. rho _λ Corresponding to the reflectivity, p, of a band containing, or in the vicinity of, the wavelength λ _nir ,ρ _red And ρ _green The reflectivity is near infrared, red wave band and green wave band. From table 1 above, a total of 11 task processing models corresponding to the vegetation indexes can be obtained in this step.

And d2, aiming at each task processing model in the task processing models respectively corresponding to the vegetation indexes, adopting a global sensitivity analysis method to analyze the sensitivity of the parameters in the task processing model so as to divide the parameters in the task processing model into a first sensitivity parameter group, a second sensitivity parameter group with the sensitivity lower than that of the first sensitivity parameter group and a third sensitivity parameter group with the sensitivity lower than that of the second sensitivity parameter group.

Wherein the parameters in the task processing model comprise parameters in response curves of the infection rate, the first conversion rate and the second conversion rate respectively to humidity, parameters in response curves of the infection rate, the first conversion rate and the second conversion rate respectively to temperature, parameters in response curves of the infection rate, the first conversion rate and the second conversion rate respectively to the growth period, and parameters in an initial infection number calculation formula.

In the embodiment, the model parameters are mainly derived from the epidemiological research results of target plant diseases, but reference values in documents cannot be directly applied to research areas due to different target plant varieties, pathogen types, research scales, research objects and the like, and have certain deviation. Therefore, it is necessary to evaluate the range of parameter values and perform parameter calibration, i.e., training of the model, based on prior knowledge. In order to successfully and effectively apply the optimization method, firstly, the sensitivity of the parameters to the output result of the model is analyzed, and the parameters are divided into a plurality of groups according to the sensitivity.

Specifically, the parameters and the related information of the parameters in each task processing model are as shown in table 2 below.

TABLE 2 parameters and information about parameters in a task processing model

As shown in table 2 above, each task processing model in the present embodiment includes 25 parameters. In this step, a global sensitivity analysis method Sobol's may be used to analyze the sensitivity of the parameters in the task processing model, and the 25 parameters are divided into three groups, for example, the high sensitivity group (the first sensitivity parameter group) includes: alpha is alpha ₀ 、β ₀ 、k _α 、σ _{A1_α} 、σ _{A2_α} 、k _β 、σ _{A1_β} 、σ _{A2_β} 、k _γ 、σ _{A1_γ} 、σ _{A2_γ} 、τ _αβ 、σ _{T1_αβ} (ii) a The medium sensitivity group (second sensitivity parameter group) includes: alpha is alpha ₀ 、β ₀ 、γ ₀ 、δ _αβ 、τ _αβ 、μ _αβ 、σ _{T1_αβ} 、σ _{T2_αβ} (ii) a The low sensitivity group (third sensitivity parameter set) includes: alpha is alpha ₀ 、β ₀ 、γ ₀ 、δ _γ 、τ _γ 、μ _γ 、σ _{T1_γ} 、σ _{T2_γ} 、n、m、k。

Optionally, the process of dividing the three parameter sets may include the following steps d21 to d 22:

step d21, inputting the training sample into the task processing model aiming at each parameter in the task processing model to obtain the sick plant rate output by the task processing model, calculating the variance of the sick plant rate output by the task processing model, and determining the first-order sensitivity, the second-order sensitivity and the total effect sensitivity corresponding to the parameter according to the parameter and the variance to obtain the first-order sensitivity, the second-order sensitivity and the total effect sensitivity corresponding to all the parameters in the task processing model respectively.

Specifically, in the field of plant disease epidemiology, the Sobol analysis has been widely used to analyze key factors affecting pathogen transmission based on the Sobol analysis SEIR model as a function:

wherein y is the model output, i.e. the disease rate DI, x _i I-th parameter, f, representing the model _ij Is x _i And x _j If d is 25, the Sobol analysis method outputs the following equations (6) to (8):

s _Ti ＝s _i +s _ij(i≠j) +…+s _1…i…d formula (8)

Wherein D (y) is the variance of y, D _i As a function f _i Variance of D _ij Is f _ij Partial variance of(s) _i Is x _i First order sensitivity of (1), reflecting the ith parameter x _i Determination of the output result, s _ij Is x _i And x _j The second-order sensitivity of (1) reflects the influence of the interaction of the two on the output result, s _Ti Is x _i Sensitivity of total effect of (1), reflecting x _i The effect of all interactions with other parameters on the results.

