A three-dimensional greening intelligent irrigation system and method for solving vertical microclimate differences

Abstract

The application discloses a kind of three-dimensional greening intelligent irrigation systems and methods for solving vertical microclimate difference, belong to accurate irrigation technical field, including multi-source perception module, data fusion and intelligent control module and partition pressure compensation execution module;Multi-source perception module is used to obtain the multidimensional information of three-dimensional greening unit, data fusion and intelligent control module and multi-source perception module communication connection.This application is by designing "height-wind speed" coupling vertical microclimate dynamic correction model, for the first time in irrigation control realizes the quantization compensation of inherent physical environment difference of three-dimensional greening, realizes water balance supply on three-dimensional space, deeply integrates high-precision weather forecast, executes "intelligent stop irrigation / reduction irrigation before rainfall" strategy, fundamentally avoids "ineffective irrigation" and "over irrigation", and it is estimated that 30%-50% irrigation water can be saved, while, accurate irrigation reduces the leaching of fertilizer, realizes water and fertilizer double saving.

Description

Technical Field

[0001] This invention relates to the field of precision irrigation technology, and in particular to a three-dimensional greening intelligent irrigation system and method for solving vertical microclimate differences. Background Technology

[0002] With the acceleration of urbanization, vertical greening has been widely used as an important means to expand urban ecological space and improve the microenvironment. However, the carriers of vertical greening (such as building facades and viaduct columns) create significant microclimate gradients in the vertical direction: typically, with increasing height, wind speed increases, sunlight intensifies, and air humidity decreases, resulting in transpiration and evaporation (transpiration) in the upper layers of plants being much greater than in the lower layers. This "vertical microclimate difference" leads to a dilemma of "drought in the upper layers and waterlogging in the lower layers" in traditional systems using a uniform irrigation model, severely restricting the landscape effect and ecological benefits of vertical greening.

[0003] Currently, vertical greening irrigation mainly relies on the following technologies: First, extensive control based on timers, which completely ignores changes in the environment and plant needs; second, simple feedback control based on a small number of soil moisture sensors, which, although improved to some extent, still has significant shortcomings: First, the point-distributed sensors cannot fully reflect the uneven moisture status of the three-dimensional substrate and are prone to failure; second, control based on soil moisture content has a lag and cannot respond to the instantaneous transpiration needs of plants; finally, existing systems generally lack direct, non-contact monitoring methods for the physiological state of plants, making it impossible to intervene before visible stress (such as wilting) occurs, and even more impossible to distinguish between water shortage stress and other stresses (such as diseases).

[0004] Furthermore, existing technologies have failed to effectively utilize meteorological forecast information for forward-looking decision-making, leading to irrigation being carried out even before rainfall, resulting in water waste and potentially causing root diseases. For high-altitude, large-area vertical greening facilities, manual inspection is costly and risky, necessitating an urgent solution that enables remote, automated, and intelligent precision maintenance.

[0005] Therefore, how to construct an intelligent irrigation system that can sense and compensate for vertical microclimate differences, integrate environmental prediction and plant physiological feedback, and achieve true "on-demand supply" has become a technical problem that urgently needs to be solved in this field. Summary of the Invention

[0006] The purpose of this invention is to solve the problems existing in the prior art by proposing a three-dimensional greening intelligent irrigation system and method to solve vertical microclimate differences.

[0007] To achieve the above objectives, the present invention adopts the following technical solution: A three-dimensional greening intelligent irrigation system that addresses vertical microclimate differences includes a multi-source sensing module, a data fusion and intelligent control module, and a zoned pressure compensation execution module. The multi-source sensing module is used to acquire multi-dimensional information of the vertical greening unit, including: Soil temperature and humidity sensor array: arranged in layers along the vertical direction of the green facade to acquire substrate moisture and temperature data for each height zone; Plant image acquisition unit: used to acquire digital images of the plant canopy covering the green facade; Micro-meteorological data acquisition unit: used to acquire real-time environmental data, including wind speed, temperature, and humidity, as well as future weather forecast information; The data fusion and intelligent control module and the multi-source sensing module are communicatively connected for: Based on the spatial location information of each altitude zone and the wind speed data, a vertical microclimate correction coefficient is calculated to compensate for differences in vertical evapotranspiration. The digital image of the plant canopy is processed to extract plant morphological features and calculate the plant wilting coefficient, which reflects its physiological stress state. By integrating the substrate moisture data, the future weather forecast information, and the plant wilting coefficient, a zonal irrigation instruction is generated through a decision algorithm; The decision algorithm is configured to perform forward-looking irrigation avoidance based on rainfall forecasts in the future weather forecast information. The partition pressure compensation execution module is connected to the data fusion and intelligent control module, and is used to implement independent and flow-controllable irrigation for different height partitions of the vertical greening facade according to the partition irrigation command.

