A method and system for intelligent multi-parameter control of sprout growth environment

Abstract

This invention discloses a multi-parameter intelligent control method and system for the growth environment of sprouts, relating to the field of intelligent agricultural cultivation technology. The method is applied to the cultivation module and includes: collecting environmental data and sprout images through sensors and a vision module; determining the growth stage and calculating a health score based on the image data; inputting the environmental data, growth stage, and health score into a prediction model to obtain growth prediction results; dynamically generating a control strategy based on the growth stage, health score, and prediction results, prioritizing the activation of a correction scheme when the health score is below a threshold; and converting the control strategy into control commands to drive the execution layer devices. This invention, by introducing a visual assessment and health measurement mechanism, solves the problem in existing technologies where environmental control is disconnected from the actual physiological state of the crop and cannot actively intervene in growth abnormalities, achieving an intelligent upgrade from "environmental control" to "crop state-driven control."

Description

Technical Field

[0001] This invention relates to the field of intelligent agricultural cultivation technology, specifically to a method and system for intelligent control of multiple parameters of the growth environment of sprouts. Background Technology

[0002] Containerized plant factories, as a modular and rapidly deployable planting solution, have demonstrated significant advantages in urban agriculture and specialty crop cultivation. Their core lies in creating optimal growth conditions for crops through the artificial control of environmental parameters such as temperature, humidity, light, and nutrients. Existing technologies generally employ control strategies based on fixed setpoints or simple environmental feedback, such as activating cooling equipment when a temperature sensor detects a value exceeding a preset threshold. However, this environmental parameter-centric control model has a fundamental flaw: it relies entirely on monitoring and responding to the physical environment, neglecting the real-time physiological state of the crop itself. Crop growth is a complex biological process, and its health status, growth stage, and future trends cannot be directly and accurately reflected by physical parameters such as temperature and humidity. Therefore, existing systems often encounter situations where environmental parameters are "up to standard," but crop growth is slow, excessive, or physiologically diseased. The system cannot identify these growth abnormalities, let alone intervene in the early stages based on crop performance. It can only passively wait for significant fluctuations in environmental parameters or visible pathological changes before reacting, often too late, leading to yield losses and resource waste. This is essentially a disconnect between "environmental control" and "crop cultivation." The core problem this application aims to solve is how to break through the existing technology's reliance on a single environmental parameter and construct a method and system that can directly sense the physiological state of crops, quantify their health status, and use this as the core driving factor for forward-looking and differentiated intelligent regulation, thereby truly realizing a paradigm shift from "controlling the environment" to "caring for crops". Summary of the Invention

[0003] To address the shortcomings of existing technologies, the present invention aims to provide a multi-parameter intelligent control method and system for the growth environment of sprouts. By introducing visual assessment and health measurement mechanisms, it solves the problems of disconnect between environmental control and the actual physiological state of crops, and the inability to actively intervene in growth abnormalities in existing technologies, thus realizing an intelligent upgrade from "environmental control" to "crop state-driven control".

[0004] To achieve the above objectives, the embodiments of this invention provide the following technical solutions:

[0005] This application provides a multi-parameter intelligent control method for the growth environment of sprouts, applied to a closed cultivation module. The method includes: real-time acquisition of environmental data including temperature, humidity, CO2 concentration, light intensity, and growth promoter inventory; simultaneously, a movable vision module navigating along a preset track to acquire image data of sprouts in each cultivation unit; determining the current growth stage of the sprouts using an image recognition model based on the image data and calculating a health score; inputting the environmental data, growth stage, and health score into a pre-trained growth prediction model, which outputs a prediction result of the sprout growth within a preset time period; dynamically generating a comprehensive control strategy based on the growth stage, health score, and prediction result, the comprehensive control strategy including a set of instructions related to environmental parameter settings, growth promoter nutrient formula and application mode, and dynamic supplemental lighting scheme; wherein, when the health score is lower than a preset health threshold for the current growth stage, the comprehensive control strategy prioritizes initiating a correction scheme; converting the comprehensive control strategy into control commands and sending them to the execution layer to drive the temperature control model, humidity control model, growth promoter precision drip irrigation system, and supplemental lighting module to work collaboratively.

[0006] Furthermore, the step of determining the current growth stage of the sprouts based on the image data using an image recognition model and calculating a health score specifically includes: extracting multiple features from the image data, including plant height, leaf spread, leaf color histogram distribution, and stem uprightness; inputting the extracted features into a convolutional neural network classifier to output the probability that the sprouts are in the germination stage, rapid growth stage, or maturity stage, and taking the stage with the highest probability value as the current growth stage of the sprouts; simultaneously, based on the features and a pre-trained health regression model, calculating a quantified health score, wherein the color of the sprouts deviating from the normal green threshold range or the stem uprightness being below the threshold range will cause the health score to be lower than the health threshold.

