CN105892287A - Crop irrigation strategy based on fuzzy judgment and decision making system - Google Patents

Abstract

The present invention relates to a crop irrigation strategy and decision-making system based on fuzzy judgment. The crop irrigation strategy includes the following steps: step S1, establishing a generation system knowledge base according to sample data; and step S2, according to the current growth environment and growth stage data of crops, Determine the current water demand for crop irrigation through fuzzy judgment; the crop irrigation strategy and crop irrigation decision-making system of the present invention can effectively establish a generation system knowledge base based on sample data to analyze the current growth environment and growth stage data of crops, and then obtain accurate irrigation water Compared with the traditional method based on soil temperature and humidity conditions, comprehensive consideration of different soil properties, growth environments, weather conditions, and different growth stages of crops can more directly reflect the moisture status of crops, and the growth environment and growth conditions of crops can be used together as crops. The basis for water demand decision-making can increase the accuracy of judgment and effectively improve the utilization rate of water resources.

Description

Crop Irrigation Strategy and Decision System Based on Fuzzy Judgment

technical field

The invention relates to pattern recognition and water-saving irrigation, especially in the crop irrigation decision system based on fuzzy judgment.

Background technique

The shortage of water resources is a major problem faced by countries all over the world. The traditional surface irrigation technology is widely used in agricultural irrigation in my country. The irrigation water utilization coefficient is lower than 50%, and the phenomenon of water waste is serious. The implementation of precision irrigation based on the characteristics of crop water demand information is the key to agricultural water saving. efficient way of irrigation. As an industry combining high-tech and agricultural production, precision irrigation precisely controls the timing, water volume, and method of irrigation according to actual conditions such as soil, climate, irrigation facilities, and crop growth, so that crop water demand can reach the best state and water resources can be realized. Efficient utilization to achieve the best economic and environmental benefits.

At present, most irrigation decisions for crops are based on soil temperature and humidity, which cannot accurately reflect the actual water demand of crops.

Therefore, in view of the above problems, it is necessary to design a crop irrigation strategy and a crop irrigation decision-making system that can accurately control the amount of irrigation to ensure better growth of crops.

Contents of the invention

The purpose of the present invention is to provide a crop irrigation strategy and a crop irrigation decision-making system to achieve the purpose of precise irrigation of crops.

In order to solve the above technical problems, the invention provides a crop irrigation strategy, comprising the steps of:

Step S1, establishing a generation system knowledge base according to the sample data;

Step S2, according to the current growth environment and growth stage data of the crops, determine the current water demand for crop irrigation through fuzzy judgment.

Further, the method for establishing and generating the system knowledge base according to the sample data in the step S1 includes:

Step S11, collecting crop growth environment and water demand data information at each growth stage as sample data, and establishing a crop water demand correlation database to construct a fuzzy information system;

Step S12, performing attribute reduction in the fuzzy information system to generate probabilistic decision rules; and

Step S13, using the neural network to train the reduction to generate the system knowledge base.

Further, in the step S2, according to the current growth environment and growth stage data of the crops, the water demand for the current crop irrigation is determined through fuzzy judgment, namely

According to the data of the current growth environment and growth stage of the crops, the water demand of the current crops is determined through the fuzzy judgment of the neural network.

Further, in the step S11, the method of collecting crop growth environment and water demand data information at each growth stage, establishing a crop water demand correlation database, and constructing a fuzzy information system includes the following steps:

Step S111, establishing the crop growth environment factor sample set U, the water demand condition attribute set C and the water demand decision attribute set V;

Step S112, said fuzzy information system construction, namely

The fuzzy information system is a quadruple: G={U, A, V, f}; where

U is the domain of discourse, which is the collection of all crop sample information;

A is an attribute set, that is, A=C∪D, where: C represents a conditional attribute, that is, a set composed of soil moisture, soil temperature, and growth stage attributes of crops; D represents a decision attribute, that is, a set composed of crop water demand attributes;

V is the union of attribute value ranges, V=∪V _a , V _a represents the value range of attribute a∈A, that is, the union of the value range sets of soil moisture, growth stage and water demand attributes;

f represents the U×A→V information function, and assigns attribute values to each object in the domain of discourse, namely x∈U, there is f(x, a)∈V _a .