[0008] As a preferred option, the vertical microclimate correction coefficient Calculated using the following model:

[0009] in, The representative wind speed of the i-th zone is obtained through actual measurement, estimation based on wind profile model, or computational fluid dynamics simulation results. Let be the height of the i-th partition relative to the reference plane; and These are the reference wind speed and reference altitude, respectively. These are calibrable model parameters. The benchmark value characterizing the vertical correction factor. Characterizing the weighting of wind speed on the enhancement of evapotranspiration. The weighting of the overall influence is highly independent of wind speed.

[0010] As a preferred embodiment, the plant image acquisition unit includes a camera with a pan-tilt-zoom function; the data fusion and intelligent control module further includes a visual perception quality control submodule, which is configured as follows: The acquired images are evaluated online, and a comprehensive quality score is calculated, which includes at least one of the following: target area coverage, average confidence of leaf segmentation, number of effective leaves, and leaf integrity index. When the overall quality score is lower than a set threshold, a control signal is generated to automatically adjust the camera's gimbal rotation angle, pitch angle, and / or optical zoom to optimize the shooting angle and reacquire images.

[0011] As a preferred embodiment, the calculation process for the plant wilting coefficient includes: Leaf-level instance segmentation and feature extraction: A deep learning model is used to process the image and segment it into multiple independent leaf regions; for each leaf region, the downward angle of its main axis and the leaf color feature value in a specific color space are calculated. Zonal-level population statistics: Statistical analysis of the characteristic values ​​of all leaf regions belonging to the same zonal region is performed to obtain population statistical characteristics including at least one of the following: proportion of wilted leaves, average leaf drooping angle, average leaf color saturation, and canopy density; Comprehensive coefficient generation: The statistical characteristics of the population are compared with the historical growth baseline of plants in the healthy state of the pre-set partition, and a quantitative plant wilting coefficient is generated through a fuzzy logic system or a weighted decision model.

[0012] As a preferred embodiment, the multi-source data fusion decision algorithm executed by the data fusion and intelligent control module adopts a three-layer cascaded decision architecture: First layer: Water demand calculation layer, which calculates the basic irrigation demand required to bring the soil moisture content to the target value based on the current soil moisture content and the predicted evapotranspiration after being corrected by the vertical microclimate correction coefficient. The second layer is the meteorological forecast layer, which receives the future weather forecast information. If it is predicted that there will be effective rainfall within a preset time window in the future, the basic irrigation demand will be reduced or reduced to zero according to the predicted rainfall, and a weather-corrected irrigation demand will be generated. The third layer is the visual feedback layer, which receives the plant wilting coefficient. If the wilting coefficient indicates that the plant is under water stress, the weather-corrected irrigation demand is adjusted upwards. If the wilting coefficient indicates that the plant is healthy but the soil data indicates that it needs water, the irrigation demand is conservatively adjusted downwards or a sensor anomaly alarm is triggered.

[0013] As a preferred embodiment, the data fusion and intelligent control module further includes a model parameter adaptive learning unit, which is configured as follows: Record the weather-corrected irrigation demand, actual irrigation volume, and changes in soil moisture content and plant wilting coefficient within a preset time period after each irrigation. Based on the deviation between the change response and the expected target, the parameters in the vertical microclimate correction coefficient model are adjusted in reverse using an optimization algorithm. , And decision thresholds and weights in a three-tiered cascaded decision architecture.

[0014] As a preferred embodiment, the partition pressure compensation execution module includes: Each branch has an independent irrigation branch corresponding to a different height zone, and each branch is equipped with a solenoid valve controlled by the data fusion and intelligent control module. Pressure-compensated drip irrigation pipes or micro-sprinklers are installed within each zone to maintain uniform water output when water supply pressure fluctuates.

[0015] A proposed method for a three-dimensional greening intelligent irrigation system that addresses vertical microclimate differences includes the following steps: S1. System initialization: Divide the vertical height into zones according to the vertical greening structure, configure irrigation hardware for each zone and set the basic water requirement parameters for the plants; S2. Multi-source data synchronous acquisition and sensing self-verification: synchronously acquire soil moisture data, plant canopy images, and meteorological data and forecasts for each zone; perform quality assessment on the plant canopy images, and if they do not meet the analysis requirements, automatically adjust the posture and parameters of the image acquisition equipment and re-acquire the data; S3. Vertical microclimate correction and plant status analysis: Calculate the vertical microclimate correction coefficient based on zone height and wind speed; analyze qualified plant images and calculate the plant wilting coefficient through multi-leaf feature statistics; S4. Multi-source fusion intelligent decision-making: Based on the current soil moisture, the vertical microclimate correction coefficient is used to predict evapotranspiration, and the future rainfall forecast is combined to make forward-looking water volume adjustments. Finally, the plant wilting coefficient is introduced for feedback correction to generate the final irrigation decisions for each zone. S5. Precise Execution by Zone: Control the execution mechanism of the corresponding zone and provide differentiated irrigation based on the decision results; S6. Effect evaluation and model optimization: Monitor the response of plants and soil after irrigation, and use the response data to adaptively update the parameters of the vertical microclimate correction model and decision rules.

