Integrated intelligent irrigation decision system based on multi-source data fusion

Abstract

The application belongs to the technical field of intelligent agricultural irrigation, and discloses an integrated intelligent irrigation decision system based on multi-source data fusion; comprising the following modules: a data acquisition module, which is used for collecting parameter data of all monitoring irrigation points in a farmland in real time.The pesticide demand prediction model is set, the average pest amount, the beneficial insect amount, the environmental temperature and the environmental humidity in the pesticide application period are input into the pesticide demand prediction model, the predicted pesticide demand of each monitoring irrigation point is obtained, the pesticide water demand is obtained through the predicted pesticide demand, the relative difference, the distance between the two grooves and the conveying effective amount are input into the aid efficiency score through the setting of the aid efficiency score, the aid efficiency score can be obtained, the highest aid efficiency score is selected, the corresponding aidable groove is assisted to the corresponding aid-needed groove, and the pesticide action time of the grooves is substantially equalized.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of intelligent agricultural irrigation technology, more particularly, the present application relates to an integrated intelligent irrigation decision-making system based on multi-source data fusion. BACKGROUND

[0002] Traditional irrigation modes have been difficult to meet the requirements of precision agriculture development due to excessive reliance on manual experience, single data dimension, and decision lag. Early intelligent irrigation systems mostly use single data source to drive decision-making, such as triggering irrigation only by relying on soil moisture threshold, which leads to the fact that the decision-making model cannot adapt to complex agricultural conditions. The introduction of multi-source data fusion technology breaks the barriers between heterogeneous data such as sensor data, remote sensing images, crop models, and historical irrigation records. Through data cleaning, spatio-temporal alignment, feature extraction, and other preprocessing procedures, a stereoscopic decision-making information base is constructed.

[0003] The current irrigation system needs to apply pesticides in the farmland at the same time when irrigating based on multiple data to assist the growth of crops. Pesticides have various types. When solid particle pesticides are released, a certain water layer needs to be set on the solid particle pesticides to allow the solid particle pesticides to continuously diffuse and play a sustained role. However, the current irrigation system is difficult to make the action time of the pesticides applied at multiple positions equal. If the action time of the pesticides is not equal, some positions may have no pesticides, and some positions may still have pesticides, which is not convenient for uniform application of pesticides. SUMMARY

[0004] To solve the problems in the background art, the present application provides an integrated intelligent irrigation decision-making system based on multi-source data fusion.

[0005] To achieve the above-mentioned purpose, the present application provides the following technical scheme: an integrated intelligent irrigation decision-making system based on multi-source data fusion, comprising the following modules:

[0006] A data acquisition module is used to acquire parameter data of all monitoring irrigation points in the farmland in real time.

[0007] An irrigation decision-making module is used to generate water demand scores of all monitoring irrigation points according to the acquired parameter data, and to make irrigation decisions according to the water demand scores.

[0008] A pesticide demand amount prediction module is used to acquire historical pesticide demand amount data and construct a pesticide demand amount prediction model, and to predict the pesticide demand amount of the monitoring irrigation points by using the pesticide demand amount prediction model.

[0009] A pesticide action time prediction module is used to obtain pesticide water demand required for the pesticide to play a role in the trench according to the pesticide demand amount, to acquire historical pesticide action time data and construct a pesticide action time prediction model, and to predict the pesticide action time in each trench by using the pesticide action time prediction model.

[0010] The trench aid module is configured to generate a set of trenches in need of aid and a set of aidable trenches according to the pesticide action time of each trench, generate an aid efficiency score of each aidable trench relative to each trench in need of aid, and cause the aidable trenches to aid the trenches in need of aid according to the aid efficiency score;

[0011] The execution module is configured to irrigate near the roots of crops in the irrigation area, and can apply pesticides and water to the trenches, and can transport water between any two trenches.

[0012] Further, the farmland is evenly divided into a plurality of irrigation areas, and a monitoring irrigation point is arranged at the center of each irrigation area, and the parameter data includes soil moisture content, air relative humidity, leaf water content and leaf area change;

[0013] The dielectric constant k is measured by TDR time domain reflectometer in 0-20cm and 20-40cm soil layers respectively, and the volume water content SWC is calculated by applying the calibration formula specific to the soil type k ;

[0014] SWC k =a×k+b;

[0015] The regression coefficients a and b are calibration parameters;

[0016] The volume water contents SWC 0-20 and SWC 20-40 of 0-20cm and 20-40cm soil layers are calculated, and a weighted average value is obtained 平均 ;

[0017] SWC 平均 =0.6×SWC 0-20 +0.4×SWC 20-40 ;

[0018] The SWC 平均 is normalized to obtain the normalized soil moisture content SWC;

[0019]

[0020] The range of soil moisture content is SWC min -SWC max , SWC min is the minimum value of soil moisture content, and SWC max is the maximum value of soil moisture content, which is suitable for precision irrigation control;

[0021] The air relative humidity RH 实测 is obtained by real-time collection of a digital temperature and humidity sensor, and RH 实测The normalized air relative humidity RH is obtained by normalization processing;

[0022]

[0023] The range of air relative humidity is RH min -RH max , RH min is the minimum value of air humidity, RH max is the maximum value of air humidity;

[0024] The reflectivity of the leaf at the wavelengths of 1450 nm and 1940 nm is measured by a portable near-infrared spectrometer, which is R 1450 and R 1940 respectively, and the leaf water content LWC 实测 is calculated using an empirical formula;

[0025]

[0026] The coefficients 46.5 and 82.2 are obtained by multiple regression fitting;

[0027] Five plants are randomly selected in the irrigation area to be irrigated, and three leaves are measured for each plant, and the average value LWC 平均 of the water content of 15 leaves is calculated, and the normalized leaf water content LWC 平均 is obtained by normalization processing on LWC min ;

[0028]

[0029] The range of leaf water content is LWC max -LWC min , LWC max is the minimum water content of the leaf surface, and LWC 实测 is the maximum water content of the leaf surface;

[0030] The effective leaf area index LAI 实测 is measured in real time by a plant canopy analyzer under cloudless weather, and the difference ΔLAI 周 between the effective leaf area index of this week and that of last week is calculated;

[0031] ΔLAI 周 = LAI 实测 (t) - LAI 实测 (t-7 days);

[0032] The daily change ΔLAI 日 of the effective leaf area index is obtained by linear interpolation;

[0033]

[0034] ΔLAI 日The normalized effective leaf area index ΔLAI is obtained by normalization processing;

[0035]

[0036] The leaf area change amount is in the range of LWI min -LWI max , LWI min is the minimum value of the effective leaf area index, and LWI max is the maximum value of the effective leaf area index.

