A precision irrigation system and method for crops based on root water demand signals

Abstract

This invention discloses a precision irrigation system and method for crops based on root water demand signals, belonging to the field of precision irrigation technology for agricultural roots. It includes a signal sensing module, an intelligent decision-making module, a precision execution module, a data storage module, and a feedback unit. The signal sensing module consists of a root image visual sensor, a stem micro-change sensor, a canopy infrared temperature sensor, and a root zone microclimate sensor array. This invention utilizes non-destructive multimodal sensing technology, comprehensively employing multi-sensor fusion sensing such as root window vision, stem flow or stem micro-changes, canopy infrared temperature, and root zone water potential to directly capture early water demand and stress signals released by root physiological activities. Furthermore, for mixed or vertical planting scenarios, it achieves specific identification of water demand signals for different crops and different vertical spaces through three-dimensional adaptive deployment and a multi-crop signal feature library.

Description

Technical Field

[0001] This invention relates to the field of precision irrigation technology for agricultural roots, and in particular to a precision irrigation system and method for crops based on root water demand signals. Background Technology

[0002] With the rapid development of intelligent agriculture, precision irrigation technology has become one of the core means to solve water shortages and improve agricultural production efficiency. Although existing precision irrigation technology has moved away from the extensive mode of traditional flood irrigation and is gradually upgrading to data-driven approaches, it has not yet broken through the traditional decision-making logic centered on the soil environment and cannot accurately match the actual water needs of crop roots. This is specifically reflected in the following: First, the sensing is delayed and indirect, leading to passive and inefficient irrigation decisions. Currently, most mainstream irrigation systems rely on soil moisture sensors. However, because soil acts as a water buffer, changes in soil moisture lag behind changes in crop water requirements. This means that changes in soil moisture cannot reflect the water absorption status of crop roots in real time, and it cannot distinguish between water evaporation and root absorption. This results in significant errors in judging the timing of root water demand. Often, irrigation is only initiated passively after crops have already suffered mild to moderate water stress and their physiological growth has been inhibited, failing to provide proactive water replenishment. Furthermore, in scenarios involving mixed or vertical planting of multiple crops, the superimposed water demand signals from different crops exacerbate the decision-making errors caused by the lag.

[0003] Secondly, the root system is the core organ for crop water absorption. Its physiological activities, such as water absorption rate, root pressure fluctuations, changes in root exudate composition, and root tip activity, release early water demand signals or stress signals, which are the most direct basis for reflecting the true water demand state of crops. However, traditional irrigation technology essentially manages soil moisture rather than responding to crop water demand signals, resulting in a disconnect between irrigation decisions and actual crop needs, making it impossible to achieve true on-demand irrigation. This is especially true when different crops are mixed or vertically planted, where their root signals differ significantly, and existing technologies cannot capture them specifically. Summary of the Invention

[0004] The purpose of this invention is to solve the problems existing in the prior art by proposing a precision irrigation system and method for crops based on root water demand signals.

[0005] To achieve the above objectives, the present invention adopts the following technical solution: A precision irrigation system for crops based on root water demand signals includes a signal sensing module, an intelligent decision-making module, a precision execution module, a data storage module, and a feedback unit. The signal sensing module consists of a root image visual sensor, a stem micro-change sensor, a canopy infrared temperature sensor, and a root microclimate sensor array. The signal sensing module captures root water requirement signals and growth environment signals of different crops in mixed or vertical planting scenarios from multiple dimensions, providing accurate data support for the intelligent decision-making module. The intelligent decision-making module consists of an embedded controller, a short-term water demand prediction AI model, and a multi-objective optimizer. The intelligent decision-making module receives real-time data from the signal sensing module and generates a three-dimensional dynamic irrigation decision based on crop type, vertical layer, and growth stage through data fusion, AI prediction, and multi-objective optimization. The precision execution module consists of a nutrient solution dynamic adjustment unit, a nutrient solution circulation system, a zoned matrix electromagnetic valve array, and a reflux recovery instant sterilization unit. The precision execution module receives three-dimensional dynamic irrigation decisions from the intelligent decision module to achieve precise irrigation and nutrient solution recycling for different crops and different vertical layers. The data storage module is equipped with a data storage device, and the feedback unit is equipped with a wireless communication module. Users can remotely monitor signal changes captured by the signal sensing module through the feedback unit and adjust irrigation decisions. The data storage module and the feedback unit record sensing data, irrigation parameters and crop growth status data through three-dimensional annotation rules, providing data support for AI model self-learning optimization, while realizing data visualization and remote control.

[0006] As a preferred embodiment, the root image visual sensor uses a high-definition anti-fog micro-root window camera, equipped with an 850nm near-infrared supplementary light, and is embedded in the side wall of the crop culture root box, maintaining a distance of 5-8cm from the root system. It is equipped with an image analysis algorithm to automatically segment the root images of the crop and extract characteristic parameters such as the average diameter of the main root, the number of newly grown white root tips, the density of fibrous roots, and the root hydration status of different crops.

[0007] As a preferred embodiment, the canopy infrared temperature sensor uses an infrared thermal imaging camera with a temperature measurement range of -20℃ to 120℃ and a measurement accuracy of ±0.5℃. It is installed on the top of a vertical frame, maintaining a distance of 1.5-2m from the canopy to ensure coverage of all crop planting zones. The difference between the temperature of the two canopy layers and the ambient air temperature is calculated using a matching algorithm to help determine the degree of water stress.

