CN110889547B - Crop growth period prediction method and device - Google Patents

Crop growth period prediction method and device Download PDF

Info

Publication number
CN110889547B
CN110889547B CN201911143448.8A CN201911143448A CN110889547B CN 110889547 B CN110889547 B CN 110889547B CN 201911143448 A CN201911143448 A CN 201911143448A CN 110889547 B CN110889547 B CN 110889547B
Authority
CN
China
Prior art keywords
information
day
sample
daily
period
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Active
Application number
CN201911143448.8A
Other languages
Chinese (zh)
Other versions
CN110889547A (en
Inventor
董学会
岳杨
黄岚
王忠义
范利锋
李进海
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
China Agricultural University
Original Assignee
China Agricultural University
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by China Agricultural University filed Critical China Agricultural University
Priority to CN201911143448.8A priority Critical patent/CN110889547B/en
Publication of CN110889547A publication Critical patent/CN110889547A/en
Application granted granted Critical
Publication of CN110889547B publication Critical patent/CN110889547B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06QINFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q10/00Administration; Management
    • G06Q10/04Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06QINFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q50/00Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
    • G06Q50/02Agriculture; Fishing; Forestry; Mining

Landscapes

  • Business, Economics & Management (AREA)
  • Engineering & Computer Science (AREA)
  • Strategic Management (AREA)
  • Human Resources & Organizations (AREA)
  • Economics (AREA)
  • Theoretical Computer Science (AREA)
  • Marketing (AREA)
  • General Physics & Mathematics (AREA)
  • General Business, Economics & Management (AREA)
  • Physics & Mathematics (AREA)
  • Tourism & Hospitality (AREA)
  • Animal Husbandry (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Agronomy & Crop Science (AREA)
  • Mining & Mineral Resources (AREA)
  • General Health & Medical Sciences (AREA)
  • Marine Sciences & Fisheries (AREA)
  • Health & Medical Sciences (AREA)
  • Primary Health Care (AREA)
  • Development Economics (AREA)
  • Game Theory and Decision Science (AREA)
  • Entrepreneurship & Innovation (AREA)
  • Operations Research (AREA)
  • Quality & Reliability (AREA)
  • Management, Administration, Business Operations System, And Electronic Commerce (AREA)

Abstract

The embodiment of the invention provides a crop growth period prediction method and a device, wherein the method comprises the following steps: acquiring weather factor information day by day of the last year of a forecast year; inputting the weather factor information of the last year of the forecast year into a preset weather factor forecast model to obtain the weather factor information of the last year of the forecast year; inputting the weather factor information of the forecast year day by day into a preset growth period forecasting model to obtain forecast days of the forecast year growth period; the preset meteorological factor prediction model is obtained by training sample daily meteorological factor information, and the preset growth period prediction model is obtained by training sample daily meteorological factor information and sample growth period data. A complete birth period prediction scheme is obtained by combining a preset meteorological factor prediction model capable of predicting day-by-day meteorological factor information and a preset birth period prediction model capable of predicting annual birth period prediction days.