In this step, the plant disease rate corresponding to the training sample output by the task processing model can be obtained by using the above formulas (1) to (5), the variance of the plant disease rate is calculated, and then the first-order sensitivity, the second-order sensitivity and the total effect sensitivity corresponding to all the parameters are obtained by calculation according to the formulas (6) to (8), which is shown in fig. 4 and is a schematic diagram of the first-order sensitivity and the total effect sensitivity corresponding to 25 parameters calculated in this embodiment. For example, the Sobol's sequence method can be used to generate 100000 sets of samples, and the sensitivity of 25 parameters to dependent variables can be analyzed.

And d22, dividing all the parameters in the task processing model into a first sensitivity parameter group, a second sensitivity parameter group and a third sensitivity parameter group according to the first-order sensitivity, the second-order sensitivity and the total effect sensitivity respectively corresponding to all the parameters in the task processing model.

Specifically, the step may first divide the parameters into three groups according to the total effect sensitivity and the first order sensitivity, and then adjust the three groups of parameters in combination with the second order sensitivity and the epidemiological significance of the parameters to ensure that the parameters having high interaction or affecting the uniform driving factors are in the same group.

D3, aiming at each task processing model in the task processing models respectively corresponding to the vegetation indexes, based on the plant diseases rate corresponding to the training samples and the marked training samples, adopting a cross validation method and a sequence least square planner, and optimizing a first sensitivity parameter group, a second sensitivity parameter group and a third sensitivity parameter group in the task processing model by taking the minimum root mean square error as a target to obtain an optimized model corresponding to the task processing model; so as to obtain optimized models corresponding to the vegetation indexes respectively.

Specifically, in order to optimize model parameters, a cross-validation method and a sequence least squares planning optimizer (slsrqp) are combined to train the parameters within a variation range, where the slsrqp is one of the most robust methods for solving the nonlinear bounded parameter optimization problem.

In this embodiment, the root mean square error (RSME) between the lesion rate DI output by the model and the real DI of the sample can be selected as the minimization objective function of the slqp method. First, three sets of parameters obtained by the Sobol sensitivity analysis were adjusted in sequence, using the default values of the parameters shown in table 2 as the starting point of the slslqp optimizer. When training a set of parameters, the values of the other parameters are set to the results of the last training, and the process is repeated several times, as shown in fig. 5 (in fig. 5, "high" represents the first sensitivity parameter set, "medium" represents the second sensitivity parameter set, and "low" represents the third sensitivity parameter set). In order to avoid overfitting, a cross-validation method is adopted. The sample is madeDividing into K-fold (optionally, this embodiment takes K ═ 3), using each subset as a test set, and using the other K-1 subsets as training sets to execute the slsrqp optimizer. The RMSE for each test set was calculated by a model optimized with the corresponding training data, and the average RMSE for K experiments, which was designated as RMSE, was calculated _CV . The cross-validation method is repeated N times (optionally, in this embodiment, N is 10) to obtain N RMSECVs, and in this embodiment, in an experiment corresponding to the smallest RMSECV, a parameter obtained by training corresponding to the smallest RMSE may be determined as a final model parameter, so as to obtain an optimized model.

And d4, determining a disease mechanism model from the optimized models respectively corresponding to the vegetation indexes.

Referring to FIG. 6, the index pair Init is shown for 11 vegetation indexes _I As can be seen from the average RMSE obtained by initializing the training model, the result of OSAVI is optimal, and therefore, optionally, the optimized model corresponding to OSAVI may be determined as the disease mechanism model in this embodiment.

For evaluating the optimal model obtained by training, namely estimating InitiI by using OSAVI, calculating the obtained RMSE and the correlation coefficient R by using the optimal parameters of the model obtained by SLSFP and cross validation method ² See FIG. 7, in which (a) is the RMSE and R obtained under the training set ² And (b) RMSE and R obtained under the test set ² 。

The embodiment can also divide the sample into different levels according to the disease incidence (for example, the level 1 is DI less than or equal to 0.3, the level 2 is 0.3< DI less than or equal to 0.6, and the level 3 is DI >0.6), and obtain the classification precision.