[0008] As a preferred embodiment, the root zone microclimate sensor array includes 3 to 5 miniature temperature, humidity, and water potential sensors arranged in each micro-planting unit. The miniature temperature, humidity, and water potential sensors are all embedded. The temperature, humidity, and water potential data of the root zone of each unit are collected in real time through the temperature, humidity, and water potential sensors to draw a three-dimensional temperature, humidity, and water potential field.

[0009] As a preferred embodiment, the embedded controller adopts an industrial-grade microcontroller with a main frequency of ≥1GHz, memory of ≥2GB, storage capacity of ≥16GB, supports multi-channel data input and output, and can simultaneously receive 8-16 channels of sensor data, covering multi-layer sensing data of various crops, with a response time of ≤100ms.

[0010] As a preferred embodiment, the short-term water requirement prediction AI model is constructed based on an LSTM network. The input layer contains 8 neurons, which correspond to crop type, water stress index, light intensity, air temperature and humidity, ventilation volume, growth stage, vertical layer height, and planting density, respectively. The hidden layer contains 32 neurons, and the output layer contains 2 neurons, which correspond to the crop's water requirement time point and water requirement intensity, respectively. The short-term water requirement prediction AI model has a built-in dedicated sub-model for cultivated crops, and uses planting data from multiple crops and multiple scenarios for initial training, and continuously learns and optimizes through feedback data.

[0011] As a preferred embodiment, the multi-objective optimizer is constructed using a genetic algorithm. The optimization objective functions include: minimizing irrigation duration, minimizing the number of pump start-ups and shutdowns, maximizing root health, and maximizing the yield of multiple crops in a coordinated manner. The multi-objective optimizer has built-in water demand conflict coordination rules to output the optimal combination of irrigation parameters and generate dynamically adjustable three-dimensional irrigation decisions.

[0012] As a preferred embodiment, the nutrient solution dynamic adjustment unit includes a central nutrient solution mother liquor tank, an acid solution tank, an alkali solution tank, a general concentrated fertilizer solution tank, a crop-specific potassium and calcium fertilizer tank, and a dedicated nitrogen and phosphorus fertilizer tank. The nutrient solution dynamic adjustment unit is equipped with 6 high-precision metering pumps that adjust the nutrient solution parameters in real time according to irrigation decisions, and the dissolved oxygen content of the crop nutrient solution is adjusted according to the different crop requirements.

[0013] A precision irrigation method for crops based on root water demand signals includes the following steps: Step S1: Data Acquisition. Various sensors in the signal sensing module collect data in real time at preset frequencies. For example, the root image vision sensor takes a picture of the crop's root system every 10 minutes and extracts characteristic parameters such as the water absorption activity of the main root and the density of the fibrous roots. The stem micro-change sensor continuously monitors the stem diameter changes of various crops through high-frequency sampling. The canopy infrared temperature sensor collects layered canopy temperature data every 3 minutes. The root microclimate sensor array collects temperature, humidity, and water potential data of each micro-planting unit every 8 minutes. All raw data are converted from analog to digital signals by the data acquisition unit and transmitted to the intelligent decision-making module. Step S2: Data preprocessing and fusion. The data preprocessing unit of the intelligent decision-making module receives various types of sensing data, removes noise interference through the Kalman filter algorithm, performs time alignment and standardization on heterogeneous data of different dimensions and frequencies, and then generates crop water stress indices for the upper and lower crops through the data fusion algorithm, thereby intuitively reflecting the real water demand status of the two crops. Step S3: Short-term water demand prediction and irrigation decision generation. The short-term water demand prediction AI model receives water stress index, environmental data and growth stage data for different crops. Combined with the built-in crop growth mechanism database, it accurately predicts the time point and water demand intensity of different crops reaching the preset stress threshold in the next 2-6 hours. The multi-objective optimizer uses a genetic algorithm for optimization. If the water demand times of two crops overlap, they are sorted by priority. Dynamic irrigation decisions are generated by adjusting the irrigation duration. Step S4: Precision irrigation execution and nutrient solution circulation. The precision execution module receives irrigation decisions. The nutrient solution dynamic adjustment unit injects crop-specific potassium and calcium fertilizer solution, nitrogen and phosphorus fertilizer solution, and acid or alkaline solution into the central mother liquor tank through a high-precision metering pump according to the decision parameters. The nutrient solution parameters are adjusted in real time. The circulation pump delivers the prepared nutrient solution to each layer of cultivation tank. The central controller controls the opening and closing of the partition matrix electromagnetic valve array in milliseconds to achieve differentiated irrigation for the tomato layer and lettuce layer. Nutrient solution that is not absorbed by the roots is collected according to crop type, flows through the filter device and UV-CLED sterilization channel to remove harmful microorganisms and impurities, and then flows back to the nutrient solution preparation branch of the corresponding crop to participate in the next cycle. Step S5: Data feedback and model self-learning optimization. The data storage and feedback unit records the sensing data, irrigation decision parameters, and crop growth status data in real time according to the three-dimensional labeling rules of crop, level, and time, and feeds the data back to the intelligent decision module. The short-term water demand prediction AI model uses the feedback data as a sample to continuously optimize the parameters of different crop-specific sub-models, update the multi-crop water demand signal feature library, adjust the water demand conflict coordination rules, and realize the intelligent iterative upgrade of the system.