[0007] Furthermore, the control strategy prioritizes the activation of a correction scheme, wherein the correction scheme includes at least one of the following: adjusting the spectral composition of the supplemental lighting module to increase the proportion of blue light to suppress excessive growth; reducing the ambient humidity setting to reduce the risk of disease; and adjusting the growth agent formula to increase the supply ratio of trace elements such as calcium and potassium to enhance the plant's resistance to stress.

[0008] Furthermore, the cultivation module is internally configured with a multi-layer cultivation rack, with each layer forming an independent cultivation unit. Based on the different growth stages and health scores of the sprouts in different cultivation units, independent comprehensive control strategies are generated and executed respectively.

[0009] Furthermore, after converting the comprehensive regulation strategy into control commands and issuing them to the execution layer to drive the temperature control model, humidity control model, growth agent precision drip irrigation system, and supplemental lighting module to work collaboratively, the method further includes: after a predetermined regulation cycle, acquiring image data again through the vision module and calculating a new health score; comparing the new health score with the health score before initiating the correction scheme to evaluate the effectiveness of the correction scheme; and based on the evaluation results, adaptively adjusting the growth prediction model or the health threshold to optimize the accuracy of subsequent decisions.

[0010] Accordingly, this application also provides a multi-parameter intelligent control system for the growth environment of sprouts, applied to a closed cultivation module. The system includes: a perception layer, including a sensor array and a vision module deployed within the cultivation module, for collecting environmental data and sprout image data; a decision layer, communicatively connected to the perception layer, the decision layer including: a growth status assessment module, for determining the growth stage of the sprouts based on image data and calculating a quantified health score; a prediction and strategy generation module, for making growth predictions based on environmental data, growth stage, and health score, and generating a comprehensive control strategy, the prediction and strategy generation module being further configured to: when the health score is lower than a preset health threshold for the current growth stage, prioritize the execution of a correction scheme in the generated comprehensive control strategy; and an execution layer, communicatively connected to the decision layer, the execution layer including a temperature control module, a humidity control module, a growth agent precision drip irrigation system, and a supplemental lighting module, the execution layer being used to convert the comprehensive control strategy into control commands.

[0011] Furthermore, the growth status assessment module is configured to: calculate the health score by analyzing the leaf color histogram distribution and stem uprightness characteristics in the image data and inputting them into the health regression model.

[0012] Furthermore, the prediction and strategy generation module pre-stores a health threshold table and a correction scheme library corresponding to different generation stages and health states; the correction scheme library includes instructions for adjusting the light spectrum, instructions for adjusting the humidity setpoint, and instructions for modifying the growth agent nutrient ratio.

[0013] Furthermore, the cultivation module is internally configured with a multi-layer cultivation rack, with each layer forming an independent cultivation unit. Each cultivation unit is equipped with independent sensing layer components and execution layer components. The decision layer is configured to perform independent health status assessment and generate comprehensive control strategies for each cultivation unit.

[0014] Furthermore, the system also includes an optimization feedback module, which is configured to: compare the changes in health scores before and after the correction scheme is executed, and dynamically adjust the health threshold or provide retraining data for the prediction and strategy generation module based on the comparison results.

[0015] The beneficial effects of this invention are as follows: By utilizing a closed cultivation module modified from a standard freight container, flexible deployment and standardized management of the cultivation scenario can be achieved; by using a multi-sensor array and a vision module to collaboratively collect data, comprehensive acquisition of growth status and environmental information can be ensured; by combining image recognition and growth prediction models, accurate judgment of growth stages and scientific prediction of future growth can be achieved; and by generating a comprehensive control strategy based on multi-dimensional data, priority can be given to solving growth health problems, ensuring that sprouts grow in a suitable environment, improving quality and survival rate, and reducing labor costs. Attached Figure Description

[0016] Figure 1 This application provides a flowchart illustrating a method for intelligent multi-parameter control of the growth environment of sprouts.

[0017] Figure 2 This application provides a schematic diagram of a multi-parameter intelligent control system for the growth environment of sprouts. Detailed Implementation

[0018] The specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings. It should be understood that the specific embodiments described herein are for illustration and explanation only and are not intended to limit the scope of the present invention.

[0019] In this invention, the terms "system" and "network" are used interchangeably. "Multiple" refers to two or more; therefore, in this invention, "multiple" can also be understood as "at least two." "And / or" describes the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A alone, A and B simultaneously, or B alone. Additionally, the character " / ", unless otherwise specified, generally indicates that the preceding and following related objects have an "or" relationship. Furthermore, it should be understood that in the description of this invention, terms such as "first" and "second" are used only for descriptive purposes and should not be construed as indicating or implying relative importance or order.

[0020] Example 1:

[0021] Sprout cultivation is highly sensitive to environmental parameters. Traditional cultivation relies heavily on manual experience for control, requiring frequent inspections to observe growth status and struggling to precisely manage the coordination of multiple parameters such as temperature and humidity. Manual judgment of growth stages is prone to error and cannot predict growth trends in advance. When abnormal growth occurs, timely intervention is difficult, resulting in inconsistent quality and high loss rates. Furthermore, traditional cultivation facilities are mostly fixed structures, lacking flexibility and hindering large-scale, standardized cultivation, thus restricting the efficient development of the sprout industry.