In addition, in the present embodiment, based on the training set and the test set, the disease mechanism model is used to predict the disease rate of wheat scab, and the undetected rate (MDR) and the accuracy rate (Acc) are calculated for the prediction result, that is, the ratio of samples that are actually at a high level but are wrongly classified as a lowest level, and the results are shown in table 3.

TABLE 3 evaluation index calculation results of training set and test set

In addition, as shown in fig. 8, a schematic diagram of the number of target plants in four states obtained by continuous prediction using a disease mechanism model in this embodiment is shown, so that this embodiment can analyze the epidemic process of diseases.

By combining the above results, the disease mechanism model provided by this embodiment can simulate the disease rate with an error lower than 0.2, and has high accuracy and precision, and can analyze the epidemic process of the disease.

The embodiments of the present application further provide a device for predicting a plant disease rate, which is described below, and the device for predicting a plant disease rate described below and the method for predicting a plant disease rate described above may be referred to in correspondence with each other.

Referring to fig. 9, a schematic structural diagram of a plant disease rate prediction apparatus provided in an embodiment of the present application is shown, and as shown in fig. 9, the plant disease rate prediction apparatus may include: a first basic data acquisition module 901, an initial prediction time determination module 902, a second basic data acquisition module 903, a process data determination module 904, and a disease incidence prediction module 905.

The first basic data obtaining module 901 is configured to obtain target plant remote sensing data and jointing stage vegetation coverage corresponding to the plurality of pixel areas, respectively, where the jointing stage vegetation coverage refers to vegetation coverage of a remote sensing image when a target plant is in jointing stage.

And the initial prediction time determining module 902 is configured to determine, according to the target plant remote sensing data corresponding to the plurality of pixel areas, green turning periods of target plants corresponding to the plurality of pixel areas, as initial prediction times corresponding to the plurality of pixel areas.

The second basic data acquiring module 903 is configured to acquire daily weather data and daily date related information corresponding to the multiple pixel areas, where the daily weather data and the daily date related information refer to the daily weather data and the daily date related information in a time range formed by the initial prediction time and the future time to be predicted.

A process data determining module 904, configured to determine, according to the remote sensing data of the target plants, the jointing vegetation coverage, the initial prediction time, the daily weather data and the information related to the daily date and day corresponding to the plurality of pixel areas, a total planting number, an initial infection number and a daily disease prevalence rate of the target plants corresponding to the plurality of pixel areas, where the target plants are in any one of a healthy state, a latent state, an infection state and an infection removal state, the initial infection number refers to the number of the target plants in the infection state at the time of initial prediction, the disease prevalence rate includes an infection rate, a first conversion rate and a second conversion rate, the infection rate refers to a probability that the target plants in the healthy state are infected by the target plants in the infection state and become the target plants in the latent state, and the first conversion rate refers to a conversion rate when the target plants in the latent state are converted into the target plants in the infection state, the second conversion rate is a conversion rate at which the target plant in an infectious state is converted into the target plant in an infection removal state.

The disease rate prediction module 905 is configured to determine the daily disease rates of the target plants corresponding to the multiple pixel areas according to the total planting number, the initial infection number, the initial prediction time, and the daily disease prevalence rate of the target plants corresponding to the multiple pixel areas, respectively.

In summary, the working principle of the plant disease rate prediction apparatus disclosed in this embodiment is the same as that of the plant disease rate prediction method disclosed in the above embodiment, and is not described herein again.

Finally, it is further noted that, herein, relational terms such as, for example, and the like may be used solely to distinguish one entity or action from another entity or action without necessarily requiring or implying any actual such relationship or order between such entities or actions. Also, the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising an … …" does not exclude the presence of other identical elements in a process, method, article, or apparatus that comprises the element.

The embodiments in the present description are described in a progressive manner, each embodiment focuses on differences from other embodiments, and the same and similar parts among the embodiments are referred to each other.