[0014] Compared with the prior art, the beneficial effects of the present invention are as follows: 1. This invention utilizes non-destructive multimodal sensing technology, integrating multiple sensors such as root window vision, stem flow or stem micro-changes, canopy infrared temperature, and root zone water potential to directly capture early water demand and stress signals released by root physiological activities. For mixed or vertical planting scenarios, it achieves specific identification of water demand signals for different crops and different vertical spaces through three-dimensional adaptive deployment and multi-crop signal feature database. This avoids the lag, indirectness, and interference from mixed planting signals in traditional soil moisture sensing, breaking through the sensing bottleneck and realizing direct, real-time, and non-destructive sensing of root water demand signals, resulting in more accurate irrigation decisions.

[0015] 2. This invention accurately captures the real water demand signals of the root system, combines them with AI models to predict the short-term water demand patterns of crops, and generates personalized irrigation decisions. This transforms irrigation decisions from responding to changes in soil moisture signals to responding to crop water demand signals, avoiding ineffective and passive irrigation, thereby achieving true on-demand irrigation, significantly improving water and fertilizer utilization efficiency, and saving resources.

[0016] 3. Based on the independent control design of the partition and the special optimization of mixed or vertical planting, the system can flexibly cope with high-density and heterogeneous planting scenarios such as vertical farms, multi-layer cultivation, and intercropping. It is especially suitable for the mixed or vertical planting needs of multiple crops in hydroponics and aeroponics. Differentiated irrigation strategies are formulated for different types of crops, different growth stages, and different spatial locations. The system can accurately control the water and air balance in the root zone, promote the healthy development of the root system, and optimize the physiological metabolism of crops, thereby achieving personalized execution, adapting to heterogeneous planting scenarios, and improving crop yield and quality.

[0017] 4. This invention constructs a unified intelligent decision-making framework. By adjusting the adaptation components of the sensing terminal and the execution terminal, it can be seamlessly applied to various mainstream cultivation modes such as soil cultivation, hydroponics, and aeroponics. At the same time, it is specifically adapted to various mixed or vertical planting scenarios of multiple crops under each mode, thereby adapting to multiple cultivation modes and mixed or vertical planting scenarios, reducing the application threshold, and facilitating large-scale promotion. Attached Figure Description

[0018] Figure 1 This is a flowchart illustrating the overall workflow of a precision irrigation system for crops based on root water demand signals, as proposed in this invention. Figure 2 This is a flowchart of the data acquisition process of the signal sensing module in a precision irrigation system for crops based on root water demand signals, as proposed in this invention. Figure 3 This is a data processing flowchart of the intelligent decision-making module in a precision irrigation system for crops based on root water demand signals, as proposed in this invention. Figure 4 This is a flowchart of a precision irrigation method for crops based on root water demand signals proposed in this invention. Detailed Implementation

[0019] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0020] It should be understood that, when used in this specification and the appended claims, the terms "comprising" and "including" indicate the presence of the described features, integrals, steps, operations, elements and / or components, but do not exclude the presence or addition of one or more other features, integrals, steps, operations, elements, components and / or collections thereof.

[0021] It should also be understood that the terminology used in this specification is for the purpose of describing particular embodiments only and is not intended to limit the invention. As used in this specification and the appended claims, the singular forms “a,” “an,” and “the” are intended to include the plural forms unless the context clearly indicates otherwise.

[0022] It should also be further understood that the term “and or or” as used in this specification and the appended claims refers to any combination of one or more of the associated listed items and all possible combinations, and includes such combinations.

[0023] Example 1, refer to Figures 1 to 3 A precision irrigation system for crops based on root water demand signals includes a signal sensing module, an intelligent decision-making module, a precision execution module, a data storage module, and a feedback unit. The signal sensing module consists of a root image visual sensor, a stem micro-change sensor, a canopy infrared temperature sensor, and a root microclimate sensor array. The signal sensing module captures root water requirement signals and growth environment signals of different crops in mixed or vertical planting scenarios from multiple dimensions, providing accurate data support for the intelligent decision-making module. Furthermore, the root image visual sensor uses a high-definition anti-fog micro-root window camera, equipped with an 850nm near-infrared supplementary light, and is embedded in the side wall of the crop culture root box, maintaining a distance of 5-8cm from the root system. It is equipped with an image analysis algorithm to automatically segment the root images of the crop and extract characteristic parameters such as the average diameter of the main root, the number of newly grown white root tips, the density of fibrous roots, and the root hydration status of different crops. The canopy infrared temperature sensor uses an infrared thermal imaging camera with a temperature measurement range of -20℃ to 120℃ and a measurement accuracy of ±0.5℃. It is installed on the top of a vertical frame, maintaining a distance of 1.5-2m from the canopy to ensure coverage of all crop planting zones. The sensor uses a matching algorithm to calculate the temperature difference between the two canopy layers and the ambient air temperature to help determine the degree of water stress. The root zone microclimate sensor array includes 3 to 5 miniature temperature, humidity, and water potential sensors arranged in each micro-planting unit. The miniature temperature, humidity, and water potential sensors are all embedded. The temperature, humidity, and water potential data of the root zone of each unit are collected in real time through the temperature, humidity, and water potential sensors to create a three-dimensional temperature, humidity, and water potential field.