Description

Crop growth period prediction method and device
Technical Field
The invention relates to the technical field of agricultural information, in particular to a crop growth period prediction method and device.
Background
The accurate prediction of the crop growth period plays an important role in helping agricultural workers to make a cultivation plan in advance, predict crop yield, and particularly determine the harvesting time of a machine, and the meteorological environment plays an extremely important role in the growth and development of crops, so that the crop growth period can be predicted according to the meteorological environment.
In the prior art, a growth period prediction method based on meteorological factors is mainly divided into two major categories, one category is a mechanism model for simulating the growth process of crops, and the other category is a data-driven model without a large amount of priori knowledge. Data-driven models have advanced significantly in this regard due to the ongoing development of deep learning in recent years. However, most of the two types of models need the support of annual daily meteorological factors, and the meteorological factors are time sequences which are very random and difficult to learn characteristics, so that it is very difficult to predict annual daily values of various meteorological factors, and it is also difficult to predict the crop growth period according to the annual daily values.
Therefore, how to effectively realize the prediction of the crop growth period has become an urgent problem to be solved in the industry.
Disclosure of Invention
Embodiments of the present invention provide a method and an apparatus for predicting a crop growth period, so as to solve the technical problems mentioned in the foregoing background art, or at least partially solve the technical problems mentioned in the foregoing background art.
In a first aspect, an embodiment of the present invention provides a method for predicting a crop growth period, including:
acquiring weather factor information day by day of the last year of a forecast year;
inputting the weather factor information of the last year of the forecast year into a preset weather factor forecast model to obtain the weather factor information of the last year of the forecast year;
inputting the weather factor information of the forecast year day by day and the growth period information of the previous year of the forecast year into a preset growth period forecast model to obtain forecast days of the growth period of the forecast year;
the preset meteorological factor prediction model is obtained by training sample daily meteorological factor information, and the preset growth period prediction model is obtained by training sample daily meteorological factor information and sample growth period data.
More specifically, after the step of deriving a predicted number of days for a predicted annual fertility period, the method further comprises:
and determining the crop harvesting time according to the predicted annual growth period predicted days and the predicted annual sowing date information.
More specifically, the preset meteorological factor prediction model specifically includes: the system comprises a daily average temperature prediction model, a daily accumulated precipitation prediction model and a daily sunshine duration prediction model.
More specifically, the preset growth period prediction model specifically includes: a seeding-emergence period prediction model, an emergence-jointing period prediction model, a jointing-emasculation period prediction model, an emasculation-milk maturity period prediction model and a milk maturity-mature period prediction model.
More specifically, before the step of inputting the daily weather factor information of the year previous to the predicted year into the preset weather factor prediction model, the method further comprises:
acquiring annual daily meteorological data information of a plurality of years of an area to be detected, and preprocessing the annual daily meteorological data information of the plurality of years of the area to be detected to obtain sample daily meteorological factor information, wherein the sample daily meteorological factor information comprises average relative humidity information, average air pressure information, average air temperature information, daily maximum air temperature information, daily minimum air temperature information, accumulated precipitation information, average wind speed information and sunshine duration information;
inputting the sample average relative humidity information, the sample average air pressure information, the sample average air temperature information, the sample daily maximum air temperature information, the sample daily minimum air temperature information, the sample accumulated precipitation amount information, the sample average air speed information and the sample sunshine hours information into a preset convolution long-and-short period codec model for training, obtaining a daily average temperature prediction model when a first preset training condition is met, obtaining a daily accumulated precipitation prediction model when a second preset training condition is met, and obtaining a daily sunshine hours prediction model when a third preset training condition is met;
and obtaining a preset meteorological factor prediction model according to the daily average temperature prediction model, the daily cumulative precipitation prediction model and the daily sunshine duration prediction model.
More specifically, before the step of inputting the weather factor information into a preset growth period prediction model, the method further comprises:
obtaining sowing-emergence period day-by-day meteorological data information, emergence-jointing period day-by-day meteorological data information, jointing-androgenesis period day-by-day meteorological data information, tasseling-milk maturity period day-by-day meteorological data information and milk maturity-mature period day-by-day meteorological data information according to sample year whole year day-by-day meteorological data information and sample year growth period information;
and training a back propagation neural network according to the sowing-emergence period diurnal meteorological data information, the emergence-jointing period diurnal meteorological data information, the jointing-emasculation period diurnal meteorological data information, the emasculation-milk maturity period diurnal meteorological data information and the milk maturity-maturity period diurnal meteorological data information respectively, and obtaining a preset growth period prediction model when a fourth preset training condition is met.
More specifically, the step of preprocessing the annual daily meteorological data information of the area to be detected for multiple years to obtain the daily meteorological factor information of the sample specifically includes:
and carrying out expansion processing on the annual daily meteorological data information of the areas to be detected for multiple years according to the dynamic sliding window to obtain the daily meteorological factor information of the samples.
In a second aspect, an embodiment of the present invention provides a device for predicting a crop growth period, including:
the acquisition module is used for acquiring the day-by-day meteorological factor information of the previous year of the forecast year;
the weather prediction module is used for inputting the day-by-day weather factor information of the previous year of the predicted year into a preset weather factor prediction model to obtain the day-by-day weather factor information of the predicted year;
the growth period prediction module is used for inputting the daily meteorological factor information of the predicted year and the growth period information of the previous year of the predicted year into a preset growth period prediction model to obtain the predicted days of the growth period of the predicted year;
the preset meteorological factor prediction model is obtained by training sample daily meteorological factor information, and the preset growth period prediction model is obtained by training sample daily meteorological factor information and sample growth period data.
In a third aspect, an embodiment of the present invention provides an electronic device, which includes a memory, a processor, and a computer program stored in the memory and executable on the processor, and when the processor executes the computer program, the steps of the crop growth period prediction method according to the first aspect are implemented.