[0016] As a preferred option, the multi-source fusion intelligent decision-making in step S4 specifically includes: The predicted values ​​of soil moisture content in each zone during a future period are calculated without irrigation. These predicted values ​​combine evapotranspiration loss enhanced by a vertical microclimate correction factor with natural rainfall replenishment from weather forecasts. The predicted values ​​are compared with the lower limit of the target moisture content for each zone to obtain the preliminary irrigation requirements. If the predicted replenishment effect of future rainfall can restore the soil moisture content to a safe range before the irrigation action takes effect, then the irrigation plan for the corresponding zone will be cancelled or reduced. The preliminary decision results were cross-validated with the plant wilting coefficient: when the wilting coefficient was high, irrigation was initiated or increased even if the soil data did not reach the water shortage threshold; when the wilting coefficient was low but the model calculated that irrigation was required, a conservative irrigation strategy was adopted and the data consistency was checked.

[0017] As a preferred option, the indicators for quality assessment of plant canopy images described in step S2; It includes at least: the pixel coverage ratio of the preset monitoring area in the image, the average confidence level obtained by the leaf segmentation model, and the proportion of leaves truncated by the image boundary; The adaptive update described in step S6 is achieved by minimizing the combined error between the predicted and measured values ​​of soil moisture content after irrigation, and between the expected improvement value and the actual change value of plant wilting coefficient, through gradient descent or Bayesian update methods.

[0018] Compared with the prior art, the beneficial effects of the present invention are as follows: 1. This invention designs a vertical microclimate dynamic correction model coupled with "height-wind speed" to quantitatively compensate for the inherent physical environmental differences in vertical greening during irrigation control. This solves the long-standing industry problems of "dry upper layers and wet lower layers" and "the upper layers dying from drought and the lower layers from flooding," achieving a balanced water supply across three-dimensional space. It deeply integrates high-precision weather forecasts and implements a "smart irrigation stop / reduction" strategy before rainfall, avoiding "ineffective irrigation" and "over-irrigation," and is expected to save 30%-50% of irrigation water. Simultaneously, precision irrigation reduces fertilizer leaching, achieving both water and fertilizer savings.

[0019] 2. This invention breaks through the limitations of traditional systems that rely on a single data source, establishing a novel paradigm of triangular verification and fusion decision-making based on "soil moisture - weather forecast - plant vision". This paradigm enables irrigation decisions to possess real-time performance, foresight, and physiological accuracy, upgrading irrigation control from a simple "response" to an intelligent closed loop of "prediction-response-verification". The multi-source data fusion mechanism forms a natural cross-validation and redundancy backup. When soil sensors fail or weather forecasts deviate locally, plant vision feedback can serve as the final corrective basis; and vice versa. This significantly reduces the risk of overall system misjudgment and mis-irrigation due to single-point failures, ensuring the stability of long-term unattended operation.

[0020] 3. This invention provides a set of replicable and scalable intelligent maintenance standard solutions for vertical greening, which helps to promote the transformation of the entire landscaping industry from extensive, experience-based maintenance to a refined, intelligent, and data-driven modern maintenance model. Through precise irrigation, it can maximize the ecological functions of vertical greening, such as carbon sequestration, cooling, dust reduction, and noise reduction, and maintain a beautiful landscape effect in the long term, thereby improving the ecological quality of the city. Attached Figure Description

[0021] Figure 1 This is a schematic diagram of the overall architecture of a three-dimensional greening intelligent irrigation system that addresses vertical microclimate differences, as proposed in this invention. Figure 2 This is a flowchart of the three-level cascaded decision-making algorithm in the intelligent irrigation system and method for vertical greening proposed in this invention to solve vertical microclimate differences; Figure 3 This is a flowchart illustrating the overall steps of an intelligent irrigation method for vertical greening that addresses differences in vertical microclimate, as proposed in this invention. Detailed Implementation

[0022] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0023] It should be understood that, when used in this specification and the appended claims, the terms "comprising" and "including" indicate the presence of the described features, integrals, steps, operations, elements and / or components, but do not exclude the presence or addition of one or more other features, integrals, steps, operations, elements, components and / or collections thereof.

[0024] It should also be understood that the terminology used in this specification is for the purpose of describing particular embodiments only and is not intended to limit the invention. As used in this specification and the appended claims, the singular forms “a,” “an,” and “the” are intended to include the plural forms unless the context clearly indicates otherwise.

[0025] It should also be further understood that the term "and / or" as used in this specification and the appended claims refers to any combination of one or more of the associated listed items and all possible combinations, and includes such combinations.