[0024] The embedded controller uses an industrial-grade microcontroller with a main frequency of ≥1GHz, memory of ≥2GB, and storage capacity of ≥16GB. It supports multi-channel data input and output, and can simultaneously receive data from 8-16 sensors, covering multi-layer sensing data of various crops, with a response time of ≤100ms.

[0025] The intelligent decision-making module consists of an embedded controller, a short-term water demand prediction AI model, and a multi-objective optimizer. The intelligent decision-making module receives real-time data from the signal sensing module and generates a three-dimensional dynamic irrigation decision based on crop type, vertical layer, and growth stage through data fusion, AI prediction, and multi-objective optimization. Furthermore, the short-term water requirement prediction AI model is built on an LSTM network. The input layer contains 8 neurons, which correspond to crop type, water stress index, light intensity, air temperature and humidity, ventilation, growth stage, vertical layer height, and planting density, respectively. The hidden layer contains 32 neurons, and the output layer contains 2 neurons, which correspond to the water requirement time point and water requirement intensity of the crop, respectively. The short-term water requirement prediction AI model has a built-in sub-model for cultivated crops. It is trained in the early stage using planting data of multiple crops and multiple scenarios, and continuously learns and optimizes itself through feedback data. Furthermore, the multi-objective optimizer is constructed using a genetic algorithm, and the optimization objective functions include: minimizing irrigation duration, minimizing the number of pump start-ups and shutdowns, maximizing root health, and maximizing the yield of multiple crops in a coordinated manner. The multi-objective optimizer has built-in water demand conflict coordination rules to output the optimal combination of irrigation parameters and generate dynamically adjustable three-dimensional irrigation decisions.

[0026] The precision execution module consists of a nutrient solution dynamic adjustment unit, a nutrient solution circulation system, a zoned matrix solenoid valve array, and a reflux recovery instant sterilization unit. The precision execution module receives three-dimensional dynamic irrigation decisions from the intelligent decision module to achieve precise irrigation and nutrient solution recycling for different crops and different vertical layers. Furthermore, the nutrient solution dynamic adjustment unit includes a central nutrient solution mother liquor tank, an acid solution tank, an alkali solution tank, a general concentrated fertilizer solution tank, a crop-specific potassium and calcium fertilizer tank, and a dedicated nitrogen and phosphorus fertilizer tank. The nutrient solution dynamic adjustment unit is equipped with 6 high-precision metering pumps that adjust the nutrient solution parameters in real time according to irrigation decisions, and the dissolved oxygen content of the crop nutrient solution is adjusted according to the different crop requirements. The data storage module is equipped with a data storage device, and the feedback unit is equipped with a wireless communication module. Users can remotely monitor signal changes captured by the signal sensing module through the feedback unit and adjust irrigation decisions. The data storage module and the feedback unit record sensing data, irrigation parameters and crop growth status data through three-dimensional annotation rules, providing data support for the self-learning optimization of the AI ​​model, while realizing data visualization and remote control. Example

[0027] This embodiment takes a deep-flow hydroponic tomato and lettuce mixed vertical planting scenario as an example. This scenario has both crop heterogeneity and vertical spatial heterogeneity, which can fully demonstrate the technical advantages and implementation details of the present invention. At the same time, the technical principle of this embodiment can also be applied to other mixed planting combinations, such as strawberry and vanilla mixed planting, cucumber and romaine lettuce mixed planting, aeroponic mixed and vertical planting scenarios. Only adaptive adjustments need to be made to the perception module and execution module, without changing the core decision-making logic. Reference Figure 4 A precision irrigation method for crops based on root water demand signals includes the following steps: Step S1: Data Acquisition. Various sensors in the signal sensing module collect data in real time at preset frequencies. For example, the root image vision sensor takes a picture of the crop's root system every 10 minutes and extracts characteristic parameters such as the water absorption activity of the main root and the density of the fibrous roots. The stem micro-change sensor continuously monitors the stem diameter changes of various crops through high-frequency sampling. The canopy infrared temperature sensor collects layered canopy temperature data every 3 minutes. The root microclimate sensor array collects temperature, humidity, and water potential data of each micro-planting unit every 8 minutes. All raw data are converted from analog to digital signals by the data acquisition unit and transmitted to the intelligent decision-making module. Based on the above, the stem micro-change sensor adopts a high-precision strain gauge sensor with a resolution of up to 0.1μm and a measurement range of ±500μm. It is fixed to a designated position on the stems of tomatoes and lettuce using a non-destructive binding method, with a flexible buffer pad wrapped around the binding point to avoid damaging the stems. The sensor outputs an analog signal, which is converted into a digital signal by a data acquisition unit and transmitted to the decision module in real time. It can accurately capture the diurnal shrinkage and expansion changes of the stems of the two crops and promptly identify early water stress signals. The shrinkage threshold for tomatoes is ≤50μm, and for lettuce it is ≤30μm. Miniature temperature, humidity, and water potential sensors are installed at a depth of 15-20 cm in the tomato layer and 5-10 cm in the lettuce layer, evenly distributed in the root growth areas of both crops. The temperature and humidity sensors measure temperature from -10℃ to 60℃ and humidity from 0 to 100%RH, with an accuracy of ±0.3℃ and ±2%RH. The water potential sensors measure from -10 to 0 bar, with an accuracy of ±0.1 bar. They can collect real-time data on temperature, humidity, and water potential in the root zone of each unit and plot three-dimensional temperature, humidity, and water potential fields.