In a fourth aspect, embodiments of the present invention provide a non-transitory computer-readable storage medium, on which a computer program is stored, the computer program, when executed by a processor, implementing the steps of the crop growth period prediction method according to the first aspect.
The embodiment of the invention provides a crop growth period prediction method and a device, a complete growth period prediction scheme is obtained by combining a preset meteorological factor prediction model capable of predicting daily meteorological factor information and a preset growth period prediction model capable of predicting annual growth period prediction days, and the pre-set growth period prediction model adds convolution structures to both the input to and state from the recursion, and filling the state values before convolution operation, so that the model has the capability of describing local features by a convolution neural network, has better performance when extracting the space-time characteristics, thereby realizing the prediction of weather factor information day by day, and then, the number of days of each growth period of the next year is obtained by predicting annual sowing date information, the number of days of each growth period of the previous year and the day-by-day value of each meteorological factor of the previous year, so that the crop growth period is predicted.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the embodiments or the prior art will be briefly described below, and it is obvious that the drawings in the following description are some embodiments of the present invention, and those skilled in the art can also obtain other drawings according to the drawings without creative efforts.
FIG. 1 is a schematic flow chart illustrating a method for predicting the growth period of a crop according to an embodiment of the present invention;
FIG. 2 is a diagram illustrating a structure of a layer with long and short convolution periods according to an embodiment of the present invention;
FIG. 3 is a diagram illustrating a long-short period layer structure according to an embodiment of the present invention;
fig. 4 is a schematic structural diagram of a convolutional long and short period codec model according to an embodiment of the present invention;
FIG. 5 is a diagram illustrating a BP neural network according to an embodiment of the present invention;
FIG. 6 is a schematic diagram of an apparatus for predicting the growth period of a plant according to an embodiment of the present invention;
fig. 7 is a schematic structural diagram of an electronic device according to an embodiment of the invention.
Detailed Description
In order to make the objects, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are some, but not all, embodiments of the present invention. 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 invention.
Fig. 1 is a schematic flow chart of a crop growth period prediction method according to an embodiment of the present invention, as shown in fig. 1, including:
step S1, acquiring weather factor information day by day of the last year of the forecast year;
step S2, inputting the weather factor information of the last year of the forecast year into a preset weather factor forecast model to obtain the weather factor information of the last year of the forecast year;
step S3, inputting the weather factor information of the forecast year day by day and the growth period information of the previous year of the forecast year into a preset growth period forecast model to obtain forecast days of the growth period of the forecast year;
the preset meteorological factor prediction model is obtained by training sample daily meteorological factor information, and the preset growth period prediction model is obtained by training sample daily meteorological factor information and sample growth period data.
Specifically, the predicted year described in the embodiments of the present invention refers to the year in which the growth period of the crop needs to be predicted.
The daily meteorological factor information described in the embodiment of the present invention refers to meteorological factor information for each day of the year, and the meteorological factor information may specifically include average relative humidity information, average air pressure information, average air temperature information, daily maximum air temperature information, daily minimum air temperature information, cumulative precipitation information, average wind speed information, and sunshine hours information.
The weather factor information on a daily basis in the predicted year described in the embodiments of the present invention is information on weather factors on each day in the predicted year.
The preset meteorological factor prediction model described in the embodiment of the present invention is specifically obtained by inputting daily meteorological factor information of a previous year of a predicted year into a preset convolution Long and Short period codec model for training, where a convolution Long and Short period ConvLSTM layer (ConvLSTM) is applied in the preset convolution Long and Short period encoder model described herein, and the ConvLSTM layer replaces a full connection structure of a conventional Long and Short period LSTM (Long Short-Term Memory, LSTM) with a convolution structure, specifically:
it=σ(Wxi*xt+Whi*ht-1+Wcict-1+bi
ft=σ(Wxf*xt+Whf*ht-1+Wcfct-1+bf)
ct=ftct-1+ittanh(Wxc*xt+Whc*ht-1+bc)
ot=σ(Wxo*xt+Who*ht-1+Wcoct+b0
ht=ottanh(ct);
wherein, σ represents a sigmoid function, i, f and o are respectively an input gate, a forgetting gate and an output gate, c and h respectively represent a long-term memory unit and a short-term memory unit, w represents a weight matrix, and "+" represents a convolution operator.
Specifically, fig. 2 is a structure diagram of a convolution Long-Short period layer described in an embodiment of the present invention, fig. 3 is a structure diagram of a Long-Short period layer described in an embodiment of the present invention, and as shown in fig. 2 and fig. 3, a common Long-Short-Term Memory (LSTM) is composed of a cell state, a forgetting gate, an input gate, and an output gate, and selective memorizing or deleting of information can be realized by controlling the LSTM through the gate. Meanwhile, in the recursive structure of the LSTM, the activation function is an identity function with the derivative of 1, and the phenomenon that the gradient disappears too fast or explodes when the propagation direction is reversed can be prevented. LSTM can better address long-term dependencies arising in time series prediction. The major improvement of the convolution Long and Short periods convoluting ConvLSTM (ConvLSTM) compared with the LSTM is that convolution structures are added to input states and state-to-state in the recursion, and state values are filled before convolution operation, so that the preset convolution Long and Short period codec model has the capability of describing local features, and has better performance when space-time features are extracted.
In the ConvLSTM codec, the ConvLSTM layer is used to learn and extract the internal key features c of the weather factors, and the LSTM decodes the key features c into day-by-day weather factors. Table 1 is a table of key parameter information for the ConvLSTM codec:
TABLE 1 Key parameter information Table for ConvLSTM codec
Figure BDA0002281543820000071
The input preset growth period prediction model described in the embodiment of the invention is obtained by training a Back Propagation (BP) neural network according to sample year all-year-round daily meteorological data information and sample year growth period information.
The embodiment of the invention combines a preset meteorological factor prediction model capable of predicting the daily meteorological factor information with a preset growth period prediction model capable of predicting the annual growth period prediction days to obtain a complete growth period prediction scheme, the preset growth period prediction model adds convolution structures to input states and states to states in the recursion, and fills state values before convolution operation, so that the model has the capability of describing local features by a convolution neural network and has better performance when extracting space-time features, thereby realizing the prediction of the daily meteorological factor information, and then the days of each growth period of the next year are obtained by predicting the annual sowing date information, the days of each growth period of the previous year and the daily values of each meteorological factor of the previous year to realize the prediction of the crop growth period.