[0037] Further, the corresponding water demand score S is obtained according to the parameter data:

[0038] S = ω1*SWC + ω2*RH + ω3*LWC + ω4*ΔLAI;

[0039] wherein ω1, ω2, ω3 and ω4 are weight coefficients, which are obtained according to historical data training;

[0040] A water demand threshold Z is set, and the water demand score S is compared with the water demand threshold Z to decide whether to trigger irrigation;

[0041] When S < Z, irrigation is not triggered;

[0042] When S ≥ Z, irrigation is triggered.

[0043] Further, factors affecting the pesticide demand of the monitored irrigation point include: pest quantity, beneficial insect quantity, environmental temperature and environmental humidity;

[0044] A pest attracting mechanism and a beneficial insect attracting mechanism are set at a certain height from the ground. The pest attracting mechanism can release pest pheromones for a certain period of time, and the beneficial insect attracting mechanism can release beneficial insect pheromones for a certain period of time. During this period, an integrated camera captures images, a YOLOv5 model is trained and deployed on an edge device, and the pest quantity and beneficial insect quantity are identified and counted in real time;

[0045] A temperature and humidity sensor is disposed inside the crop canopy to collect environmental temperature and humidity in real time, and to calculate daily average temperature and humidity;

[0046] Historical pesticide demand data is obtained and a pesticide demand prediction model is constructed. The process of predicting the pesticide demand of the monitored irrigation point using the pesticide demand prediction model includes:

[0047] A pesticide application cycle is set, and historical pesticide demand data of a single monitored irrigation point in different pesticide application cycles is obtained. The pesticide and fertilizer demand data includes the average pest quantity, beneficial insect quantity, environmental temperature, environmental humidity of the single monitored irrigation point in different pesticide application cycles, and the historical pesticide demand of the single monitored irrigation point used in the pesticide application cycle.

[0048] According to the pest amount, beneficial insect amount, environmental temperature, environmental humidity and corresponding historical pesticide demand amount of different historical pesticide demand amount data corresponding to the monitoring irrigation points, a pesticide demand amount prediction set is generated, and the pesticide demand amount prediction set is divided into a first training set and a first test set;

[0049] A first convolutional neural network is constructed, the pest amount, beneficial insect amount, environmental temperature and environmental humidity in different historical pesticide demand amount data in the first training set are taken as input data of the first convolutional neural network, and the corresponding historical pesticide demand amount in the first training set is taken as output data of the first convolutional neural network;

[0050] The first convolutional neural network is trained to obtain a first initial convolutional neural network, the first initial convolutional neural network is verified by using the first test set, and the first initial convolutional neural network with a first test error threshold value less than or equal to a preset first test error threshold value is output as a pesticide demand amount prediction model;

[0051] Every time a pesticide application cycle is reached, the average pest amount, beneficial insect amount, environmental temperature and environmental humidity of each monitoring irrigation point in the pesticide application cycle are input into the pesticide demand amount prediction model to obtain the predicted pesticide demand amount of each monitoring irrigation point.

[0052] Further, the process of deriving the pesticide water demand required for the pesticide to play a role in the trench from the pesticide demand amount comprises:

[0053] The pesticide water demand refers to the need to set a certain depth of water layer on the pesticide, and the use of the pesticide is assisted by setting a certain depth of water layer, such as a certain depth of water layer can promote pesticide dissolution and diffusion, ensure effective absorption of crops, prevent pesticide loss and volatilization, and adjust the soil environment and microbial activity. The optimal range of the depth of the water layer is M-N, and the median of M and N is U;

[0054] The farmland is set as a dry field, the pesticide used is a solid particle pesticide, a plurality of trenches are dug in each irrigation area to be irrigated, the trenches are located on one side of the crop roots, the length, width and depth of the trench are fixed values, that is, the bottom area D of the trench is a fixed value, the predicted pesticide demand amount required by each monitoring irrigation point is divided by the number of trenches therein to obtain the pesticide amount required by each trench, the volume K of the pesticide can be obtained according to the required pesticide amount, the volume of a single solid pesticide particle and the weight of the pesticide, the depth Q of the pesticide in the trench can be obtained according to the bottom area of the trench, that is, the optimal depth range of water in the trench is (Q+M)-(Q+N), and the pesticide water demand in a single trench is (D×Q-K)+D×U according to the bottom area of the trench, the volume K of the pesticide and the depth K of the pesticide.

[0055] Further, the pesticide action duration refers to the duration of the pesticide in the groove cooperating with water to act, and factors affecting the pesticide action duration in the groove include: fertilizer demand, pesticide water demand, groove area directly exposed to sunlight, and groove water infiltration speed;

[0056] The historical fertilizer demand is obtained by acquiring historical data of the fertilizer used in a single groove, and the fertilizer demand is obtained by calculating the average of the historical fertilizer demand;

[0057] The pesticide water demand in a single groove is obtained by the above calculation;

[0058] The groove area directly exposed to sunlight is obtained by taking a photo of the groove with a high-resolution device to ensure that the shadow is clear, identifying the shadow area through color space conversion and brightness contrast analysis, and segmenting the groove contour with a YOLOv8 model, and then subtracting the shadow area from the total area of the groove;

[0059] A certain amount of pesticide and water is added to the groove, the pesticide will diffuse into the water, and the volume reduction of the water is measured within a certain time, and the water infiltration speed of the groove is obtained by dividing the water reduction by the time.