[0026] Example, refer to Figures 1 to 3A three-dimensional greening intelligent irrigation system that addresses vertical microclimate differences includes a multi-source sensing module, a data fusion and intelligent control module, and a zoned pressure compensation execution module. The multi-source sensing module is used to acquire multi-dimensional information of the vertical greening unit, specifically including: Soil temperature and humidity sensor array: arranged in layers along the vertical direction of the green facade, used to collect real-time data on the volumetric moisture content and temperature of the substrate in each height zone. Multiple measuring points are set in each zone to obtain the average value, thereby improving the representativeness of the data. Based on the spatial location information of each altitude zone and the wind speed data, a vertical microclimate correction coefficient is calculated to compensate for differences in vertical evapotranspiration. Plant image acquisition unit: used to acquire digital images of the plant canopy covering the green facade; It mainly includes network cameras with pan-tilt and optical zoom functions, which are installed on the opposite building or on a dedicated bracket to cover the target green area.

[0027] The data fusion and intelligent control module also includes a visual perception quality regulation submodule, which is configured as follows: The acquired images are evaluated online, and a comprehensive quality score is calculated, which includes at least one of the following: target area coverage, average confidence of leaf segmentation, number of effective leaves, and leaf integrity index. When the overall quality score is lower than a set threshold, a control signal is generated to automatically adjust the camera's gimbal rotation angle, pitch angle, and / or optical zoom to optimize the shooting angle and reacquire images.

[0028] Micro-weather data acquisition unit: Used to acquire real-time environmental data and future weather forecast information. It may include a micro-weather station deployed on site (measuring temperature, humidity, wind speed, wind direction, solar radiation, etc.) and access authoritative meteorological services through a network interface (such as API) to obtain high-precision rainfall, temperature, and wind speed forecasts for the next 6-72 hours; The data fusion and intelligent control module, which communicates with the multi-source sensing module, is the "brain" of the system. Its core function is to run a series of algorithm models, including: Vertical microclimate correction model: Based on the spatial height information of each zone and real-time or forecasted wind speed, a vertical microclimate correction coefficient is calculated. This is used to quantify and compensate for the additional evapotranspiration demand of the partition due to its location. A specific and preferred mathematical model is as follows:

[0029] in, The representative wind speed for the i-th zone is obtained through actual measurement, estimation based on a wind profile model, or computational fluid dynamics simulation (measured by a small anemometer installed at that height, or based on ground wind speed). (Estimated using boundary layer wind profile formula) The height of the i-th partition relative to the reference plane (measured from the ground or reference surface); and These are the reference wind speed and the reference height (usually 2 meters). These are calibrable model parameters. Characterizing the weighting of wind speed on the enhancement of evapotranspiration. The weighting of the overall influence is highly independent of wind speed. The baseline value characterizing the vertical correction coefficient is the basic calibration factor (usually ≈1). and It can be determined through regression fitting of initial experimental data and can be optimized through self-learning during operation.

[0030] The digital image of the plant canopy is processed to extract plant morphological features and calculate the plant wilting coefficient, which reflects its physiological stress state. By integrating the substrate moisture data, the future weather forecast information, and the plant wilting coefficient, a zonal irrigation instruction is generated through a decision algorithm; The decision algorithm is configured to perform forward-looking irrigation avoidance based on rainfall forecasts in the future weather forecast information. The calculation process for the plant wilting coefficient includes: Leaf-level instance segmentation and feature extraction: A deep learning model is used to process the image and segment it into multiple independent leaf regions; for each leaf region, the downward angle of its main axis and the leaf color feature value in a specific color space are calculated. Zonal-level population statistics: Statistical analysis of the characteristic values ​​of all leaf regions belonging to the same zonal region is performed to obtain population statistical characteristics including at least one of the following: proportion of wilted leaves, average leaf drooping angle, average leaf color saturation, and canopy density; Comprehensive coefficient generation: The statistical characteristics of the population are compared with the historical growth baseline of plants in the healthy state of the pre-set partition, and a quantitative plant wilting coefficient is generated through a fuzzy logic system or a weighted decision model.

[0031] It is worth noting that the plant visual analysis and wilting coefficient calculation model includes: First, a visual perception quality control submodule; this submodule performs online evaluation of the images transmitted from the camera and calculates the overall quality score (QS). The overall quality score (QS) can be calculated based on indicators such as the coverage ratio of the target monitoring area in the image, image sharpness, and the average confidence level obtained through a pre-trained leaf segmentation model.

[0032] If the QS (Quality Segmentation) is below the threshold, the submodule will automatically generate control commands to adjust the camera's gimbal angle and zoom level until a high-quality image meeting the analysis requirements is obtained. Next, for images of acceptable quality, a deep learning-based instance segmentation algorithm (such as Mask R-CNN) is used to identify and segment a large number of individual leaves from the image. For each leaf, its morphological features (such as the downward angle of the leaf's main axis relative to the direction of gravity) and color features (such as saturation in HSV space or the calculated greenness index) are calculated. ).

[0033] Then, a population statistical analysis was performed on the characteristics of all leaves in the same zone to calculate indicators such as the proportion of wilted leaves (the proportion of leaves with a drooping angle exceeding the threshold), average leaf color saturation, and canopy density (the ratio of the total leaf area to the planting area).