Further, in the step S1, attribute reduction will be performed in the fuzzy information system, namely

Judging attribute similarity, reducing conditional attributes, and obtaining reduced conditional attribute set C ₁ , the method includes the following steps:

Step S121, define attribute similarity Among them: X, Y ∈ A, X and Y are any two attributes in the domain of discourse, that is, A is the union of conditional attributes and decision attributes; ind( ) is the equivalent division class of domain U about attribute ; n is the number of equivalence division classes;

Step S122, calculating the similarity S(c _i , D) between the condition attribute and the decision attribute among the attributes, where c _i , c _j ∈ C, i=1...m, i≠j, m is the number of condition attributes ;According to the value of similarity S(c _i , D), arrange the condition attributes in descending order C={c ₁ ,c ₂ ,...,c _m }; where

S(c ₁ , D)≥S(c ₂ , D)≥…≥S(c _m , D);

Step S123, define the reduced attribute set C ₁ =Φ, and calculate the similarity S(ci,c _j ) between two conditional attributes in the attribute, where c _i ,c _j ∈C,i, _j = 1...m, i≠j, i<j;

if S(c _i , c _j )≥S(c _i , D) then C ₁ =C c _j ; and

Calculate S(C ₁ , D), if S(C ₁ , D)=S(C, D) then C ₁ =Cc _j , else C ₁ =C;

Step S124, until the traversal of each condition attribute is completed;

In step S125, the reduction condition attribute set C ₁ is obtained.

Further, the method for using the neural network to train the reduction in the step S13 includes the following steps:

Step S131, sending the conditionally reduced sample data into the input layer of the probabilistic neural network, and the number of neuron nodes is equal to the dimension of the conditional attribute set;

Step S133, setting the number of neuron nodes in the hidden layer to be equal to the number of samples, and the neuron uses a Gaussian kernel function to determine the input-output relationship:

${\mathrm{<span\; class="notranslate">\; \&Phi;</span>}}_{\mathrm{<span\; class="notranslate">\; t</span>}\mathrm{<span\; class="notranslate">\; k</span>}}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>\frac{\mathrm{<span\; class="notranslate">\; 1</span>}}{{<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 2</span>}\mathrm{<span\; class="notranslate">\; \&pi;</span>}<span\; class="notranslate">\; )</span>}^{\frac{\mathrm{<span\; class="notranslate">\; d</span>}}{\mathrm{<span\; class="notranslate">\; 2</span>}}}}\mathrm{<span\; class="notranslate">\; exp</span>}<span\; class="notranslate">\; \&lsqb;</span><span\; class="notranslate">\; -</span>\frac{<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; -</span>{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; t</span>}\mathrm{<span\; class="notranslate">\; k</span>}}<span\; class="notranslate">\; )</span>{<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; -</span>{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; t</span>}\mathrm{<span\; class="notranslate">\; k</span>}}<span\; class="notranslate">\; )</span>}^{\mathrm{<span\; class="notranslate">\; T</span>}}}{{\mathrm{<span\; class="notranslate">\; \&sigma;</span>}}^{\mathrm{<span\; class="notranslate">\; 2</span>}}}<span\; class="notranslate">\; \&rsqb;</span>$

Where t=1,...M,k=1,...N _t ; x is the input vector that needs to make a decision, with d attributes; the value of d is the conditional attribute dimension obtained after reduction; x _tk is the hidden layer Vector, which is the k-th neuron of the t-th class sample in the hidden layer; M is the total number of training samples, which is equal to the dimension of the decision attribute; Nt is the number of samples of the t-th class; and σ is the smoothing factor, σ ∈(0,∞);

In step S133, the weighted layer performs weighted average on the outputs of hidden neurons belonging to the same class:

${\mathrm{<span\; class="notranslate">\; v</span>}}_{\mathrm{<span\; class="notranslate">\; t</span>}}<span\; class="notranslate">\; =</span>\frac{\mathrm{<span\; class="notranslate">\; 1</span>}}{{\mathrm{<span\; class="notranslate">\; N</span>}}_{\mathrm{<span\; class="notranslate">\; t</span>}}}\underset{\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\overset{{\mathrm{<span\; class="notranslate">\; N</span>}}_{\mathrm{<span\; class="notranslate">\; t</span>}}}{<span\; class="notranslate">\; \&Sigma;</span>}}{\mathrm{<span\; class="notranslate">\; \&Phi;</span>}}_{\mathrm{<span\; class="notranslate">\; t</span>}\mathrm{<span\; class="notranslate">\; k</span>}}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; )</span>$

where t=1,...M;

Step S134, calculating the category with the maximum posterior probability density value ρ(x)=argmax(t) in all neuron outputs as the output result of the neural network;

Step S135, taking the output result of the neural network as the decision value of soil water demand, and constructing a decision table.