[0060] The historical pesticide action duration data is acquired and a pesticide action duration prediction model is constructed, and the process of predicting the pesticide action duration in each groove by using the pesticide action duration prediction model includes:

[0061] The historical pesticide action duration data of a single groove is acquired, and the historical pesticide action duration data includes the fertilizer demand, pesticide water demand, groove area directly exposed to sunlight, groove water infiltration speed of the single groove, and the historical pesticide action duration of the single groove;

[0062] According to the fertilizer demand, pesticide water demand, groove area directly exposed to sunlight, groove water infiltration speed, and corresponding historical pesticide action duration in different historical pesticide action duration data, a pesticide action duration prediction set is generated, and the pesticide action duration prediction set is divided into a second training set and a second test set;

[0063] A second convolutional neural network is constructed, the fertilizer demand, pesticide water demand, groove area directly exposed to sunlight, and groove water infiltration speed in different historical water demand data in the second training set are taken as input data of the second convolutional neural network, and the corresponding historical pesticide action duration in the second training set is taken as output data of the second convolutional neural network;

[0064] The second convolutional neural network is trained to obtain a second initial convolutional neural network, the second test set is used to verify the model of the second initial convolutional neural network, and the second initial convolutional neural network with a second test error threshold value less than or equal to a preset second test error threshold value is output as a pesticide action duration prediction model;

[0065] The fertilizer requirement, pesticide water requirement, direct sunlight area of each trench, and trench water infiltration speed are input into the pesticide action time prediction model to obtain the predicted pesticide action time of each trench.

[0066] Further, the process of generating the set of trenches needing assistance and the set of trenches capable of assistance according to the pesticide action time of each trench includes:

[0067] The predicted pesticide action time of each trench is obtained by the pesticide action time prediction model in the irrigation area, the average of the predicted pesticide action time is calculated according to the predicted pesticide action time of each trench, the trenches with a predicted pesticide action time greater than the average are included in the set of trenches capable of assistance, and the trenches with a predicted pesticide action time less than the average are included in the set of trenches needing assistance.

[0068] The trenches capable of assistance refer to the trenches with a predicted pesticide action time longer than the average, and the trenches needing assistance refer to the trenches with a predicted pesticide action time shorter than the average. The predicted pesticide action time of each trench is equalized by making the trenches capable of assistance assist the trenches needing assistance, so as to uniformly apply pesticides subsequently. The assistance refers to the water in the trenches capable of assistance being transported into the trenches needing assistance, the water containing diffused pesticides, and the predicted pesticide action time of the trenches capable of assistance having a time difference with the average. The time difference is converted into the water amount of assistance by calculation.

[0069] Further, the process of generating the assistance efficiency score of each trench capable of assistance relative to each trench needing assistance and making the trenches capable of assistance assist the trenches needing assistance according to the assistance efficiency score includes:

[0070] The assistance efficiency score refers to the efficiency score of the trenches capable of assistance assisting the trenches needing assistance, and the trenches needing assistance are assisted by the trenches capable of assistance with the highest assistance efficiency score according to the assistance efficiency score.

[0071] The assistance efficiency score related data includes the relative difference, the distance between the two trenches, and the effective transport amount.

[0072] The absolute value of the difference between the predicted pesticide action time of the trenches capable of assistance and the average is the first difference, the absolute value of the difference between the predicted pesticide action time of the trenches needing assistance and the average is the second difference, and the absolute value of the first difference minus the second difference is the relative difference.

[0073] The effective transport amount refers to the loss of water during the transport between the trenches. The effective transport amount can be obtained by subtracting the loss amount from the total transport amount.

[0074] The assistance efficiency score P can be obtained as follows:

[0075]

[0076] wherein:

[0077] alpha1, alpha2 and alpha3 are weight coefficients, which are trained according to historical data;

[0078] E is a relative difference value, E 可 is the predicted pesticide action duration of the aidable trench, f is the distance between two trenches, f max is the distance between the two farthest trenches in the irrigation area, G is the delivery effective amount, G 总 is the total delivery amount;

[0079] a aid efficiency score of each aidable trench relative to each aided trench is obtained, the aidable trench corresponding to the highest aid efficiency score is selected to aid the aided trench corresponding thereto.

[0080] Further, the execution module comprises an irrigation device arranged in the irrigation area, the irrigation device comprising an irrigation pipe network, a plurality of water outlets being arranged on the irrigation pipe network, each water outlet being opposite to the root of a corresponding crop, the water amount allocated to each crop by the irrigation device according to the predicted water consumption is used to irrigate the root of each crop, realizing large-scale precision irrigation, the execution module further comprises a pesticide application device and a delivery device, the pesticide application device can add pesticides into the trench and uniformly lay the pesticides on the bottom of the trench, water is also added into the trench through the water outlets on the irrigation pipe network, realizing the joint action of water and pesticides, the delivery device can deliver water between any two trenches, and the mixture of water and pesticides is delivered from a suitable trench to a corresponding trench according to the aid efficiency score.

[0081] The integrated intelligent irrigation decision system based on multi-source data fusion has the following technical effects and advantages:

[0082] (1) By setting a pesticide demand prediction model, the average pest amount, beneficial insect amount, environmental temperature and environmental humidity in the pesticide application period are input into the pesticide demand prediction model to obtain the predicted pesticide demand of each monitoring irrigation point, the pesticide water demand is obtained through the predicted pesticide demand, the relative difference value, the distance between two trenches and the delivery effective amount are input into the aid efficiency score to obtain the aid efficiency score, the aidable trench corresponding to the highest aid efficiency score is selected to aid the aided trench corresponding thereto, and the pesticide action duration of the trench is substantially equalized.