[0028] Step S2: Data preprocessing and fusion. The data preprocessing unit of the intelligent decision-making module receives various types of sensing data, removes noise interference through the Kalman filter algorithm, performs time alignment and standardization on heterogeneous data of different dimensions and frequencies, and then generates crop water stress indices for the upper and lower crops through the data fusion algorithm, thereby intuitively reflecting the real water demand status of the two crops. Based on the above, the calculation formulas involved in the Kalman filtering process are as follows: Equations of state:

[0029] in, For the first The system status value at any given time represents the actual signals from the sensors, such as changes in stem diameter, crown temperature difference, and root zone water potential. For the first System state values ​​at any given time; The state transition matrix is ​​set to 1 in this invention because the sensor signal is stable in the short term and there are no significant state changes. To control the input, it is set to 0 in this invention, requiring no additional control intervention; To control the input matrix, which quantifies the effect of controllable external intervention on the actual signal system state of the sensor, this invention defaults to... This means there is no active control input; non-zero values ​​are obtained only through experimental calibration when it is necessary to actively adjust sensor signals or system status. value; For process noise, obey , The process noise covariance matrix is ​​calibrated based on sensor type. Observation equation:

[0030] in, For the first Time-based sensor observations, such as measured values ​​from stem micro-change sensors and measured values ​​from canopy infrared temperature sensors; The observation matrix is ​​set to 1 in this invention, and the observed values ​​are directly related to the state values; To observe noise, obey , The noise covariance matrix is ​​determined by the sensor accuracy parameters to observe it; Filtering update equation:

[0031] in, For the first The optimal estimated value after filtering at time step; Kalman gain is used to balance the weights of predicted and observed values. For the first Time-estimation error covariance matrix; It is the identity matrix; The weighted calculation formula for the Crop Water Stress Index (WSI) during data fusion is as follows:

[0032] The normalization formulas for the data from each sensor are as follows:

[0033] The final WSI quantization formula is as follows:

[0034] in, The normalized water stress index is located in the range [0, 1]. The larger the value, the more urgent the crop's water demand. The quantified water stress index, with a value of 0-10, is the core basis for decision-making. , , and These are sensor weighting coefficients, summing to 1. The calibration value in this invention is: (Root system vision) (Slight changes in the stem) (Canopy Infrared) (Root zone water potential); The normalized values ​​of the root system visual sensor are obtained by collecting root system images (every 10 minutes or one shot) using a root system image visual sensor. The average root diameter, number of new white root tips, and grayscale values ​​of water potential status are extracted by the image segmentation algorithm and then normalized (0-10 minutes). The normalized value of the stem micro-change sensor is obtained by comparing the daily stem contraction amount (resolution 0.1μm) collected by the stem micro-change sensor (high frequency sampling) with the preset normal contraction amount and normalizing it (0-10 points). The normalized value of the canopy infrared sensor is obtained by collecting canopy temperature (every 5 minutes or times) from the canopy infrared temperature sensor and combining it with the ambient air temperature collected by the root zone microclimate sensor. The difference is calculated and then normalized (0-10 minutes). The normalized value of the root zone microclimate sensor is obtained by collecting the water potential of the root zone medium (measurement range -10 to 0 bar) by the root zone microclimate sensor array (8 minutes or times), comparing it with the suitable water potential range for crops, and normalizing it (0-10 minutes). For the first Raw measurement values ​​from the sensor; For the first The minimum effective threshold of the sensor, such as the minimum measurable value of stem micro-change in this invention, is 0.1 μm; For the first The maximum effective threshold of the sensor, such as the crown temperature difference moderate stress threshold of 3.5℃ in this invention.

[0035] Step S3: Short-term water demand prediction and irrigation decision generation. The short-term water demand prediction AI model receives water stress index, environmental data and growth stage data for different crops. Combined with the built-in crop growth mechanism database, it accurately predicts the time point and water demand intensity of different crops reaching the preset stress threshold in the next 2-6 hours. The multi-objective optimizer uses a genetic algorithm for optimization. If the water demand times of two crops overlap, they are sorted by priority. Dynamic irrigation decisions are generated by adjusting the irrigation duration. Based on the above, the prediction formulas for the hidden layer, output layer, and water requirement time in the short-term water requirement prediction process for different crops are as follows: Hidden layer state update:

[0036] in, For the LSTM hidden layer Real-time output, integrating historical and current data features; For the LSTM hidden layer Output at all times; For the first The time-input vector, in this invention, includes: history Light intensity, air temperature and humidity, ventilation volume and crop growth stage are encoded (collected by the sensing module, called by the data storage unit and encoded according to preset rules); It stores long-term characteristic information for the cell state; or The forget gate or input gate output is located in the interval [0, 1] and controls feature forgetting and updating; Water demand intensity prediction for output layer:

[0037] in, For the first Predicted water demand intensity at any given time (unit: L / m²·h); or or The weight matrix is ​​obtained through training with multi-crop planting data; or or The bias vector is obtained through optimization during training. These are activation functions, such as the sigmoid function and the hyperbolic tangent function; The maximum water demand intensity is set to 5 L / m²·h in this invention to suit facility agriculture scenarios; Water demand time point forecast:

[0038] in, The forecast time points for water demand (unit: h) range 2-6 hours; for The trigger threshold is set to 6 points in this invention, corresponding to the critical value of moderate stress. This refers to the current moment.