The previous description of the disclosed embodiments is provided to enable any person skilled in the art to make or use the present application. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the generic principles defined herein may be applied to other embodiments without departing from the spirit or scope of the application. Thus, the present application is not intended to be limited to the embodiments shown herein but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for predicting the plant disease rate is characterized by comprising the following steps:

acquiring remote sensing data of target plants and jointing-stage vegetation coverage corresponding to a plurality of pixel areas respectively, wherein the jointing-stage vegetation coverage refers to vegetation coverage under a remote sensing image when the target plants are in jointing stage;

determining the green turning periods of the target plants corresponding to the pixel areas respectively according to the target plant remote sensing data corresponding to the pixel areas respectively, and taking the green turning periods as initial prediction time corresponding to the pixel areas respectively;

acquiring daily meteorological data and daily date related information which correspond to the multiple pixel areas respectively, wherein the daily meteorological data and the daily date related information refer to the daily meteorological data and the daily date related information in a time range formed by the initial prediction time and the future time to be predicted;

determining the total planting quantity, the initial infection quantity and the daily disease prevalence rate of target plants corresponding to the pixel areas according to the remote sensing data of the target plants, the jointing vegetation coverage, the initial prediction time, the daily meteorological data and the daily date related information corresponding to the pixel areas, wherein the target plants are in any one of a healthy state, a latent state, an infection state and an infection removal state, the initial infection quantity refers to the quantity of the target plants in the infection state in the initial prediction, the disease prevalence rate comprises an infection rate, a first conversion rate and a second conversion rate, the infection rate refers to the probability that the target plants in the healthy state are infected by the target plants in the infection state to become the target plants in the latent state, and the first conversion rate refers to the conversion rate when the target plants in the latent state are converted into the target plants in the infection state A rate at which the target plant in the infection state is transformed into the target plant in the infection removal state;

and determining the daily plant diseases of the target plants corresponding to the pixel areas according to the total planting number, the initial infection number, the initial prediction time and the daily disease prevalence rate of the target plants corresponding to the pixel areas.

2. The method of predicting the incidence of plant diseases according to claim 1, wherein the daily weather data includes daily temperature data and daily humidity data;

the determining of the total planting quantity, the initial infection quantity and the disease prevalence rate of each day of the target plants respectively corresponding to the plurality of pixel areas according to the target plant remote sensing data, the jointing vegetation coverage, the initial prediction time, the weather data of each day and the date information of each day respectively corresponding to the plurality of pixel areas comprises:

determining the total planting quantity of target plants corresponding to the pixel areas according to the jointing vegetation coverage degree corresponding to the pixel areas;

determining the initial infection quantity of the target plants corresponding to the pixel areas according to the target plant remote sensing data and the daily temperature data corresponding to the pixel areas respectively and by combining the initial prediction time corresponding to the pixel areas respectively;

determining the flowering periods of the target plants corresponding to the pixel areas respectively according to the target plant remote sensing data and the daily temperature data corresponding to the pixel areas respectively;

and determining the daily disease prevalence rate of the target plant corresponding to the pixel areas respectively according to the daily temperature data, the daily humidity data and the daily date related information corresponding to the pixel areas respectively and by combining the flowering periods corresponding to the pixel areas respectively.

3. The method for predicting the plant disease rate according to claim 2, wherein the determining the total planting number of the target plants corresponding to the plurality of pixel areas according to the jointing vegetation coverage degree corresponding to the plurality of pixel areas comprises:

determining a linear regression coefficient according to the jointing vegetation coverage degree respectively corresponding to the pixel areas;

and determining the total planting quantity of the target plants corresponding to the pixel areas according to the linear regression coefficient and the jointing vegetation coverage degree corresponding to the pixel areas.

4. The method for predicting the plant disease rate according to claim 3, wherein the determining of the initial infection number of the target plant corresponding to each of the plurality of pixel areas according to the target plant remote sensing data and the daily temperature data corresponding to each of the plurality of pixel areas in combination with the initial prediction time corresponding to each of the plurality of pixel areas comprises:

for each of the plurality of pixel areas:

determining a vegetation index of a green turning period corresponding to the pixel area according to the target plant remote sensing data and the initial prediction time corresponding to the pixel area;

acquiring the vegetation coverage of the green turning stage corresponding to the pixel area according to the initial prediction time corresponding to the pixel area, wherein the vegetation coverage of the green turning stage refers to the vegetation coverage of the remote sensing image when the target plant is in the green turning stage;

calculating average temperature data of the target plant in the range from the wintering period to the green turning period according to the daily temperature data corresponding to the pixel area, and taking the average temperature data as the average temperature data corresponding to the pixel area;

obtaining average temperature data, a green-turning vegetation index and green-turning vegetation coverage which correspond to the pixel areas respectively;

and determining the initial infection number of the target plants corresponding to the pixel areas according to the average temperature data, the vegetation index in the green turning period and the vegetation coverage in the green turning period corresponding to the pixel areas.