In yet another aspect, the present invention also provides a crop irrigation decision-making system.

The crop irrigation decision-making system includes: a collection module for collecting data on the current growth environment and growth stage of the crops, and a server connected to the corresponding data collection module of the crops; the server is suitable for adjusting water consumption for crop irrigation.

Further, the server is adapted to use the crop irrigation strategy to regulate water consumption.

The beneficial effect of the present invention is that the crop irrigation strategy and the crop irrigation decision-making system of the present invention can effectively establish a generation system knowledge base according to the sample data to analyze the current growth environment and growth stage data of the crops, and then obtain accurate water consumption for irrigation. Compared with the soil temperature and humidity conditions, comprehensive consideration of different soil quality, growth environment, weather conditions and different growth stages of crops can more directly reflect the water status of crops, and the growth environment and growth conditions of crops can be used as the basis for crop water demand decision-making , can increase the judgment accuracy and effectively improve the utilization rate of water resources.

Description of drawings

The present invention will be further described below in conjunction with the accompanying drawings and embodiments.

Fig. 1 is the block flow diagram of crop irrigation strategy of the present invention;

Fig. 2 is a functional block diagram of the crop irrigation decision-making system of the present invention.

detailed description

The present invention is described in further detail now in conjunction with accompanying drawing. These drawings are all simplified schematic diagrams, which only illustrate the basic structure of the present invention in a schematic manner, so they only show the configurations related to the present invention.

Example 1

As shown in Figure 1, the present embodiment provides a crop irrigation strategy, including the following steps:

Step S1, establishing a generative system knowledge base according to the sample data; and

Step S2, according to the current growth environment and growth stage data of the crops, determine the current water demand for crop irrigation through fuzzy judgment.

Wherein, the sample data includes but not limited to crop growth environment data such as crop growth soil type, soil humidity, soil temperature, air relative humidity, air temperature, atmospheric light intensity, wind speed, and sunshine hours.

Specifically, the method for establishing and generating the system knowledge base according to the sample data in the step S1 includes:

Step S11, collecting crop growth environment and water demand data information at each growth stage as sample data, and establishing a crop water demand correlation database to construct a fuzzy information system;

Step S12, performing attribute reduction in the fuzzy information system to generate probabilistic decision rules; and

Step S13, using the neural network to train the reduction to generate the system knowledge base.

Specifically, in the step S2, according to the current growth environment and growth stage data of the crops, the current water demand for crop irrigation is determined through fuzzy judgment, namely

According to the data of the current growth environment and growth stage of the crops, the water demand of the current crops is determined through the fuzzy judgment of the neural network.

Specifically, in the step S11, the method of collecting crop growth environment and water demand data information at each growth stage, and establishing a crop water demand correlation database to construct a fuzzy information system includes the following steps:

Step S111, establishing the crop growth environment factor sample set U, the water demand condition attribute set C and the water demand decision attribute set V;

Step S112, said fuzzy information system construction, namely

The fuzzy information system is a quadruple: G={U, A, V, f}; where

U is the domain of discourse, which is the collection of all crop sample information;

A is an attribute set, that is, A=C∪D, where: C represents a conditional attribute, that is, a set composed of soil moisture, soil temperature, and growth stage attributes of crops; D represents a decision attribute, that is, a set composed of crop water demand attributes;

V is the union of attribute value ranges, V=∪V _a , V _a represents the value range of attribute a∈A, that is, the union of the value range sets of soil moisture, growth stage and water demand attributes, and a represents a certain conditional attribute ;

f represents the U×A→V information function, and assigns attribute values to each object in the domain of discourse, namely x∈U, there is f(x, a)∈V _a . x represents an object in the domain of discourse U, specifically a sample or an object to be decided. f assigns an information value to each attribute of each object in the domain of discourse.