[0039] Step S4: Precision irrigation execution and nutrient solution circulation. The precision execution module receives irrigation decisions. The nutrient solution dynamic adjustment unit injects crop-specific potassium and calcium fertilizer solution, nitrogen and phosphorus fertilizer solution, and acid or alkaline solution into the central mother liquor tank through a high-precision metering pump according to the decision parameters. The nutrient solution parameters are adjusted in real time. The circulation pump delivers the prepared nutrient solution to each layer of cultivation tank. The central controller controls the opening and closing of the partition matrix electromagnetic valve array in milliseconds to achieve differentiated irrigation for the tomato layer and lettuce layer. Nutrient solution that is not absorbed by the roots is collected according to crop type, flows through the filter device and UV-CLED sterilization channel to remove harmful microorganisms and impurities, and then flows back to the nutrient solution preparation branch of the corresponding crop to participate in the next cycle. Based on the above, the nutrient solution dynamic adjustment unit includes one central nutrient solution stock tank (capacity 1000L), two acid solution tanks (capacity 100L), two alkali solution tanks (capacity 100L), two general-purpose concentrated fertilizer solution tanks (capacity 200L), one tomato-specific potassium and calcium fertilizer tank (capacity 200L), and one lettuce-specific nitrogen and phosphorus fertilizer tank (capacity 200L); it is equipped with six high-precision metering pumps (flow range 0.1-10L or h, accuracy ±0.01L or h), which can adjust the nutrient solution parameters in real time according to irrigation decisions, with parameter fluctuations not exceeding ±0.1mS or cm (EC value) and ±0.2 (pH value); the dissolved oxygen content of the nutrient solution in the tomato layer is controlled at 7-8mg or L, and in the lettuce layer at 8-10mg or L; The circulating pumps in the nutrient solution circulation system and the partitioned matrix solenoid valve array are made of stainless steel, with a rated flow rate of 10-20 m³ or h and flow rate regulation function; the water flow velocity in the tomato layer cultivation trough is controlled at 0.8-1.0 L or min, and in the lettuce layer at 0.5-0.7 L or min; the slope of the cultivation trough is adjusted according to the vertical layer, 1.5% for the tomato layer and 1.0% for the lettuce layer; the solenoid valve array adopts a matrix layout, with each micro-planting unit corresponding to one independent solenoid valve, with a response time ≤10 ms, to achieve differentiated irrigation; one-way valves and 5μm independent filters are installed in the circulation branches of different crops to avoid cross-contamination; The reflux recovery and instant sterilization unit uses a stainless steel collection tank with a 15° tilt angle and is designed with separate channels according to crop type for easy collection of nutrient solution. The UV-CLED sterilization channel is 1m long and equipped with 10 sets of UV-CLED lamps (wavelength 254nm, power 10W or more), with a sterilization efficiency of ≥99%. The reflux pipe is made of PVC and is arranged according to crop type, equipped with 2 reflux pumps (flow rate 5-10m³ or h) to ensure that the nutrient solution for tomatoes and lettuce is quickly returned to the corresponding dispensing branch.

[0040] Step S5: Data feedback and model self-learning optimization. The data storage and feedback unit records the sensing data, irrigation decision parameters, and crop growth status data in real time according to the three-dimensional labeling rules of crop, level, and time, and feeds the data back to the intelligent decision module. The short-term water demand prediction AI model uses the feedback data as a sample to continuously optimize the parameters of different crop-specific sub-models, update the multi-crop water demand signal feature library, adjust the water demand conflict coordination rules, and realize the intelligent iterative upgrade of the system. Based on the above, the optimization formulas involved in optimizing water conservation, energy conservation, and root zone health objectives are as follows: Fitness function of genetic algorithm:

[0041] Its sub-objective functions are as follows: Water conservation goals:

[0042] Energy saving target:

[0043] Root domain health goals:

[0044] in, This is the overall fitness value, located in the range [0, 1]. The larger the value, the better the optimization effect. , and The target weights, summed to 1, are the calibration values ​​in this invention: =0.4、 =0.3、 =0.3; The target function value for water conservation is located in the interval [0, 1], reflecting the proportion of water conservation. The actual irrigation water volume of the invention system (unit: L) is collected by the flow sensor of the precision execution module (resolution 0.01L). The amount of irrigation water used in traditional irrigation methods (unit: L); The value of the energy-saving objective function is located in the interval [0, 1], reflecting the energy-saving ratio; The actual energy consumption of the invention system (unit: kWh) is collected by the energy consumption monitoring module of the actuators such as water pumps and solenoid valves (resolution 0.01 kWh). Energy consumption for traditional irrigation methods (unit: kWh); The root domain health objective function value is located in the interval [0, 1], reflecting the root domain water potential fit. The actual root domain water potential (unit: bar) is directly measured by the water potential sensor in the root domain microclimate sensor array (accuracy ±0.1 bar). The optimal root zone water potential for crops (unit: bar) is calibrated to -1 to -3 bar in this invention.