5. The method for predicting the plant disease rate according to claim 4, wherein the determining the daily disease prevalence rate of the target plants corresponding to the pixel areas according to the daily temperature data, the daily humidity data and the daily date related information corresponding to the pixel areas and the flowering periods corresponding to the pixel areas respectively comprises:

determining the response values of the daily infection rate, the first conversion rate and the second conversion rate respectively corresponding to the pixel areas to the humidity according to the daily humidity data respectively corresponding to the pixel areas and the pre-established response curves of the infection rate, the first conversion rate and the second conversion rate respectively corresponding to the pixel areas to the humidity, and respectively taking the response values as a first response value, a second response value and a third response value of each day;

determining response values of the daily infection rate, the first conversion rate and the second conversion rate respectively corresponding to the pixel areas to the temperature according to the daily temperature data respectively corresponding to the pixel areas and pre-established response curves of the infection rate, the first conversion rate and the second conversion rate respectively corresponding to the pixel areas to the temperature, and respectively using the response values as a fourth response value, a fifth response value and a sixth response value of each day;

according to the date-of-day related information and flowering period corresponding to the multiple pixel areas respectively and the pre-established response curves of the infection rate, the first conversion rate and the second conversion rate to the growth period respectively, determining the response values of the infection rate, the first conversion rate and the second conversion rate of the multiple pixel areas corresponding to the multiple pixel areas to the growth period respectively as a seventh response value, an eighth response value and a ninth response value of each day;

determining daily infection rates of the target plants corresponding to the multiple pixel areas respectively according to the daily first response value, the daily fourth response value and the daily seventh response value;

determining daily first conversion rates of the target plants corresponding to the pixel areas respectively according to the daily second response value, the daily fifth response value and the daily eighth response value;

and determining a second daily conversion rate of the target plant corresponding to each of the plurality of pixel areas according to the third response value, the sixth response value and the ninth response value of each day.

6. The method for predicting the plant disease rate according to claim 5, wherein the determining the daily disease rate of the target plants corresponding to the plurality of pixel areas according to the total planting number, the initial infection number, the initial prediction time and the daily disease prevalence rate of the target plants corresponding to the plurality of pixel areas respectively comprises:

for each of the plurality of pixel areas:

determining the number of target plants in the pixel area in the healthy, latent, infectious and infection removal states every day according to the total planting number, the initial infectious number, the initial prediction time and the disease prevalence rate of the target plants corresponding to the pixel area every day;

determining the daily plant disease rate of the target plants corresponding to the pixel area according to the number of the target plants in the healthy, latent, infectious and infection-removing states of the pixel area every day;

so as to obtain the daily plant disease rate of the target plants corresponding to the plurality of pixel areas respectively.

7. The method for predicting the plant disease rate according to any one of claims 4 to 6, wherein the initial infection rates of the target plants corresponding to the pixel areas are determined according to the average temperature data, the vegetation index at the turning stage and the vegetation coverage at the turning stage corresponding to the pixel areas, the daily disease prevalence rates of the target plants corresponding to the pixel areas are determined according to the daily temperature data, the daily humidity data and the daily date related information corresponding to the pixel areas, the flowering stages corresponding to the pixel areas are combined, the daily disease rate of the target plants corresponding to the pixel areas is determined according to the total planting number, the initial infection number, the initial prediction time and the daily disease prevalence rate of the target plants corresponding to the pixel areas, the method comprises the following steps:

processing daily temperature data, daily humidity data, daily date related information, average temperature data, striking stage vegetation index, striking stage vegetation coverage, flowering stage, total planting quantity and initial prediction time which correspond to the pixel areas respectively by utilizing a disease mechanism model obtained in advance to obtain daily plant disease rates of target plants contained in the pixel areas respectively and output by the disease mechanism model;

the disease mechanism model is obtained by training a training sample by taking training daily temperature data, training daily humidity data, training daily date related information, training average temperature data, training green-turning vegetation index, training green-turning vegetation coverage, training flowering time, total planting quantity of training target plants and training initial prediction time which correspond to the pixel areas respectively, and training by taking marked plant diseases corresponding to the training sample as a training label.

