Intelligent crop testing system

By integrating rain shelters, rain collection equipment, and environmental monitoring equipment, along with intelligent control equipment, the problem of inaccurate rainwater reuse and irrigation in agricultural experimental systems in humid southern regions has been solved. This has enabled efficient and precise rainwater utilization and irrigation management, and improved the accuracy and stability of experimental data.

CN121176288BActive Publication Date: 2026-11-10HOHAI UNIV +1
View PDF 1 Cites 0 Cited by

Patent Information

Application Number
CN202511090811.X
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2025-08-05
Publication Date
2026-11-10
Estimated Expiration
2045-08-05

AI Technical Summary

Technical Problem

Existing agricultural experimental systems lack intelligent rain shelter, rainwater reuse, and intelligent irrigation mechanisms in humid southern regions, resulting in resource waste and inaccurate irrigation, making it difficult to meet the needs of meteorological disturbance response and experimental water use.

Method used

It integrates rain shelters, rain collection equipment, environmental monitoring equipment, and control equipment to achieve rainwater collection, treatment, and intelligent irrigation. It monitors meteorological data and crop growth data through sensors to dynamically optimize irrigation strategies.

Benefits of technology

It effectively avoids interference from sudden rainfall, saves water resources, improves irrigation precision and efficiency, adapts to weather changes, meets the water demand for experiments, and enhances the accuracy and stability of experimental data.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN121176288B_ABST
    Figure CN121176288B_ABST
Patent Text Reader

Abstract

This invention relates to the field of agricultural technology, specifically to an intelligent crop testing system. The intelligent crop testing system includes: a rain shelter, used to provide rain protection for crops within the shelter under the control of a control device, and to transfer rainwater to a rain collection device; a rain collection device, used to collect rainwater and, under the control of the control device, to treat the collected rainwater to meet irrigation standards; environmental monitoring equipment, used to detect environmental data within the rain shelter and to transmit the environmental data to the control device; the environmental data includes at least one of temperature data, relative humidity data, wind speed data, wind direction data, and radiation data; and a control device, used to determine a target irrigation strategy for the crops based on the environmental data; and, according to the target irrigation strategy, to control the irrigation device to draw rainwater from the rain collection device to irrigate the crops. This system overcomes the limitations of existing agricultural testing systems with their single function, providing more efficient and reliable technical support for agricultural research.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This invention relates to the field of agricultural technology, and more specifically to an intelligent crop testing system. Background Technology

[0002] In agricultural production and research, climate uncertainty and the uneven spatial and temporal distribution of water resources are significant factors affecting crop growth and experimental accuracy. Particularly in the humid regions of southern my country, frequent but spatially complex rainfall presents numerous challenges to the development and implementation of irrigation regimes in agricultural experiments. On the one hand, sudden rainfall can easily disrupt experimental conditions, affecting the accuracy of crop water balance and yield assessment; on the other hand, although the total rainfall in the humid southern regions is relatively large, the seasonal and uneven distribution of rainfall means that experimental water still faces periodic shortages.

[0003] Existing agricultural experimental systems are limited in function and lack systematic rainwater collection and reuse, leading to resource waste. Irrigation relies heavily on experience-based settings, lacking intelligent analysis and dynamic control, making precise water management difficult. Although some research has introduced sensors, crop models, and AI algorithms, these are mostly applied to large-scale field production and cannot meet the needs of small-scale experiments in humid southern regions. In particular, there is a lack of effective mechanisms for meteorological disturbance response, rainwater reuse, and coordinated optimization of irrigation. There is an urgent need for intelligent crop experimental systems that integrate intelligent rain shelter, rainwater reuse, model-driven approaches, and intelligent decision-making. Summary of the Invention

[0004] In view of this, the present invention provides an intelligent crop testing system to solve the problem of the urgent need for an intelligent crop testing system that integrates intelligent rain avoidance, rainwater reuse, model-driven and intelligent decision-making.

[0005] In a first aspect, the present invention provides an intelligent crop testing system, comprising a rain shelter, a rain collection device, an environmental monitoring device, an irrigation device, and a control device. The rain shelter is connected to the rain collection device, the irrigation device is connected to the rain collection device, the environmental monitoring device is installed in the rain shelter, and the environmental monitoring device is connected to the control device, wherein:

[0006] Rain shelters are used to provide rain protection for crops inside the shelter under the control of control equipment and to transfer rainwater to rain collection equipment;

[0007] Rainwater harvesting equipment is used to collect rainwater and treat the collected rainwater under the control of control equipment so that the rainwater meets irrigation standards;

[0008] Environmental monitoring equipment is used to detect environmental data inside the rain shelter and transmit the environmental data to control equipment; the environmental data includes at least one of temperature data, relative humidity data, wind speed data, wind direction data, and radiation data;

[0009] The control equipment is used to determine the target irrigation strategy for crops based on environmental data; and to control the irrigation equipment to draw rainwater from the rainwater collection equipment to irrigate the crops according to the target irrigation strategy.

[0010] In one optional embodiment, a rainfall sensor and a wind speed sensor are installed above the rain shelter. The rainfall sensor is communicatively connected to a control device. An automatic roller blind transparent film is installed around the perimeter of the upper surface of the rain shelter. An elastic retraction mechanism is installed along the ridge line of the upper surface of the rain shelter. The automatic roller blind transparent film is connected to a drive motor, which is communicatively connected to the control device. Telescopic support rods are installed at the bottom of both sides of the rain shelter body. Water channels are installed between the arches of the rain shelter and at the edge of the roof.

[0011] The control device is used to receive rainfall data transmitted by the rainfall sensor and combine it with the obtained weather forecast data. When the rainfall reaches the preset rainfall threshold and / or within a preset time before the rainfall, it controls the drive motor to drive the automatic roller shutter transparent film to a position at a first preset height above the ground, and controls the elastic shrink mechanism to shrink to provide rain protection for crops in the rain shelter, and transmits rainwater to the rain collection equipment through the water delivery channel.

[0012] The control device is also used to receive wind speed data transmitted by the wind speed sensor. When the wind speed is greater than the preset wind speed threshold, it controls the drive motor to raise the automatic roller shutter transparent film to a position at a second preset height above the ground; and activates the telescopic support rod; wherein the second preset height is greater than the first preset height.

[0013] In one optional embodiment, the rainwater harvesting equipment includes a sedimentation tank, a water quality monitoring tank, and a storage tank. The sedimentation tank and the water quality monitoring tank are connected via a first pipe and a second pipe. Filter screens are installed on the first and second pipes in front of one end of the sedimentation tank. A pumping valve is installed at the end of the first pipe near the sedimentation tank. A first submersible pump is installed at the end of the second pipe extending into the water quality monitoring tank. A water quality sensor and water treatment equipment are installed inside the water quality monitoring tank, both of which are communicatively connected to a control device. A first water intake is located on the cover of the water quality monitoring tank. The water quality monitoring tank is connected to the storage tank via a smart valve and a third pipe. A second submersible pump is located inside the storage tank and is connected to an outlet pipe, which is connected to irrigation equipment.

[0014] Water quality sensors are used to detect the current water quality data of rainwater in the water quality monitoring tank and transmit the current water quality data to the control equipment;

[0015] The control equipment is used to compare the current water quality data with the irrigation standard water quality data corresponding to crops, and based on the comparison results, control the water treatment equipment to treat the rainwater in the water quality monitoring pool.

[0016] In one optional implementation, the intelligent crop testing system further includes a canopy recognition device and a radar sensor, both of which are communicatively connected to the control device; wherein;

[0017] Canopy recognition equipment is used to photograph and / or scan the canopy of crops to generate canopy data, which includes at least one of chlorophyll content, leaf area index, canopy height, and canopy closure.

[0018] Radar sensors are used to scan the root system of crops to determine root data, including root distribution density, root length, and root activity.

[0019] Control equipment is used to output target irrigation strategies for crops based on environmental data, canopy data, and root system data.

[0020] In one alternative implementation, a control device is used to identify canopy data and root data to establish an initial physiological causal relationship between the canopy and the root system; to identify environmental data and determine initial latent stress signals from the environmental data; the latent stress signals are used to characterize non-overt physiological risks that pose a potential threat to crop growth;

[0021] The initial physiological causal relationship and initial implicit stress signal are input into the preset irrigation strategy model, and the target irrigation strategy corresponding to the crop is output.

[0022] In one optional implementation, the control device is used to acquire the corresponding growth stage of the crop, and determine the current growth stage of the crop and the weight information corresponding to the current growth stage based on the growth stage.

[0023] Based on the weight information corresponding to the current growth stage, calculate the target water transfer efficiency coefficient for the crop.

[0024] Based on the target water conduction efficiency coefficient, an initial physiological causal relationship between the canopy and the root system is established.

[0025] In one alternative implementation, the control device is used to perform sliding window analysis on various data in the environmental data and calculate the rate of change of the corresponding indicators for each data.

[0026] Based on the collection locations corresponding to each data point in the environmental data, determine the spatial gradient corresponding to each data point;

[0027] The environmental data is converted into the frequency domain by Fourier transform, and the frequency domain characteristics corresponding to the environmental data are determined.

[0028] The rate of change of each data point, the spatial gradient of each data point, and the frequency domain characteristics of the environmental data are input into the dynamic mapping model.

[0029] The dynamic mapping model determines the weight information corresponding to the rate of change of the index, the spatial gradient, and the frequency domain features, respectively; based on the weight information corresponding to the rate of change of the index, the spatial gradient, and the frequency domain features, a feature weight matrix is ​​constructed.

[0030] Based on the feature weight matrix, the matching degree between the input feature combination and the preset feature combination template is calculated. Based on the matching degree of the input feature combination and the feature over-standard degree of each individual feature in the input feature combination, the initial hidden stress signal corresponding to the input feature combination is determined.

[0031] In one alternative implementation, the control device is used to calculate the basic irrigation demand corresponding to the crop based on an initial physiological causal relationship.

[0032] Based on the relationship between initial latent stress signals and basic irrigation demand, the corresponding risk regulation amount for crops is calculated;

[0033] Based on the basic irrigation demand and risk control, determine the initial irrigation amount for the crops.

[0034] Based on the root system data of the crops, determine the root system characteristics of the crops; based on the root system characteristics, determine the initial irrigation period and the initial irrigation method.

[0035] Based on the initial crop irrigation amount, initial irrigation period, and initial irrigation method, an initial irrigation strategy is determined; the initial irrigation strategy is then calibrated to obtain the target irrigation strategy.

[0036] In one alternative implementation, the control device is used to define a state space and an action space based on an initial irrigation strategy; the action space is used to adjust the various parameters included in the initial irrigation strategy to generate various irrigation actions.

[0037] For each irrigation action, environmental data, canopy data, root data, soil data corresponding to the crop, and irrigation action are input into the preset crop water production function model. The preset crop water production function model outputs the simulated yield and simulated water use efficiency corresponding to each action.

[0038] Based on simulated yield and simulated water use efficiency, a dynamic reward function corresponding to the initial irrigation strategy is constructed; the dynamic reward function includes at least one of yield reward, water use efficiency reward and experimental data stability reward.

[0039] The dynamic reward function is used to evaluate each irrigation action in the action space;

[0040] Based on the evaluation results, the initial irrigation strategy was adjusted to obtain the target irrigation strategy.

[0041] In one optional implementation, the dynamic reward function includes a yield reward, a water use efficiency reward, and a test data stability reward; wherein: the yield reward is as follows:

[0042]

[0043] Where R1 is the production bonus, Y 模拟 To simulate output; Y 潜在 S represents the potential yield of a crop under optimal conditions. 胁迫 The stress correction coefficient is related to environmental data; w1 is the first weight information.

[0044] Water use efficiency rewards are as follows:

[0045]

[0046] Where R2 is the water use efficiency bonus, WUE 模拟 To simulate water use efficiency (simulated yield / water consumption); WUE 基准 The baseline water use efficiency for crops; G 雨水 w1 represents the rainwater reuse gain; w2 represents the second weighting information.

[0047] The experimental data stability bonus is as follows:

[0048]

[0049] Where, σ Y The standard deviation coefficient of the simulated output for a continuously preset number of times; σ WUE The standard deviation coefficient of water use efficiency is the number of consecutive preset simulations; w3 is the third weighting information;

[0050] The dynamic reward function is as follows:

[0051] R 总 =R1+R2+R3

[0052] Among them, R 总 This is a dynamic reward function.