On the basis of the above embodiment, after the step of obtaining the predicted annual growth period predicted days, the method further comprises:
obtaining the date of each growth period according to the predicted annual growth period days and the sowing date, and then determining the crop harvesting time.
Specifically, in the embodiment of the invention, the date suitable for seeding in the forecast year is judged according to the weather factor information day by day of the forecast year and the specific type of the crop, so as to obtain the seeding date information of the forecast year.
The number of days of the growth period of the predicted year described in the embodiment of the present invention refers to information on the number of days of different growth periods within the predicted year, and after the sowing date information of the predicted year is obtained, the information on each growth period can be determined according to the specific date of the sowing date and the number of days of each growth period determined in advance, and then the crop harvesting time can be determined.
According to the embodiment of the invention, the accurate estimation of the crop growth period is realized under the condition of calibrating and predicting the weather factor information day by day of the predicted year, so that the time suitable for crop harvesting is determined according to the sowing date information of the predicted year.
On the basis of the above embodiment, the preset meteorological factor prediction model specifically includes: the system comprises a daily average temperature prediction model, a daily accumulated precipitation prediction model and a daily sunshine duration prediction model.
Prior to the step of inputting the day-by-day meteorological factor information for the year prior to the predicted year into a preset meteorological factor prediction model, the method further comprises:
acquiring annual daily meteorological data information of a plurality of years of a to-be-detected area, and preprocessing the annual daily meteorological data information of the plurality of years of the to-be-detected area to obtain sample daily meteorological factor information, wherein the sample daily meteorological factor information comprises sample average relative humidity information, sample average air pressure information, sample average air temperature information, sample daily maximum air temperature information, sample daily minimum air temperature information, sample accumulated precipitation information, sample average wind speed information and sample sunshine hours information;
and inputting the sample average relative humidity information, the sample average air pressure information, the sample average air temperature information, the sample daily maximum air temperature information, the sample daily minimum air temperature information, the sample accumulated precipitation amount information, the sample average air speed information and the sample sunshine hours information into a preset convolution long-and-short period codec model for training, wherein the daily average temperature prediction model is obtained when a first preset training condition is met, the daily accumulated precipitation prediction model is obtained when a second preset training condition is met, and the daily sunshine hours prediction model is obtained when a third preset training condition is met.
Specifically, the meteorological factor information described in the embodiment of the present invention may specifically include average relative humidity information, average air pressure information, average air temperature information, daily maximum air temperature information, daily minimum air temperature information, cumulative precipitation information, average wind speed information, and sunshine hours information.
Therefore, according to the annual daily meteorological data information of a plurality of years of the region to be measured, carrying out abnormal value filling and (0, 1) data normalization operation on the obtained meteorological data information to obtain sample daily meteorological factor information, respectively inputting the sample average relative humidity information, the sample average air pressure information, the sample average air temperature information, the sample daily maximum air temperature information, the sample daily minimum air temperature information, the sample accumulated precipitation amount information, the sample average air speed information and the sample sunshine information into a preset convolution long-and-short period codec model for training, when a first preset training condition is met, obtaining a daily average temperature prediction model, when a second preset training condition is met, obtaining a daily accumulated precipitation prediction model, when a third preset training condition is met, obtaining a daily sunshine number prediction model.
The first preset training condition described in the embodiment of the present invention is specifically that the preset number of training rounds is 700 rounds, and the number of early stops is set to 20 rounds; the second preset training condition is specifically that the number of the preset training rounds is 500 rounds, and the number of the early-stopping rounds is set to 20 rounds; the third preset training condition is specifically that the number of training rounds is 800, and the number of early-stop rounds is 30.
The preset meteorological factor prediction model described in the embodiment of the invention specifically comprises an encoder and a decoder, wherein the encoder uses a ConvLSTM layer to input a sequence (x)1,...xT) Encoding into a fixed-length hidden state tensor c, decoding c into an output sequence (y) using LSTM as decoder1,...,yT′) And the final output result is obtained through the full connection layer.
The input required by the ConvLSTM codec is a five-dimensional vector, the specific structure of the five-dimensional vector is the total number of samples participating in model training, the daily meteorological factors in a year are divided into a plurality of time steps, the dimension of each characteristic subsequence, the length of each time step and the number of the characteristic subsequences
The encoder and decoder are connected in a way that the output of the last cell of ConvLSTM is used as a fixed-length vector c, and then the fixed-length vector c is flattened into a unary vector, and the unary vector is transmitted to the LSTM decoder after being repeated for n times, wherein the repetition time n is the same as the length of a time step in the structure.
The embodiment of the invention trains according to the annual daily meteorological data information of the area to be measured for multiple years to respectively obtain a daily average temperature prediction model, a daily cumulative precipitation prediction model and a daily cumulative precipitation prediction model, thereby forming a preset meteorological factor prediction model and effectively realizing the daily meteorological factor prediction.
On the basis of the above embodiment, the preset growth period prediction model specifically includes: a seeding-emergence period prediction model, an emergence-jointing period prediction model, a jointing-emasculation period prediction model, an emasculation-milk maturity period prediction model and a milk maturity-mature period prediction model.
Before the step of inputting the weather factor information into a preset growth period prediction model, the method further comprises:
obtaining sowing-emergence period day-by-day meteorological data information, emergence-jointing period day-by-day meteorological data information, jointing-androgenesis period day-by-day meteorological data information, tasseling-milk maturity period day-by-day meteorological data information and milk maturity-mature period day-by-day meteorological data information according to sample year whole year day-by-day meteorological data information and sample year growth period information;
and training a back propagation neural network according to the sowing-emergence period diurnal meteorological data information, the emergence-jointing period diurnal meteorological data information, the jointing-emasculation period diurnal meteorological data information, the emasculation-milk maturity period diurnal meteorological data information and the sowing milk maturity-maturity period diurnal meteorological data information respectively, and obtaining a preset growth period prediction model when a second preset training condition is met.