8. The method for predicting the plant disease rate according to claim 7, wherein the process of obtaining the disease mechanism model includes:

acquiring task processing models corresponding to a plurality of pre-constructed vegetation indexes;

aiming at each task processing model in the task processing models respectively corresponding to the vegetation indexes, analyzing the sensitivity of parameters in the task processing model by adopting a global sensitivity analysis method, to divide the parameters in the task processing model into a first sensitivity parameter set, a second sensitivity parameter set having a lower sensitivity than the first sensitivity parameter set, and a third sensitivity parameter set having a lower sensitivity than the second sensitivity parameter set, wherein the parameters in the task processing model comprise parameters in response curves of the infection rate, the first conversion rate and the second conversion rate respectively to humidity, and parameters in the response curves of the infection rate, the first conversion rate and the second conversion rate respectively to the temperature, and the infection rate, the first conversion rate and the second conversion rate are respectively parameters in a response curve of the birth period and parameters in an initial infection quantity calculation formula;

aiming at each task processing model in the task processing models respectively corresponding to the vegetation indexes, based on the training samples and the marked plant diseases rate corresponding to the training samples, optimizing the first sensitivity parameter group, the second sensitivity parameter group and the third sensitivity parameter group in the task processing model by adopting a cross validation method and a sequence least square planner and taking the minimum root mean square error as a target to obtain an optimized model corresponding to the task processing model; obtaining optimized models corresponding to the vegetation indexes respectively;

and determining the disease mechanism model from the optimized models respectively corresponding to the vegetation indexes.

9. The method of claim 8, wherein the analyzing the sensitivity of the parameters in the task processing model by the global sensitivity analysis method to divide the parameters in the task processing model into a first sensitivity parameter group, a second sensitivity parameter group with a lower sensitivity than the first sensitivity parameter group, and a third sensitivity parameter group with a lower sensitivity than the second sensitivity parameter group comprises:

inputting the training sample into the task processing model aiming at each parameter in the task processing model to obtain the sick plant rate output by the task processing model, calculating the variance of the sick plant rate output by the task processing model, and determining the first-order sensitivity, the second-order sensitivity and the total effect sensitivity corresponding to the parameter according to the parameter and the variance so as to obtain the first-order sensitivity, the second-order sensitivity and the total effect sensitivity corresponding to all the parameters in the task processing model respectively;

and dividing all the parameters in the task processing model into the first sensitivity parameter group, the second sensitivity parameter group and the third sensitivity parameter group according to the first-order sensitivity, the second-order sensitivity and the total effect sensitivity which correspond to all the parameters in the task processing model respectively.

10. A device for predicting a plant disease rate, comprising:

the first basic data acquisition module is used for acquiring target plant remote sensing data and jointing stage vegetation coverage which correspond to the plurality of pixel areas respectively, wherein the jointing stage vegetation coverage refers to vegetation coverage under a remote sensing image when a target plant is in jointing stage;

the second basic data acquisition module is used for acquiring the weather data and the date-of-day related information which correspond to the pixel areas respectively, wherein the weather data and the date-of-day related information refer to the weather data and the date-of-day related information in a time range formed by the initial prediction time and the future time to be predicted;

a process data determining module, configured to determine, according to target plant remote sensing data, jointing vegetation coverage, initial prediction time, daily meteorological data and daily date related information corresponding to the plurality of pixel areas, a total planting number, an initial infection number, and a daily disease prevalence rate of target plants corresponding to the plurality of pixel areas, where the target plants are in any one of healthy, latent, infected, and infection-removing states, the initial infection number is the number of target plants in the infection state at the initial prediction, the disease prevalence rate includes an infection rate, a first conversion rate, and a second conversion rate, and the infection rate is a probability that the target plants in the healthy state are infected by the target plants in the infection state and become the target plants in the latent state, the first conversion rate is the conversion rate when the target plant in the latent state is converted into the target plant in the infection state, and the second conversion rate is the conversion rate when the target plant in the infection state is converted into the target plant in the infection removal state;

and the disease rate prediction module is used for determining the daily disease rate of the target plants corresponding to the pixel areas according to the total planting number, the initial infection number, the initial prediction time and the daily disease prevalence rate of the target plants corresponding to the pixel areas.