[0053] The intelligent crop experiment system provided in this application includes a rain shelter that effectively protects crops from rain, preventing sudden rainfall from interfering with experimental conditions, reducing the impact of rainwater on crop water balance and yield assessment, and ensuring the accuracy of experimental data. Simultaneously, environmental monitoring equipment monitors environmental data such as temperature and humidity inside the shelter in real time, providing a stable and monitorable environmental basis for the experiment and solving the problem of fluctuating experimental conditions caused by frequent rainfall in humid regions. Rainwater harvesting equipment collects rainwater and treats it to irrigation standards, converting abundant rainwater resources into experimental irrigation water, reducing dependence on external water sources and alleviating the problem of periodic water shortages in the region. This rainwater reuse mechanism saves water resources and reduces experimental costs, meeting the development needs of water-saving agriculture. Canopy recognition equipment generates canopy data such as chlorophyll content and leaf area index through imaging and scanning, intuitively reflecting the crop's photosynthetic capacity and growth vitality. Radar sensors scan and acquire root data such as root distribution density, length, and vitality, revealing the crop's water absorption capacity and distribution characteristics. The control equipment combines two types of data to accurately determine the match between the crop's "canopy water demand" and "root water supply capacity," avoiding the supply-demand imbalance caused by traditional irrigation relying solely on experience or a single indicator. This makes the target irrigation strategy more aligned with the actual physiological state of the plant. The control equipment also links and analyzes canopy data, root data, and environmental data such as temperature and humidity acquired by environmental monitoring equipment. This multi-dimensional collaborative mechanism ensures crop water needs are met while preventing water waste, significantly improving the efficiency of water use in humid regions. Furthermore, canopy and root data change in real time with crop growth, allowing the control equipment to continuously optimize the target irrigation strategy based on this dynamic data. This dynamic adjustment mechanism reduces the interference of fixed irrigation strategies on experimental conditions, ensuring that crops at different growth stages are in a suitable water environment, improving the stability and accuracy of experimental data. It can dynamically optimize irrigation regimes, outputting scientific recommendations for irrigation time and volume, achieving precision irrigation tailored to different crops, weather conditions, and soil moisture, significantly improving irrigation accuracy and water resource utilization efficiency. The intelligent crop testing system integrates the functions of rain shelters, rainwater harvesting equipment, environmental monitoring equipment, irrigation equipment, and control equipment, overcoming the shortcomings of existing agricultural testing systems with limited functionality. It particularly establishes effective mechanisms for meteorological disturbance response, rainwater reuse, and coordinated irrigation optimization, perfectly adapting to the specific needs of small-scale experiments in humid regions and providing more efficient and reliable technical support for agricultural research. Attached Figure Description

[0054] To more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the specific embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0055] Figure 1 This is a schematic diagram of the structure of an intelligent crop testing system according to an embodiment of the present invention;

[0056] Figure 2 This is a schematic diagram of the structure of a rain shelter according to an embodiment of the present invention;

[0057] Figure 3 This is a schematic diagram of the structure of a rainwater harvesting device according to an embodiment of the present invention;

[0058] Figure 4 This is a schematic diagram of the structure of a water treatment device according to an embodiment of the present invention;

[0059] Figure 5 This is a schematic diagram of the structure of another intelligent crop testing system according to an embodiment of the present invention;

[0060] Figure 6 This is a schematic diagram of the process by which a control device determines a target irrigation strategy according to an embodiment of the present invention;

[0061] Figure 7 This is a schematic diagram illustrating the setting of differentiated time decay rates for different types of features according to an embodiment of the present invention;

[0062] The system includes: 1. Rain shelter; 2. Rain collection equipment; 3. Environmental monitoring equipment; 4. Irrigation equipment; 5. Control equipment; 11. Rainfall sensor; 12. Wind speed sensor; 13. Automatic roller blind transparent film; 14. Elastic retraction mechanism; 15. Drive motor; 16. Telescopic support rod; 17. Water conveying trough; 21. Sedimentation tank; 22. Water quality monitoring tank; 23. Water storage tank; 24. First pipeline; 25. Second pipeline; 26. Filter screen; 27. Pumping valve; 28. First submersible pump; 29. ​​Water quality sensor; 210. Water treatment equipment; 211. Intelligent valve; 212. Second submersible pump; 213. Outlet pipe; 214. Detailed Implementation

[0063] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0064] This invention provides an intelligent crop testing system, such as Figure 1As shown, the intelligent crop testing system includes a rain shelter 1, a rain collection device 2, an environmental monitoring device 3, an irrigation device 4, and a control device 5. The rain shelter 1 is connected to the rain collection device 2, the irrigation device 4 is connected to the rain collection device 2, the environmental monitoring device 3 is installed in the rain shelter 1, and the environmental monitoring device 3 is connected to the control device 5.

[0065] Rain shelter 1 is used to provide rain shelter for crops inside the rain shelter 1 under the control of control device 5 and to transfer rainwater to rain collection device 2;

[0066] Rainwater collection device 2 is used to collect rainwater and, under the control of control device 5, treat the collected rainwater to meet irrigation standards.

[0067] Environmental monitoring equipment 3 is used to detect environmental data inside the rain shelter 1 and transmit the environmental data to control equipment 5; the environmental data includes at least one of temperature data, relative humidity data, wind speed data, wind direction data, and radiation data;

[0068] Control device 5 is used to determine the target irrigation strategy for crops based on environmental data; and to control irrigation device 4 to draw rainwater from rainwater collection device 2 to irrigate crops according to the target irrigation strategy.

[0069] Specifically, the rain shelter 1 serves as the basic protective facility for the experiment. Its core function is to shield the crops inside from natural rainfall, preventing sudden precipitation from disrupting the crop's water balance and the stability of the experimental conditions. Simultaneously, through its structure (such as a gentle slope design), the shelter guides rainwater to the rainwater collection device 2, achieving orderly collection and transmission of rainwater, laying the foundation for subsequent reuse. The rainwater collection device 2 is responsible for the systematic treatment of the rainwater transmitted from the rain shelter 1: firstly, it removes impurities such as sediment and suspended solids through filtration and purification processes; then, it adjusts water quality parameters such as pH, temperature, and salinity to meet the "Standards for Irrigation Water Quality" (GB5084-2021); finally, it is stored in the reservoir 23, providing a qualified water source for the irrigation equipment 4, effectively utilizing the abundant rainfall resources in the south, and alleviating the pressure on experimental water supply.

[0070] Environmental monitoring equipment 3 is installed inside the rain shelter 1 to collect key environmental data such as temperature, relative humidity, wind speed, wind direction, and radiation in real time, and transmit the data to control equipment 5. These data directly reflect the microclimate conditions inside the shelter and are an important basis for judging the crop growth environment and formulating irrigation strategies, ensuring that irrigation decisions are in line with actual environmental conditions.

[0071] Irrigation equipment 4 is connected to the water storage tank 23 of rainwater harvesting equipment 2. Under the command of control equipment 5, it extracts treated rainwater through devices such as submersible pumps and flexibly irrigates crops using drip irrigation, micro-sprinkler irrigation, or other methods according to experimental needs. Its precise control characteristics can reduce water waste and achieve targeted delivery and efficient utilization of water.

[0072] When the control device 5 detects that the rainfall has reached a preset rainfall threshold, it controls the rain shelter 1 to provide shelter for the crops. While the rain shelter 1 is providing shelter for the crops, it also transmits the collected rainwater to the rainwater collection device 2. The rainwater collection device 2 collects the rainwater and, under the control of the control device 5, processes the collected rainwater to meet irrigation standards.

[0073] In addition, the control device 5 receives environmental data detected by the environmental monitoring device 3 inside the rain shelter 1, and determines the target irrigation strategy for the crops based on the environmental data; and controls the irrigation device 4 to draw rainwater from the rain collection device 2 to irrigate the crops according to the target irrigation strategy.

[0074] The intelligent crop testing system provided in this application embodiment features a rain shelter 1 that effectively protects crops from rain, preventing sudden rainfall from interfering with experimental conditions, reducing the impact of rainwater on crop water balance and yield assessment, and ensuring the accuracy of experimental data. Simultaneously, environmental monitoring equipment 3 monitors environmental data such as temperature and humidity inside the shelter in real time, providing a stable and monitorable environmental foundation for the experiment and solving the problem of fluctuating experimental conditions caused by frequent rainfall in humid regions. Rainwater harvesting equipment 2 collects rainwater and treats it to irrigation standards, converting abundant rainwater resources into experimental irrigation water, reducing dependence on external water sources and alleviating the problem of periodic water shortages in the region. This rainwater reuse mechanism saves water resources and reduces experimental costs, meeting the development needs of water-saving agriculture. Control equipment 5 determines target irrigation strategies based on environmental monitoring data, overcoming the limitations of traditional irrigation relying on experience. It can dynamically optimize irrigation regimes, outputting scientific suggestions for irrigation time and volume, achieving precise irrigation tailored to different crops, weather conditions, and soil moisture, significantly improving irrigation accuracy and water resource utilization efficiency. The intelligent crop testing system integrates the functions of a rain shelter (1), rain collection equipment (2), environmental monitoring equipment (3), irrigation equipment (4), and control equipment (5), overcoming the shortcomings of existing agricultural testing systems with limited functionality. It particularly establishes an effective mechanism for optimizing meteorological disturbance response, rainwater reuse, and irrigation linkage, perfectly adapting to the specific needs of small-scale experiments in humid areas and providing more efficient and reliable technical support for agricultural research.

[0075] In one optional embodiment of this application, such as Figure 2 As shown, a rain sensor 11 and a wind speed sensor 12 are installed above the rain shelter 1. The rain sensor 11 is communicatively connected to the control device 5. An automatic roller blind transparent film 13 is provided around the upper surface of the rain shelter 1. An elastic retraction mechanism 14 is provided along the ridge line of the upper surface of the rain shelter 1. The automatic roller blind transparent film 13 is connected to the drive motor 15. The drive motor 15 is communicatively connected to the control device 5. Telescopic support rods 16 are installed at the bottom of both sides of the rain shelter 1. Water channels 17 are provided between the arches of the rain shelter 1 and at the edge of the roof.

[0076] The control device 5 is used to receive rainfall data transmitted by the rainfall sensor 11 and combine it with the obtained weather forecast data. When the rainfall reaches the preset rainfall threshold and / or within a preset time before the rainfall, it controls the drive motor 15 to drive the automatic roller shutter transparent film 13 to descend to a position at a first preset height above the ground, and controls the elastic shrink mechanism 14 to shrink to provide rain protection for the crops in the rain shelter 1, and transmits the rainwater to the rain collection device 2 through the water delivery trough 17.

[0077] The control device 5 is also used to receive wind speed data transmitted by the wind speed sensor 12, and when the wind speed is greater than the preset wind speed threshold, control the drive motor 15 to drive the automatic roller shutter transparent film 13 to rise to a position of a second preset height above the ground; and activate the telescopic support rod 16; wherein the second preset height is greater than the first preset height.

[0078] The rain sensor 11 is installed on the top of the rain shelter 1 to monitor rainfall and intensity in real time, and communicates with the control device 5 to provide a measured basis for rain shelter decisions. The wind speed sensor 12 is located on the roof to collect real-time wind speed data; when the wind speed exceeds a preset threshold, a wind protection mechanism is triggered. An automatic roller blind transparent film 13 is arranged around the upper surface of the rain shelter 1, driven by a drive motor 15 to raise and lower; it is translucent to ensure adequate lighting inside the shelter; it can be lowered to a first preset height from the ground and raised to a second preset height. An elastic retraction mechanism 14 is set along the ridgeline of the roof, working in conjunction with the retraction / expansion of the roller blind film to enhance the sealing during rain sheltering and the cushioning during wind protection.

[0079] The retractable support rods 16 are installed at the bottom of both sides of the shed and are securely fixed using pre-embedded metal bases. They are normally retracted and extend and secure during strong winds, enhancing the overall stability of the shed. During the initial shed construction, workers dig trenches of appropriate depth, typically 50 to 80 centimeters, at the designated support rod installation points, according to the shed's design specifications. Custom-made metal bases, shaped like a "T," are then placed into the trenches. The horizontal section is wide and has multiple anchor holes, while the vertical section connects to the support rods. High-strength concrete is poured into the trenches to ensure the metal base forms a stable, integrated structure with the ground, effectively resisting lateral pressure and wind forces from the shed. An adjustable hinge device is provided at the vertical connection end of the metal base. The support rods are connected to the base via this hinge, allowing the angle of the support rods to be flexibly adjusted within a certain range during extension and retraction, based on the actual stress requirements of the shed and the ground slope, ensuring optimal support performance. For example, if the ground in the area where the shed is located has a certain slope, the angle of the support rod can be adjusted by hinges to keep it in the optimal support state that is always perpendicular or nearly perpendicular to the ground.

[0080] In addition, to further enhance the reliability of the support system, auxiliary connection points are installed at the upper part of both sides of the shed. When the support rods are extended to the working state, their tops can be connected to these auxiliary connection points via high-strength ropes or adjustable metal rods. These ropes or rods have a certain degree of elasticity and adjustable length, which can share some of the lateral tension when the support rods support the shed. At the same time, they can automatically adjust the tension according to the deformation of the shed under different weather conditions to ensure the stability of the shed structure.