Specifically, the growth period described in the embodiments of the present invention includes a seeding-emergence period, a emergence-jointing period, a jointing-emasculation period, an emasculation-milk stage, a milk stage-maturation period, and a seeding-whole period prediction, and the seeding-maturation period prediction means a prediction value obtained by adding all growth periods.
Correspondingly, the encoder and decoder are connected in such a way that the output of the last cell of ConvLSTM is taken as a fixed-length vector c, and then flattened into a unary vector, and is transferred to the LSTM decoder after repeating n times, where the repetition time n is the same as the length of the time step in the structure.
Respectively training a back propagation neural network according to the sowing-emergence period diurnal meteorological data information, the emergence-emergence period diurnal meteorological data information, the jointing-emasculation period diurnal meteorological data information, the tasseling-milk maturity period diurnal meteorological data information and the milk maturity-maturity period diurnal meteorological data information, obtaining a sowing-emergence period prediction model, an emergence-emergence period prediction model, an joints-emasculation period prediction model, an tasseling-milk maturity period prediction model and a milk maturity-maturity period prediction model when a fourth preset training condition is met, thereby forming a preset growth period prediction model, and adding the prediction days of the five models to obtain the total prediction days of the total growth period of the prediction year.
On the basis of the above embodiment, the step of preprocessing the year-round day-by-day meteorological data information of the multiple years of the area to be detected to obtain the sample day-by-day meteorological factor information specifically includes:
and carrying out expansion processing on the annual daily meteorological data information of the areas to be detected for multiple years according to the dynamic sliding window to obtain the daily meteorological factor information of the samples.
Specifically, the size of the dynamic sliding window described in the embodiment of the present invention may be 365, which is the same as the number of days included in the normal year.
The dynamic sliding window in the embodiment of the invention is used for expanding the data volume, and is convenient for smooth training.
Fig. 4 is a schematic structural diagram of a convolutional long and short period codec model described in an embodiment of the present invention, as shown in fig. 4, the convolutional long and short period codec model includes a ConvLSTM layer, a vector flattening layer, a vector stacking layer, an LSTM layer, and two fully-connected layers, where activation functions of the ConvLSTM layer, the LSTM layer, and the fully-connected layers all use Scaled Explicit Linear Units (SELU) to reduce errors generated during training and improve a convergence rate of the model, and a formula thereof is shown as follows.
Figure BDA0002281543820000111
Fig. 5 is a schematic structural diagram of a BP neural network according to an embodiment of the present invention, as shown in fig. 5, wherein a specific method of input preprocessing is as follows:
for example, given that the duration days of each growth period in 2007 are 12 days, 34 days, 24 days, 41 days and 31 days, and the predicted sowing date sequence in 2008 is 123, the date sequence values in 2008 are 124 + 135 days, 136 + 169 days, 170 + 193 days, 194 + 234 days and 234 + 265 days respectively establish the input structures of 12 + 3, 34 + 3, 24 + 3, 41 + 3 and 31 + 3, wherein the first dimension means days, and the second dimension means three meteorological factors (temperature, sunlight and precipitation). At the same time, the corresponding real number of days of growth period is taken as output. Since the number of days in each growth period varies every year, based on the longest day in the growth period in many years, the sequence with the length shorter than that of the sequence is filled with 0 at the back part to ensure that the model obtains the sequence input with the same length. For example, in 1993 + 2013, the number of days of the sowing-emergence stage is 8 days at the shortest and 17 days at the longest, and the average temperature, the sunshine duration and the accumulated precipitation are used as characteristics, the longest day in the stage corresponds to the data size of 17 x 3, the shortest day corresponds to the data size of 8 x 3, 0 needs to be filled in the rear part of the data to be 17 x 3, so that the input structure of the final corresponding network is 20 x 17 3, and the output structure is 20 x 1. At the moment, a group of input and output are obtained for BP neural network training of a seeding-emergence stage, the meaning of the first dimension of input data is year, the meaning of the second dimension is days, and the third dimension is three meteorological factors. The first dimension of the corresponding output data means the year and the second dimension the corresponding number of days of the birth date.
In another embodiment of the invention, in order to verify the performance of the preset meteorological factor prediction model and the preset growth period prediction model, a neural network framework is constructed by using Python 3.5+ Keras 2.2.4. During training, an Adam algorithm with a learning rate of 0.0001 is adopted as an optimizer. The experiments were carried out in the environment of Windows10, Intercore i5-9400F 2.9Ghz, 16G RAM, NVIDIA GeForce GTX 1660.
The meteorological data of the Dandong region in 1981 and 2016 and the spring corn growth period data of 1993 and 2013 are both from a China meteorological data network. The meteorological factors used in the prediction of the next year day by day temperature, illumination and precipitation in the previous year comprise: average relative humidity, average air pressure, average air temperature, daily maximum air temperature, daily minimum air temperature, cumulative precipitation, average wind speed, and sunshine hours.
The experimental performance indexes are as follows: to verify the effectiveness of the meteorological factor model and the growth period model, the performance of different codecs was evaluated using Mean Absolute Error (MAE) and Root Mean Square Error (RMSE) and the predicted performance of each growth period model was evaluated using MAE, RMSE and R-Square (R-squared), the equations of which are shown below. The smaller the MAE and RMSE values are, the better the performance of the model is represented, and the closer the R-side value is to 1, the more accurate the model prediction is.
Figure BDA0002281543820000121
Figure BDA0002281543820000122
Figure BDA0002281543820000123
Wherein m is the total number of samples,
Figure BDA0002281543820000124
to predict value, yiIn order to be the true value of the value,
Figure BDA0002281543820000125
for sample mean, SST is the sum of the squares of the differences between the raw data and the mean, and SSE is the sum of the squares of the errors.
In this embodiment, the proposed ConvLSTM codec is compared with other common codec methods, the comparison objects include an LSTM codec model and a CNN-LSTM codec model, and table 2 shows the comparison results of temperature prediction of multiple models, as shown in table 2, taking the annual daily temperature prediction in dengton as an example, it can be seen that the ConvLSTM codec model has the best MAE and RMSE on the test set in 2014-2016, and the se reaches 2.602 ℃ and 3.456 ℃ respectively.
TABLE 2 comparison of temperature predictions for multiple models
Figure BDA0002281543820000126
This example predicts and evaluates the number of days that each growth stage of spring corn lasts in the Dandong region of 1993 and 2013. Taking the total growth period (sowing-mature period) as an example, and taking the total growth period as a prediction result table in table 3, as shown in table 3, it can be seen that the MAE and RMSE between the predicted value and the true value of the total growth period in 1993-2013 are both below 3.0d, and the R side reaches 0.790, which proves that the model has a better prediction effect.
TABLE 3 Total growth period prediction results
Figure BDA0002281543820000131