Specifically, in the step S1, attribute reduction will be performed in the fuzzy information system, namely

Judging attribute similarity, reducing conditional attributes, and obtaining reduced conditional attribute set C ₁ , the method includes the following steps:

Step S121, define attribute similarity Among them: X, Y∈A, X and Y are any two attributes in the domain of discourse, that is, A is the union of conditional attributes and decision attributes, where X and Y can both be conditional attributes or decision attributes; ind( ) is the equivalence division class of domain U with respect to attribute ; n is the number of equivalence division classes;

Step S122, calculate the similarity S(c _i , D) between the condition attribute and the decision attribute among the attributes, where c _i , c _j ∈ C, i=1...m, i≠j, m is the number of condition attributes ;According to the value of similarity S(c _i , D), arrange the condition attributes in descending order C={c ₁ ,c ₂ ,...c _m }; where

S(c ₁ , D)≥S(c ₂ , D)≥…≥S(c _m , D);

Step S123, define the reduced attribute set C ₁ =Φ, and calculate the similarity S(ci,c _j ) between two conditional attributes in the attribute, where c _i ,c _j ∈C,i, _j = 1...m, i≠j, i<j;

if S(c _i , c _j )≥S(c _i , D) then C ₁ =C c _j ; and

Calculate S(C ₁ , D), if S(C ₁ , D)=S(C,D) then C ₁ =Cc _j , else C ₁ =C;

Specifically, after the condition attributes are sorted in descending order according to the similarity between the condition attributes and the decision attributes, the similarity between the two condition attributes is compared. If the similarity between two conditional attributes is greater than the similarity between a conditional attribute and a decision attribute, it is considered that there is a high possibility of redundancy between the two conditional attributes, and the relationship between the conditional attribute and the decision attribute The condition attribute with smaller similarity is regarded as redundant attribute, and the remaining condition attributes are regarded as reduced condition attributes. If the similarity between the conditional attributes after the reduction and the conditional attributes before the reduction is equal, then the redundancy of the redundant attributes is considered unnecessary. Otherwise, the reduced condition attribute is still not the original condition attribute.

Step S124, until the traversal of each condition attribute is completed;

In step S125, the reduction condition attribute set C ₁ is obtained.

Specifically, in the step S13, the method for using the neural network to train the reduction includes the following steps:

Step S131, sending the conditionally reduced sample data into the input layer of the probabilistic neural network, and the number of neuron nodes is equal to the dimension of the conditional attribute set;

Step S133, setting the number of neuron nodes in the hidden layer to be equal to the number of samples, and the neuron uses a Gaussian kernel function to determine the input-output relationship:

${\mathrm{<span\; class="notranslate">\; \&Phi;</span>}}_{\mathrm{<span\; class="notranslate">\; t</span>}\mathrm{<span\; class="notranslate">\; k</span>}}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>\frac{\mathrm{<span\; class="notranslate">\; 1</span>}}{{<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 2</span>}\mathrm{<span\; class="notranslate">\; \&pi;</span>}<span\; class="notranslate">\; )</span>}^{\frac{\mathrm{<span\; class="notranslate">\; d</span>}}{\mathrm{<span\; class="notranslate">\; 2</span>}}}}\mathrm{<span\; class="notranslate">\; exp</span>}<span\; class="notranslate">\; \&lsqb;</span><span\; class="notranslate">\; -</span>\frac{<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; -</span>{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; t</span>}\mathrm{<span\; class="notranslate">\; k</span>}}<span\; class="notranslate">\; )</span>{<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; -</span>{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; t</span>}\mathrm{<span\; class="notranslate">\; k</span>}}<span\; class="notranslate">\; )</span>}^{\mathrm{<span\; class="notranslate">\; T</span>}}}{{\mathrm{<span\; class="notranslate">\; \&sigma;</span>}}^{\mathrm{<span\; class="notranslate">\; 2</span>}}}<span\; class="notranslate">\; \&rsqb;</span>$

Where t=1,...M,k=1,...N _t ; x is the input vector that needs to make a decision, with d attributes; the value of d is the conditional attribute dimension obtained after reduction; x _tk is the hidden layer Vector, which is the k-th neuron of the t-th class sample in the hidden layer; M is the total number of training samples, which is equal to the dimension of the decision attribute; Nt is the number of samples of the t-th class; and σ is the smoothing factor, σ ∈(0,∞);

In step S133, the weighted layer performs weighted average on the outputs of hidden neurons belonging to the same class:

${\mathrm{<span\; class="notranslate">\; v</span>}}_{\mathrm{<span\; class="notranslate">\; t</span>}}<span\; class="notranslate">\; =</span>\frac{\mathrm{<span\; class="notranslate">\; 1</span>}}{{\mathrm{<span\; class="notranslate">\; N</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}}\underset{\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\overset{{\mathrm{<span\; class="notranslate">\; N</span>}}_{\mathrm{<span\; class="notranslate">\; t</span>}}}{<span\; class="notranslate">\; \&Sigma;</span>}}{\mathrm{<span\; class="notranslate">\; \&Phi;</span>}}_{\mathrm{<span\; class="notranslate">\; t</span>}\mathrm{<span\; class="notranslate">\; k</span>}}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; )</span>$

Among them t=1,...M; Φ _tk is the output value of the previous step; v _t is the weighted layer to do the weighted average of the output of the hidden neurons belonging to the same class in the hidden layer;

Step S134, calculating the category with the largest posterior probability density value ρ(x)=argmax(v _t ) among all neuron outputs as the output result of the neural network;

Step S135, taking the output result of the neural network as the decision value of soil water demand, and constructing a decision table.