[0081] Water delivery channels 17 are distributed between each arch and at the edge of the roof, with a slope design (e.g., ≤2°) to guide the collected rainwater to the rainwater collection device 2, so as to realize the utilization of rainwater resources.

[0082] Specifically, the rainfall sensor 11 can monitor rainfall data and transmit the monitored rainfall data to the control device 5. Based on the received rainfall data and acquired weather forecast data, the control device 5 can control the drive motor 15 to lower the automatic roller shutter transparent film 13 to a position at a first preset height above the ground when the rainfall reaches a preset rainfall threshold and / or within a preset time period before rainfall. The first preset height can be 0.5m or 0.3m above the ground; this embodiment does not specifically limit the first preset height. Furthermore, the control device 5 controls the elastic contraction mechanism 14 to contract, reducing the risk of deformation of the shed due to water pressure impact, thus providing shelter for crops inside the rain shelter 1, and transmitting rainwater to the rain collection device 2 via the water trough 17.

[0083] In addition, the control device 5 can receive wind speed data transmitted by the wind speed sensor 12. Then, the control device 5 compares the wind speed data with a preset wind speed threshold. When the wind speed is greater than the preset wind speed threshold, the control device 5 controls the drive motor 15 to raise the automatic roller blind transparent film 13 to a position at a second preset height above the ground; and activates the telescopic support rod 16.

[0084] For example, the control device 5 controls the drive motor 15 to drive the automatic roller shutter transparent film 13 to automatically raise the bottom 1.2m of the roller shutter, forming a "ventilation and pressure relief channel". At the same time, the telescopic support rods 16 on both sides of the shed are activated to enhance the structural stability and prevent strong winds from overturning the shed.

[0085] The intelligent crop testing system provided in this application embodiment uses a rain shelter 1 to acquire rainfall data in real time via a rainfall sensor 11. Combined with weather forecast information, when the rainfall reaches a preset threshold, the control device 5 drives the automatic roller shutter transparent film 13 to descend to a first preset height, while simultaneously controlling the elastic contraction mechanism 14 to contract, achieving precise rain protection for the crops. This mechanism effectively avoids interference from sudden rainfall in humid areas on experimental conditions, reduces the impact of rainwater on crop water balance and yield assessment, and ensures the accuracy of experimental data. When the wind speed sensor 12 detects a wind speed greater than a preset threshold, the control device 5 drives the automatic roller shutter transparent film 13 to rise to a higher second preset height and activates the retractable support rod 16. This design reduces the impact of strong winds on the shelter structure, while increasing the height of the roller shutter enhances ventilation inside the shelter, preventing damage to the shelter structure under high wind speeds. It adapts to windy and sudden gust climates, ensuring the safety of system operation. Furthermore, the water channels 17 installed between the arches of the rain shelter 1 and along the edge of the roof can efficiently collect rainwater during the rain sheltering process and transfer it to the rain collection device 2. Combined with the structural design of the shelter, this further increases the amount of rainwater collected. This function makes full use of abundant rainfall resources, providing sufficient water for subsequent rainwater purification and reuse, reducing the pressure on experimental water use, and achieving efficient water resource utilization. The height adjustment of the automatic roller blind transparent film 13 and the linkage between the elastic contraction mechanism 14 and the telescopic support rod 16 can ensure the rain sheltering effect during rainfall, and can also maintain the relative uniformity between the inside and outside environment (such as ventilation and lighting conditions) when there is no wind or low wind speed by raising the roller blind, avoiding the formation of extreme microclimates inside the shelter, ensuring the coordination between experimental conditions and the natural environment, and improving the representativeness of experimental data. Through the above-mentioned rain shelter 1, a dynamic response to meteorological factors such as rainfall and wind speed is achieved. While ensuring experimental stability, improving wind resistance safety, and enhancing rainwater utilization efficiency, it also takes into account the coordination of the experimental environment, fully adapting to the complex and changeable climate characteristics of humid regions, and providing reliable environmental control support for agricultural experiments.

[0086] In one alternative implementation, such as Figure 3As shown, the rainwater collection device 2 includes a sedimentation tank 21, a water quality monitoring tank 22, and a storage tank 23. The sedimentation tank 21 and the water quality monitoring tank 22 are connected by a first pipe 24 and a second pipe 25. A filter screen 26 is installed in front of one end of the first pipe 24 and the second pipe 25, near the sedimentation tank 21. A pumping valve 27 is installed at the end of the first pipe 24 near the sedimentation tank 21. A first submersible pump 28 is installed at the end of the second pipe 25 extending into the water quality monitoring tank 22. Water quality sensor 29 and water treatment equipment 210 are installed inside pool 22. Both water quality sensor 29 and water treatment equipment 210 are communicatively connected to control equipment 5. A first water intake is located on the cover of water quality monitoring pool 22. Water quality monitoring pool 22 is connected to water storage pool 23 via intelligent valve 211 and third pipe 214. Water storage pool 23 contains a second submersible pump 212, which is connected to outlet pipe 213, which is connected to irrigation equipment 4. Wherein:

[0087] Water quality sensor 29 is used to detect the current water quality data of rainwater in water quality monitoring pool 22 and transmit the current water quality data to control device 5;

[0088] Control device 5 is used to compare the current water quality data with the irrigation standard water quality data corresponding to crops. Based on the comparison results, control water treatment device 210 to treat the rainwater in water quality monitoring pond 22.

[0089] The sedimentation tank 21 features a stepped design with a pre-set bottom slope. The first stage uses honeycomb-shaped inclined tube packing material to intercept silt particles with a diameter greater than 0.1 mm (accounting for more than 60% of rainwater impurities) using shallow sedimentation principles. A height-adjustable baffle (0-30 cm) is installed on the side. During the rainy season when water volume is high, the baffle is raised to extend the retention time (≥30 minutes), while during the dry season when water volume is low, the baffle is lowered to accelerate flow (≤15 minutes).

[0090] Specifically, 1. The dual-pipeline design of sedimentation tank 21 → water quality monitoring tank 22.

[0091] First Pipeline 24 (Conventional Water Transmission): Path: Sedimentation Tank 21 → Filter Screen 26 (Preliminary Sludge Removal) → Pumping Valve 27 (Controlled by Control Device 5) → First Pipeline 24 → Water Quality Monitoring Tank 22. Function: When the rainwater turbidity is low (e.g., <10 NTU), Control Device 5 controls the opening of Pumping Valve 27, relying on gravity / low-power pumping to achieve "energy-saving water transmission".

[0092] Second Pipeline 25 (Emergency / High Turbidity Water Transfer): Path: Sedimentation Tank 21 → Filter Screen 26 (Coarse Filtration) → First Submersible Pump 28 (Control Device 5 controls the first submersible pump 28 to actively pump water) → Second Pipeline 25 → Water Quality Monitoring Tank 22. Function: When rainwater turbidity is high (e.g., >20 NTU) or the water level in Sedimentation Tank 21 is low (insufficient gravity water transfer), Control Device 5 controls the start of the first submersible pump 28 to forcibly pump water into the monitoring tank, preventing Sedimentation Tank 21 from overflowing.

[0093] 2. Water supply chain from water quality monitoring pool 22 to water storage pool 23 to irrigation equipment 4.

[0094] Water quality monitoring tank 22 internal treatment: After rainwater enters the monitoring tank, water quality sensor 29 detects it in real time, and water treatment equipment 210 (such as filtration, disinfection, and pH adjustment modules) purifies it as needed. Water conveyance to storage tank 23: Path: Water quality monitoring tank 22 → Smart valve 211 (control device 5 controls the on / off state and flow rate of smart valve 211) → Third pipeline 214 → Storage tank 23. Function: After the water quality meets the standards, smart valve 211 opens, and the purified rainwater is stored in storage tank 23 as a backup water source for irrigation.

[0095] Irrigation water delivery: Path: reservoir 23 → second submersible pump 212 (control device 5 controls the second submersible pump 212 to pressurize) → outlet pipe 213 → irrigation equipment 4 (drip irrigation, sprinkler irrigation, etc.) Function: According to the crop's water demand instructions, the second submersible pump 212 starts and delivers water from the reservoir 23 to the field.

[0096] Filter screen 26 is installed at the inlet of the first pipe 24 and the second pipe 25 (on the side of the sedimentation tank 21), which is equivalent to the "first line of defense". Function: It intercepts large particulate impurities (such as leaves and clumps of silt) in rainwater, protecting the subsequent water pumping valve 27, the first submersible pump 28, and the water quality sensor 29 from clogging / wearing.

[0097] Water quality sensor 29 is deployed within the water quality monitoring tank 22 (usually in the lower middle part of the tank to avoid interference from surface floating matter). Detection parameters include pH, turbidity, conductivity, heavy metals (such as cadmium and lead), and microorganisms (such as E. coli), providing real-time output of "current water quality data." Its core value lies in providing the control device 5 with a basis for determining whether the water quality meets standards; it serves as the "eyes" of the system's intelligence.

[0098] For example, such as Figure 4 As shown, the main installation frame of the water treatment equipment 210 is a modular shell adapted to the water quality monitoring pool 22: the vertical or horizontal box made of corrosion-resistant materials (such as fiberglass, 304 stainless steel) can be fixed to the inner wall of the water quality monitoring pool 22 or suspended in the pool. The bottom of the shell is equipped with a hollow support to facilitate rainwater circulation.

[0099] Inlet: Connected to the rainwater area in the water quality monitoring pool 22 via a pipe, with a built-in one-way valve to prevent backflow, and a coarse filter screen (5-10mm aperture) is installed near the inlet to initially intercept large particles such as leaves and pebbles.

[0100] Water outlet: Located on the other side of the shell, the treated water flows back to the water quality monitoring tank 22 (or is transported to the water storage tank 23 through the intelligent valve 211). A fine filter screen (pore size 1-2mm) is installed at the water outlet to prevent internal parts of the equipment from falling off and polluting the water.

[0101] The core treatment units of the water treatment equipment 210 (distributed sequentially according to the water flow direction):

[0102] 1) Pretreatment equipment is used to remove suspended impurities and colloidal substances from rainwater, reducing the burden on subsequent treatment.

[0103] Cyclone separation zone: After the water flows in, it forms a vortex through the guide plate, and uses centrifugal force to separate mud, sand and fine particles. The bottom is equipped with a sedimentation tank that can be cleaned regularly.

[0104] Precision filtration components: Employ stepped filter screens (gradually decreasing from 50μm to 10μm) or pleated filter cartridges to further trap suspended particles, algae, etc. in the water. The filter screens are removable, washable, or replaceable.

[0105] 2) Water quality conditioning equipment: This equipment addresses potential pH imbalances (such as low pH due to acid rain) or excessive ion levels in rainwater. It includes: an acid-base adjustment unit containing a pH sensor (communicating with control device 5), a reagent storage tank (such as sodium hydroxide solution or dilute hydrochloric acid), and a metering pump. It automatically adds reagents based on real-time pH values ​​to adjust the water quality to a suitable range for crops (typically pH 6.5-8.0).

[0106] 3) Purification and Disinfection Equipment: Eliminates microorganisms (bacteria, viruses, protozoa) and trace organic matter in the water, ensuring irrigation safety. Includes: Ultraviolet (UV) Disinfection Unit: Built-in UV-C lamp array (wavelength 254nm). When water flows around the lamps, the ultraviolet light destroys the DNA structure of microorganisms, achieving a sterilization rate of over 99%. A quartz sleeve is installed on the outside of the lamps to prevent contamination, and it is linked to the control equipment 5 to monitor the lamp's operating status (e.g., lifespan, intensity). Activated Carbon Adsorption Zone: Filled with granular activated carbon to adsorb pesticide residues, humic acid, odors, etc., in the water, improving the taste of the water (if rainwater is used for foliar spraying, it can reduce irritation to crops).

[0107] 4) Deep homogenization equipment: Ensures stable water quality after treatment and avoids local fluctuations in indicators. Includes: Buffer tank: After the water flows in, it forms turbulence through baffles, so that the water quality is uniformly mixed. The tank is equipped with a stirring device (such as a small submersible pump) to enhance the homogenization effect.

[0108] The intelligent crop testing system provided in this application embodiment uses a sedimentation tank 21 for preliminary filtration of sediment. A water quality sensor 29 collects data in real time, and the control device 5 records the treatment strategy and water quality changes, forming a complete data chain of "water input - treatment process - crop response." It can purify rainwater to irrigation standards suitable for crops (such as controlling pH, turbidity, heavy metals, etc.), avoiding root damage, disease growth, and yield and quality decline caused by poor-quality water irrigation, and providing a stable water environment for crop growth. Furthermore, from rainwater collection (sedimentation tank 21 - monitoring tank water delivery), water treatment (automatic module scheduling) to irrigation water delivery (storage tank 23 - irrigation equipment 4), the entire process is executed in conjunction with the control device 5, eliminating the need for frequent manual operation. Compared to traditional rainwater harvesting systems that require manual detection and water replacement, it can save more than 70% of maintenance manpower and reduce the risk of water quality fluctuations due to human error.