[0083] (2) by setting the water demand score, according to the soil moisture content, air relative humidity, leaf water content and leaf area change of the crops in the irrigation area within a certain time, the above change is input into the water demand score, the water demand score of each irrigation area can be obtained, according to the comparison with the water demand threshold, whether the irrigation instruction needs to be triggered is decided, whether irrigation is needed is judged through multi-source data fusion, and the intelligentization of irrigation decision is realized. BRIEF DESCRIPTION OF DRAWINGS

[0084] Figure 1 It is a schematic diagram of the system structure of the application. DETAILED DESCRIPTION

[0085] The technical solutions in the embodiments of the application will be clearly and completely described below with reference to the drawings in the embodiments of the application. Obviously, the described embodiments are only part of the embodiments of the application, rather than all the embodiments. Based on the embodiments in the application, all other embodiments obtained by those skilled in the art without creative labor fall within the scope of protection of the application.

[0086] Referring to Figure 1 The integrated intelligent irrigation decision system based on multi-source data fusion comprises the following modules:

[0087] The data acquisition module is used for acquiring parameter data of all monitoring irrigation points in the farmland in real time.

[0088] The irrigation decision module is used for generating water demand scores of all monitoring irrigation points according to the acquired parameter data, and deciding to irrigate according to the water demand scores.

[0089] The pesticide demand amount prediction module is used for acquiring historical pesticide demand amount data and constructing a pesticide demand amount prediction model, and predicting the pesticide demand amount of the monitoring irrigation point by using the pesticide demand amount prediction model.

[0090] The pesticide action duration prediction module is used for obtaining pesticide water demand required for the pesticide to play a role in the trench according to the pesticide demand amount, acquiring historical pesticide action duration data and constructing a pesticide action duration prediction model, and predicting the pesticide action duration in each trench by using the pesticide action duration prediction model.

[0091] The trench aid module is used for generating a set of trenches needing aid and a set of aidable trenches according to the pesticide action duration of each trench, generating an aid efficiency score of each aidable trench relative to each trench needing aid, and making the aidable trenches aid the trenches needing aid according to the aid efficiency score.

[0092] The execution module is used for irrigating near the roots of the crops in the irrigation area, and can discharge pesticides and water into the trenches, and can transport water between any two trenches.

[0093] It needs to be further explained that in the specific implementation process, the farmland is evenly divided into a plurality of irrigation areas, a monitoring irrigation point is arranged at the center of each irrigation area, and the parameter data includes soil moisture content, air relative humidity, leaf water content and leaf area change amount;

[0094] The dielectric constant k is measured by a TDR time domain reflectometer in the 0-20cm and 20-40cm soil layers respectively, and the volume water content SWC is calculated by applying a specific calibration formula for the soil type k ;

[0095] SWC k =a×k+b;

[0096] The regression coefficients a and b are calibration parameters;

[0097] The volume water contents SWC 0-20 and SWC 20-40 of the 0-20cm and 20-40cm soil layers are calculated, and a weighted average value is obtained to obtain SWC 平均 ;

[0098] SWC 平均 =0.6×SWC 0-20 +0.4×SWC 20-40 ;

[0099] The SWC 平均 is normalized to obtain the normalized soil moisture content SWC;

[0100]

[0101] Taking sandy soil as an example, a is 0.12, b is 5.8, the range of soil moisture content is SWC min -SWC max , specifically 5%-15%, SWC min is the minimum value of soil moisture content, and SWC max is the maximum value of soil moisture content, which is suitable for precise irrigation control. When k measured in the 0-20cm soil layer is 20 and k measured in the 20-40cm soil layer is 25, SWC 平均 is 8.48%, and SWC is 0.348;

[0102] The air relative humidity RH 实测 is obtained by real-time collection by a digital temperature and humidity sensor, and the RH 实测 is normalized to obtain the normalized air relative humidity RH;

[0103]

[0104] The range of air relative humidity is RHmin -RH max , specifically 30%-90%, RH min is the minimum value of air humidity, RH max is the maximum value of air humidity, when the measured RH 实测 is 52%, then RH is 0.37;

[0105] The reflectivity of the leaf at 1450nm and 1940nm wavelengths is measured by a portable near-infrared spectrometer, which are R 1450 and R 1940 , respectively. The leaf water content LWC 实测 is calculated using an empirical formula;

[0106]

[0107] The coefficients 46.5 and 82.2 are obtained by multiple regression fitting;

[0108] 5 plants are randomly selected in the irrigation area, and 3 leaves are measured for each plant. The average value of the water content of 15 leaves LWC 平均 is calculated, and the LWC 平均 is normalized to obtain the normalized leaf water content LWC

[0109]

[0110] The range of leaf water content is LWC min -LWC max , and the leaf water content of healthy leaves is specifically 75%-95%, LWC min is the minimum leaf water content, and LWC max is the maximum leaf water content. When the measured LWC 平均 is 88.5%, then LWC is 0.68;

[0111] The effective leaf area index LAI 实测 is measured in real time by a plant canopy analyzer on a cloudless day, and the difference between this week's effective leaf area index and last week's is calculated ΔLAI 周 ;

[0112] ΔLAI 周 = LAI 实测 (t)- LAI 实测 (t-7 days);

[0113] The daily change of effective leaf area index ΔLAI 日 is obtained by linear interpolation;

[0114]

[0115] The ΔLAI 日The normalized effective leaf area index ΔLAI is obtained by normalization processing;

[0116]

[0117] The leaf area change amount is in the range of LWI min LWI max , specifically -0.2 to +0.2 cm2 / plant·day, LWI min is the minimum value of the effective leaf area index, LWI max is the maximum value of the effective leaf area index, and a negative value indicates a decrease in leaf area. When ΔLAI 日 is 0.086 cm2 / plant·day, ΔLAI is 0.72.

[0118] It needs to be further explained that in the specific implementation process, the water requirement score S is obtained according to the parameter data:

[0119] S = ω1×SWC + ω2×RH + ω3×LWC + ω4×ΔLAI;

[0120] wherein ω1, ω2, ω3 and ω4 are weight coefficients, which are obtained according to historical data training;

[0121] Different ω1, ω2, ω3 and ω4 need to be set at each stage of plant growth. For example, ω1, ω2, ω3 and ω4 are 0.25, 0.2, 0.3 and 0.25 respectively at the vegetative growth stage, ω1, ω2, ω3 and ω4 are 0.3, 0.15, 0.25 and 0.3 respectively at the reproductive growth stage, and ω1, ω2, ω3 and ω4 are 0.35, 0.25, 0.2 and 0.2 respectively at the mature stage.