[0034] Finally, these statistical indicators are compared with the baseline established by the plants in the historical health state of the region, and fused with a rule set or machine learning model (such as support vector machine or random forest) to output a plant wilting coefficient between 0 and 1. The lower the value, the more severe the stress.

[0035] The partition pressure compensation execution module is connected to the data fusion and intelligent control module, and is used to implement independent and flow-controllable irrigation for different height partitions of the vertical greening facade according to the partition irrigation command.

[0036] As a preferred embodiment, the multi-source data fusion decision algorithm executed by the data fusion and intelligent control module adopts a three-layer cascaded decision architecture: First layer: Water demand calculation layer, which calculates the basic irrigation demand required to bring the soil moisture content to the target value based on the current soil moisture content and the predicted evapotranspiration after being corrected by the vertical microclimate correction coefficient. Specifically, based on the current soil moisture content, and after the above... The revised predicted evapotranspiration (the reference evapotranspiration ET0 can be calculated using the Penman-Monteith or Hargreaves formula, and then multiplied by the plant coefficient) and ), calculate the basic irrigation demand required to maintain soil moisture content within the target range over a future period (e.g., 24 hours).

[0037] The second layer is the meteorological forecast layer, which receives the future weather forecast information. If it is predicted that there will be effective rainfall within a preset time window in the future, the basic irrigation demand will be reduced or reduced to zero according to the predicted rainfall, and a weather-corrected irrigation demand will be generated. Specifically: Read future rainfall forecasts; if there is "effective rainfall" (e.g., predicted rainfall > 5 mm) within the time window when the forecast takes effect in the next irrigation cycle, assess the effect of the rainfall on soil moisture replenishment; if the rainfall is sufficient to restore the soil moisture content to above the target lower limit, set the basic irrigation demand for that area to zero ("stop irrigation"); if the rainfall can only partially replenish the soil moisture, reduce the demand proportionally ("reduce irrigation") and generate weather-corrected irrigation demand.

[0038] The third layer is the visual feedback layer, which receives the plant wilting coefficient. If the wilting coefficient indicates that the plant is under water stress, the weather-corrected irrigation demand is adjusted upwards. If the wilting coefficient indicates that the plant is healthy but the soil data indicates that it needs water, the irrigation demand is conservatively adjusted downwards or a sensor anomaly alarm is triggered.

[0039] Specifically, the system incorporates the calculated plant wilting coefficient and its trend. If the coefficient for a particular zone remains consistently low or decreases rapidly (indicating that the plant is under water stress), the system will trigger an irrigation alarm or perform protective irrigation, even if the weather-corrected irrigation demand is zero or very small. Conversely, if the plant wilting coefficient is high (indicating healthy plants), and the model calculates a large demand, the system will adopt a conservative strategy (e.g., irrigating only 50% of the demand) and simultaneously trigger a data consistency check alarm, indicating potential sensor anomalies.

[0040] The data fusion and intelligent control module also includes a model parameter adaptive learning unit, which is configured as follows: Record the weather-corrected irrigation demand, actual irrigation volume, and changes in soil moisture content and plant wilting coefficient within a preset time period after each irrigation. Based on the deviation between the change response and the expected target, the parameters in the vertical microclimate correction coefficient model are adjusted in reverse using an optimization algorithm. , And decision thresholds and weights in a three-tiered cascaded decision architecture.

[0041] The partitioned pressure compensation execution module executes irrigation operations according to the instructions output by the data fusion and intelligent control module, including: Irrigation zone network: The green facade is divided into several independently controlled horizontal irrigation zones along its height, and each zone has an independent water supply branch pipe.

[0042] Control and Execution Unit: Solenoid valves are installed on the branch pipes of each zone, and the opening and closing of the solenoid valves are directly controlled by the control module. For precise water volume control, flow meters can be installed for closed-loop feedback. Within each zone, pressure-compensated drip irrigation pipes or pressure-compensated micro-sprinklers are evenly distributed. This design ensures that even if water pressure varies at different locations in the network due to height differences, the flow rate at each outlet remains essentially consistent. This forms the hardware foundation for achieving uniform vertical irrigation and maintains water uniformity when supply pressure fluctuates.

[0043] A proposed method for a three-dimensional greening intelligent irrigation system that addresses vertical microclimate differences includes the following steps: S1. System Initialization and Zoning Modeling: Based on the physical structure and plant configuration of the vertical greening carrier, it is divided into N vertical height zones. Corresponding sensors and actuators are configured for each zone, and the target soil moisture content range, plant type and basic water requirement coefficient are set in the control module.

[0044] S2. Multi-source data synchronous acquisition and sensing self-verification: At preset time points, soil moisture data, plant canopy images, and meteorological data and forecast information for each zone are synchronously triggered and acquired. After image acquisition, quality assessment is performed immediately; if the image is unqualified (e.g., key areas are obscured or blurred), the camera position is automatically adjusted and the image is retaken to ensure the reliability of the data input to the analysis module.