[0109] In one optional embodiment of this application, such as Figure 5 As shown, the intelligent crop testing system also includes a canopy recognition device 6 and a radar sensor 7, both of which are communicatively connected to the control device 5; wherein;

[0110] The canopy recognition device 6 is used to photograph and / or scan the canopy of crops to generate canopy data, which includes at least one of chlorophyll content, leaf area index, canopy height, and canopy closure.

[0111] Specifically, the canopy recognition device 6 may include a multispectral camera (400-1000nm band) and a lidar (LiDAR).

[0112] Plant chlorophyll exhibits strong absorption of blue light (400-500nm), strong reflection of red light (600-700nm), and high reflectivity of near-infrared light (700-1000nm). Multispectral cameras can collect reflectivity data from these three key wavelengths and input it into vegetation index algorithms (such as NDVI (Normalized Difference Vegetation Index) and RVI (Ratio Difference Vegetation Index)) to calculate chlorophyll content. For example, healthy leaves, due to their high chlorophyll content, have weak red light reflection and strong near-infrared reflection, resulting in higher NDVI values; leaves lacking chlorophyll have significantly lower NDVI values.

[0113] In addition, the multispectral camera can also acquire multispectral images. Based on the crop area covered by the multispectral image, the control device 5 can calculate the total leaf area per unit area (LAI) by statistically analyzing the ratio of "leaf-covered pixels" to "total pixels" and combining this with a geometric correction model (to eliminate lens distortion and terrain effects).

[0114] LiDAR (Light Detection and Ranging) emits laser pulses and records the time difference between transmission and reception. By combining this time with the speed of light, it calculates the distance to the top and bottom of the canopy, and the difference represents the canopy height. For example, if a laser hits the highest leaf and returns in 0.1 milliseconds, and hits the ground and returns in 0.2 milliseconds, the canopy height can be calculated using this time difference.

[0115] Furthermore, during lidar scanning, the ratio of "points with laser echoes" (reflected by leaves and branches) to "points without echoes" (penetrating to the ground) is statistically analyzed. Combined with the area of ​​the scanned region, the canopy closure can be calculated. The principle is similar to "outlining the canopy with point cloud data." Regardless of how the leaves overlap, as long as the laser can reach the vegetation, it is included in the coverage area, thus allowing the canopy closure to be calculated.

[0116] Multispectral cameras, with their narrow-band detection (focusing only on the critical 400-1000nm range), can filter out irrelevant spectra (such as ultraviolet and long-wave infrared), making leaf reflectance data more stable under different lighting conditions. Lidar, relying on "time-difference ranging," is completely unaffected by lighting conditions, accurately scanning canopy structures even on cloudy days or at dusk. The combination of these two technologies is equivalent to "using spectroscopy to observe physiology and using lasers to observe structure," ensuring stable data regardless of changes in lighting conditions.

[0117] Radar sensor 7 is used to scan the root system of crops and determine the root system data of crops, including root distribution density, root length and root activity.

[0118] Specifically, radar sensor 7 can be a ground-penetrating radar (GPR) array, with a predetermined number of detection units deployed in different areas within the greenhouse. The GPR array transmits high-frequency electromagnetic pulses (typically 100-1000MHz) that penetrate the soil. When the pulses encounter interfaces between different media, such as roots, they are reflected and captured by the receiving antenna. Due to the differences in electromagnetic properties (such as dielectric constant) between the roots and the surrounding soil, the strength and propagation time of the reflected signal will vary accordingly.

[0119] For root distribution density, the GPR array can acquire radar images of the soil profile through multi-channel synchronous scanning. The quantity and density of root reflection signals in the image can reflect the root distribution density per unit volume of soil—the denser the reflection signals, the denser the root distribution.

[0120] In determining root length, the extension length of the root system can be calculated by continuously tracking the root reflection signals in radar images and combining this with the propagation speed of electromagnetic waves in the soil. Thicker taproots and thinner lateral roots can be distinguished and their lengths measured separately due to their different reflection signal characteristics, and then the total root length can be obtained by summing these measurements.

[0121] Root activity can be indirectly determined by analyzing changes in the amplitude of reflected signals. Roots with higher activity have stronger cell activity and relatively stable water content, resulting in a more significant difference in dielectric constant between them and the soil, and thus a larger reflected signal amplitude. Roots with lower activity, on the other hand, have a weaker reflected signal amplitude. This can be used to assess root activity.

[0122] This ground-penetrating radar (GPR) array can perform non-destructive and efficient scanning of crop roots without damaging the soil and root system, providing important root data support for the formulation of subsequent irrigation strategies.

[0123] Control device 5 is used to output the target irrigation strategy for crops based on environmental data, canopy data and root system data.

[0124] In one optional embodiment of this application, such as Figure 6 As shown, the aforementioned "control device 5, used to output the target irrigation strategy corresponding to the crop based on environmental data, canopy data, and root data" may include the following steps:

[0125] Step S101: Identify canopy data and root data to establish an initial physiological causal relationship between the canopy and the root system.

[0126] Specifically, step S101 above may include the following steps:

[0127] Step S1011: Obtain the corresponding growth stage of the crop, and determine the current growth stage of the crop and the weight information corresponding to the current growth stage based on the growth stage.

[0128] Specifically, control device 5 can receive crop growth stages input by the user, as well as crop growth stages sent by other devices. Control device 5 can acquire crop images, identify the crop images, and determine the corresponding canopy features, stem features, and reproductive organ features. Canopy features include leaf area index (LAI), leaf morphology (e.g., leaf length, width, aspect ratio, leaf margin shape (e.g., long, thin leaves in rice seedlings, wide, and thick leaves in mature stages)), and leaf color; stem features include stem height, stem thickness, and number of internodes; reproductive organ features include flowering status: number of flowers and open proportion; fruit / seed features: number of fruits, size, color (e.g., the color gradient of tomatoes from green to red), and seed plumpness (e.g., changes in seed morphology during wheat grain filling). Then, based on the crop's corresponding canopy features, stem features, and reproductive organ features, it compares them with the canopy features, stem features, and reproductive organ features corresponding to each stage in the growth stage classification to determine the current growth stage of the crop.

[0129] Then, based on the correspondence between the current growth stage of the crop and the weight information, the weight information corresponding to the current growth stage is determined.

[0130] Step S10112: Calculate the target water transfer efficiency coefficient for the crop based on the weight information corresponding to the current growth stage.

[0131] Specifically, the control device 5 can derive the following formula based on canopy data (leaf area index, chlorophyll content, canopy height) and environmental data (radiation data, humidity data, wind speed data): Canopy transpiration = 0.622 × radiation intensity × LAI × (saturated vapor pressure difference) / (air pressure × (1 + 0.34 × wind speed)) (Note: In high humidity, the saturated vapor pressure difference is small, and the transpiration automatically decreases, reflecting the influence of the environment on water consumption).

[0132] Control device 5 can derive the following based on root data (root distribution density, root length, root activity) and soil data (field holding capacity) using a root water absorption model: Root water absorption = Root activity × Root distribution density × (Soil moisture content - Wilting coefficient) × 0.8. (Note: In heavy red soil, the soil moisture content is high but the available water is low; the water absorption calculation is automatically reduced by 15% to adapt to soil characteristics.)

[0133] The initial water transfer efficiency coefficient is calculated as follows: Initial water transfer efficiency coefficient = (canopy transpiration / root water absorption) × optimal ratio, with an environmental correction factor introduced. The optimal ratio can be calculated based on historical experience. When humidity > 80%, the initial water transfer efficiency coefficient is multiplied by 0.9 (high humidity inhibits transpiration, reducing transfer efficiency); when wind speed < 0.5 m / s, the initial water transfer efficiency coefficient is multiplied by 0.85 (still wind environment hinders water vapor diffusion, reducing transfer efficiency).

[0134] After the control device 5 calculates the initial water conduction efficiency coefficient, it can use the initial water conduction efficiency coefficient multiplied by the weight information corresponding to the current growth stage to calculate the target water conduction efficiency coefficient for the crop.

[0135] Step S10113: Based on the target water conduction efficiency coefficient, establish the initial physiological causal relationship between the canopy and the root system.

[0136] Specifically, control device 5 can divide the target water conduction efficiency coefficient into three dynamic threshold ranges: "cooperative balance interval," "root restriction interval," and "canopy restriction interval," providing quantitative triggering conditions for association rules. The specific division logic is as follows: Cooperative balance interval: target water conduction efficiency coefficient × 0.9 to target coefficient × 1.1 (canopy water consumption matches root water supply, no intervention required); Root restriction interval: ≤ target water conduction efficiency coefficient × 0.9 (root water absorption capacity is insufficient, unable to meet canopy demand); Canopy restriction interval: ≥ target water conduction efficiency coefficient × 1.1 (canopy transpiration / water consumption is limited, root water supply is relatively excessive).

[0137] For example: if the target coefficient for the grain-filling stage of rice is 0.8, then the equilibrium range is 0.72 to 0.88. If it is lower than 0.72, it is considered root limitation; if it is higher than 0.88, it is considered canopy limitation.

[0138] Control device 5 establishes a mapping relationship between canopy characteristics (chlorophyll content, leaf area index), root characteristics (root distribution density, root length, and root activity) and target water conduction efficiency coefficient ranges, forming a callable association rule base to visualize physiological relationships. Control device 5 can calculate the current water conduction efficiency coefficient of the crop by dividing canopy transpiration by root water uptake. Then, based on the preset rule base, it establishes an initial physiological causal relationship between the canopy and the root system.

[0139] For example, the core entries in the rule base (taking southern rice as an example): Root restriction interval (current water conduction efficiency coefficient ≤ 0.72): If the canopy chlorophyll content is normal (NDVI ≥ 0.7) and leaf area index meets the standard (LAI = 4-5), but root activity < 60% and root length < 30cm, then the association rule is "insufficient root activity → potential canopy water shortage → need to improve root water absorption capacity (e.g., adjust irrigation method to drip irrigation to promote deep root development)"; Canopy restriction interval (current water conduction efficiency coefficient ≥ 0.88): If If the root length is >40cm and the vitality is ≥80%, but the canopy chlorophyll content is low (NDVI <0.6) and the leaf area index is small (LAI <3), then the association rule is "weak canopy photosynthetic capacity → relatively excessive root water supply → need to improve canopy growth conditions (such as supplementing nitrogen to enhance chlorophyll synthesis)". In the synergistic balance interval (0.72 < current water conduction efficiency coefficient <0.88): when both canopy and root characteristics meet the standards, the association rule is "canopy water consumption - root water supply synergistic → maintain the current management strategy and continuously monitor coefficient fluctuations".

[0140] Optionally, control device 5 can also combine environmental data (temperature data, humidity data, wind speed data, wind direction data, and radiation data) to calibrate the association rules in real time, avoiding false associations caused by environmental stress and enhancing the reliability of physiological causal relationships. The calibration logic is as follows:

[0141] High humidity environment (humidity > 85%): Root activity measurement values ​​may be low (due to lack of oxygen). If the current water transfer efficiency coefficient is in the root limitation range, a "humidity correction coefficient" (e.g., 0.9) needs to be added, and the actual current water transfer efficiency coefficient (original current water transfer efficiency coefficient × 0.9) needs to be recalculated. If the correction results in a balance range, the association rule is adjusted to "high humidity environment leads to a false decrease in root activity → no increase in irrigation, priority given to ventilation to reduce humidity".

[0142] High radiation environment (radiation > 800W / m) 2 A sudden increase in the canopy transpiration rate may cause the coefficient to briefly enter the canopy limitation range. At this time, a "radiation correction factor" (such as 1.1) is introduced. After correction, if the current water conduction efficiency coefficient returns to the equilibrium range, the association rule is adjusted to "short-term strong transpiration leads to supply and demand imbalance → temporarily shorten irrigation interval to avoid canopy water shortage".

[0143] Through the above steps, the initial physiological causal relationship is transformed from an abstract "cooperative relationship" into a dynamic model of "quantitative threshold + rule base + environmental calibration + visual trigger". This model can accurately reflect the inherent growth law of crops and adapt to the impact of the complex environment of the humid southern region on the canopy and root system, providing a feasible decision-making basis for subsequent irrigation strategy optimization.

[0144] Step S102: Identify the environmental data and determine the initial latent stress signal from the environmental data.

[0145] Among them, latent stress signals are used to characterize non-overt physiological risks that pose a potential threat to crop growth.

[0146] Specifically, step S102 above may include the following steps:

[0147] Step S1021: Perform sliding window analysis on each data point in the environmental data and calculate the rate of change of the corresponding indicators for each data point.