[0122] Taking tomato as an example, if the tomato is in the reproductive growth stage, SWC 平均 is 8.48%, SWC is 0.348, RH 实测 is 52%, RH is 0.37, LWC 平均 is 88.5%, LWC is 0.68, and ΔLAI 日 is 0.086 cm2 / plant·day, ΔLAI is 0.72;

[0123] S = 0.3×0.348 + 0.15×0.37 + 0.25×0.68 + 0.3×0.72 = 0.5495.

[0124] A water requirement threshold Z is set, Z is specifically 0.7, and the water requirement score S is compared with the water requirement threshold Z to decide whether to trigger irrigation;

[0125] When S < Z, irrigation is not triggered;

[0126] When S≥Z, irrigation is triggered;

[0127] When S=0.5495<0.7, the monitoring irrigation point will not trigger irrigation.

[0128] It should be further explained that in the specific implementation process, the factors affecting the pesticide demand of the monitoring irrigation point include: pest quantity, beneficial insect quantity, environmental temperature and environmental humidity;

[0129] The pest attracting mechanism and the beneficial insect attracting mechanism are arranged at a certain height from the ground. The pest attracting mechanism can release pest pheromones for a certain period of time, and the beneficial insect attracting mechanism can release beneficial insect pheromones for a certain period of time. During this period, the integrated camera captures images, and the YOLOv5 model is trained and deployed on the edge device to identify and count the pest quantity and beneficial insect quantity in real time.

[0130] The temperature and humidity sensor is arranged in the crop canopy to collect environmental temperature and humidity in real time, and calculate the daily average temperature and humidity;

[0131] The historical pesticide demand data is obtained and a pesticide demand prediction model is constructed. The process of predicting the pesticide demand of the monitoring irrigation point using the pesticide demand prediction model includes:

[0132] The pesticide application cycle is set, and the historical pesticide demand data of a single monitoring irrigation point in different pesticide application cycles is obtained. The pesticide fertilizer demand data includes the average pest quantity, beneficial insect quantity, environmental temperature, environmental humidity of the single monitoring irrigation point in different pesticide application cycles, and the historical pesticide demand of the single monitoring irrigation point in the pesticide application cycle.

[0133] According to the pest quantity, beneficial insect quantity, environmental temperature, environmental humidity and corresponding historical pesticide demand of the corresponding monitoring irrigation point in different historical pesticide demand data, a pesticide demand prediction set is generated, and it is divided into a first training set and a first test set.

[0134] A first convolutional neural network is constructed, and the pest quantity, beneficial insect quantity, environmental temperature and environmental humidity in the different historical pesticide demand data in the first training set are used as the input data of the first convolutional neural network. The corresponding historical pesticide demand in the first training set is used as the output data of the first convolutional neural network.

[0135] The first convolutional neural network is trained to obtain a first initial convolutional neural network. The first test set is used to verify the model of the first initial convolutional neural network, and the first initial convolutional neural network with an output less than or equal to a preset first test error threshold is output as a pesticide demand prediction model.

[0136] Every time a pesticide application cycle is reached, the average pest amount, beneficial insect amount, environmental temperature and environmental humidity of each monitoring irrigation point in the pesticide application cycle are input into the pesticide demand prediction model to obtain the predicted pesticide demand of each monitoring irrigation point.

[0137] It needs to be further explained that, in the specific implementation process, the process of obtaining the pesticide water demand required for the pesticide to function in the trench from the pesticide demand includes:

[0138] The pesticide water demand refers to the need to set a certain depth of water layer on the pesticide, which assists the use of the pesticide, such as promoting the dissolution and diffusion of the pesticide, ensuring the effective absorption of crops, preventing the loss and volatilization of the pesticide, and adjusting the soil environment and microbial activity. The optimal range of the depth of the water layer is M-N, and the median of M and N is U.

[0139] The farmland is set as a dry field, the pesticide used is a solid particle pesticide, a plurality of trenches are dug in each irrigation area to be irrigated, the trenches are located on one side of the crop roots, the length, width and depth of the trench are fixed values, that is, the bottom area D of the trench is a fixed value, the predicted pesticide demand required by each monitoring irrigation point is divided by the number of trenches therein to obtain the pesticide amount required by each trench, the volume K of the pesticide can be obtained according to the required pesticide amount, the volume and weight of a single solid pesticide particle, and the depth Q of the pesticide in the trench can be obtained according to the bottom area of the trench, that is, the optimal depth range of water in the trench is (Q+M)-(Q+N), and the pesticide water demand in a single trench is (D×Q-K)+D×U according to the bottom area of the trench, the volume K of the pesticide and the depth K of the pesticide.

[0140] It needs to be further explained that the pesticide action duration refers to the duration of the pesticide functioning in the trench with water, and the factors affecting the pesticide action duration in the trench include: fertilizer demand, pesticide water demand, sunlight direct radiation area of the trench and water infiltration speed of the trench.

[0141] The historical fertilizer demand is obtained by obtaining the historical data of the fertilizer used in a single trench, and the fertilizer demand is obtained by calculating the average of the historical fertilizer demand.

[0142] The pesticide water demand in a single trench is obtained by the above calculation.

[0143] The trench contour is segmented by using the YOLOv8 model, and the sunlight direct radiation area of the trench is obtained by subtracting the shadow area from the total area of the trench by taking a photo of the trench with a high-resolution device to ensure clear shadows, and by color space conversion and brightness contrast analysis.

[0144] A certain amount of pesticide and water is added into the groove, the pesticide will diffuse into the water, and the volume reduction of the water is measured within a certain time, and the water reduction divided by time can obtain the water seepage speed of the groove.