[0022] like Figure 1 As shown in the figure, this application provides a multi-parameter intelligent control method for the growth environment of sprouts, applied to a closed cultivation module. The method includes: real-time acquisition of environmental data including temperature, humidity, CO2 concentration, light intensity, and growth agent inventory; simultaneously, a movable vision module navigating along a preset track to acquire image data of sprouts in each cultivation unit; based on the image data, determining the current growth stage of the sprouts using an image recognition model and calculating a health score; inputting the environmental data, the growth stage, and the health score into a pre-trained growth prediction model, and the growth prediction model outputs... The system generates a prediction of the growth of sprouts within a preset time period. Based on the growth stage, the health score, and the prediction results, it dynamically generates a comprehensive control strategy. This comprehensive control strategy includes a set of instructions related to environmental parameter settings, growth agent nutrient formula and application mode, and dynamic supplemental lighting. When the health score is lower than the preset health threshold for the current growth stage, the comprehensive control strategy prioritizes the activation of a correction scheme. The comprehensive control strategy is then converted into control instructions and sent to the execution layer to drive the temperature control model, humidity control model, growth agent precision drip irrigation system, and supplemental lighting module to work collaboratively.

[0023] In another possible embodiment, a standard freight container is first converted into a sealed cultivation module, with a suitable installation structure built inside to ensure stable control of environmental parameters such as temperature and humidity. A multi-sensor array, including temperature, humidity, CO2 concentration, light, and growth agent level sensors, is deployed within the cultivation module. High-definition industrial cameras are installed above and to the sides of the cultivation area as vision modules to ensure comprehensive image acquisition of the sprouts. After system startup, the sensor array collects various environmental data in real time and transmits it to the data processing node. The vision modules capture images of the sprouts at a preset cycle (e.g., every 2 hours) and upload them synchronously. The image data, after preprocessing, is input into a trained image recognition model, which outputs the current growth stage and health score of the sprouts. Subsequently, the environmental data, growth stage results, and health score are input into a pre-trained growth prediction model. This model, based on historical cultivation data and real-time parameters, outputs predictions of the sprouts' growth rate, leaf development, and other parameters for the next 12-24 hours. The system determines whether the health score meets the standard based on the standard health threshold corresponding to the current growth stage. If it does not meet the standard, it prioritizes retrieving the correction scheme and combines the prediction results to generate a comprehensive control strategy covering temperature control, humidity control, growth agent drip irrigation, and supplemental lighting parameters. Finally, the control strategy is converted into control commands that can be recognized by each execution module and sent to the execution layer through the communication bus to drive the corresponding devices to coordinate actions and complete one control cycle.

[0024] By utilizing sealed cultivation modules modified from standard freight containers, flexible deployment and standardized management of cultivation scenarios are achieved; data is collected collaboratively by multi-sensor arrays and vision modules to ensure comprehensive acquisition of growth status and environmental information; image recognition and growth prediction models are combined to achieve accurate judgment of growth stages and scientific prediction of future growth conditions; and comprehensive control strategies are generated based on multi-dimensional data to prioritize solving growth health issues, ensuring that sprouts grow in a suitable environment, improving quality and survival rate, and reducing labor costs.

[0025] Traditionally, the assessment of the growth status of sprouts relies heavily on manual observation of intuitive characteristics such as plant height and color. This approach is highly subjective, lacks quantification, makes it difficult to accurately distinguish subtle growth differences, and fails to systematically correlate multiple characteristics for health assessment. This leads to frequent misjudgments of growth stages and missed diagnoses of health problems, which in turn affects the targeting and effectiveness of subsequent control measures.

[0026] In the embodiments of this application, the step of determining the current growth stage of the sprouts based on the image data using an image recognition model and calculating a health score specifically includes: extracting multiple features from the image data, including plant height, leaf spread, leaf color histogram distribution, and stem uprightness; inputting the extracted features into a convolutional neural network classifier to output the probability that the sprouts are in the germination stage, rapid growth stage, or maturity stage, and taking the stage with the highest probability value as the current growth stage of the sprouts; simultaneously, calculating a quantified health score based on the features and a pre-trained health regression model, wherein the color of the sprouts deviating from the normal green threshold range or the stem uprightness being lower than the threshold range will cause the health score to be lower than the health threshold.