[0045] Example 3: The precision irrigation system for crops based on root water demand signals of the present invention can be adaptively adjusted for the following two planting scenarios, and the adjustments are as follows: Adjustment plans for aeroponic or vertical farming scenarios such as mixing strawberries, basil, and lettuce: Sensing module adjustments: Sensing units are deployed vertically in layers, with one set of sensors configured independently for every two layers; a light sensor is added to the top strawberry layer, with a sampling frequency of 3 minutes per sample to enhance the identification of root browning characteristics; the binding method of the stem micro-change sensor is optimized in the middle basil layer to adapt to slender stems; and the density of root zone humidity sensors is increased in the bottom lettuce layer; water requirement signal baselines are constructed for the three crops, such as the stem shrinkage threshold of basil seedlings being ≤35μm.

[0046] Module adjustments: Each layer is equipped with independent branch pipes and solenoid valve arrays. The atomization pressure for the strawberry layer is 0.4-0.5MPa, and the spray angle is 45° downward. The atomization pressure for the basil layer is 0.35-0.4MPa, and the angle is vertically downward. The atomization pressure for the lettuce layer is 0.3-0.35MPa, and the angle is 60° upward. New strawberry-specific fertilizer (high potassium) liquid tanks and basil-specific fertilizer (high nitrogen) liquid tanks are added to precisely adjust the EC values ​​(1.6-1.8mS or cm for strawberries, 1.0-1.2mS or cm for basil, and 1.2-1.4mS or cm for lettuce). Nutrient solutions are collected separately and sterilized and reconstituted according to crop requirements before reuse.

[0047] Adjustment plans for substrate or soil-based mixed or vertical planting scenarios, such as mixed planting of cucumbers and romaine lettuce: Sensing module adjustments: Sensing units are deployed vertically in layers. A stem flow meter is added to the top cucumber layer and installed at the base of the stem, with a sampling frequency of 1 minute. The substrate water potential sensor is enhanced in the bottom lettuce layer, with a sampling frequency of 5 minutes. Root visual perception adopts an insertable root canal camera to avoid damaging the substrate structure. Image algorithms are optimized to identify the differences in water requirement characteristics between deep cucumber roots and shallow lettuce roots.

[0048] Module adjustments: The cucumber layer uses a micro-spraying system with a spray particle size of 100-150μm, while the lettuce layer uses a drip irrigation system with a dripper flow rate of 2-3L or h; the upper layer irrigation time is shortened by 20% and the frequency is increased by 30% to adapt to environments with strong light and rapid evaporation; the linked biological agent injection unit precisely injects potassium fertilizer into the cucumber layer and focuses on nitrogen fertilizer supply into the lettuce layer; the greenhouse substrate culture retains the return and recycling unit to recover nutrient solution according to crop, while the return and recycling unit can be eliminated in field soil culture.

[0049] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. A precision irrigation system for crops based on root water demand signals, characterized in that, It includes a signal sensing module, an intelligent decision-making module, a precise execution module, a data storage module, and a feedback unit; The signal sensing module consists of a root image visual sensor, a stem micro-change sensor, a canopy infrared temperature sensor, and a root microclimate sensor array. The signal sensing module captures root water requirement signals and growth environment signals of different crops in mixed or vertical planting scenarios from multiple dimensions, providing accurate data support for the intelligent decision-making module. The intelligent decision-making module consists of an embedded controller, a short-term water demand prediction AI model, and a multi-objective optimizer. The intelligent decision-making module receives real-time data from the signal sensing module and generates a three-dimensional dynamic irrigation decision based on crop type, vertical layer, and growth stage through data fusion, AI prediction, and multi-objective optimization. The precision execution module consists of a nutrient solution dynamic adjustment unit, a nutrient solution circulation system, a zoned matrix electromagnetic valve array, and a reflux recovery instant sterilization unit. The precision execution module receives three-dimensional dynamic irrigation decisions from the intelligent decision module to achieve precise irrigation and nutrient solution recycling for different crops and different vertical layers. The data storage module is equipped with a data storage device, and the feedback unit is equipped with a wireless communication module. Users can remotely monitor signal changes captured by the signal sensing module through the feedback unit and adjust irrigation decisions. The data storage module and the feedback unit record sensing data, irrigation parameters and crop growth status data through three-dimensional annotation rules, providing data support for AI model self-learning optimization, while realizing data visualization and remote control.

2. The precision irrigation system for crops based on root water demand signals according to claim 1, characterized in that, The root image visual sensor uses a high-definition anti-fog micro-root window camera, equipped with an 850nm near-infrared supplementary light, and is embedded in the side wall of the crop culture root box, maintaining a distance of 5-8cm from the root system. It is equipped with an image analysis algorithm to automatically segment the root images of the crop and extract characteristic parameters such as the average diameter of the main root, the number of newly grown white root tips, the density of fibrous roots, and the root hydration status of different crops.

3. The precision irrigation system for crops based on root water demand signals according to claim 1, characterized in that, The canopy infrared temperature sensor uses an infrared thermal imaging camera with a temperature measurement range of -20℃ to 120℃ and a measurement accuracy of ±0.5℃. It is installed on the top of a vertical frame, maintaining a distance of 1.5-2m from the canopy to ensure coverage of all crop planting zones. The sensor uses a matching algorithm to calculate the temperature difference between the two canopy layers and the ambient air temperature to help determine the degree of water stress.

4. A precision irrigation system for crops based on root water demand signals according to claim 1, characterized in that, The root zone microclimate sensor array includes 3 to 5 miniature temperature, humidity, and water potential sensors arranged in each micro-planting unit. The miniature temperature, humidity, and water potential sensors are all embedded. The temperature, humidity, and water potential data of the root zone of each unit are collected in real time through the temperature, humidity, and water potential sensors to draw a three-dimensional temperature, humidity, and water potential field.