[0045] S3. Vertical Microclimate Correction and Vegetation Status Analysis: Based on current wind speed data and the altitude of each zone, the vertical microclimate correction model is used to calculate the vertical microclimate correction for each zone. Meanwhile, deep learning analysis is performed on qualified plant images, and the wilting coefficient of each region is calculated through multi-leaf feature statistics.

[0046] S4. Multi-source fusion intelligent decision-making: Execute the three-layer cascaded decision-making algorithm. First, combine the current soil moisture and... The revised evapotranspiration forecast is used to calculate future water deficit. Secondly, the impact of future rainfall forecasts is overlaid to proactively adjust the initial irrigation plan. Finally, a plant wilting coefficient is introduced as the "final arbiter" to verify and correct the aforementioned results, generating precise instructions for each zone regarding "whether to irrigate" and "how much to irrigate."

[0047] S5. Precise Execution by Zone: The control module converts irrigation commands into opening duration signals for the corresponding zone's solenoid valves, driving the execution module to work. Pressure-compensated irrigation devices ensure uniform irrigation within the same zone.

[0048] S6. Effect Evaluation and Model Optimization: After irrigation, the system does not enter dormancy but instead initiates effect monitoring. Over the next few hours, soil moisture and plant images are continuously collected to observe recovery. The recovery data is compared with the initial decision-making expectations, and the resulting error signal is used to drive the adaptive learning unit to optimize model parameters online, forming a continuous improvement loop of "decision-making plus execution plus learning."

[0049] The specific implementation is as follows: Taking a 20-meter-high modular vertical green wall in an office building as an example. Step 1: System Initialization and Partition Modeling The green wall is 20 meters high and 15 meters wide, divided into four equal irrigation zones along its height: Zone 1 (0-5 meters), Zone 2 (5-10 meters), Zone 3 (10-15 meters), and Zone 4 (15-20 meters). The main plants are *Trachelospermum jasminoides* and *Euonymus fortunei*. Three soil temperature and humidity sensors are evenly embedded within the planting modules of each zone. A high-performance pan-tilt-zoom camera is installed on the roof of the building opposite the green wall, and a miniature weather station is set up in an open area on the rooftop. An industrial edge computing gateway is installed in the control cabinet as the core of the data fusion and intelligent control module. The main irrigation pipe is connected to the building's water supply system and branches into four branch pipes, each equipped with a solenoid valve and pressure-compensating drip irrigation tape, connecting to the corresponding zone.

[0050] In the control system, the following settings are configured for each zone: target soil volumetric moisture content range of 18%-25%; vegetation coefficient. Set the value to 0.8; initialize the vertical microclimate correction model parameters. The camera's initial view is preset to cover the entire wall.

[0051] Step 2, Data Collection and Self-Verification (triggered daily at 6 AM): The system initiates the data acquisition process. Soil sensor data: average moisture content of zone 1 is 22%, zone 2 is 20%, zone 3 is 18%, and zone 4 is 16%. Weather station data shows current wind speed is 3 m / s (2 meters high) and temperature is 25℃.

[0052] The meteorological API returned a forecast of no rainfall for the next 24 hours. After the camera captured a panoramic image, the quality assessment submodule detected that the average leaf segmentation confidence level in the canopy image of partition 4 (top) was only 0.65 (below the threshold of 0.8).

[0053] The submodule then controls the gimbal to aim the lens at partition 4 and zoom appropriately to retake the image. The confidence level of the second image reached 0.88, indicating acceptable quality.

[0054] Step 3, Vertical Correction and Visual Analysis: Correction factor calculation: Assuming the wind speed U4 in zone 4 (H=17.5 meters) is estimated to be approximately 5.2 m / s based on the wind profile formula, calculate its correction factor: .

[0055] Similarly, the calculation yields partition 1. It is approximately 1.10; it can be seen that the top evaporation demand has been revised to nearly twice that of the bottom.

[0056] Visual analysis: Analysis of close-up images of partition 4 revealed that the instance segmentation model identified 152 valid leaves. Calculations showed that 45 of these leaves had a drooping angle greater than 40 degrees, indicating a high proportion of wilted leaves. The average greenness index decreased by 15% compared to last week. The overall plant wilting coefficient was calculated to be 0.45 (under moderate stress), while the wilting coefficient of the bottom area was 0.85 (healthy).

[0057] Step 4, Multi-source fusion decision-making: Water demand calculation layer: based on current water content and economic conditions. The revised evapotranspiration forecast shows that: Zone 4 needs to replenish water by 5mm (approximately 7.5L / m²) in the next 24 hours to reach the target lower limit; Zone 1 only needs to replenish water by 1mm.

[0058] Weather forecast: No rain is expected in the future, therefore no forecast adjustments will be made.