[0027] In another possible embodiment, the sprout images acquired by the vision module are first preprocessed, including denoising, cropping, grayscale conversion, and image enhancement to highlight the main features of the sprouts. Image segmentation algorithms are used to separate the sprouts from the background. Through pixel coordinate conversion and edge detection techniques, the plant height (vertical distance from the root to the growing point), leaf spread (horizontal distance between the tips of the largest leaves), and stem uprightness (angle between the stem's central axis and the vertical direction) are calculated. Simultaneously, color histograms of the leaf areas are extracted, and the distribution characteristics of the RGB and HSV color spaces are analyzed to select color parameters strongly correlated with health status. All extracted features are standardized and then divided into two groups, which are input into a convolutional neural network classifier and a health regression model, respectively. The convolutional neural network classifier is trained on a large number of sprout images labeled with growth stages (germination, rapid growth, and maturity). It outputs the probability of the sprout being in each growth stage through feature matching, and the stage with the highest probability is taken as the judgment result. The health regression model is trained on samples labeled with health levels and corresponding features. It maps the input features to a quantitative score of 0-100. If the leaf color deviates from the preset normal green range (such as yellowish or brownish) or the stem uprightness angle exceeds 15°, the model will reduce the score to below the health threshold according to the degree of deviation and finally output an accurate health score.

[0028] By extracting multi-dimensional image features of sprouts, the quantitative and precise judgment of growth stage and health score can be achieved. The morphological features such as plant height and leaf spread are combined with health-related features such as leaf color and stem uprightness to comprehensively reflect the growth status of sprouts. With the help of professional processing of convolutional neural network classifiers and regression models, the accuracy of growth stage judgment and the credibility of health score are improved, providing reliable data support for subsequent precise regulation.

[0029] When sprouts exhibit growth health problems, traditional management often employs uniform, routine measures without developing targeted solutions for specific health issues. For example, when facing different problems such as excessive growth or disease risk, the management methods are singular, which not only fails to quickly improve the growth status but may also exacerbate health problems and prolong the recovery period due to inappropriate measures.

[0030] In the embodiments of this application, the control strategy prioritizes the activation of a correction scheme, wherein the correction scheme includes at least one of the following: adjusting the spectral composition of the supplemental lighting module to increase the proportion of blue light to suppress excessive growth; reducing the environmental humidity setting value to reduce the risk of disease; and adjusting the growth agent formula to increase the supply ratio of trace elements such as calcium and potassium to enhance the plant's resistance to stress.

[0031] In another possible embodiment, the system pre-sets a correction scheme library corresponding to different health problems. When the health score is lower than the health threshold for the current growth stage, it combines the feature data output by the image recognition model to preliminarily determine the type of health problem. If abnormally tall sprouts with thin stems are detected, it is determined to be an etiolation problem. The system controls the spectral adjustment unit of the supplemental lighting module to increase the proportion of blue light in the total spectrum from the usual 15%-20% to 30%-35%, while maintaining stable light intensity. Blue light inhibits the longitudinal growth of sprouts and promotes thicker stems. If the decrease in health score is due to water-soaked spots on the leaves or the humidity sensor detects persistently high ambient humidity, it is determined to be a risk of disease. The ambient humidity setpoint is reduced by 5%-10%, the dehumidification unit of the humidity control module is activated, and internal ventilation is enhanced to accelerate leaf water evaporation and reduce conditions for disease growth. If yellowing leaf edges and easily broken stems are detected, it indicates insufficient stress resistance. The formulation parameters of the precision drip irrigation system for growth regulators should be adjusted, increasing the supply ratio of micronutrients such as calcium and potassium by 20%-30%. The adjusted growth regulators will then be delivered to the roots of the seedlings via precision drip irrigation, enhancing the plant's cell wall strength and stress resistance. Multiple corrective measures can be combined and activated based on the specific health issues.

[0032] Specialized correction programs have been developed to address common health problems in sprouts. By adjusting the supplemental light spectrum, environmental humidity, and growth agent formula, these programs precisely address issues such as excessive growth, disease risk, and poor stress resistance. The various correction programs can be flexibly combined according to the actual health condition to quickly improve the health of sprouts, shorten recovery time, and ensure continuous growth.

[0033] Traditional cultivation modules often employ a uniform control strategy. Even with multi-layer cultivation racks, the lack of independent monitoring of the environment and growth status of each layer leads to uneven growth of sprouts in different layers due to subtle differences in light, ventilation, etc. This makes it impossible to provide targeted control, and plans can only be formulated based on the worst-growing areas, resulting in wasted resources and an inability to guarantee the overall cultivation effect.

[0034] In the embodiments of this application, the cultivation module is configured with a multi-layer cultivation rack, with each layer forming an independent cultivation unit. Based on the different growth stages and health scores of the sprouts in different cultivation units, independent comprehensive control strategies are generated and executed respectively.