[0145] The process of obtaining historical pesticide action duration data and constructing a pesticide action duration prediction model includes:

[0146] Obtain historical pesticide action duration data of a single trench, which includes the fertilizer demand of the single trench, the pesticide water demand, the area of the trench directly exposed to sunlight, the water seepage speed of the trench, and the historical pesticide action duration of the single trench;

[0147] According to the fertilizer demand, pesticide water demand, area of trench directly exposed to sunlight, water seepage speed of the trench and corresponding historical pesticide action duration in different historical pesticide action duration data, a pesticide action duration prediction set is generated, and it is divided into a second training set and a second test set;

[0148] Construct a second convolutional neural network, and take the fertilizer demand, pesticide water demand, area of trench directly exposed to sunlight and water seepage speed of the trench in different historical water consumption data in the second training set as the input data of the second convolutional neural network, and take the corresponding historical pesticide action duration in the second training set as the output data of the second convolutional neural network;

[0149] Train the second convolutional neural network to obtain a second initial convolutional neural network, and use the second test set to verify the model of the second initial convolutional neural network, and output the second initial convolutional neural network with a second test error threshold less than or equal to the preset second test error threshold as the pesticide action duration prediction model;

[0150] Input the fertilizer demand, pesticide water demand, area of trench directly exposed to sunlight and water seepage speed of the trench of each trench into the pesticide action duration prediction model to obtain the predicted pesticide action duration of each trench.

[0151] It needs to be further explained that, in the specific implementation process, the process of generating the set of trenches needing assistance and the set of trenches that can be assisted according to the pesticide action duration of each trench includes:

[0152] Obtain the predicted pesticide action duration of each trench by the pesticide action duration prediction model in the area to be irrigated, calculate the average of the predicted pesticide action duration, and include the trenches with predicted pesticide action duration greater than the average into the set of trenches that can be assisted, and include the trenches with predicted pesticide action duration less than the average into the set of trenches that need assistance;

[0153] The assistable trench refers to a trench with a predicted pesticide action time longer than the average, and the trench in need of assistance refers to a trench with a predicted pesticide action time shorter than the average. The predicted pesticide action time of each trench is equalized by assisting the assistable trench to the trench in need of assistance, so as to uniformly apply pesticide subsequently. The assistance refers to transporting water in the assistable trench to the trench in need of assistance, and the water contains diffused pesticide. There is a time difference between the predicted pesticide action time of the assistable trench and the average. The time difference is converted into the water volume of the assistance transportation through calculation.

[0154] It needs to be further explained that, in the specific implementation process, the process of generating the assistance efficiency score of each assistable trench relative to each trench in need of assistance and assisting the assistable trench to the trench in need of assistance according to the assistance efficiency score includes:

[0155] The assistance efficiency score refers to the efficiency score of the assistable trench assisting the trench in need of assistance, so that the trench in need of assistance is assisted by the assistable trench with the highest assistance efficiency score according to the assistance efficiency score;

[0156] The assistance efficiency score related data includes: the relative difference, the distance between the two trenches and the effective transportation volume;

[0157] The absolute value of the difference between the predicted pesticide action time of the assistable trench and the average is the first difference, the absolute value of the difference between the predicted pesticide action time of the trench in need of assistance and the average is the second difference, and the absolute value of the first difference minus the second difference is the relative difference;

[0158] The effective transportation volume refers to the loss of water in the transportation process between the trenches. The effective transportation volume can be obtained by subtracting the loss from the total transportation volume;

[0159] Then the assistance efficiency score P can be obtained:

[0160]

[0161] Wherein:

[0162] α1, α2 and α3 are weight coefficients, which are obtained by training according to historical data, and specifically 0.4, 0.3 and 0.3 respectively;

[0163] E is the relative difference, E 可 is the predicted pesticide action time of the assistable trench, f is the distance between the two trenches, f max is the distance between the two farthest trenches in the irrigation area, and G is the effective transportation volume, G 总 is the total transportation volume;

[0164] If E is 1, E 可 is 10, f is 2, f max is 20, and G is 20, G总 is 25;

[0165] P=0.4x(1-0.1)+0.3x(1-0.1)+0.3x0.8=0.87.

[0166] The assistance efficiency score of each assistable trench relative to each trench in need of assistance is obtained, and the one with the highest assistance efficiency score is selected to assist the corresponding trench in need of assistance.

[0167] It should be further explained that in the specific implementation process, the execution module includes an irrigation device arranged in the irrigation area, the irrigation device includes an irrigation pipe network, a plurality of water outlets are arranged on the irrigation pipe network, each water outlet is directly opposite the root of the corresponding crop, and the water amount distributed to each crop by the irrigation device according to the predicted water consumption is used to irrigate the root of each crop, so that large-scale precision irrigation is realized. The execution module further includes a pesticide application device and a conveying device. The pesticide application device can add pesticides into the trench and uniformly lay the pesticides on the bottom of the trench. Water is also added into the trench through the water outlets on the irrigation pipe network, so that the water and the pesticides jointly act. The conveying device can convey water between any two trenches. According to the assistance efficiency score, a suitable trench is selected to convey the mixture of water and pesticides to the corresponding trench.

[0168] The above merely describes specific embodiments of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art can easily think of changes or replacements within the technical range disclosed in the present application, which should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.

[0169] Finally, the above merely describes preferred embodiments of the present application and is not used to limit the present application. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application should be included in the protection scope of the present application.