[0059] Visual feedback layer: The wilting coefficient (0.45) and high water demand (5mm) of zone 4 corroborate each other, confirming the necessity of irrigation. Due to its severe stress, the decision is to give 110% irrigation amount (i.e. 5.5mm) to accelerate recovery. Zone 1 has a high wilting coefficient (0.85) but a low calculated water demand (1mm). The system confirms that its condition is good and executes the original amount of 1mm.

[0060] Step 5, execute by partition: The control module issues the following instructions: the solenoid valve of zone 4 is opened, and the irrigation time is converted to 11 minutes based on the dripper flow rate; the solenoid valve of zone 1 is opened for 2 minutes; zones 2 and 3 are executed sequentially according to their respective calculations; and the pressure-compensated drippers ensure that water drips evenly throughout the zone.

[0061] Step 6, Learning Optimization Three hours after irrigation, the system monitored the soil again. The soil moisture content in zone 4 had risen to 19%, and the proportion of wilted leaves had decreased to 20%, but the recovery rate was slightly lower than the model expected. Based on this, the adaptive learning unit made slight adjustments to the soil moisture content in zone 4. In the model The value (fine-tuned from 0.08 to 0.082) makes the estimation of top-level evapotranspiration more accurate in the next decision.

[0062] As can be seen from the above embodiments, the system of the present invention can automatically identify the special water demand of the top area and obtain key evidence through image self-adjustment. It can accurately implement differentiated irrigation of "more water at the top and less water at the bottom" on rainless days. In long-term operation, the system parameters are continuously optimized by itself, and the irrigation strategy becomes more and more precise.

[0063] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. A three-dimensional greening intelligent irrigation system that addresses vertical microclimate differences, characterized in that, It includes a multi-source sensing module, a data fusion and intelligent control module, and a partitioned pressure compensation execution module; The multi-source sensing module is used to acquire multi-dimensional information of the vertical greening unit, including: Soil temperature and humidity sensor array: arranged in layers along the vertical direction of the green facade to acquire substrate moisture and temperature data for each height zone; Plant image acquisition unit: used to acquire digital images of the plant canopy covering the green facade; Micro-meteorological data acquisition unit: used to acquire real-time environmental data, including wind speed, temperature, and humidity, as well as future weather forecast information; The data fusion and intelligent control module and the multi-source sensing module are communicatively connected for: Based on the spatial location information of each altitude zone and the wind speed data, a vertical microclimate correction coefficient is calculated to compensate for differences in vertical evapotranspiration. The digital image of the plant canopy is processed to extract plant morphological features and calculate the plant wilting coefficient, which reflects its physiological stress state. By integrating the substrate moisture data, the future weather forecast information, and the plant wilting coefficient, a zonal irrigation instruction is generated through a decision algorithm; The decision algorithm is configured to perform proactive irrigation avoidance based on rainfall forecasts in the future weather forecast information. The partition pressure compensation execution module is connected to the data fusion and intelligent control module, and is used to implement independent and flow-controllable irrigation for different height partitions of the vertical greening facade according to the partition irrigation command.

2. The intelligent irrigation system for vertical greening to address vertical microclimate differences according to claim 1, characterized in that, The vertical microclimate correction coefficient Calculated using the following model: in, The representative wind speed of the i-th zone is obtained through actual measurement, estimation based on wind profile model, or computational fluid dynamics simulation results. Let be the height of the i-th partition relative to the reference plane; and These are the reference wind speed and reference altitude, respectively. For calibrable model parameters The benchmark value characterizing the vertical correction factor. Characterizing the weighting of wind speed on the enhancement of evapotranspiration. The weighting of the overall influence is highly independent of wind speed.

3. The intelligent irrigation system for vertical greening to address vertical microclimate differences according to claim 1, characterized in that, The plant image acquisition unit includes a camera with a pan-tilt-zoom function; the data fusion and intelligent control module further includes a visual perception quality control submodule, which is configured as follows: The acquired images are evaluated online, and a comprehensive quality score is calculated, which includes at least one of the following: target area coverage, average confidence of leaf segmentation, number of effective leaves, and leaf integrity index. When the overall quality score is lower than a set threshold, a control signal is generated to automatically adjust the camera's gimbal rotation angle, pitch angle, and optical zoom to optimize the shooting angle and reacquire images.

4. The intelligent irrigation system for vertical greening to address vertical microclimate differences according to claim 1, characterized in that, The calculation process for the plant wilting coefficient includes: Leaf-level instance segmentation and feature extraction: A deep learning model is used to process the image and segment it into multiple independent leaf regions; for each leaf region, the downward angle of its main axis and the leaf color feature value in a specific color space are calculated. Zonal-level population statistics: Statistical analysis of the characteristic values ​​of all leaf regions belonging to the same zonal region is performed to obtain population statistical characteristics including at least one of the following: proportion of wilted leaves, average leaf drooping angle, average leaf color saturation, and canopy density; Comprehensive coefficient generation: The statistical characteristics of the population are compared with the historical growth baseline of plants in the healthy state of the pre-set partition, and a quantitative plant wilting coefficient is generated through a fuzzy logic system or a weighted decision model.