Example 2

On the basis of embodiment 1, this embodiment 2 also provides a crop irrigation decision-making system, including: a collection module for collecting data on the current growth environment and growth stage of the crops, and a server connected to the corresponding data collection module of the crops; The server is adapted to regulate the amount of water used for irrigation of crops.

Specifically, the collection module includes: a soil moisture sensor, a soil temperature sensor, an air humidity sensor, an air temperature sensor, a light intensity sensor and a wind speed sensor, to respectively collect the soil type, soil humidity, soil temperature, relative air humidity, air Temperature, atmospheric light intensity, wind speed, sunshine hours and other crop growth environment data, and the acquired growth environment data is transmitted to the server through the wireless network.

Each sensor can use a controller as a transmission node to summarize the data of each sensor and send it to the server, and collect and summarize each data through the controller and send it wirelessly, which can be realized by using existing technologies.

Wherein, the server is adapted to use the crop irrigation strategy as described in Embodiment 1 to regulate water consumption.

Specifically, the server makes fuzzy decisions based on the crop growth environment data and the current growth stage, judges the current water demand of the crops, and determines the opening time of the crop irrigation solenoid valves. The server sends the opening information of the battery valve to the controller through the wireless network, controls the opening and closing of the electromagnetic valve, and finally realizes the precise control of the water demand of the crop.

Inspired by the above-mentioned ideal embodiment according to the present invention, through the above-mentioned description content, relevant workers can make various changes and modifications within the scope of not departing from the technical idea of the present invention. The technical scope of the present invention is not limited to the content in the specification, but must be determined according to the scope of the claims.

Claims (8)