Claims

1. An integrated intelligent irrigation decision system based on multi-source data fusion, characterized in that, The method comprises the following modules: a data acquisition module for acquiring parameter data of all monitoring irrigation points in the farmland in real time, the parameter data comprising soil moisture content, air relative humidity, leaf water content and leaf area change amount; an irrigation decision module for generating water demand scores of all monitoring irrigation points according to the acquired parameter data, and deciding to irrigate according to the water demand scores; a pesticide demand amount prediction module for obtaining historical pesticide demand amount data and constructing a pesticide demand amount prediction model, and predicting the pesticide demand amount of the monitoring irrigation points by using the pesticide demand amount prediction model; a pesticide action duration prediction module for obtaining pesticide water demand amount required for the pesticide to play a role in the trench according to the pesticide demand amount, the pesticide water demand amount referring to the need to set a certain depth of water layer on the pesticide to assist the use of the pesticide, obtaining historical pesticide action duration data and constructing a pesticide action duration prediction model, and predicting the pesticide action duration in each trench by using the pesticide action duration prediction model, the pesticide action duration referring to the duration for which the pesticide plays a role in the trench in cooperation with water; a trench assistance module for generating a set of trenches requiring assistance and a set of assistable trenches according to the pesticide action duration of each trench, generating an assistance efficiency score of each assistable trench relative to each trench requiring assistance, and causing the assistable trenches to assist the trenches requiring assistance according to the assistance efficiency score; an execution module for irrigating near the roots of crops in the to-be-irrigated area, and for applying pesticides and water into the trenches, and for transporting water between any two trenches; The assistable trenches refer to the trenches with a predicted pesticide action duration longer than the average, and the trenches requiring assistance refer to the trenches with a predicted pesticide action duration shorter than the average. By causing the assistable trenches to assist the trenches requiring assistance, the predicted pesticide action duration of each trench is equalized, so as to uniformly apply pesticides subsequently. Assistance refers to transporting water in the assistable trenches into the trenches requiring assistance, the water containing diffused pesticides. There is a duration difference between the predicted pesticide action duration of the assistable trenches and the average, and the duration difference is converted into the amount of water to be transported by assistance through calculation.

2. The integrated smart irrigation decision system based on multi-source data fusion according to claim 1, wherein, The farmland is evenly divided into a plurality of to-be-irrigated areas, and a monitoring irrigation point is arranged at the center of each to-be-irrigated area; The dielectric constant k was measured by TDR time domain reflectometer in 0-20 cm and 20-40 cm soil layers, respectively, and the volumetric water content was calculated by using the calibration formula specific to soil type ; The regression coefficients a and b are calibration parameters; Calculate the volumetric water content of both soil layers 0-20 cm and 20-40 cm and and take a weighted average to get ; To normalize the soil water content SWC; The range of soil moisture content is - , is the minimum value of soil moisture content, is the maximum value of soil moisture content, which is applicable to precision irrigation control; The relative humidity of the air is collected in real time by a digital temperature and humidity sensor. ,right Normalization is performed to obtain the normalized relative humidity (RH). the range of the relative humidity of the air is - , the minimum value of the air humidity, the maximum value of the air humidity; Leaf reflectance at 1450 nm and 1940 nm wavelengths were measured by a portable near infrared spectrometer, respectively and Leaf water content was calculated using an empirical formula ; The coefficients 46.5 and 82.2 are obtained by multiple regression fitting; Randomly select 5 plants in the area to be irrigated, measure 3 leaves of each plant, and calculate the average moisture content of 15 leaves , and the normalized leaf water content LWC is obtained by normalization ​ The range of leaf water content is - , is the minimum leaf water content, is the maximum leaf water content; Real-time measurement of the effective leaf area index by means of a plant canopy analyzer in cloudless weather , calculating the difference between this week's effective leaf area index and last week's ; The daily variation of effective leaf area index is obtained by linear interpolation ; To The normalized effective leaf area index is obtained by normalizing ; The leaf area change amount ranges from - , is the minimum value of the effective leaf area index, is the maximum value of the effective leaf area index.

3. The integrated smart irrigation decision system based on multi-source data fusion according to claim 2, characterized in that, According to each parameter data, a corresponding water demand score S is obtained: wherein, , , and are weight coefficients, trained from historical data; A water demand threshold Z is set, and the water demand score S is compared with the water demand threshold Z to decide whether to trigger irrigation; When S < Z, irrigation is not triggered; When S ≥ Z, irrigation is triggered.

4. The integrated smart irrigation decision system based on multi-source data fusion according to claim 3, characterized in that, Factors affecting the pesticide demand amount of the monitoring irrigation point include the amount of pests, the amount of beneficial insects, the environmental temperature and the environmental humidity; Pest and beneficial insect attractors are arranged at a certain height from the ground, the pest attractor can release pest pheromones for a certain period of time, the beneficial insect attractor can release beneficial insect pheromones for a certain period of time, and an integrated camera captures images during this period, a YOLOv5 model is trained and deployed on an edge device to identify and count the amount of pests and beneficial insects in real time; The temperature and humidity sensor setting part is arranged in the crop canopy to collect the environmental temperature and humidity in real time and calculate the daily average temperature and humidity; The process of obtaining historical pesticide demand data and constructing a pesticide demand prediction model, and predicting the pesticide demand of the monitoring irrigation point using the pesticide demand prediction model includes: Setting a pesticide application cycle, obtaining historical pesticide demand data of a single monitoring irrigation point in different pesticide application cycles, the pesticide demand data including the average pest amount, beneficial insect amount, environmental temperature, and environmental humidity of the single monitoring irrigation point in different pesticide application cycles, and the historical pesticide demand of the single monitoring irrigation point in the pesticide application cycle; According to the pest amount, beneficial insect amount, environmental temperature, environmental humidity, and corresponding historical pesticide demand of the corresponding monitoring irrigation point in different historical pesticide demand data, a pesticide demand prediction set is generated, and it is divided into a first training set and a first test set; A first convolutional neural network is constructed, the pest amount, beneficial insect amount, environmental temperature, and environmental humidity in different historical pesticide demand data in the first training set are taken as input data of the first convolutional neural network, and the corresponding historical pesticide demand in the first training set is taken as output data of the first convolutional neural network; The first convolutional neural network is trained to obtain a first initial convolutional neural network, the first test set is used to verify the model of the first initial convolutional neural network, and the first initial convolutional neural network with an output less than or equal to a preset first test error threshold is taken as the pesticide demand prediction model; Every time a pesticide application cycle is reached, the average pest amount, beneficial insect amount, environmental temperature, and environmental humidity of each monitoring irrigation point in the pesticide application cycle are input into the pesticide demand prediction model to obtain the predicted pesticide demand of each monitoring irrigation point.

5. The integrated smart irrigation decision system based on multi-source data fusion according to claim 4, characterized in that, The process of obtaining the pesticide water demand required for the pesticide to play a role in the trench from the pesticide demand includes: The optimal range of water layer depth is M-N, and the median of M and N is U; The farmland is set as a dry field, the pesticide used is a solid particle pesticide, a plurality of trenches are dug in each irrigation area to be irrigated, the trenches are located on one side of the crop roots, the length, width, and depth of the trenches are fixed values, that is, the bottom area D of the trench is a fixed value, the predicted pesticide demand required by each monitoring irrigation point is divided by the number of trenches therein to obtain the pesticide amount required by each trench, the volume K of the pesticide can be obtained according to the required pesticide amount, the volume of a single solid pesticide particle, and the weight, and the depth Q of the pesticide in the trench can be obtained according to the bottom area of the trench, that is, the optimal depth range of water in the trench is (Q+M)-(Q+N), and the pesticide water demand in a single trench is (D×Q-K)+D×U according to the bottom area of the trench, the volume K of the pesticide, and the depth K of the pesticide.

6. The integrated smart irrigation decision system based on multi-source data fusion according to claim 5, characterized in that, The factors affecting the duration of the pesticide in the trench include: fertilizer demand, pesticide water demand, sunlight direct radiation area of the trench, and water seepage speed of the trench; The historical fertilizer demand is obtained by obtaining the historical data of the fertilizer used in a single trench, and the average of the historical fertilizer demand is calculated to obtain the fertilizer demand; The pesticide water demand in a single trench is obtained through the above calculation; By taking a trench photo with a high-resolution device, ensuring that the shadow is clear, identifying the shadow area through color space conversion and brightness contrast analysis, and segmenting the trench contour with a YOLOv8 model, the area of the trench directly exposed to sunlight can be obtained by subtracting the shadow area from the total area of the trench; A certain amount of pesticide and water is added to the trench, the pesticide will diffuse into the water, and the volume reduction of the water is measured within a certain time, and the water reduction is divided by the time to obtain the water seepage speed of the trench; The process of obtaining historical pesticide action duration data and constructing a pesticide action duration prediction model includes: Obtain the historical pesticide action duration data of a single trench, which includes the fertilizer demand, pesticide water demand, trench direct sunlight area, trench water seepage speed, and historical pesticide action duration of the single trench; According to the fertilizer demand, pesticide water demand, trench direct sunlight area, trench water seepage speed, and corresponding historical pesticide action duration in different historical pesticide action duration data, a pesticide action duration prediction set is generated, and it is divided into a second training set and a second test set; A second convolutional neural network is constructed, the fertilizer demand, pesticide water demand, trench direct sunlight area, and trench water seepage speed in different historical water consumption data in the second training set are used as input data of the second convolutional neural network, and the corresponding historical pesticide action duration in the second training set is used as output data of the second convolutional neural network; The second convolutional neural network is trained to obtain a second initial convolutional neural network, and the second test set is used to verify the model of the second initial convolutional neural network, and the second initial convolutional neural network with a second test error threshold less than or equal to a preset second test error threshold is output as a pesticide action duration prediction model; The fertilizer demand, pesticide water demand, trench direct sunlight area, and trench water seepage speed of each trench are input into the pesticide action duration prediction model to obtain the predicted pesticide action duration of each trench.

7. The integrated smart irrigation decision system based on multi-source data fusion according to claim 6, characterized in that, The process of generating a set of trenches that need assistance and a set of trenches that can assist according to the pesticide action duration of each trench includes: The predicted pesticide action duration of each trench is obtained by the pesticide action duration prediction model in the area to be irrigated, and the average of the predicted pesticide action duration is calculated, the trenches with a predicted pesticide action duration greater than the average are included in the set of trenches that can be assisted, and the trenches with a predicted pesticide action duration less than the average are included in the set of trenches that need assistance.

8. The integrated smart irrigation decision system based on multi-source data fusion according to claim 7, characterized in that, The process of generating an assistance efficiency score of each assistable trench relative to each trench that needs assistance and assisting the trenches that need assistance according to the assistance efficiency score includes: The assistance efficiency score refers to the efficiency score of the assistable trench assisting the trench that needs assistance, and the trench that needs assistance is assisted by the assistable trench with the highest assistance efficiency score according to the assistance efficiency score; Collect assistance efficiency score related data, including relative difference, distance between two trenches, and delivery effective amount; The absolute value of the difference between the predicted pesticide action duration of the assistable trench and the average is a first difference value, the absolute value of the difference between the predicted pesticide action duration of the trench needing assistance and the average is a second difference value, and the absolute value of the first difference value minus the second difference value is a relative difference value; The effective amount of delivery refers to the loss of water during the delivery process between the trenches, and the effective amount of delivery can be obtained by subtracting the loss amount from the total amount of delivery; The assistance efficiency score P of each assistable trench relative to each trench needing assistance is obtained: Wherein: , and are weight coefficients, trained according to historical data; E is the relative difference, is the predicted pesticide action duration of the aidable trench, f is the distance between two trenches, is the distance between the two most distant trenches within the area to be irrigated, G is the delivered effective amount, is the delivered total amount; The assistance efficiency score P of each assistable trench relative to each trench needing assistance is obtained:

9. The integrated smart irrigation decision system based on multi-source data fusion according to claim 8, characterized in that, The assistance efficiency score P of each assistable trench relative to each trench needing assistance is obtained: The execution module includes an irrigation device arranged in the irrigation area, the irrigation device includes an irrigation pipe network, and a plurality of water outlets are arranged on the irrigation pipe network, each water outlet is directly opposite the root of the corresponding crop, the water amount allocated to each crop by the irrigation device according to the predicted water consumption is used to irrigate the root of each crop, large-scale precision irrigation is realized, the execution module further includes a pesticide application device and a delivery device, the pesticide application device can add pesticides into the trench and uniformly lay the pesticides on the bottom of the trench, water is also added into the trench through the water outlets on the irrigation pipe network, water and pesticides jointly act, and the delivery device can deliver water between any two trenches, and the mixture of water and pesticides is delivered from a suitable trench to a corresponding trench according to the assistance efficiency score.

Images