[0148] Specifically, control device 5 can first identify various data points in the environmental data and determine the corresponding change characteristics of each data point. Then, based on the change characteristics of each data point, it can dynamically adjust the size of the sliding window to avoid lag or noise sensitivity issues caused by a fixed window.

[0149] For rapidly changing environmental data (such as wind speed and radiation intensity), a short window (5 minutes) is used to calculate the rate of change of the index per unit time (such as the rise and fall of wind speed every 5 minutes) and capture instantaneous fluctuations.

[0150] For slowly changing data (such as temperature and humidity), a long window (30 minutes) is used. Noise is smoothed by weighted averaging (0.7 weight for recent data and 0.3 weight for older data), and the trend change rate (such as the cumulative change in humidity over 30 minutes) is calculated.

[0151] Advantages: For example, during the plum rain season in southern China, a short window can capture sudden changes such as "humidity rising by 10% within 5 minutes" in a timely manner, while a long window can filter out "short-term fluctuations in humidity" and focus on the trend of "continuous high humidity for 30 minutes", providing more accurate rate of change data for subsequent stress signal identification.

[0152] Step S1022: Determine the spatial gradient corresponding to each data point based on the collection location of each data point in the environmental data.

[0153] Specifically, control device 5 can generate a spatial distribution heat map of the environmental data based on the collection locations corresponding to various data points in the environmental data (such as five points: east, west, south, north, and center inside the greenhouse), using Kriging interpolation. It then calculates the "point-to-point" local gradient (such as the temperature difference between two adjacent points) and the "region-to-region" global gradient (such as the humidity difference between the edge and center of the greenhouse). Next, it assigns a high weight (0.6) to the local gradient and a low weight (0.4) to the global gradient, thereby determining the spatial gradient corresponding to each data point. Priority is given to local environmental differences in densely planted crop areas (such as a region with significantly higher humidity due to insufficient ventilation).

[0154] Advantages: In greenhouse experiments in southern regions, it can accurately identify the hidden stress of "local high temperature and high humidity in the corner of the greenhouse", avoiding the neglect of local risks due to the normal global gradient.

[0155] Step S1023: Convert the environmental data into the frequency domain using Fourier transform to determine the frequency domain characteristics corresponding to the environmental data.

[0156] Specifically, control device 5 can convert environmental data (such as hourly recorded wind speed and radiation) into the frequency domain and identify the dominant frequency of periodic fluctuations (such as the 24-hour cycle of diurnal temperature difference and the 10-minute cycle of gusts). Then, it sets a "crop adaptation frequency library" (such as the radiation cycle for crop photosynthesis adaptation being 6-12 hours), filters out natural fluctuations that match the frequencies in the library, and retains abnormal frequencies (such as high-frequency fluctuations with sudden increases and decreases in radiation within 1 hour) as stress-related frequency domain features.

[0157] Advantages: It can distinguish between "normal diurnal humidity changes" and "sudden high humidity stress", providing a basis for identifying non-periodic environmental disturbances.

[0158] Step S1024: Input the rate of change of the indicators corresponding to each data point, the spatial gradient corresponding to each data point, and the frequency domain features corresponding to the environmental data into the dynamic mapping model.

[0159] Specifically, the control device 5 can input the rate of change of the indicators corresponding to each data point, the spatial gradient corresponding to each data point, and the frequency domain characteristics corresponding to the environmental data into the dynamic mapping model.

[0160] Step S1025: The dynamic mapping model determines the weight information corresponding to the rate of change of the index, the spatial gradient, and the frequency domain features.

[0161] Specifically, the dynamic mapping model can calculate the rate of change of indicators, spatial gradient, and frequency domain features based on a bidirectional LSTM network, and the basic correlation between these features and different stress types (such as high humidity, drought, and strong radiation).

[0162] Specifically, the control device 5 can initialize a bidirectional LSTM network in the dynamic mapping model, which includes a forward LSTM (capturing temporal dependencies from the past to the present) and a backward LSTM (capturing temporal dependencies from the future to the present). Each layer has 64 hidden units, and dropout (0.3) is used to prevent overfitting.

[0163] The input features are the rate of change of indicators arranged in time series (such as the rate of change of humidity every 10 minutes), spatial gradient (such as the spatial difference of temperature every 10 minutes), and frequency domain features (such as the frequency of radiation fluctuations every 10 minutes), forming an input matrix of shape [time step T, feature dimension D] (for example, T=60, D=9, corresponding to 10 hours × 6 features).

[0164] Forward LSTM computation: Features are processed sequentially from the first time step to the Tth time step, and the forward hidden state h_forward(t) is output, reflecting "how past features affect the current state". For example, if the humidity change rate increases sharply at time t-2, the forward LSTM will pass this information to time t, strengthening the recognition of the "persistent high humidity" trend.

[0165] Inverse LSTM computation: Features are processed in reverse from the T-th time step to the first time step, and the inverse hidden state h_backward(t) is output, reflecting the "backward impact of future features on the current state". For example, if a sudden drop in radiation occurs at time t+1, the inverse LSTM will feed this information back to time t to help determine the risk of combined stress of "high humidity + low radiation".

[0166] By concatenating the forward hidden state and the backward hidden state, we obtain a feature vector containing the full temporal dependence: h(t) = [h_forward(t)||h_backward(t)] (with a dimension of 128, i.e., 64+64). This vector contains both historical trends and predictions of potential future changes.

[0167] Initialize the event memory unit (EMU) as a dynamically updatable matrix EMU with shape [K, 128], where K is the maximum number of historical stress events to store (e.g., K = 50), and each row corresponds to the core feature of a historical stress event (extracted from h(t) output by the bidirectional LSTM).

[0168] Stress events of different stress types (such as "high humidity stress" and "drought stress") are selected from historical databases. The temporal features from 2 hours before the event to the time of occurrence are extracted. The hidden state h_event at the peak of the event is calculated using a bidirectional LSTM and stored in the EMU as the "feature fingerprint" of the event. For example, the feature fingerprint of the high humidity stress event is h_humid = [0.82, 0.15, ..., 0.31] (128 dimensions), which contains the temporal pattern of "continuous increase in humidity change rate + decrease in wind speed spatial gradient".

[0169] For the hidden state h(t) of the current bidirectional LSTM output, calculate its cosine similarity with the feature fingerprints of all historical events in the EMU: sim(k,t)=cosine_similarity(h(t),EMU(k))(k=1,2,...,K).

[0170] This allows us to calculate the basic correlation between the rate of change of the indicators, the spatial gradient, and the frequency domain features and different stress types (such as high humidity, drought, and strong radiation). Based on the basic correlation, we determine the initial weight information corresponding to the rate of change of the indicators, the spatial gradient, and the frequency domain features.

[0171] Based on the timestamps corresponding to the rate of change of the index, the spatial gradient, and the frequency domain characteristics, the time decay factor is calculated according to the principle of "prioritizing recent data and decaying older data".

[0172] For example, the rate of change of indicators, spatial gradient, and frequency domain features are divided by time granularity into: Immediate window (0-1 hour): data weight is retained at 100%, reflecting the latest environmental changes; Short-term window (1-6 hours): data weight decays linearly, e.g., data from 3 hours ago has a weight of 70%; Medium-term window (6-24 hours): data weight decays exponentially, e.g., data from 12 hours ago has a weight of 30%; Long-term window (>24 hours): data weight is fixed at 10%, providing only background reference. Then, differentiated time decay rates are set for different types of features. For example, such as... Figure 7As shown: Rapidly changing characteristics (such as humidity change rate): The half-life is set to 1 hour, meaning that the weight of data collected 2 hours ago is decayed to 25% of the initial value; Slowly changing characteristics (such as soil temperature spatial gradient): The half-life is set to 6 hours, and the weight of data collected 6 hours ago is retained at 50%; Periodic characteristics (such as radiation frequency domain characteristics): Corrected according to the diurnal cycle, the weight of radiation data collected at night is automatically reduced to 10%. Then, based on the time decay function, the time decay coefficients corresponding to the index change rate, spatial gradient, and frequency domain characteristics are calculated respectively.

[0173] For example, (1) Instant window (0≤Δt≤Δt1): data freshness is the highest, and the weight is kept at the maximum value (100%): w(Δt)=1.0; where w(Δt) is the time decay coefficient, Δt: the time difference between the feature timestamp and the current time (unit: hours, Δt≥0).

[0174] (2) Short-term window (Δt1<Δt≤Δt2): Linear decay is adopted to balance real-time performance and stability, ensuring a smooth transition from 1.0 to the starting point of the medium-term window:

[0175]

[0176] Among them, w lt Long-term window base weights (feature type specific, such as humidity w) lt =5%). Example: Δt1=1h, Δt2=6h, w lt When the slope is 15%, the slope is (1-0.15) / (6-1)=0.17, and the decay is 17% per hour.

[0177] (3) Mid-term window (Δt2<Δt≤Δt3): Exponential decay is adopted to quickly reduce the impact of outdated data, based on the feature-specific decay coefficient λ:

[0178] Example: When the temperature characteristic λ = 0.347 (half-life 2h) and Δt = 8h (Δt2 = 6h), the decay term is e^(-0.347×2) = 0.5, so w = 0.15×0.5 = 0.075.

[0179] (4) Long-term window (Δt>Δt3): The weight is fixed at half of the long-term base value and is superimposed with a periodic correction factor p (such as day / night / seasonal influence, 0.8≤p≤1.2).

[0180] Where Δt is the time difference between the feature timestamp and the current time (unit: hours, Δt≥0); w(Δt) is the time decay coefficient of the feature at the time difference Δt (0≤w≤1); Δt1, Δt2, and Δt3 are all time window thresholds (e.g., Δt1=1h, Δt2=6h, Δt3=24h); λ is the feature-specific decay coefficient (calculated from the half-life t1 / 2: λ=ln2 / t1 / 2)w lt Long-term window base weights (feature type specific, such as humidity w) lt =5%).

[0181] Next, the dynamic mapping model can calculate the spatial proportion of the influence of index change rate, spatial gradient, and frequency domain features. For example, global features (covering more than 80% of the area): weight increased by 30%, such as the average humidity change rate of the entire greenhouse; regional features (covering 30%-80% of the area): weight maintained at the baseline value, such as the temperature spatial gradient in the eastern half of the greenhouse; local features (covering <30% of the area): weight decreased by 20%, such as wind speed data from corner sensors. The dynamic mapping model can assign importance coefficients to the index change rate, spatial gradient, and frequency domain features of different regions. For example, core growth area (such as the crop canopy area): feature weight × 1.5; buffer zone (such as ventilation openings at the edge of the greenhouse): feature weight × 0.8; edge area (such as equipment storage area): feature weight × 0.5. Based on the spatial proportion of the influence of index change rate, spatial gradient, and frequency domain features, and the corresponding importance coefficients, the spatial attenuation coefficients corresponding to the index change rate, spatial gradient, and frequency domain features are obtained.

[0182] Then, the initial weight information corresponding to the rate of change of the index, the spatial gradient, and the frequency domain features are corrected using the time decay coefficient and the spatial decay coefficient, respectively, to obtain the weight information corresponding to the rate of change of the index, the spatial gradient, and the frequency domain features, respectively.

[0183] Step S1026: Construct a feature weight matrix based on the weight information corresponding to the index change rate, spatial gradient, and frequency domain features.

[0184] Specifically, based on the characteristics of three types of data—indicator change rate, spatial gradient, and frequency domain features—a "three-dimensional hierarchical matrix" is constructed, with the dimensions defined as follows:

[0185] The first dimension (rows) is the subdivision of feature types (e.g., the rate of change of indicators includes sub-features such as the rate of change of humidity and the rate of change of temperature); the second dimension (columns) is the crop growth stage (e.g., seedling stage, tillering stage, reproductive stage, maturity stage); the third dimension (layers) is the environmental scene label (e.g., sunny day, cloudy day, rainy day, windy day, etc.).

[0186] The weight information of the indicator change rate, spatial gradient, and frequency domain features is filled into the matrix according to the three-dimensional coordinates of "sub-feature-reproductive stage-environmental scenario". The core logic is: sub-feature weight mapping: for example, the weight of "humidity change rate" in the "reproductive stage-rainy day" scenario is 0.6 (taken from the weight calculation result), so the element value of the matrix coordinate (humidity change rate, reproductive stage, rainy day) is set to 0.6. Similarly, other sub-features are filled (such as the temperature spatial gradient having a weight of 0.4 in the "seedling stage-sunny day" scenario, with the corresponding coordinate element value being 0.4).

[0187] Normalize the column vector of each "reproductive stage × environmental scenario" to ensure that the sum of the weights of all sub-features under the same stage and scenario is 1. Formula: w_ij'=w_ij / Σw_ij (i is the sub-feature, j is the stage-scenario combination) to avoid the excessive weight of a certain type of feature.