In summary, the method of the present invention uses the ConvLSTM codec to predict the average temperature per day, the cumulative rainfall per day and the sunshine duration per day of the predicted year, and based on the average temperature per day, the cumulative rainfall per day and the sunshine duration per day of the predicted year, the seeding-emergence period, the emergence-jointing period, the jointing-emasculation period, the tasking-milk maturation period, the milk maturation-maturation period and the duration of the total growth period (seeding-maturation period) of the predicted year are predicted. Experimental results show that the ConvLSTM codec is superior to the LSTM codec and the CNN-LSTM codec in MAE and RMSE, and the spring corn growth period prediction method suitable for machine harvesting provided by the invention has better performance.
Fig. 6 is a schematic structural diagram of a device for predicting the growth period of a crop according to an embodiment of the present invention, as shown in fig. 6, including: the system comprises an acquisition module 610, a weather prediction module 620 and a growth period prediction module 630, wherein the acquisition module 610 is used for acquiring day-by-day weather factor information of the last year of a predicted year; the weather prediction module 620 is configured to input the daily weather factor information of the previous year of the predicted year into a preset weather factor prediction model to obtain the daily weather factor information of the predicted year; the growth period prediction module 630 is configured to input the daily meteorological factor information of the predicted year into a preset growth period prediction model to obtain predicted days of the growth period of the predicted year; the preset meteorological factor prediction model is obtained by training sample daily meteorological factor information, and the preset growth period prediction model is obtained by training sample daily meteorological factor information and sample growth period data.
The apparatus provided in the embodiment of the present invention is used for executing the above method embodiments, and for details of the process and the details, reference is made to the above embodiments, which are not described herein again.
The embodiment of the invention combines a preset meteorological factor prediction model capable of predicting the daily meteorological factor information with a preset growth period prediction model capable of predicting the annual growth period prediction days to obtain a complete growth period prediction scheme, the preset growth period prediction model adds convolution structures to input states and states to states in the recursion, and fills state values before convolution operation, so that the model has the capability of describing local features by a convolution neural network and has better performance when extracting space-time features, thereby realizing the prediction of the daily meteorological factor information, and then the days of each growth period of the next year are obtained by predicting the annual sowing date information, the days of each growth period of the previous year and the daily values of each meteorological factor of the previous year to realize the prediction of the crop growth period.
Fig. 7 is a schematic structural diagram of an electronic device according to an embodiment of the present invention, and as shown in fig. 7, the electronic device may include: a processor (processor)710, a communication Interface (Communications Interface)720, a memory (memory)730, and a communication bus 740, wherein the processor 710, the communication Interface 720, and the memory 730 communicate with each other via the communication bus 740. Processor 710 may call logic instructions in memory 730 to perform the following method: acquiring weather factor information day by day of the last year of a forecast year; inputting the weather factor information of the last year of the forecast year into a preset weather factor forecast model to obtain the weather factor information of the last year of the forecast year; inputting the weather factor information of the forecast year day by day into a preset growth period forecasting model to obtain forecast days of the forecast year growth period; the preset meteorological factor prediction model is obtained by training sample daily meteorological factor information, and the preset growth period prediction model is obtained by training sample daily meteorological factor information and sample growth period data.
In addition, the logic instructions in the memory 730 can be implemented in the form of software functional units and stored in a computer readable storage medium when the software functional units are sold or used as independent products. Based on such understanding, the technical solution of the present invention may be embodied in the form of a software product, which is stored in a storage medium and includes instructions for causing a computer device (which may be a personal computer, a server, or a network device) to execute 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), a magnetic disk or an optical disk, and other various media capable of storing program codes.
An embodiment of the present invention discloses a computer program product, which includes a computer program stored on a non-transitory computer readable storage medium, the computer program including program instructions, when the program instructions are executed by a computer, the computer can execute the methods provided by the above method embodiments, for example, the method includes: acquiring weather factor information day by day of the last year of a forecast year; inputting the weather factor information of the last year of the forecast year into a preset weather factor forecast model to obtain the weather factor information of the last year of the forecast year; inputting the weather factor information of the forecast year day by day into a preset growth period forecasting model to obtain forecast days of the forecast year growth period; the preset meteorological factor prediction model is obtained by training sample daily meteorological factor information, and the preset growth period prediction model is obtained by training sample daily meteorological factor information and sample growth period data.
Embodiments of the present invention provide a non-transitory computer-readable storage medium storing server instructions, where the server instructions cause a computer to execute the method provided in the foregoing embodiments, for example, the method includes: acquiring weather factor information day by day of the last year of a forecast year; inputting the weather factor information of the last year of the forecast year into a preset weather factor forecast model to obtain the weather factor information of the last year of the forecast year; inputting the weather factor information of the forecast year day by day into a preset growth period forecasting model to obtain forecast days of the forecast year growth period; the preset meteorological factor prediction model is obtained by training sample daily meteorological factor information, and the preset growth period prediction model is obtained by training sample daily meteorological factor information and sample growth period data.
The above-described embodiments of the apparatus are merely illustrative, and the units described as separate parts may or may not be physically separate, and parts displayed as units may or may not be physical units, may be located in one place, or may be distributed on a plurality of network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution of the present embodiment. One of ordinary skill in the art can understand and implement it without inventive effort.
Through the above description of the embodiments, those skilled in the art will clearly understand that each embodiment can be implemented by software plus a necessary general hardware platform, and certainly can also be implemented by hardware. With this understanding in mind, the above-described technical solutions may be embodied in the form of a software product, which can be stored in a computer-readable storage medium such as ROM/RAM, magnetic disk, optical disk, etc., and includes instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to execute the methods described in the embodiments or some parts of the embodiments.
Finally, it should be noted that: the above examples are only intended to illustrate the technical solution of the present invention, but not to limit it; although the present 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 solutions described in the foregoing embodiments may still be modified, or some technical features may be equivalently replaced; and such modifications or substitutions do not depart from the spirit and scope of the corresponding technical solutions of the embodiments of the present invention.