[0035] In another possible embodiment, a multi-layered steel structure cultivation rack is constructed within the container-converted cultivation module. Each unit is equipped with a miniaturized sensor array (including temperature, humidity, and CO2 concentration sensors) and an independent vision camera to ensure accurate acquisition of environmental data and sprout images within that unit. In the execution layer, each cultivation unit is equipped with an independent small variable frequency fan, a local humidifier, a growth agent drip irrigation branch pipeline, and LED supplemental lighting, enabling independent control of each unit's equipment. During system operation, the vision module of each unit acquires images of the sprouts in its layer, which are independently analyzed by an image recognition model to output the corresponding growth stage and health score for each layer. After receiving the environmental data, growth stage, and health score from each unit, the decision-making system generates independent comprehensive control strategies for each unit's specific situation. For example, if the upper unit is growing rapidly due to more sunlight and is in a fast growth phase, while the lower unit is growing more slowly and is in the germination phase, then a strategy of higher CO2 concentration and appropriate supplemental lighting is formulated for the upper unit, while a strategy of higher humidity and extended supplemental lighting time is formulated for the lower unit. Subsequently, the corresponding execution devices of each unit receive dedicated control commands and take action, achieving hierarchical independent control.

[0036] The multi-layered cultivation racks within the cultivation module are designed as independent cultivation units, enabling precise monitoring and independent control of each unit. Personalized strategies are developed to address the growth differences of sprouts in different units, avoiding resource waste and uneven growth caused by "one-size-fits-all" control. This ensures that sprouts in each unit are in the optimal growth environment, improving the consistency of overall cultivation quality.

[0037] After the implementation of traditional sprout vegetable control measures, there is a lack of effective effect evaluation and feedback mechanisms, making it impossible to judge in a timely manner whether the control has achieved the expected results. If the control is ineffective, it will delay the best intervention time. Furthermore, it is impossible to optimize the control model and judgment criteria based on the actual cultivation effect, resulting in subsequent control always being in an "experience-driven" state, making it difficult to continuously improve the accuracy.

[0038] In the embodiments of this application, after the comprehensive regulation strategy is converted into control commands and sent to the execution layer to drive the temperature control model, humidity control model, growth agent precision drip irrigation system, and supplemental lighting module to work together, the method further includes: after a predetermined regulation cycle, acquiring image data again through the vision module and calculating a new health score; comparing the new health score with the health score before the correction scheme is initiated to evaluate the effectiveness of the correction scheme; and based on the evaluation results, adaptively adjusting the growth prediction model or the health threshold to optimize the accuracy of subsequent decisions.

[0039] In another possible embodiment, after the execution layer completes a comprehensive control strategy, the system records the time when the correction scheme is initiated and the health score at that time, while presetting the evaluation period (e.g., 20 hours for sprouts in the germination stage and 10 hours for the rapid growth stage). After the evaluation period is reached, the vision module automatically collects image data of the corresponding sprouts and calculates a new health score. The system compares the new score with the score before the correction scheme was initiated. If the score increases by ≥20%, the correction scheme is deemed effective; if the increase is between 10% and 20%, it is deemed basically effective; if the increase is <10% or the score decreases, it is deemed ineffective. Based on the evaluation results, if the results are effective or basically effective, the system will supplement the training dataset of the growth prediction model with the parameters and growth data of this regulation as positive samples to fine-tune the model. If the results are ineffective, the system will analyze the environmental data and image features to determine whether the deviation in the regulation direction is caused by an unreasonable health threshold setting. The system will then appropriately adjust the health threshold for the current growth stage (e.g., lower the health threshold for the germination period from 80 points to 75 points to better reflect reality). At the same time, the data from this regulation will be used as negative samples to input into the model to optimize the model's decision-making logic and ensure that the subsequent generated regulation strategies are more accurate.

[0040] Example 2:

[0041] like Figure 2As shown in the figure, this application provides a multi-parameter intelligent control system for the growth environment of sprouts, applied to a closed cultivation module. The system includes: a perception layer, including a sensor array and a vision module deployed within the cultivation module, for collecting environmental data and sprout image data; a decision layer, communicatively connected to the perception layer, the decision layer including: a growth status assessment module, for determining the growth stage of the sprouts based on image data and calculating a quantitative health score; a prediction and strategy generation module, for making growth predictions based on environmental data, growth stage, and health score, and generating a comprehensive control strategy, the prediction and strategy generation module is further configured to: when the health score is lower than a preset health threshold for the current growth stage, prioritize the execution of a correction scheme in the generated comprehensive control strategy; and an execution layer, communicatively connected to the decision layer, the execution layer including a temperature control module, a humidity control module, a growth agent precision drip irrigation system, and a supplemental lighting module, the execution layer for converting the comprehensive control strategy into control commands.

[0042] In another possible embodiment, the system's hardware and software architecture is built according to functional requirements. The perception layer uses a high-precision, low-power sensor array, with the temperature sensor accuracy set to ±0.1℃ and the humidity sensor accuracy to ±1%RH. The vision module uses an industrial camera with at least 2 megapixels and is equipped with LED supplementary lighting. The decision layer uses an industrial-grade embedded motherboard as the core control unit, running a Linux operating system and installing customized system software. The software integrates a growth status assessment module and a prediction and strategy generation module, and is configured with Ethernet and RS485 communication interfaces to achieve bidirectional communication with the perception and execution layers. The execution layer's temperature control module uses a variable frequency air conditioner and a heating element working together, the humidity control module consists of a humidifier and a dehumidifier, the growth agent precision drip irrigation system is equipped with multiple peristaltic pumps and flow sensors, and the supplementary lighting module uses an adjustable spectrum LED light panel. During system assembly, the sensors and cameras of the perception layer are connected to the decision layer via an RS485 bus, and the devices of the execution layer are connected to the decision layer via relay modules and communication interfaces. After the system is started, the perception layer uploads data to the decision layer in real time. The growth status assessment module processes the image data and outputs the growth stage and health score. The prediction and strategy generation module combines environmental data to complete the growth prediction and regulation strategy generation. If the health score is lower than the threshold, it is included in the correction scheme first. Then the decision layer converts the strategy into control instructions and sends them to the execution layer through the communication line to drive each device to act according to the instructions, so as to realize the coordinated operation of each level.

[0043] By clearly dividing the system into a perception layer, a decision-making layer, and an execution layer, the system achieves professional division of labor and efficient collaboration among its functional modules. The perception layer comprehensively collects data, the decision-making layer accurately analyzes and generates strategies, and the execution layer efficiently implements control instructions. Smooth communication between layers ensures the efficient operation of the control process. The system is compatible with containerized cultivation modules, enhancing deployment flexibility and providing stable and reliable technical support for the large-scale and standardized cultivation of sprouts.

[0044] Traditional growth status assessments often focus on the morphological characteristics of sprouts, neglecting key features directly related to health status, such as leaf color and stem uprightness. This results in one-sided assessments that fail to accurately reflect the growth and health status of sprouts, thus affecting the effectiveness of control measures.

[0045] In an embodiment of this application, the growth status assessment module is configured to: analyze the leaf color histogram distribution and stem uprightness characteristics in the image data and input them into the health regression model to calculate the health score.

[0046] In another possible embodiment, the growth status assessment module presets the RGB value range of normal leaf color (e.g., R: 80-100, G: 150-180, B: 60-80) and the normal stem uprightness threshold (angle ≤15° with the vertical direction). When the image data acquired by the vision module is transmitted to this module, the leaf and stem regions are first extracted using image segmentation technology. For the leaf region, color histogram distribution data is extracted, the average RGB value of the region is calculated, and compared with the preset normal range to statistically analyze the percentage of pixels deviating from the color. For the stem region, edge detection and line fitting techniques are used to determine the stem's central axis and calculate its angle with the vertical direction. These two feature parameters—the percentage of pixels deviating from the color and the stem uprightness angle—are combined with morphological features such as plant height and leaf spread, and input into a pre-trained health regression model. Based on the weight coefficients of each feature, the model comprehensively quantifies and scores the health status of the sprouts. The higher the percentage of color deviation and the larger the stem uprightness angle, the greater the corresponding score deduction, ultimately outputting a score that accurately reflects the health status.

[0047] The growth status assessment module focuses on two core health-related characteristics: leaf color and stem uprightness. Through professional model-based quantitative analysis, the health score is made more consistent with the actual health status of the sprouts, and the assessment results are more valuable, providing a precise basis for decision-makers to formulate targeted control strategies.

[0048] In the embodiments of this application, the prediction and strategy generation module pre-stores a health threshold table and a correction scheme library corresponding to different generation stages and health states; the correction scheme library includes instructions for adjusting the light spectrum, instructions for adjusting the humidity setpoint, and instructions for modifying the growth agent nutrient ratio.

[0049] In another possible embodiment, a health threshold table, validated through extensive testing, is pre-imported into the storage unit of the prediction and strategy generation module. This table clearly divides the stage into three phases: germination, rapid growth, and maturity. Each phase corresponds to a different health score threshold (e.g., 80 points for germination, 75 points for rapid growth, and 70 points for maturity). Simultaneously, a correction scheme library is established, storing corresponding control instructions categorized by health problem type. For example, instructions to adjust the light spectrum include parameter settings for different blue light percentages (e.g., 25%-40%), instructions to adjust the humidity setpoint cover humidity ranges for different growth stages (e.g., 75%-85% for germination, 65%-75% for rapid growth), and instructions to modify the nutrient ratio of growth promoters specify the exact addition ratios of trace elements such as calcium and potassium. When the module receives growth stage and health score data, it first queries the health threshold table to determine if a correction scheme needs to be activated. If so, it retrieves the corresponding instructions from the scheme library based on the health problem type, adjusts the parameters in conjunction with real-time environmental data, and integrates them into a complete comprehensive control strategy.

[0050] By pre-stored health threshold tables and correction scheme libraries, standardized basis is provided for growth status judgment and regulation strategy formulation. The health threshold tables set specific standards for different growth stages, and the correction scheme library covers solutions to a variety of health problems, ensuring that decision-makers can quickly and systematically generate regulation strategies, thereby improving the consistency and effectiveness of regulation.

[0051] In the embodiments of this application, the cultivation module is internally configured as a multi-layer cultivation rack, with each layer forming an independent cultivation unit. The cultivation unit is equipped with independent sensing layer elements and execution layer elements. The decision layer is configured to perform independent health status assessment and generate comprehensive control strategies for each cultivation unit.

[0052] In another possible embodiment, within each cultivation unit of the multi-layer cultivation rack, a set of miniaturized sensing elements is independently deployed, including a composite sensor integrating temperature, humidity, and CO2 concentration, and a fixed-focus industrial camera. The sensor is installed in the middle of the unit, and the camera is fixed at the top center of the unit to ensure the accuracy of the collected data. Regarding the actuators, each unit is equipped with a small variable frequency fan, a miniature humidifier, an independent set of LED supplemental lighting strips, and a growth agent drip irrigation branch. The supplemental lighting strips are arranged along both sides of the unit, and the drip irrigation branch is evenly distributed above the cultivation trays. The core control unit of the decision layer establishes data receiving links with the sensing elements of each unit through a multi-channel communication module, and simultaneously establishes control links with the actuators of each unit through a multi-channel relay module. During system operation, the decision layer receives data uploaded by each sensing element sequentially according to the unit order. The growth status assessment module processes the image data of each unit independently, and the prediction and strategy generation module generates control strategies for the parameters of each unit, subsequently issuing corresponding control commands to the actuators of each unit, achieving independent control of each unit without interference.

[0053] By equipping each cultivation unit with independent sensing and execution elements, combined with the independent processing capabilities of the decision-making layer, precise monitoring, independent evaluation, and dedicated control of each unit can be achieved. This effectively solves the problem of uneven growth caused by differences in the microenvironment in multi-layer cultivation, improves the growth quality of sprouts in each layer, and ensures the overall cultivation effect.

[0054] In embodiments of this application, the system further includes an optimization feedback module, which is configured to: compare the changes in health scores before and after the correction scheme is executed, and dynamically adjust the health threshold or provide retraining data for the prediction and strategy generation module based on the comparison results.

[0055] In another possible embodiment, an optimization feedback module is added to the system. This module establishes a data interaction channel with the prediction and strategy generation module and the growth status assessment module of the decision-making layer. After the correction scheme is executed and the evaluation cycle is completed, the optimization feedback module obtains the health scores before and after correction from the growth status assessment module, calculates the magnitude of score change, and generates an evaluation report. If the score improves significantly (e.g., ≥20%), the optimization feedback module organizes the environmental data, growth stage, control strategy, and score changes of this cultivation into positive sample data and sends it to the prediction and strategy generation module for model retraining and parameter optimization. If the score improvement is not significant or decreases, the module analyzes whether it is due to an unreasonable health threshold setting. Combining historical cultivation data of the current growth stage, it calculates a reasonable threshold adjustment range, generates threshold adjustment suggestions, and submits them to the prediction and strategy generation module. After system confirmation, the health threshold table is updated. At the same time, the relevant data of this ineffective control is used as negative samples input into the prediction and strategy generation module to help the model avoid similar decision biases and achieve continuous optimization of the system's control capabilities.

[0056] By setting up an optimization feedback module as the system's "self-upgrade" unit, the system can accurately assess the control effect by comparing the changes in health before and after the implementation of the correction scheme. This allows for dynamic optimization of the health threshold and model training data, enabling the system's decision-making capabilities to continuously improve and adapt to the needs of sprout cultivation in different varieties and environments.

[0057] The optional embodiments of the present invention have been described in detail above with reference to the accompanying drawings. However, the embodiments of the present invention are not limited to the specific details in the above embodiments. Within the scope of the technical concept of the embodiments of the present invention, various simple modifications can be made to the technical solutions of the embodiments of the present invention, and these simple modifications all fall within the protection scope of the embodiments of the present invention.

[0058] It should also be noted that the various specific technical features described in the above embodiments can be combined in any suitable manner without contradiction. To avoid unnecessary repetition, the embodiments of the present invention will not describe the various possible combinations separately.

[0059] Furthermore, various different implementations of the present invention can be combined arbitrarily, as long as they do not violate the spirit of the present invention, they should also be regarded as the content disclosed in the present invention.

Claims

1. A method for intelligent multi-parameter control of the growth environment of sprouts, characterized in that, The method, applied to a closed cultivation module, includes: Real-time collection of environmental data including temperature, humidity, CO2 concentration, light intensity and growth agent inventory; at the same time, a movable vision module cruises along a preset track to collect image data of the sprouts in each cultivation unit. Based on the image data, the current growth stage of the sprouts is determined by an image recognition model, and a health score is calculated and generated. The environmental data, the growth stage, and the health score are input into a pre-trained growth prediction model, which outputs a prediction result of the growth of sprouts within a preset time period. Based on the growth stage, the health score, and the prediction results, a comprehensive regulation strategy is dynamically generated. The comprehensive regulation strategy includes a set of instructions related to environmental parameter settings, growth agent nutrient formula and application mode, and dynamic supplemental lighting scheme. When the health score is lower than the preset health threshold for the current growth stage, the comprehensive regulation strategy prioritizes the activation of a correction scheme. The comprehensive regulation strategy is converted into control commands and sent to the execution layer to drive the temperature control model, humidity control model, growth agent precision drip irrigation system and supplemental lighting module to work together.

2. The multi-parameter intelligent control method for the growth environment of sprouts according to claim 1, characterized in that, The process of determining the current growth stage of the sprouts based on the image data using an image recognition model and calculating a health score specifically includes: The image data is subjected to multi-feature extraction, and the features include plant height, leaf spread, leaf color histogram distribution, and stem uprightness. The extracted features are input into a convolutional neural network classifier, which outputs the probability that the sprouts are in the germination period, rapid growth period or maturity period, and the stage with the highest probability value is taken as the current growth stage of the sprouts. Meanwhile, based on the features and the pre-trained health regression model, a quantitative health score is calculated, wherein the color of the sprouts deviates from the normal green threshold range or the stem uprightness is lower than the threshold range, which will cause the health score to be lower than the health threshold.

3. The method according to claim 1, wherein, The control strategy prioritizes the activation of a correction scheme, wherein the correction scheme includes at least one of the following: Adjust the spectral composition of the supplementary lighting module to increase the proportion of blue light in order to suppress excessive growth; Lower the ambient humidity setting to reduce the risk of disease. Adjust the growth regulator formula to increase the supply ratio of trace elements such as calcium and potassium to enhance the plant's resistance to stress.

4. The multi-parameter intelligent control method for the growth environment of sprouts according to claim 1, characterized in that, The cultivation module is internally configured with a multi-layer cultivation rack, with each layer forming an independent cultivation unit. Based on the different growth stages and health scores of the sprouts in different cultivation units, independent comprehensive control strategies are generated and executed.

5. The method for intelligent multi-parameter control of the growth environment of sprouts according to claim 1, characterized in that, After converting the comprehensive regulation strategy into control commands and issuing them to the execution layer to drive the temperature control model, humidity control model, growth agent precision drip irrigation system, and supplemental lighting module to work collaboratively, the method further includes: After the predetermined adjustment period, image data is collected again through the vision module, and a new health score is calculated. The new health score is compared with the health score before the correction program was initiated to evaluate the effectiveness of the correction program; Based on the evaluation results, the growth prediction model or the health threshold is adaptively adjusted to optimize the accuracy of subsequent decisions.

6. A multi-parameter intelligent control system for the growth environment of sprouts, characterized in that, The system, applied to a closed-loop cultivation module, includes: The perception layer includes a sensor array and a vision module deployed within the cultivation module, used to collect environmental data and sprout image data; A decision layer, communicatively connected to the perception layer, comprising: The growth status assessment module is used to determine the growth stage of sprouts based on image data and calculate a quantitative health score. The prediction and strategy generation module is used to predict growth based on environmental data, growth stage and health score, and generate a comprehensive control strategy. The prediction and strategy generation module is also configured to: when the health score is lower than the preset health threshold of the current growth stage, prioritize the execution of the correction scheme in the generated comprehensive control strategy. The execution layer is communicatively connected to the decision-making layer. The execution layer includes a temperature control module, a humidity control module, a growth agent precision drip irrigation system, and a supplemental lighting module. The execution layer is used to convert the comprehensive regulation strategy into control commands.

7. The intelligent multi-parameter control of the growth environment of sprouts according to claim 6, characterized in that, The growth status assessment module is configured as follows: The health score is calculated by analyzing the leaf color histogram distribution and stem uprightness characteristics in the image data and inputting them into the health regression model.

8. The intelligent multi-parameter control method for the growth environment of sprouts according to claim 6, characterized in that, The prediction and strategy generation module pre-stores a health threshold table and a correction scheme library corresponding to different generation stages and health states. The calibration scheme library includes instructions for adjusting the light spectrum, instructions for adjusting the humidity setpoint, and instructions for modifying the growth agent nutrient ratio.

9. The intelligent multi-parameter control method for the growth environment of sprouts according to claim 6, characterized in that, The cultivation module is internally configured with a multi-layer cultivation rack, with each layer forming an independent cultivation unit. The cultivation unit is equipped with independent sensing layer components and execution layer components. The decision-making layer is configured to perform independent health status assessments and generate comprehensive control strategies for each breeding unit.

10. The multi-parameter intelligent regulation and control method for the growth environment of sprouts according to claim 6, characterized in that, The system also includes an optimization feedback module, which is configured to: compare the changes in health scores before and after the correction scheme is implemented, and dynamically adjust the health threshold or provide retraining data for the prediction and strategy generation module based on the comparison results.

Images