1. A crop irrigation strategy, characterized in that, comprises the steps: Step S1, establishing a generative system knowledge base according to the sample data; and Step S2, according to the current growth environment and growth stage data of the crops, determine the current water demand for crop irrigation through fuzzy judgment. 2. The crop irrigation strategy according to claim 1, characterized in that, The method for establishing and generating the system knowledge base according to the sample data in the step S1 includes: Step S11, collecting crop growth environment and water demand data information at each growth stage as sample data, and establishing a crop water demand correlation database to construct a fuzzy information system; Step S12, performing attribute reduction in the fuzzy information system to generate probabilistic decision rules; and Step S13, using the neural network to train the reduction to generate the system knowledge base. 3. The crop irrigation strategy according to claim 1 or 2, characterized in that, In the step S2, according to the current growth environment and growth stage data of the crops, the water demand for the current crop irrigation is determined through fuzzy judgment, namely According to the data of the current growth environment and growth stage of the crops, the water demand of the current crops is determined through the fuzzy judgment of the neural network. 4. The crop irrigation strategy according to claim 3, characterized in that, In the step S11, the method of collecting crop growth environment and water demand data information at each growth stage, and establishing a crop water demand correlation database to construct a fuzzy information system includes the following steps: Step S111, establishing the crop growth environment factor sample set U, the water demand condition attribute set C and the water demand decision attribute set V; Step S112, said fuzzy information system construction, namely The fuzzy information system is a quadruple: G={U, A, V, f}; where U is the domain of discourse, which is the collection of all crop sample information; A is an attribute set, that is, A=C∪D, where: C represents a conditional attribute, that is, a set composed of soil moisture, soil temperature, and growth stage attributes of crops; D represents a decision attribute, that is, a set composed of crop water demand attributes; V is the union of attribute value ranges, V=∪V _a , V _a represents the value range of attribute a∈A, that is, the union of the value range sets of soil moisture, growth stage and water demand attributes; f represents the U×A→V information function, and assigns attribute values to each object in the domain of discourse, namely There is f(x, a)∈V _a . 5. The crop irrigation strategy according to claim 4, characterized in that, in the step S1, attribute reduction will be carried out in the fuzzy information system, namely Judging attribute similarity, reducing conditional attributes, and obtaining reduced conditional attribute set C ₁ , the method includes the following steps: Step S121, define attribute similarity Among them: X, Y∈A, X and Y are any two attributes in the domain of discourse, that is, A is the union of conditional attributes and decision attributes; ind( ) is the equivalent division of domain U about a certain attribute Classification; n is the number of equivalence division classes; Step S122, calculating the similarity S(c _i , D) between the condition attribute and the decision attribute among the attributes, where c _i , c _j ∈ C, i=1...m, i≠j, m is the number of condition attributes ; Step S123, define the reduced attribute set C ₁ =Φ, and calculate the similarity S(ci,c _j ) between two conditional attributes in the attribute, where c _i ,c _j ∈C,i, _j = 1...m, i≠j, i<j; Step S124, until the traversal of each condition attribute is completed; In step S125, the reduction condition attribute set C ₁ is obtained. 6. The crop irrigation strategy according to claim 5, characterized in that, In the step S13, the method for training the reduction using a neural network comprises the following steps: Step S131, sending the conditionally reduced sample data into the input layer of the probabilistic neural network, and the number of neuron nodes is equal to the dimension of the conditional attribute set; Step S133, setting the number of neuron nodes in the hidden layer to be equal to the number of samples, and the neuron uses a Gaussian kernel function to determine the input-output relationship: ${\mathrm{<span\; class="notranslate">\; \&Phi;</span>}}_{\mathrm{<span\; class="notranslate">\; tk</span>}}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>\frac{\mathrm{<span\; class="notranslate">\; 1</span>}}{{<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 2</span>}\mathrm{<span\; class="notranslate">\; \&pi;</span>}<span\; class="notranslate">\; )</span>}^{\frac{\mathrm{<span\; class="notranslate">\; d</span>}}{\mathrm{<span\; class="notranslate">\; 2</span>}}}}\mathrm{<span\; class="notranslate">\; exp</span>}<span\; class="notranslate">\; [</span><span\; class="notranslate">\; -</span>\frac{<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; -</span>{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; tk</span>}}<span\; class="notranslate">\; )</span>{<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; -</span>{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; tk</span>}}<span\; class="notranslate">\; )</span>}^{\mathrm{<span\; class="notranslate">\; T</span>}}}{{\mathrm{<span\; class="notranslate">\; \&sigma;</span>}}^{\mathrm{<span\; class="notranslate">\; 2</span>}}}<span\; class="notranslate">\; ]</span>$ Where t=1,...M,k=1,...N _t ; x is the input vector that needs to make a decision, with d attributes; the value of d is the conditional attribute dimension obtained after reduction; x _tk is the hidden layer Vector, is the k-th neuron of the t-th class sample in the hidden layer; M is the total number of classes of the training samples, which is equal to the dimension of the decision attribute; Nt is the number of samples of the t-th class; and σ is the sliding factor, σ ∈(0,∞); In step S133, the weighted layer performs weighted average on the outputs of hidden neurons belonging to the same class: ${\mathrm{<span\; class="notranslate">\; v</span>}}_{\mathrm{<span\; class="notranslate">\; t</span>}}<span\; class="notranslate">\; =</span>\frac{\mathrm{<span\; class="notranslate">\; 1</span>}}{{\mathrm{<span\; class="notranslate">\; N</span>}}_{\mathrm{<span\; class="notranslate">\; t</span>}}}\underset{\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\overset{{\mathrm{<span\; class="notranslate">\; N</span>}}_{\mathrm{<span\; class="notranslate">\; t</span>}}}{<span\; class="notranslate">\; \&Sigma;</span>}}{\mathrm{<span\; class="notranslate">\; \&Phi;</span>}}_{\mathrm{<span\; class="notranslate">\; t</span>}\mathrm{<span\; class="notranslate">\; k</span>}}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; )</span>$ where t=1,...M; Step S134, calculating the category with the maximum posterior probability density value ρ(x)=argmax(t) in all neuron outputs as the output result of the neural network; Step S135, taking the output result of the neural network as the decision value of soil water demand, and constructing a decision table. 7. A crop irrigation decision-making system, characterized in that it comprises: a collection module for collecting data on the current growth environment and growth stages of the crops, and a server connected to the corresponding data collection module of the crops; The server is adapted to regulate the amount of water used for irrigation of crops. 8. The crop irrigation decision-making system according to claim 7, wherein the server is adapted to use the crop irrigation strategy according to claim 1 to regulate water consumption.