[0188] For any two sub-features (such as "humidity change rate" and "humidity spatial gradient"), calculate the Pearson correlation coefficient r of their historical data. If |r| ≥ 0.7 (strong correlation), the correlation coefficient is set to 1.1; if |r| ≤ 0.3 (weak correlation), the correlation coefficient is set to 0.9. The weights of strongly correlated sub-features are enhanced bidirectionally, for example: humidity change rate weight = original weight × 1.1, humidity spatial gradient weight = original weight × 1.1; for weakly correlated sub-features (such as "light intensity frequency domain feature" and "wind speed change rate"), the weights are weakened bidirectionally: light intensity frequency domain feature weight = original weight × 0.9, wind speed change rate weight = original weight × 0.9.

[0189] For each sub-feature, a calibration factor is calculated based on the deviation rate δ between the current environmental data and the historical mean (δ = (current value - mean) / mean): if δ ≥ 20% (data is abnormally high), calibration factor = 1 + 0.1 × δ; if δ ≤ -20% (data is abnormally low), calibration factor = 1 + 0.05 × |δ|; under normal circumstances, calibration factor = 1.0.

[0190] The calibration factor is multiplied with the original matrix as the fourth dimension to generate the feature weight matrix.

[0191] Step S1027: Based on the feature weight matrix, calculate the matching degree between the input feature combination and the preset feature combination template. Based on the matching degree corresponding to the input feature combination and the feature over-standard degree corresponding to each individual feature in the input feature combination, determine the initial latent stress signal corresponding to the input feature combination.

[0192] The preset feature combination template is a "benchmark feature set" constructed based on historical stress event data. The template dimensions are consistent with the feature weight matrix, including three-dimensional labels of "sub-feature - growth stage - environmental scenario" (e.g., "drought latent stress template" under the "seedling stage - sunny day" scenario); feature threshold ranges: within each template, the normal fluctuation range and stress warning range of each sub-feature are clearly defined (e.g., in the drought template, the warning range for "soil moisture change rate" is ≤-0.2 / day, and the warning range for "temperature spatial gradient" is ≥0.5℃ / meter); template weight distribution: the ideal weight percentage of each sub-feature within the template is indicated (e.g., in the drought template, the weight of "moisture change rate" should be 40%, and "light frequency domain features" should be 25%), serving as a benchmark for subsequent matching degree calculations.

[0193] Control device 5 preprocesses the real-time collected input feature combinations (such as 12 sub-feature data including humidity change rate and temperature spatial gradient under the current "seedling stage - sunny day" scenario). The original values ​​of each sub-feature are converted into standardized values ​​in the [0,1] interval, using the formula: Standardized value = (Original value - Historical minimum value) / (Historical maximum value - Historical minimum value). For example, if the original humidity change rate is -0.15 / day, the historical minimum is -0.3 / day, and the maximum is 0.2 / day, then the standardized value is (-0.15 + 0.3) / (0.2 + 0.3) = 0.3. Then, the input feature combinations are labeled with "fertility stage + environmental scenario" (e.g., the current scenario matches "seedling stage - sunny day") to retrieve the corresponding weight data from the feature weight matrix.

[0194] Let the standardized vector of the input feature combination be X = [x1, x2, ..., x]. n The standard vector of the preset template is Y = [y1, y2, ..., y]. n (The ideal values ​​of each sub-feature in the template), the weight vector corresponding to the "seedling stage - sunny day" scenario in the feature weight matrix is ​​W = [w1, w2, ..., w n (e.g., humidity change rate weight w1 = 0.4, temperature spatial gradient weight w2 = 0.3, etc.), then the weighted cosine similarity is calculated as the matching degree.

[0195] Set a matching threshold θ (e.g., θ = 0.7). When the matching degree between the input feature combination and a certain template is ≥ θ, it is determined to be a "potential matching template"; otherwise, the template is excluded.

[0196] For each sub-feature in the input feature combination, calculate its "degree of deviation from the normal range". Specifically: if the sub-feature value in the input feature combination is higher than the upper limit of the normal range of the template (e.g., temperature spatial gradient = 0.6℃ / m, template upper limit = 0.5℃ / m), it is judged as "positive deviation"; if it is lower than the lower limit (e.g., humidity change rate = -0.3 / day, template lower limit = -0.2 / day), it is judged as "negative deviation".

[0197] The deviation degree index is calculated using a two-factor model of "deviation degree-weight". The formula is: deviation degree index = (|feature value - interval critical value| / interval width) × corresponding weight. For example, if the humidity change rate = -0.3 / day (template lower limit = -0.2 / day, interval width = 0.4 / day) and the corresponding weight = 0.4, then: deviation degree index = (|-0.3-(-0.2)| / 0.4) × 0.4 = (0.1 / 0.4) × 0.4 = 0.1. The higher the index, the greater the contribution of the feature to the stress (value range [0,1]).

[0198] By combining the matching degree and the degree of exceeding the standard index, the initial latent stress signal is determined through "double threshold filtering + hierarchical sorting".

[0199] First layer of filtering: Matching degree threshold screening: Only retain the preset template with a matching degree ≥ θ (e.g., 0.7) (for example, if the input feature combination has a matching degree of 0.82 with the "drought latent stress template" and a matching degree of 0.51 with the "pest and disease template", then only the drought template will be retained).

[0200] Second layer of filtering: Total exceedance verification: Calculate the sum of exceedance indices S of all sub-features in the input feature combination. If S≥0.5 (indicating that multiple features exceed the limit at the same time), proceed to the next step; otherwise, it is judged as "no significant stress signal".

[0201] Finally, the signal strength is determined by the product of the matching degree (M) and the total excess (S), using the formula: Signal Strength = M × S, and is divided into three levels: Strong signal: signal strength ≥ 0.6 (e.g., M = 0.82, S = 0.8 → 0.656); Medium signal: 0.3 ≤ signal strength < 0.6 (e.g., M = 0.75, S = 0.5 → 0.375); Weak signal: 0.1 ≤ signal strength < 0.3 (e.g., M = 0.72, S = 0.35 → 0.252).

[0202] Finally, the initial latent stress signal was determined, and the output format was "stress type + signal intensity + key exceedance features", for example: "drought latent stress (strong signal): the main contributing features are humidity change rate (exceedance index 0.32) and temperature spatial gradient (exceedance index 0.28)".

[0203] Step S103: Input the initial physiological causal relationship and the initial latent stress signal into the preset irrigation strategy model, and output the target irrigation strategy corresponding to the crop.

[0204] Specifically, step S103 above may include the following steps:

[0205] Step S1031: Based on the initial physiological causal relationship, calculate the basic irrigation requirements corresponding to crops.

[0206] Specifically, based on the initial physiological causal relationship between the canopy and the root system (such as "the target water conduction efficiency coefficient needs to be maintained at 0.8±0.1"), and combined with the current canopy transpiration rate (derived from leaf temperature inversion) and root water absorption capacity (based on root vigor and length in the root data), the basic irrigation demand is calculated.

[0207] Calculation formula: Basic irrigation demand = (canopy transpiration rate × correction coefficient) / (target water transfer efficiency coefficient).

[0208] Correction coefficient: dynamically adjusted according to the growth stage (correction coefficient setting rules: seedling stage: 0.9 (low water requirement, avoid excessive growth); tillering stage / vegetative growth stage: 1.0 (normal water requirement); jointing stage / flowering stage: 1.2 (peak water requirement, extra 10% reserve); grain filling stage / maturity stage: 1.1 (stable water requirement, while ensuring full grains).

[0209] In this study, infrared thermal imagers were used to collect canopy leaf temperatures. Combined with environmental parameters (air temperature, humidity, and wind speed), the transpiration rate was calculated using a "leaf-air temperature difference model": Canopy transpiration rate = g × (leaf temperature - air temperature) × wind speed correction factor; where g is a crop-specific coefficient (0.8 for rice and 0.7 for wheat), and the wind speed correction factor is 1.2 when the wind speed is <1 m / s (weak transpiration, need to amplify the effect of temperature difference) and 0.9 when the wind speed is >3 m / s (strong transpiration, need to reduce the effect).

[0210] Example: Leaf temperature 30℃, air temperature 28℃, wind speed 2m / s (correction factor 1.0), rice g = 0.8, then canopy transpiration rate = 0.8 × (30-28) × 1.0 = 1.6 mm / h, which is converted to daily transpiration rate = 1.6 × 10 (effective light hours) = 16 mm / d (this is an example, and the actual rate needs to be adjusted according to crop characteristics).

[0211] Step S1032: Based on the relationship between the initial latent stress signal and the basic irrigation demand, calculate the risk regulation amount corresponding to the crop.

[0212] Specifically, the control device 5 classifies the initial latent stress signal into 5 risk levels (levels 1-5) based on the intensity (strong / medium / weak) and type (high humidity / drought / nutrient imbalance). The higher the level, the greater the regulation amount.

[0213] Risk regulation amount = basic irrigation demand × stress response coefficient × risk level coefficient.

[0214] Among them, the stress response coefficient is 1.3 for drought stress (requiring increased irrigation), 0.7 for high humidity stress (requiring reduced irrigation), and 1.0 for nutrient imbalance (maintenance amount but adjustment method); the risk level coefficient is 0.1 for level 1, 0.2 for level 2, and so on (0.5 for level 5).

[0215] Example: If the basic irrigation amount is 6.47 mm / d and a "moderate drought stress (risk level 3)" occurs, then the risk regulation amount = 6.47 × 1.3 × 0.3 ≈ 2.5 mm / d (an additional 2.5 mm of irrigation is required).

[0216] Step S1033: Based on the basic irrigation demand and risk control amount, determine the initial irrigation amount for the crops.

[0217] Specifically, the control device 5 can calculate the initial crop irrigation amount based on the following formula: Initial crop irrigation amount = basic irrigation demand ± risk control amount (+ for drought, - for high humidity).

[0218] Constraints: Upper limit: not exceeding 80% of the soil field water holding capacity (e.g., if the field water holding capacity of sandy soil is 20%, then the soil moisture after irrigation should be ≤16%); Lower limit: not lower than the crop wilting coefficient (e.g., if the wilting coefficient of rice is 10%, then the soil moisture after irrigation should be ≥10%).

[0219] If the calculation result exceeds the upper / lower limit, it will be automatically corrected to the upper / lower limit value, and the "adjustment deviation" will be recorded (e.g., if the irrigation should be 8mm, but the upper limit is 7mm, the deviation +1mm will be included in the next irrigation).

[0220] Step S1034: Determine the root system characteristics of the crop based on the root system data; determine the initial irrigation period and initial irrigation method based on the root system characteristics.

[0221] Specifically, control device 5 identifies root system data and extracts key features from it: taproot depth (e.g., 30cm) determines irrigation water penetration depth; lateral root distribution density (e.g., 20 roots / cm²)... 3 The irrigation coverage area is determined by the root system's peak activity period (e.g., 9-11 am). The optimal irrigation time is determined by the root system's peak activity period.

[0222] Then, match the peak root activity with environmental conditions (such as avoiding high temperature evaporation at noon). For example, if the peak root activity is between 9 and 11 o'clock and the wind speed is <2m / s (weak evaporation) at this time, then set the irrigation period to 9:00-10:00. If there is cloudy or rainy weather, postpone it to 15-16 o'clock in the afternoon (to avoid the superposition of high humidity stress).

[0223] Choose the method based on the root distribution characteristics: For plants with deep taproots and few lateral roots (such as corn): use drip irrigation (20cm deep) to specifically moisten the taproot area; For plants with shallow and dense roots (such as vegetables): use a combination of sprinkler irrigation and surface flood irrigation to balance coverage and surface moisture; For plants with low root activity (such as during the maturity stage): add "pulse irrigation" (stop for 5 minutes every 30 minutes) to promote root water absorption efficiency.

[0224] Step S1035: Determine the initial irrigation strategy based on the initial crop irrigation amount, initial irrigation period, and initial irrigation method.

[0225] Specifically, the control device 5 determines the initial irrigation strategy based on the initial crop irrigation amount, the initial irrigation period, and the initial irrigation method.

[0226] Step S1036: The initial irrigation strategy is calibrated to obtain the target irrigation strategy.

[0227] Specifically, step S1036 above may include the following steps:

[0228] Step a1: Define the state space and action space based on the initial irrigation strategy.

[0229] The action space is used to adjust the various parameters included in the initial irrigation strategy and generate various irrigation actions.

[0230] Specifically, define the state space (S)1, which includes key parameters of the initial irrigation strategy and real-time environmental feedback: Initial strategy parameters: Initial irrigation amount (e.g., 0.5m³). 3 / mu), initial time period (e.g., 6:00 AM), initial method (e.g., drip irrigation); environmental feedback data: soil moisture (0-100cm soil moisture content), canopy data (chlorophyll content, LAI), root data (vitality, distribution density), and latent stress signals (e.g., high humidity and low light risk level) 24 hours after implementation.

[0231] Define the action space (A) 1. Set the calibrable dimensions and adjustment range of the initial strategy, and optimize key features. Main actions: 1. Irrigation volume adjustment: ±5%, ±10%, ±15% (5% step size to adapt to sudden increases and decreases in soil moisture caused by heavy rainfall); 2. Irrigation time adjustment: ±1 hour (e.g., from 6:00 to 5:00 or 7:00 to avoid the high humidity period of morning dew during the rainy season); 3. Irrigation mode switching: drip irrigation → micro-sprinkler (or vice versa) to meet the "water control + oxygenation" needs when the root system is hypoxic. Fine-tuning actions: for detailed parameters such as the interval of pulse irrigation (±5 minutes) and the distribution of irrigation points (local density / sparseness).

[0232] Each action comes with a "cost coefficient" (e.g., flood irrigation cost coefficient 1.2, drip irrigation cost coefficient 0.8) to prevent high-cost actions from being over-selected.

[0233] Step a2: For each irrigation action, input the environmental data, canopy data, root data, soil data corresponding to the crop, and irrigation action into the preset crop water production function model. The preset crop water production function model outputs the simulated yield and simulated water use efficiency corresponding to each action.

[0234] Specifically, based on traditional crop water production functions (such as the Jensen model), a pre-defined crop water production function model with a three-layer dynamic framework of "input layer - treatment layer - output layer" is constructed. The input layer receives environmental data (temperature, humidity, radiation, etc.), canopy data, root data, soil data (0-30cm soil moisture, field capacity, etc.), and irrigation parameters (irrigation amount, time period, method). The treatment layer embeds a "stress memory module" and a "stage-sensitive module" to achieve dynamic responses to cumulative effects and growth stages. The output layer outputs simulated yield, water use efficiency, and additional indicators (potential stress probability, yield components). The core formula of the traditional crop water production function is based on:

[0235] Relative output = ∏(1-(1-ETa / ETm)^λ_i).

[0236] Where ETa is the actual evapotranspiration, ETm is the maximum evapotranspiration, and λ_i is the sensitivity coefficient of the reproductive stage (base value), which is dynamically corrected by the processing layer.

[0237] The "stress memory module" in the processing layer calculates individual stress indices based on the environmental conditions after each irrigation: High humidity stress index: S_wet = (soil moisture - suitable humidity) / field capacity (effective when humidity exceeds suitable, maximum 0.5); Drought stress index: S_dry = (suitable humidity - soil moisture) / wilting coefficient (effective when humidity is below suitable, maximum 0.5). For example, if the soil moisture is 30% after the first irrigation (suitable humidity 25%, field capacity 35%), then S_wet1 = (30-25) / 35 ≈ 0.14.

[0238] Cumulative Stress Index and Memory Factor: The cumulative stress index is generated by weighting and summing the previous three stress indices: S_total=0.5×S_1+0.3×S_2+0.2×S_3 (recent stress has a higher weight).

[0239] The stress memory factor M is defined as: M = 1 + 0.2 × S_total (when S_total = 0.5, M = 1.1, which means amplifying the stress effect by 10%).

[0240] The stress memory factor is embedded in the treatment layer to correct the evapotranspiration calculation: ETa' = ETa × M (ETa' increases when high humidity accumulates, simulating the decrease in water absorption capacity caused by root hypoxia).

[0241] The "stage-sensitive module" in the processing layer assigns values ​​to the basic sensitivity coefficients: It sets basic λ_i values ​​for different growth stages (e.g., 0.2 for seedling stage, 0.3 for tillering stage, 0.4 for reproductive stage, and 0.1 for maturity stage) to reflect natural sensitivity differences. It introduces the crop variety's "water sensitivity index" K_s (K_s = 0.8 for drought-resistant varieties and K_s = 1.2 for moisture-loving varieties), and modifies it as: λ_i' = λ_i × K_s. For example, the λ_i' of drought-resistant maize during the reproductive stage is 0.4 × 0.8 = 0.32, lower than that of moisture-loving rice (0.4 × 1.2 = 0.48).

[0242] Further adjustments are made based on the current environmental stress index (0-10 points): λ_i”=λ_i'×(1+0.05×stress index). If the reproductive period encounters a stress score of 3, the λ_i” of maize is 0.32×(1+0.05×3)=0.37, and its sensitivity increases with the increase of stress.

[0243] Cumulative effect parameter processing: The cumulative effect of the first 3 irrigations is converted into a calculable value through "weighted summation + state transformation": the degree of root hypoxia caused by continuous high humidity: hypoxia index = 0.6×S_wet1 + 0.3×S_wet2 + 0.1×S_wet3 (range 0-0.5).

[0244] The degree of root shrinkage caused by prolonged drought: Shrinkage index = 0.6 × S_dry1 + 0.3 × S_dry2 + 0.1 × S_dry3 (range 0-0.5).

[0245] Variety parameters such as drought resistance coefficient and water sensitivity index are encoded into "variety feature vectors". For example, the feature vector of drought-resistant wheat is [0.9 (drought resistance coefficient), 0.8 (water sensitivity index), 1.1 (recovery coefficient)], which is used for specific correction of each stage of the model.

[0246] Irrigation action parameter conversion: Convert discrete parameters such as irrigation method and time period into continuous coefficients: Irrigation method coefficient: drip irrigation 0.8, sprinkler irrigation 1.0, flood irrigation 1.2 (reflecting the difference in water use efficiency).

[0247] Time period coefficients: 0.9 for morning (6-9 am), 1.2 for noon (10-2 pm), and 1.0 for evening (3-6 pm) (reflecting differences in evaporation loss).

[0248] Actual evapotranspiration (ETa) calculation: ETa = Irrigation volume × Time period coefficient × Irrigation method coefficient + Effective rainfall - Deep seepage.

[0249] Example: Irrigation amount = 20mm (base amount 16.7mm, increased by 20%), time period is noon (coefficient 1.2), flood irrigation (coefficient 1.2), effective rainfall 0mm, deep seepage = 20×1.2×1.2×0.15 (flood irrigation seepage rate 15%) = 4.32mm, then ETa = 20×1.2×1.2-4.32 = 28.8-4.32 = 24.48mm.

[0250] Two of the first three irrigations were high-humidity. S_total=0.14(first time)+0.08(second time)+0=0.22, then M=1+0.2×0.22=1.044, after correction ETa'=24.48×1.044≈25.56mm.

[0251] The crop is in the reproductive stage, with a basic λ_i = 0.4, and the variety is moisture-loving rice (K_s = 1.2). The current stress index is 3 points, so λ_i” = 0.4 × 1.2 × (1 + 0.05 × 3) = 0.48 × 1.15 = 0.552.

[0252] The core benchmark of the pre-defined crop water production function model is: relative yield = ∏(1-(1-ETa' / ETm)^λ_i”), assuming ETm = 28mm, then 1-ETa' / ETm = 1-25.56 / 28≈0.087.

[0253] Simulated yield = 1 - (0.087)^0.552 ≈ 0.78 (Benchmark yield 1090 kg / mu, simulated yield = 1090 × 0.78 ≈ 850 kg / mu, +5% compared to benchmark).

[0254] Simulated water use efficiency = Simulated yield / (Irrigation amount × 10) (1 mm / m) 2 =10m 3 / mu), that is, 850 / (20×10)=4.25kg / m 3 .

[0255] Root rot risk: Based on the hypoxia index and current soil moisture, the root rot risk = 0.5 × hypoxia index + 0.5 × (soil moisture - field capacity) / field capacity (effective when exceeding field capacity). In the example, the hypoxia index is 0.22 and the soil moisture is 32% (field capacity 30%), then the root rot risk = 0.5 × 0.22 + 0.5 × (32 - 30) / 30 ≈ 0.11 + 0.03 = 0.14. Taking 0.3 (moderate) as in the original example may integrate other stress factors.

[0256] Step a3: Based on the simulated yield and simulated water use efficiency, construct the dynamic reward function corresponding to the initial irrigation strategy.

[0257] The dynamic reward function includes at least one of the following: yield reward, water use efficiency reward, and experimental data stability reward.

[0258] In one optional implementation, the dynamic reward function includes a yield reward, a water use efficiency reward, and a test data stability reward; wherein: the yield reward is as follows:

[0259]

[0260] Where R1 is the production bonus, Y 模拟 To simulate output; Y 潜在 S represents the potential yield of a crop under optimal conditions. 胁迫 The stress correction coefficient is related to environmental data; w1 is the first weight information.

[0261] Water use efficiency rewards are as follows:

[0262]

[0263] Where R2 is the water use efficiency bonus, WUE 模拟 To simulate water use efficiency (simulated yield / water consumption); WUE 基准 The baseline water use efficiency for crops; G 雨水 w1 represents the rainwater reuse gain; w2 represents the second weighting information.

[0264] The experimental data stability bonus is as follows:

[0265]

[0266] Where, σ Y The standard deviation coefficient of the simulated output for a continuously preset number of times; σ WUE The standard deviation coefficient of water use efficiency is the number of consecutive preset simulations; w3 is the third weighting information;

[0267] The dynamic reward function is as follows:

[0268] R 总 =R1+R2+R3

[0269] Among them, R 总 This is a dynamic reward function.

[0270] Step a4: Evaluate each irrigation action in the action space based on the dynamic reward function.

[0271] Specifically, the control device 5 evaluates each irrigation action based on the reward value corresponding to each irrigation action in the action space, using a dynamic reward function.

[0272] Step a5: Based on the evaluation results, adjust the initial irrigation strategy to obtain the target irrigation strategy.

[0273] Specifically, control device 5 maps the evaluation results to three core adjustment dimensions, forming a three-dimensional decision-making space: Spatiotemporal dimension: adjusting irrigation time (e.g., advancing / delaying the irrigation period) and spatial distribution (e.g., locally increasing / decreasing irrigation points); Intensity dimension: adjusting irrigation volume (±10% / ±20%) and frequency (increasing / decreasing the number of irrigations); Mode dimension: switching irrigation modes (drip irrigation → sprinkler irrigation → flood irrigation) or mixed modes (e.g., drip irrigation + pulse irrigation). Each dimension has a set elastic threshold (e.g., irrigation volume adjustment should not exceed 80% of field capacity) to avoid excessive adjustments leading to resource waste.

[0274] Based on the stress risk level in the assessment results, the decision threshold is automatically adjusted: High-risk scenarios (e.g., root rot risk ≥ 0.6): trigger the "safety priority" mode, tighten the threshold (e.g., reduce the irrigation volume limit to 60% of field capacity), and prioritize reducing the risk; Low-risk scenarios (e.g., water use efficiency ≥ 1.5 kg / m³): 3 ): Triggers "efficiency priority" mode, relaxes thresholds (e.g., allows irrigation volume to briefly exceed 80% of field capacity) to pursue higher yields; Medium-risk scenario: Maintains default thresholds and adopts a "balanced optimization" strategy. Example: If the assessment shows a root rot risk of 0.7 (high risk), the original plan of flood irrigation + 20% water volume is adjusted to drip irrigation + 10% water volume, and the irrigation time is moved from noon to the morning (to reduce evaporation).

[0275] Then, irrigation parameters (time, amount, method) are encoded into chromosomes (e.g., [8:00, 7.5mm, drip irrigation]). The desirable characteristics of different strategies are randomly combined (e.g., combining the time of strategy A with the water volume of strategy B), and a certain parameter of the chromosome is adjusted slightly (e.g., irrigation volume ±5%). The adjusted chromosomes are evaluated, and based on the reward function values ​​in the evaluation results, the top 30% of high-quality strategies are retained for the next generation. This process is repeated until the reward function values ​​corresponding to each chromosome stabilize within a preset range. From this, the optimal chromosome is determined, and the target irrigation strategy is generated based on the irrigation parameters corresponding to the optimal chromosome.

[0276] The intelligent crop testing system provided in this application uses a canopy recognition device to generate canopy data such as chlorophyll content and leaf area index through imaging and scanning, which intuitively reflects the crop's photosynthetic capacity and growth vitality. A radar sensor scans and acquires root data such as root distribution density, length, and vitality, revealing the crop's water absorption capacity and distribution characteristics. The control device 5 combines these two types of data to accurately determine the matching degree between the crop's "canopy water demand" and "root water supply capacity," avoiding the supply-demand imbalance caused by traditional irrigation relying solely on experience or a single indicator, and making the target irrigation strategy more closely aligned with the actual physiological state of the crop.