5. A precision irrigation system for crops based on root water demand signals according to claim 1, characterized in that, The embedded controller uses an industrial-grade microcontroller with a main frequency of ≥1GHz, memory of ≥2GB, storage capacity of ≥16GB, supports multi-channel data input and output, and can simultaneously receive 8-16 channels of sensor data, covering multi-layer sensing data of various crops, with a response time of ≤100ms.

6. A precision irrigation system for crops based on root water demand signals according to claim 1, characterized in that, The short-term water requirement prediction AI model is built on an LSTM network. The input layer contains 8 neurons, which correspond to crop type, water stress index, light intensity, air temperature and humidity, ventilation, growth stage, vertical layer height, and planting density, respectively. The hidden layer contains 32 neurons, and the output layer contains 2 neurons, which correspond to the crop's water requirement time point and water requirement intensity, respectively. The short-term water requirement prediction AI model has a built-in dedicated sub-model for cultivated crops. It is trained using planting data from multiple crops and multiple scenarios, and continuously learns and optimizes itself through feedback data.

7. A precision irrigation system for crops based on root water demand signals according to claim 1, characterized in that, The multi-objective optimizer is constructed using a genetic algorithm. The optimization objective functions include: minimizing irrigation duration, minimizing the number of pump start-ups and shutdowns, maximizing root health, and maximizing the yield of multiple crops in a coordinated manner. The multi-objective optimizer has built-in water demand conflict coordination rules to output the optimal combination of irrigation parameters and generate dynamically adjustable three-dimensional irrigation decisions.

8. A precision irrigation system for crops based on root water demand signals according to claim 1, characterized in that, The nutrient solution dynamic adjustment unit includes a central nutrient solution mother liquor tank, an acid solution tank, an alkali solution tank, a general concentrated fertilizer solution tank, a crop-specific potassium and calcium fertilizer tank, and a dedicated nitrogen and phosphorus fertilizer tank. The nutrient solution dynamic adjustment unit is equipped with 6 high-precision metering pumps that adjust the nutrient solution parameters in real time according to irrigation decisions, and the dissolved oxygen content of the crop nutrient solution is adjusted according to the different crop requirements.

9. A method for precision irrigation of crops based on root water demand signals, applied to a precision irrigation system for crops based on root water demand signals as described in any one of claims 1-8, characterized in that, Includes the following steps: Step S1: Data Acquisition. Various sensors in the signal sensing module collect data in real time at preset frequencies. For example, the root image vision sensor takes a picture of the crop's root system every 10 minutes and extracts characteristic parameters such as the water absorption activity of the main root and the density of the fibrous roots. The stem micro-change sensor continuously monitors the stem diameter changes of various crops through high-frequency sampling. The canopy infrared temperature sensor collects layered canopy temperature data every 3 minutes. The root microclimate sensor array collects temperature, humidity, and water potential data of each micro-planting unit every 8 minutes. All raw data are converted from analog to digital signals by the data acquisition unit and transmitted to the intelligent decision-making module. Step S2: Data preprocessing and fusion. The data preprocessing unit of the intelligent decision-making module receives various types of sensing data, removes noise interference through the Kalman filter algorithm, performs time alignment and standardization on heterogeneous data of different dimensions and frequencies, and then generates crop water stress indices for the upper and lower crops through the data fusion algorithm, thereby intuitively reflecting the real water demand status of the two crops. Step S3: Short-term water demand prediction and irrigation decision generation. The short-term water demand prediction AI model receives water stress index, environmental data and growth stage data for different crops. Combined with the built-in crop growth mechanism database, it accurately predicts the time point and water demand intensity of different crops reaching the preset stress threshold in the next 2-6 hours. The multi-objective optimizer uses a genetic algorithm for optimization. If the water demand times of two crops overlap, they are sorted by priority. Dynamic irrigation decisions are generated by adjusting the irrigation duration. Step S4: Precision irrigation execution and nutrient solution circulation. The precision execution module receives irrigation decisions. The nutrient solution dynamic adjustment unit injects crop-specific potassium and calcium fertilizer solution, nitrogen and phosphorus fertilizer solution, and acid or alkaline solution into the central mother liquor tank through a high-precision metering pump according to the decision parameters. The nutrient solution parameters are adjusted in real time. The circulation pump delivers the prepared nutrient solution to each layer of cultivation tank. The central controller controls the opening and closing of the partition matrix electromagnetic valve array in milliseconds to achieve differentiated irrigation for the tomato layer and lettuce layer. Nutrient solution that is not absorbed by the roots is collected according to crop type, flows through the filter device and UV-CLED sterilization channel to remove harmful microorganisms and impurities, and then flows back to the nutrient solution preparation branch of the corresponding crop to participate in the next cycle. Step S5: Data feedback and model self-learning optimization. The data storage and feedback unit records the sensing data, irrigation decision parameters, and crop growth status data in real time according to the three-dimensional labeling rules of crop, level, and time, and feeds the data back to the intelligent decision module. The short-term water demand prediction AI model uses the feedback data as a sample to continuously optimize the parameters of different crop-specific sub-models, update the multi-crop water demand signal feature library, adjust the water demand conflict coordination rules, and realize the intelligent iterative upgrade of the system.

Images