Claims (8)

1. A method for predicting the growth period of a crop, comprising:
acquiring weather factor information day by day of the last year of a forecast year;
inputting the weather factor information of the last year of the forecast year into a preset weather factor forecast model to obtain the weather factor information of the last year of the forecast year;
inputting the weather factor information of the forecast year day by day and the growth period information of the previous year of the forecast year into a preset growth period forecast model to obtain forecast days of the growth period of the forecast year;
the preset meteorological factor prediction model is obtained by training sample daily meteorological factor information, and the preset growth period prediction model is obtained by training sample daily meteorological factor information and sample growth period data;
wherein, prior to the step of inputting the diurnal meteorological factor information for the year prior to the predicted year into a preset meteorological factor prediction model, the method further comprises:
acquiring annual daily meteorological data information of a plurality of years of a to-be-detected area, and preprocessing the annual daily meteorological data information of the plurality of years of the to-be-detected area to obtain sample daily meteorological factor information, wherein the sample daily meteorological factor information comprises sample average relative humidity information, sample average air pressure information, sample average air temperature information, sample daily maximum air temperature information, sample daily minimum air temperature information, sample accumulated precipitation information, sample average wind speed information and sample sunshine hours information;
inputting the sample average relative humidity information, the sample average air pressure information, the sample average air temperature information, the sample daily maximum air temperature information, the sample daily minimum air temperature information, the sample accumulated precipitation amount information, the sample average air speed information and the sample sunshine hours information into a preset convolution long-and-short period codec model for training, obtaining a daily average temperature prediction model when a first preset training condition is met, obtaining a daily accumulated precipitation prediction model when a second preset training condition is met, and obtaining a daily sunshine hours prediction model when a third preset training condition is met;
obtaining a preset meteorological factor prediction model according to the daily average temperature prediction model, the daily cumulative precipitation prediction model and the daily sunshine duration prediction model;
wherein, before the step of inputting the weather factor information day by day of the predicted year and the birth period information of the year previous to the predicted year into a preset birth period prediction model, the method further comprises:
obtaining sowing-emergence period day-by-day meteorological data information, emergence-jointing period day-by-day meteorological data information, jointing-androgenesis period day-by-day meteorological data information, tasseling-milk maturity period day-by-day meteorological data information and milk maturity-mature period day-by-day meteorological data information according to sample year whole year day-by-day meteorological data information and sample year growth period information;
and training a back propagation neural network according to the sowing-emergence period diurnal meteorological data information, the emergence-jointing period diurnal meteorological data information, the jointing-emasculation period diurnal meteorological data information, the emasculation-milk maturity period diurnal meteorological data information and the milk maturity-maturity period diurnal meteorological data information respectively, and obtaining a preset growth period prediction model when a fourth preset training condition is met.
2. The method of crop growth stage prediction according to claim 1, characterized in that after the step of obtaining predicted annual growth stage prediction days, the method further comprises:
and determining the crop harvesting time according to the predicted annual growth period predicted days and the predicted annual sowing date information.
3. The method for predicting the crop growth period according to claim 1, wherein the preset meteorological factor prediction model specifically comprises: the system comprises a daily average temperature prediction model, a daily accumulated precipitation prediction model and a daily sunshine duration prediction model.
4. The method for predicting the crop growth period according to claim 1, wherein the preset growth period prediction model specifically comprises: a seeding-emergence period prediction model, an emergence-jointing period prediction model, a jointing-emasculation period prediction model, an emasculation-milk maturity period prediction model, a milk maturity-maturity period prediction model and a seeding-maturity period prediction model.
5. The method for predicting the crop growth period according to claim 1, wherein the step of preprocessing the yearly daily meteorological data information of the plurality of years of the area to be detected to obtain the sample daily meteorological factor information specifically comprises:
and carrying out expansion processing on the annual daily meteorological data information of the areas to be detected for multiple years according to the dynamic sliding window to obtain the daily meteorological factor information of the samples.
6. A crop growth period prediction apparatus, comprising:
the acquisition module is used for acquiring the day-by-day meteorological factor information of the previous year of the forecast year;
the weather prediction module is used for inputting the day-by-day weather factor information of the previous year of the predicted year into a preset weather factor prediction model to obtain the day-by-day weather factor information of the predicted year;
the growth period prediction module is used for inputting the daily meteorological factor information of the predicted year and the growth period information of the previous year of the predicted year into a preset growth period prediction model to obtain the predicted days of the growth period of the predicted year;
the preset meteorological factor prediction model is obtained by training sample daily meteorological factor information, and the preset growth period prediction model is obtained by training sample daily meteorological factor information and sample growth period data;
the apparatus is further configured to:
acquiring annual daily meteorological data information of a plurality of years of a to-be-detected area, and preprocessing the annual daily meteorological data information of the plurality of years of the to-be-detected area to obtain sample daily meteorological factor information, wherein the sample daily meteorological factor information comprises sample average relative humidity information, sample average air pressure information, sample average air temperature information, sample daily maximum air temperature information, sample daily minimum air temperature information, sample accumulated precipitation information, sample average wind speed information and sample sunshine hours information;
inputting the sample average relative humidity information, the sample average air pressure information, the sample average air temperature information, the sample daily maximum air temperature information, the sample daily minimum air temperature information, the sample accumulated precipitation amount information, the sample average air speed information and the sample sunshine hours information into a preset convolution long-and-short period codec model for training, obtaining a daily average temperature prediction model when a first preset training condition is met, obtaining a daily accumulated precipitation prediction model when a second preset training condition is met, and obtaining a daily sunshine hours prediction model when a third preset training condition is met;
obtaining a preset meteorological factor prediction model according to the daily average temperature prediction model, the daily cumulative precipitation prediction model and the daily sunshine duration prediction model;
the apparatus is further configured to:
obtaining sowing-emergence period day-by-day meteorological data information, emergence-jointing period day-by-day meteorological data information, jointing-androgenesis period day-by-day meteorological data information, tasseling-milk maturity period day-by-day meteorological data information and milk maturity-mature period day-by-day meteorological data information according to sample year whole year day-by-day meteorological data information and sample year growth period information;
and training a back propagation neural network according to the sowing-emergence period diurnal meteorological data information, the emergence-jointing period diurnal meteorological data information, the jointing-emasculation period diurnal meteorological data information, the emasculation-milk maturity period diurnal meteorological data information and the milk maturity-maturity period diurnal meteorological data information respectively, and obtaining a preset growth period prediction model when a fourth preset training condition is met.
7. An electronic device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, wherein the processor executes the program to perform the steps of the method for predicting the growth period of a crop as claimed in any one of claims 1 to 5.
8. A non-transitory computer readable storage medium having stored thereon a computer program, wherein the computer program when executed by a processor implements the steps of the method for crop growth prediction according to any one of claims 1 to 5.
CN201911143448.8A 2019-11-20 2019-11-20 Crop growth period prediction method and device Active CN110889547B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201911143448.8A CN110889547B (en) 2019-11-20 2019-11-20 Crop growth period prediction method and device