5. The intelligent irrigation system for vertical greening to address vertical microclimate differences according to claim 1, characterized in that, The multi-source data fusion decision-making algorithm executed by the data fusion and intelligent control module adopts a three-level cascaded decision-making architecture: First layer: Water demand calculation layer, which calculates the basic irrigation demand required to bring the soil moisture content to the target value based on the current soil moisture content and the predicted evapotranspiration after being corrected by the vertical microclimate correction coefficient. The second layer is the meteorological forecast layer, which receives the future weather forecast information. If it is predicted that there will be effective rainfall within a preset time window in the future, the basic irrigation demand will be reduced or reduced to zero according to the predicted rainfall, and a weather-corrected irrigation demand will be generated. The third layer is the visual feedback layer, which receives the plant wilting coefficient. If the wilting coefficient indicates that the plant is under water stress, the weather-corrected irrigation demand is adjusted upwards. If the wilting coefficient indicates that the plant is healthy but the soil data indicates that it needs water, the irrigation demand is conservatively adjusted downwards or a sensor anomaly alarm is triggered.

6. The intelligent irrigation system for vertical greening to address vertical microclimate differences according to claim 1, characterized in that, The data fusion and intelligent control module also includes a model parameter adaptive learning unit, which is configured as follows: Record the weather-corrected irrigation demand, actual irrigation volume, and changes in soil moisture content and plant wilting coefficient within a preset time period after each irrigation. Based on the deviation between the change response and the expected target, the parameters in the vertical microclimate correction coefficient model are adjusted in reverse using an optimization algorithm. , And decision thresholds and weights in a three-tiered cascaded decision architecture.

7. The intelligent irrigation system for vertical greening to address vertical microclimate differences according to claim 1, characterized in that, The partition pressure compensation execution module includes: Each branch has an independent irrigation branch corresponding to a different height zone, and each branch is equipped with a solenoid valve controlled by the data fusion and intelligent control module. Pressure-compensated drip irrigation pipes or micro-sprinklers are installed within each zone to maintain uniform water output when water supply pressure fluctuates.

8. The method for a three-dimensional greening intelligent irrigation system for solving vertical microclimate differences according to any one of claims 1-7, characterized in that, Includes the following steps: S1. System initialization: Divide the vertical height into zones according to the vertical greening structure, configure irrigation hardware for each zone and set the basic water requirement parameters for the plants; S2. Multi-source data synchronous acquisition and sensing self-verification: synchronously acquire soil moisture data, plant canopy images, and meteorological data and forecasts for each zone; perform quality assessment on the plant canopy images, and if they do not meet the analysis requirements, automatically adjust the posture and parameters of the image acquisition equipment and re-acquire the data; S3. Vertical microclimate correction and plant status analysis: Calculate the vertical microclimate correction coefficient based on zone height and wind speed; analyze qualified plant images and calculate the plant wilting coefficient through multi-leaf feature statistics; S4. Multi-source fusion intelligent decision-making: Based on the current soil moisture, the vertical microclimate correction coefficient is used to predict evapotranspiration, and the future rainfall forecast is combined to make forward-looking water volume adjustments. Finally, the plant wilting coefficient is introduced for feedback correction to generate the final irrigation decisions for each zone. S5. Precise Execution by Zone: Control the execution mechanism of the corresponding zone and provide differentiated irrigation based on the decision results; S6. Effect evaluation and model optimization: Monitor the response of plants and soil after irrigation, and use the response data to adaptively update the parameters of the vertical microclimate correction model and decision rules.

9. The method according to claim 8, characterized in that, The multi-source fusion intelligent decision-making process described in step S4 specifically includes: The predicted values ​​of soil moisture content in each zone during a future period are calculated without irrigation. These predicted values ​​combine evapotranspiration loss enhanced by a vertical microclimate correction factor with natural rainfall replenishment from weather forecasts. The predicted values ​​are compared with the lower limit of the target moisture content for each zone to obtain the preliminary irrigation requirements. If the predicted replenishment effect of future rainfall can restore the soil moisture content to a safe range before the irrigation action takes effect, then the irrigation plan for the corresponding zone will be cancelled or reduced. The preliminary decision results were cross-validated with the plant wilting coefficient: when the wilting coefficient was high, irrigation was initiated or increased even if the soil data did not reach the water shortage threshold; when the wilting coefficient was low but the model calculated that irrigation was required, a conservative irrigation strategy was adopted and the data consistency was checked.

10. The method according to claim 9, characterized in that, The indicators for quality assessment of plant canopy images described in step S2; It includes at least: the pixel coverage ratio of the preset monitoring area in the image, the average confidence level obtained by the leaf segmentation model, and the proportion of leaves truncated by the image boundary; The adaptive update described in step S6 is achieved by minimizing the combined error between the predicted and measured values ​​of soil moisture content after irrigation, and between the expected improvement value and the actual change value of plant wilting coefficient, through gradient descent or Bayesian update methods.

Images