[0277] Control device 5 acquires information on the crop's growth stages, determines the current growth stage and its corresponding weight, and focuses subsequent analysis on the crop's key growth needs. This targeted approach avoids the crudeness of a "uniform analysis throughout the entire growth period," ensuring that the initial physiological causal relationship aligns with the crop's growth dynamics in real time, improving the timeliness and accuracy of the correlation. Based on the current growth stage and weight information, the target water conduction efficiency coefficient (characterizing the degree of matching between canopy water demand and root water supply capacity) is calculated for each growth stage. This coefficient transforms the physiological relationship between the canopy and roots from a qualitative description into a quantitative indicator, making the correlation analysis more objective and operable, and providing a clear quantitative basis for subsequent irrigation strategy formulation. The initial physiological causal relationship established based on the target water conduction efficiency coefficient can dynamically adapt to the physiological characteristics of different crop growth stages. This dynamic relationship ensures that irrigation strategies can be adjusted according to crop growth stages, avoiding "one-size-fits-all" management, and is particularly adaptable to the significant changes in water demand during the crop growth cycle in humid regions, improving the targeting and effectiveness of irrigation regulation.

[0278] Control device 5 analyzes and calculates the rate of change of indicators through a sliding window, determines the spatial gradient based on the collection location, and extracts frequency domain features through Fourier transform, comprehensively analyzing environmental data from three dimensions: time, space, and frequency. This multi-dimensional analysis breaks through the limitations of traditional single-indicator monitoring, capturing complex signals in humid environments and providing richer evidence for the identification of latent stresses. The dynamic mapping model determines the weight information of the rate of change of indicators, spatial gradient, and frequency domain features and constructs a feature weight matrix, which can dynamically adjust the importance of each feature according to the environmental characteristics of humid regions. This dynamic adaptation ensures a high degree of fit between feature analysis and the regional environment. Then, based on a preset feature combination template, the matching degree is calculated, and the initial latent stress signal is determined by combining the degree of exceedance of individual features, realizing a dual verification of "overall pattern recognition + local feature verification". This mechanism can effectively identify latent stress signals unique to humid regions. By capturing these signals in advance, sufficient response time can be provided for experimental management, reducing the interference of environmental fluctuations on crop growth and experimental data, and ensuring the stability and reliability of agricultural experiments under complex climatic conditions.

[0279] Control device 5 calculates the basic irrigation demand based on the initial physiological causal relationship, ensuring that the irrigation amount matches the inherent law of crop "canopy water consumption - root water absorption". The basic irrigation demand is dynamically adjusted as both grow synergistically, avoiding the supply-demand imbalance caused by traditional calculations using fixed coefficients, and providing a scientific benchmark for subsequent regulation. Risk regulation is calculated based on the relationship between initial latent stress signals and basic irrigation demand, enabling targeted responses to potential physiological risks. The initial crop irrigation amount obtained by integrating basic irrigation demand and risk regulation not only meets the basic water requirements for crop growth but also incorporates considerations for environmental stress response. This ensures both growth and avoids waterlogging, significantly improving water resource utilization efficiency. Root characteristics are determined based on root data, thus matching the initial irrigation time and method: shallow-rooted crops are irrigated in the early morning to reduce evaporation; deep-rooted crops are irrigated with micro-sprinklers for extended periods to promote water infiltration. This "root adaptation" mechanism makes water easier to absorb, avoiding waste caused by a disconnect between irrigation and root distribution. Control device 5 defines the state space and action space based on the initial irrigation strategy, providing a clear range of variables for strategy optimization. By pre-setting a crop water production function model, various data and irrigation actions are input, and simulated yield and water use efficiency are output, allowing for prediction of action effects without actual execution. This process reduces the manpower and material resources consumed in field trials, while avoiding interference with crop growth from improper irrigation. It is particularly suitable for experimental scenarios in humid regions with rainy seasons, providing a safe and efficient evaluation basis for strategy optimization. A dynamic reward function, constructed based on simulated yield, water use efficiency, and experimental data stability, can comprehensively evaluate the merits of irrigation actions. This multi-dimensional reward mechanism avoids the limitations of single-objective optimization, guiding strategies towards a synergistic development of "high yield, water conservation, and reliable data." After evaluating irrigation actions in the action space through the dynamic reward function, the adjustment direction corresponding to high reward values ​​is prioritized for iterative optimization of the initial irrigation strategy. This process enables the target irrigation strategy to continuously adapt to environmental changes and crop growth dynamics. From state action definition and simulation evaluation to reward feedback and strategy adjustment, a complete closed-loop optimization mechanism is formed, allowing the initial irrigation strategy to gradually evolve into the target irrigation strategy. This mechanism breaks away from the traditional static management model that relies on experience. Through data-driven dynamic optimization, it continuously improves the scientific nature and effectiveness of irrigation strategies in agricultural experiments in humid southern regions, providing stable and reliable water management support for experimental research.

[0280] Although embodiments of the invention have been described in conjunction with the accompanying drawings, those skilled in the art can make various modifications and variations without departing from the spirit and scope of the invention, and such modifications and variations all fall within the scope defined by the appended claims.

Claims

1. An intelligent crop testing system, characterized in that, The intelligent crop testing system includes a rain shelter, rain collection equipment, environmental monitoring equipment, irrigation equipment, and control equipment. The rain shelter is connected to the rain collection equipment, and the irrigation equipment is also connected to the rain collection equipment. The environmental monitoring equipment is installed inside the rain shelter and is connected to the control equipment. Rain shelters are used to provide rain protection for crops inside the shelter under the control of control equipment and to transfer rainwater to rain collection equipment; Rainwater harvesting equipment is used to collect rainwater and treat the collected rainwater under the control of control equipment so that the rainwater meets irrigation standards; Environmental monitoring equipment is used to detect environmental data inside the rain shelter and transmit the environmental data to control equipment; the environmental data includes at least one of temperature data, humidity data, wind speed data, wind direction data, and radiation data; The control equipment is used to determine the target irrigation strategy for crops based on environmental data; and to control the irrigation equipment to draw rainwater from the rainwater collection equipment to irrigate the crops according to the target irrigation strategy. The control equipment is used to identify canopy and root data and establish initial physiological causal relationships between the canopy and roots; identify environmental data and determine initial latent stress signals from the environmental data; latent stress signals are used to characterize non-obvious physiological risks that pose a potential threat to crop growth; input the initial physiological causal relationships and initial latent stress signals into a preset irrigation strategy model and output the target irrigation strategy corresponding to the crop. The control equipment is used to perform sliding window analysis on various data points in the environmental data, calculate the rate of change of each corresponding indicator; determine the spatial gradient of each data point based on its acquisition location; convert the environmental data into the frequency domain using Fourier transform to determine the frequency domain features of the environmental data; input the rate of change of each data point, the spatial gradient of each data point, and the frequency domain features of the environmental data into a dynamic mapping model; the dynamic mapping model determines the weight information corresponding to the rate of change of indicators, the spatial gradient, and the frequency domain features; construct a feature weight matrix based on the weight information corresponding to the rate of change of indicators, the spatial gradient, and the frequency domain features; calculate the matching degree between the input feature combination and the preset feature combination template based on the feature weight matrix; and determine the initial latent stress signal corresponding to the input feature combination based on the matching degree of the input feature combination and the degree of feature overshoot of each individual feature in the input feature combination. The control equipment is used to calculate the basic irrigation demand of crops based on initial physiological causal relationships; calculate the risk regulation amount of crops based on the relationship between initial latent stress signals and basic irrigation demand; determine the initial irrigation amount of crops based on basic irrigation demand and risk regulation amount; determine the root characteristics of crops based on root data; determine the initial irrigation period and initial irrigation method based on root characteristics; determine the initial irrigation strategy based on the initial irrigation amount, initial irrigation period, and initial irrigation method; and calibrate the initial irrigation strategy to obtain the target irrigation strategy. The system includes a control device for defining a state space and an action space based on an initial irrigation strategy. The action space is used to adjust various parameters included in the initial irrigation strategy to generate different irrigation actions. For each irrigation action, environmental data, canopy data, root data, soil data corresponding to the crop, and the irrigation action itself are input into a preset crop water production function model. The preset crop water production function model outputs the simulated yield and simulated water use efficiency corresponding to each action. Based on the simulated yield and simulated water use efficiency, a dynamic reward function corresponding to the initial irrigation strategy is constructed. The dynamic reward function includes at least one of yield reward, water use efficiency reward, and experimental data stability reward. Each irrigation action in the action space is evaluated based on the dynamic reward function. Based on the evaluation results, the initial irrigation strategy is adjusted to obtain the target irrigation strategy. The dynamic reward function includes yield reward, water use efficiency reward, and experimental data stability reward; among which, the yield reward is as follows: ; in, R 1 is a production bonus. Y 模拟 To simulate production output; Y 潜在 The potential yield of crops under optimal conditions; S 胁迫 This is a stress correction factor, which is related to environmental data; w 1 represents the first weighted information; Water use efficiency rewards are as follows: ; in, R 2 is a water use efficiency bonus, WUE 模拟 To simulate water use efficiency, i.e., simulated yield / water consumption; WUE 基准 The baseline water use efficiency for crops; G 雨水 Gain for rainwater reuse; w 2 represents the second weighting information; The experimental data stability bonus is as follows: ; in, σ Y The standard deviation coefficient of the simulated output for a continuous preset number of times; σ WUE The standard deviation coefficient of water use efficiency is calculated for a continuously preset number of simulations. w 3 represents the third weighting information; The dynamic reward function is as follows: ; in, R It is always a dynamic reward function.

2. The intelligent crop testing system according to claim 1, characterized in that, A rain sensor and a wind speed sensor are installed above the rain shelter. The rain sensor is communicatively connected to the control device. An automatic roller blind transparent film is installed around the perimeter of the upper surface of the rain shelter. An elastic retraction mechanism is installed along the ridge line of the upper surface of the rain shelter. The automatic roller blind transparent film is connected to a drive motor, which is communicatively connected to the control device. Telescopic support rods are installed at the bottom of both sides of the bottom edge of the rain shelter. Water channels are installed between the arches of the rain shelter and at the edge of the roof. Wherein: The control device is used to receive rainfall data transmitted by the rainfall sensor and combine it with the acquired weather forecast data. When the rainfall reaches a preset rainfall threshold and / or within a preset time before the rainfall, the device controls the drive motor to drive the automatic roller shutter transparent film to descend to a position at a first preset height above the ground, and controls the elastic shrinking mechanism to shrink to provide rain protection for the crops inside the rain shelter, and transmits the rainwater to the rain collection device through the water conveying trough. The control device is also used to receive wind speed data transmitted by the wind speed sensor, and when the wind speed is greater than a preset wind speed threshold, control the drive motor to drive the automatic roller blind transparent film to rise to a position of a second preset height above the ground; and activate the telescopic support rod; wherein the second preset height is greater than the first preset height.

3. The intelligent crop testing system according to claim 1, characterized in that, The rainwater harvesting equipment includes a sedimentation tank, a water quality monitoring tank, and a storage tank. The sedimentation tank and the water quality monitoring tank are connected via a first pipe and a second pipe. Filter screens are installed in front of one end of the sedimentation tank on both the first and second pipes. A pumping valve is installed in the first pipe near the sedimentation tank. A first submersible pump is installed in the second pipe extending into the water quality monitoring tank. A water quality sensor and water treatment equipment are installed in the water quality monitoring tank, both of which are communicatively connected to the control equipment. A first water intake is located on the cover of the water quality monitoring tank. The water quality monitoring tank is connected to the storage tank via a smart valve and a third pipe. A second submersible pump is located inside the storage tank and is connected to an outlet pipe, which is connected to the irrigation equipment. The water quality sensor is used to detect the current water quality data of rainwater in the water quality monitoring pool and transmit the current water quality data to the control device. The control device is used to compare the current water quality data with the irrigation standard water quality data corresponding to the crop, and control the water treatment device to treat the rainwater in the water quality monitoring pool based on the comparison result.

4. The intelligent crop testing system according to claim 1, characterized in that, The intelligent crop testing system also includes a canopy recognition device and a radar sensor, both of which are communicatively connected to the control device. in; The canopy recognition device is used to photograph and / or scan the canopy of the crop to generate canopy data, which includes at least one of chlorophyll content, leaf area index, canopy height, and canopy closure. The radar sensor is used to scan the root system of the crop and determine the root system data of the crop, including root distribution density, root length and root activity. The control device is used to output the target irrigation strategy corresponding to the crop based on the environmental data, the canopy data, and the root system data.

5. The intelligent crop testing system according to claim 4, characterized in that, The control device is used to acquire the growth stage corresponding to the crop, and determine the current growth stage of the crop and the weight information corresponding to the current growth stage based on the growth stage. Based on the weight information corresponding to the current growth stage, calculate the target water transfer efficiency coefficient for the crop. Based on the target water conduction efficiency coefficient, an initial physiological causal relationship between the canopy and the root system is established.

Citation Information

Patent Citations

  • Multi-functional integrated experimental shed

    CN107926374A