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201911143448.8A CN110889547B (en) 2019-11-20 2019-11-20 Crop growth period prediction method and device

Publications (2)

Publication Number Publication Date
CN110889547A CN110889547A (en) 2020-03-17
CN110889547B true CN110889547B (en) 2022-04-15

Family

ID=69748150

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201911143448.8A Active CN110889547B (en) 2019-11-20 2019-11-20 Crop growth period prediction method and device

Country Status (1)

Country Link
CN (1) CN110889547B (en)

Families Citing this family (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN111428924A (en) * 2020-03-20 2020-07-17 北京禾水源智慧农装科技股份有限公司 Plant growth period prediction method and device, electronic equipment and storage medium
CN111915096B (en) * 2020-08-14 2021-03-09 中国科学院地理科学与资源研究所 Crop yield early-stage forecasting technology based on crop model, remote sensing data and climate forecasting information
CN112001543A (en) * 2020-08-24 2020-11-27 浙江省气候中心 Crop growth period prediction method based on ground temperature and related equipment
CN112947635B (en) * 2021-01-15 2022-04-29 江南大学 Multi-model wheat seedling growth cabin optimal parameter prediction method based on Kalman filter
CN113988376B (en) * 2021-09-29 2023-08-29 南京物链云农业科技有限公司 Rice growth period prediction method, system and device
CN113902215B (en) * 2021-11-01 2024-05-07 北京飞花科技有限公司 Method for forecasting cotton delay type cold damage dynamic state
CN114467530B (en) * 2022-01-25 2023-06-20 深圳市五谷网络科技有限公司 Fruit tree girdling method, fruit tree girdling device, terminal equipment and computer readable storage medium
CN117744861A (en) * 2023-12-08 2024-03-22 中化现代农业有限公司 Method and device for predicting physical period, electronic equipment and storage medium

Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN108665107A (en) * 2018-05-15 2018-10-16 中国农业大学 Crop yield prediction technique and system
CN109558939A (en) * 2018-10-29 2019-04-02 广东奥博信息产业股份有限公司 A kind of crop growth model selection method neural network based and device
CN110264018A (en) * 2019-07-09 2019-09-20 北京兴农丰华科技有限公司 A kind of During Growing Period of Winter Wheat prediction technique based on effective accumulated temperature over the years

Family Cites Families (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2018009482A1 (en) * 2016-07-05 2018-01-11 Sostena, Inc. System and method for crop management

Patent Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN108665107A (en) * 2018-05-15 2018-10-16 中国农业大学 Crop yield prediction technique and system
CN109558939A (en) * 2018-10-29 2019-04-02 广东奥博信息产业股份有限公司 A kind of crop growth model selection method neural network based and device
CN110264018A (en) * 2019-07-09 2019-09-20 北京兴农丰华科技有限公司 A kind of During Growing Period of Winter Wheat prediction technique based on effective accumulated temperature over the years

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
Using Deep Learning to Predict Plant Growth and Yield in Greenhouse Environments;Bashar Alhnaity;《arXiv.org》;20190701;原文第1-6页 *

Also Published As

Publication number Publication date
CN110889547A (en) 2020-03-17

Similar Documents

Publication Publication Date Title
CN110889547B (en) Crop growth period prediction method and device
CN110378435B (en) Apple leaf disease identification method based on convolutional neural network
CN107480775B (en) Pond dissolved oxygen prediction method based on data restoration
US11030528B1 (en) Convolutional neural network pruning method based on feature map sparsification
US20230128989A1 (en) Systems and methods of data preprocessing and augmentation for neural network climate forecasting models
CN111243736B (en) Survival risk assessment method and system
CN111563431A (en) Plant leaf disease and insect pest identification method based on improved convolutional neural network
CN109002917A (en) Total output of grain multidimensional time-series prediction technique based on LSTM neural network
CN111696101A (en) Light-weight solanaceae disease identification method based on SE-Inception
CN108416460B (en) Blue algae bloom prediction method based on multi-factor time sequence-random depth confidence network model
CN109886496B (en) Agricultural yield prediction method based on meteorological information
CN112070241A (en) Plant growth prediction method, device and equipment based on machine learning model
CN112686376A (en) Node representation method based on timing diagram neural network and incremental learning method
CN115829162A (en) Crop yield prediction method, device, electronic device and medium
WO2022039675A1 (en) Method and apparatus for forecasting weather, electronic device and storage medium thereof
CN110046756B (en) Short-term weather forecasting method based on wavelet denoising and Catboost
CN112884238A (en) Photovoltaic power generation power prediction method and device
CN113947182A (en) Traffic flow prediction model construction method based on double-stage stack graph convolution network
CN114694379B (en) Traffic flow prediction method and system based on self-adaptive dynamic graph convolution
CN114169505A (en) Method and system for training yield prediction model and related equipment
CN116844041A (en) Cultivated land extraction method based on bidirectional convolution time self-attention mechanism
CN114942951A (en) Fishing vessel fishing behavior analysis method based on AIS data
CN114841412A (en) Method for predicting pH value of sea cucumber growing water
CN114881286A (en) Short-time rainfall prediction method based on deep learning
CN117665825A (en) Radar echo extrapolation prediction method, system and storage medium

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant