Wild-to-home-planting cultivation system and method for Chenopodium japonicum based on ecological simulation

Abstract

The embodiment of the invention discloses an ecological simulation-based wild-to-home cultivation system and method for Chenopodium japonicum. The system comprises a data acquisition module used for acquiring the original data of the cultivation environment of Chenopodium japonicum; the data processing module is used for removing abnormal data, obtaining cultivation environment data, judging the matching degree of the cultivation environment data and cultivation environment ideal data, and calculating a single-parameter relative deviation value and an environment comprehensive deviation value; the decision generation module is used for obtaining a state early warning signal according to the single-parameter relative deviation value and the parameter regulation priority, identifying a leaf health level, a plant growth health index and a plant health state level, obtaining a candidate regulation strategy according to a reinforcement learning strategy optimization algorithm, and determining an optimal regulation strategy according to a multi-objective optimization algorithm; and the decision execution module is used for cooperatively regulating and controlling the cultivation equipment according to the optimal regulation and control strategy. According to the method, the accuracy and intelligence of wild-to-home cultivation of the rhododendron dauricum can be improved, and the survival rate of the home-planted rhododendron dauricum is greatly increased.

Description

Technical Field

[0001] The present application relates to the technical field of artificial cultivation of Panax notoginseng, and is related to, but not limited to, a system and method for converting wild Panax notoginseng to domestic cultivation based on ecological simulation. Background Art

[0002] As a member of the genus Panax notoginseng in the Asteraceae family, Panax notoginseng holds a prominent position in Traditional Chinese Medicine. In recent years, with in-depth research on its medicinal value, Panax notoginseng has demonstrated unique therapeutic effects in the treatment of specialized diseases such as lung disease, endocrinology, and skin disease, resulting in a growing market demand for Panax notoginseng. However, Panax notoginseng primarily grows in specific ecological environments such as hillside forest margins, stream banks, and roadside wastelands at altitudes between 1,600 and 3,300 meters. Its growth cycle is long and extremely sensitive to environmental conditions. Furthermore, due to long-term overharvesting, the survival of Panax notoginseng in the wild faces severe challenges. The sharp decline in its wild resources not only threatens the species' survival but also makes it difficult to meet the growing demand for its medicinal properties. Therefore, achieving sustainable resource utilization through artificial cultivation is particularly important.

[0003] In the existing technology, artificial cultivation of Panax notoginseng mostly adopts simple greenhouse methods, combined with conventional agricultural methods for management, and simulates the natural environment by manually adjusting the shade net, timed irrigation and fertilization, etc. A few high-end cultivation systems introduce sensors to monitor environmental parameters in real time, but the existing cultivation technology can only achieve feedback adjustment of a single parameter, and lacks the coordinated simulation of the entire ecological elements of Panax notoginseng growth, resulting in inaccurate environmental simulation and inability to accurately match the complex ecological conditions required for the wild growth of Panax notoginseng, thereby resulting in a decline in the quality of the medicinal material and a decrease in the survival rate; and the existing cultivation technology mostly adopts timed or manual experience management, and cannot dynamically adjust the cultivation strategy according to real-time environmental changes and plant growth stages, affecting the yield and quality stability of Panax notoginseng; in addition, the existing cultivation technology has a low degree of intelligence, lacks coordination and linkage between equipment, and has weak data collection and analysis capabilities, making it difficult to achieve efficient resource utilization and effective cost control. The above problems result in a matching degree between the domestic artificial cultivation environment and the wild ecological environment of less than 60%, and a survival rate of Panax notoginseng of less than 65%.

[0004] Therefore, there is an urgent need for a more intelligent artificial cultivation system to solve the problems existing in existing artificial cultivation technology, such as inaccurate environmental simulation, lack of dynamic regulation and low level of intelligence, so as to improve the accuracy and intelligence of the cultivation of wild-cultivated Panax notoginseng, thereby greatly improving the survival rate of domesticated Panax notoginseng and ultimately realizing large-scale cultivation of Panax notoginseng. Summary of the Invention

[0005] The embodiments of the present application provide a system and method for converting wild Panax notoginseng to domestic cultivation based on ecological simulation.

[0006] The technical solution of the embodiment of the present application is implemented as follows: In a first aspect, the embodiments of the present application provide a wild to cultivated system of Gynostemma pentaphyllum based on ecological simulation, the system comprising a data acquisition module, a data processing module, a decision generation module and a decision execution module, wherein: The data acquisition module is configured to acquire original data of a cultivation environment of Gynostemma pentaphyllum, the original data of the cultivation environment comprising soil environment data, meteorological environment data and microbial data; the data processing module is configured to repair and remove abnormal data and noise data in the original data of the cultivation environment to obtain cultivation environment data, judge a matching degree between the cultivation environment data and ideal cultivation environment data through dynamic time warping and Bayesian network inference, and calculate a single parameter relative deviation value and an environment comprehensive deviation value of the cultivation environment data and the ideal cultivation environment data; the decision generation module is configured to obtain a state early warning signal according to the single parameter relative deviation value and a parameter control priority, identify a leaf health grade, a plant growth health index and a plant health state grade according to a convolutional neural network image classification model, a random forest algorithm and plant physiological data, combine the environment comprehensive deviation value and the plant growth health index, obtain a candidate control strategy according to a reinforcement learning strategy optimization algorithm, and determine an optimal control strategy according to a multi-objective optimization algorithm verification of the candidate control strategy; and the decision execution module is configured to coordinate control of irrigation and fertilization control equipment, light and temperature and humidity control equipment and ventilation and gas regulation equipment according to the optimal control strategy, and monitor and adjust a running state of the execution equipment in real time.

[0007] The technical scheme provided in the application collects the original data of the cultivation environment of the radix heterophyllae through a data acquisition module, the original data of the cultivation environment including soil environment data, meteorological environment data and microbial data, all kinds of environment data being collected in full dimension in real time to provide data support for precise regulation and control and realize early warning of insect pests; the data processing module repairs and removes abnormal data and noise data in the original data of the cultivation environment, improves the reliability of data repair, obtains the cultivation environment data, judges the matching degree of the cultivation environment data and ideal cultivation environment data through dynamic time warping and Bayesian network inference, quantifies the dynamic environment matching degree, facilitates accurate identification of environmental deviation, calculates the single-parameter relative deviation value and the environmental comprehensive deviation value of the cultivation environment data and the ideal cultivation environment data, calculates the environmental comprehensive deviation value according to the parameter sensitivity weighting, determines the control priority, and avoids system shock caused by simultaneous adjustment of multiple parameters; in the decision generation module, the state warning signal is obtained according to the single-parameter relative deviation value and the parameter control priority, the leaf health grade, the plant growth health index and the plant health state grade are identified according to the convolutional neural network image classification model, the random forest algorithm and the plant physiological data, the dynamic evaluation of the plant growth state is realized, the candidate control strategy is obtained according to the reinforcement learning strategy optimization algorithm combined with the environmental comprehensive deviation value and the plant growth health index, the high cost-effective control action is selected through the reward function, the efficiency of artificial decision is improved by 80%, the candidate control strategy is verified according to the multi-objective optimization algorithm, the three targets of "maximizing the plant growth health index, minimizing the resources and minimizing the environmental fluctuation" are coordinated, and the optimal control strategy is determined; in the decision execution module, the irrigation and fertilization control equipment, the light and temperature and humidity control equipment and the ventilation and gas regulation equipment are cooperatively regulated and controlled according to the optimal control strategy, the running state of the execution equipment is monitored and adjusted in real time, millisecond-level linkage response of the equipment is realized, multiple equipment cooperatively completes the regulation and control within 3 minutes, the efficiency of artificial response is improved by 80%, accurate resource control is realized, resource waste is avoided, and the stability of the cultivation environment is strengthened.

[0008] Optionally, the data acquisition module comprises a soil data acquisition unit, a meteorological data acquisition unit and a microbial data acquisition unit, wherein: the soil data acquisition unit is used to acquire soil temperature data and soil humidity data according to a capacitive soil temperature and humidity sensor, remove the influence of soil thermal conductivity according to a temperature compensation algorithm, acquire soil microbial activity data and rhizosphere microenvironment data through a bionic soil probe, acquire soil fertility data and soil electrochemical data through a spectral soil fertility sensor and a pH-conductivity composite sensor, and generate a soil nutrient spatial distribution map through a space-time interpolation algorithm; the meteorological data acquisition unit is used to acquire air flow data according to an ultrasonic wind speed and direction sensor, identify carbon dioxide concentration data through a dual-beam infrared carbon dioxide sensor, acquire leaf surface water film thickness through a leaf humidity sensor, acquire plant spectral data through a multispectral imaging sensor, and acquire ultraviolet radiation data through an ultraviolet intensity probe; the microbial data acquisition unit is used to analyze the abundance of dominant flora through a portable fluorescent quantitative polymerase chain reaction instrument, and identify pest species and quantity according to a YOLOv5 algorithm through a sex attractant-image recognition composite sensor.

[0009] Optionally, the data processing module comprises a data cleaning unit, an environment matching unit and a deviation analysis unit, wherein: the data cleaning unit is used to identify abnormal data in continuous data and periodic abnormal data in time series data by using a statistical algorithm, identify abnormal data in high-dimensional data by using a machine learning algorithm, repair single-point missing abnormal data by using a forward-backward filling method, repair continuous missing abnormal data based on a Kalman filtering method, repair spatially inconsistent abnormal data by using a Kriging interpolation method for weighted calculation of missing values, remove noise data in high-frequency fluctuation data by using a moving average filtering method, remove noise data in low-frequency slowly varying data by using a median filtering method, and obtain the cultivation environment data; the environment matching unit is used to perform elastic matching through dynamic time warping for the cultivation environment data and the ideal cultivation environment data, calculate a bending path distance, construct a multi-parameter dependency graph through Bayesian network reasoning, infer environmental suitability through a conditional probability table, and finally obtain the matching degree; and the deviation analysis unit is used to calculate a single-parameter absolute deviation value of the cultivation environment data and the ideal cultivation environment data, obtain a single-parameter relative deviation value according to the single-parameter absolute deviation value, and calculate an environmental comprehensive deviation value according to the single-parameter relative deviation value and a cultivation environment parameter weight.

[0010] Optionally, the decision generation module comprises a state warning unit, a health index generation unit and a strategy generation unit, wherein: the state warning unit is configured to determine parameter control priority according to the sensitivity and importance of the current growth stage cultivation environment parameters, determine a state warning signal level according to the single parameter relative deviation value, and determine the state warning signal according to the parameter control priority and the state warning signal level; the health index generation unit is configured to identify the leaf health grade through transfer learning according to a convolutional neural network image classification model and a leaf image recognition result, normalize six environmental indicators and six soil nutrient indicators, assign weights to the twelve indicators through a random forest algorithm, linearly combine the normalized indicator values and the weight values to obtain the plant growth health index, and obtain the plant health state grade according to the leaf health grade and the plant growth health index; and the strategy generation unit is configured to combine the environmental comprehensive deviation value and the plant growth health index, randomly select and execute an initial action from a basic control action group according to a reinforcement learning strategy optimization algorithm, observe the feedback of the plant growth health index, retain high-reward actions through a reward function evaluation, form a stable strategy library and screen the candidate control strategy, verify the candidate control strategy according to a multi-objective optimization algorithm, and determine the optimal control strategy according to the constraint conditions.

[0011] Optionally, the decision execution module is further configured to calculate crop water requirement based on a soil water potential model, dynamically adjust an irrigation plan in combination with weather forecast data, and adjust the amount of fertilization according to soil fertility sensor data and growth stage threshold values, using a proportional-integral-derivative control algorithm.

[0012] In a second aspect, the embodiments of the present application provide a wild to domesticated cultivation method of Gynostemma pentaphyllum based on ecological simulation. The method is applied to a wild to domesticated cultivation system of Gynostemma pentaphyllum, the system comprises a data acquisition module, a data processing module, a decision generation module and a decision execution module, and the method comprises the following steps: collecting original data of a cultivation environment of Gynostemma pentaphyllum, wherein the original data of the cultivation environment comprises soil environment data, meteorological environment data and microbial data; repairing and removing abnormal data and noise data in the original data of the cultivation environment to obtain cultivation environment data; judging a matching degree between the cultivation environment data and ideal cultivation environment data by dynamic time warping and Bayesian network inference, and calculating a single parameter relative deviation value and an environment comprehensive deviation value of the cultivation environment data and the ideal cultivation environment data; obtaining a state early warning signal according to the single parameter relative deviation value and a parameter control priority, identifying a leaf health grade, a plant growth health index and a plant health state grade according to a convolutional neural network image classification model, a random forest algorithm and plant physiological data, combining the environment comprehensive deviation value and the plant growth health index, obtaining a candidate control strategy according to a reinforcement learning strategy optimization algorithm, verifying the candidate control strategy according to a multi-objective optimization algorithm, and determining an optimal control strategy; and cooperatively controlling irrigation and fertilization control equipment, light and temperature and humidity control equipment and ventilation and gas regulation equipment according to the optimal control strategy, and monitoring and adjusting a running state of the execution equipment in real time.

[0013] In a third aspect, the embodiments of the present application provide an electronic device, comprising a memory and a processor, wherein the memory stores a computer program capable of running on the processor, and the processor implements the steps of the wild to domesticated cultivation method of Gynostemma pentaphyllum based on ecological simulation when executing the program.

[0014] In a fourth aspect, the embodiments of the present application provide a computer readable storage medium, which stores a computer program capable of being executed by a processor to implement the steps of the wild to domesticated cultivation method of Gynostemma pentaphyllum based on ecological simulation.

[0015] The technical scheme provided by the embodiments of the present application has at least the following beneficial effects: The present application provides a system and method for the cultivation of wild-to-domesticated Panax notoginseng based on ecological simulation. The data acquisition module collects the original data of the cultivation environment of Panax notoginseng, and the original data of the cultivation environment include soil environment data, meteorological environment data and microbial data. Various environmental data are collected in real time in all dimensions to provide data support for precise regulation and realize early warning of insect pests. The data processing module repairs and removes abnormal data and noise data in the original data of the cultivation environment, thereby improving the reliability of data repair and obtaining the cultivation environment data. The matching degree between the cultivation environment data and the ideal cultivation environment data is judged through dynamic time warping and Bayesian network reasoning, and the dynamic environment matching degree is quantified to facilitate accurate identification of environmental deviations. The relative deviation value of single parameters and the comprehensive environmental deviation value of the cultivation environment data and the ideal cultivation environment data are calculated, and the comprehensive environmental deviation value is calculated according to the parameter sensitivity weight, so as to clarify the regulation priority and avoid system oscillation caused by simultaneous adjustment of multiple parameters. In the decision generation module, the status is obtained according to the relative deviation value of single parameters and the parameter regulation priority. The system generates dynamic early warning signals and identifies leaf health levels, plant growth health indexes, and plant health status levels based on convolutional neural network image classification models, random forest algorithms, and plant physiological data, enabling dynamic assessment of plant growth status. Combined with the comprehensive environmental deviation value and the plant growth health index, the system obtains candidate control strategies based on the reinforcement learning strategy optimization algorithm. The reward function is used to screen cost-effective control actions, improving efficiency by 80% compared to manual decision-making. The system verifies candidate control strategies based on a multi-objective optimization algorithm to ensure the coordination of the three objectives of "maximizing plant growth health index, minimizing resources, and minimizing environmental fluctuations" and determine the optimal control strategy. In the decision-making execution module, the system coordinates and controls irrigation and fertilization control equipment, lighting and temperature and humidity control equipment, and ventilation and gas regulation equipment according to the optimal control strategy. The system monitors and adjusts the operating status of the execution equipment in real time, achieving millisecond-level linkage response of the equipment. Control of multiple devices can be completed within 3 minutes, improving efficiency by 80% compared to manual response. It also achieves precise resource control, ensures zero resource waste, and enhances the stability of the cultivation environment. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the following briefly introduces the drawings required for describing the embodiments. Obviously, the drawings described below are only some embodiments of the present application. Those skilled in the art can also derive other drawings based on these drawings without inventive work, among which: Figure 1 A schematic diagram of a wild-to-domestic cultivation system for Panax notoginseng based on ecological simulation provided in an embodiment of the present application; Figure 2 A flow chart of a method for converting wild ginseng to domestic cultivation based on ecological simulation provided in an embodiment of the present application; Figure 3 A hardware entity schematic diagram of an electronic device is provided for the embodiments of the present application. DETAILED DESCRIPTION

[0017] To make the objectives, technical solutions and advantages of the embodiments of the present application clearer, the technical solutions in the embodiments of the present application will be described below in conjunction with the accompanying drawings of the embodiments of the present application. Obviously, the described embodiments are some of the embodiments of the present application but not all the embodiments of the present application. The following embodiments are used to illustrate the present application but not to limit the scope of the present application. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative work fall within the scope of the present application.

[0018] In the following description, “some embodiments” are related to a subset of all possible embodiments, but it can be understood that “some embodiments” can be the same subset or different subsets of all possible embodiments and can be combined with each other without conflict.

[0019] It should be noted that the terms “first\second\third” involved in the embodiments of the present application are only to distinguish similar objects and do not represent a specific order of the objects. It can be understood that “first\second\third” can be interchanged with a specific order or sequence as allowed, so that the embodiments of the present application described herein can be implemented in an order other than that illustrated or described herein.

[0020] Those skilled in the art can understand that, unless otherwise defined, all terms (including technical terms and scientific terms) used herein have the same meaning as generally understood by those of ordinary skill in the art to which the embodiments of the present application belong. It should also be understood that terms such as those defined in a general dictionary should be understood to have meanings consistent with those in the context of the prior art, and unless specifically defined as such herein, should not be interpreted to have idealized or overly formal meanings.

[0021] The embodiments of the present application will be further described below in conjunction with the accompanying drawings.

[0022] In view of the problems existing in the wild-to-cultivated cultivation of Shouyangchun based on ecological simulation in the field of artificial cultivation of Shouyangchun, the embodiments of the present application provide a wild-to-cultivated cultivation system and method of Shouyangchun based on ecological simulation.

[0023] The technical solutions of the present application will be introduced below. First, the system embodiments of the present application will be introduced.

[0024] Please refer to Figure 1 which shows a schematic diagram of a wild-to-cultivated cultivation system of Shouyangchun based on ecological simulation provided by the embodiments of the present application, as shown inFigure 1 As shown, the system comprises a data acquisition module 01, a data processing module 02, a decision generation module 03 and a decision execution module 04. The data acquisition module 01, the data processing module 02, the decision generation module 03 and the decision execution module 04 are connected in sequence. The data acquisition module 01 is used to acquire the original data of the cultivation environment of the roots of the plant, and the original data of the cultivation environment includes soil environment data, meteorological environment data and microbial data; the data processing module 02 is used to repair and remove abnormal data and noise data in the original data of the cultivation environment, to obtain cultivation environment data, to judge the matching degree of the cultivation environment data and ideal cultivation environment data through dynamic time warping and Bayesian network inference, and to calculate the single parameter relative deviation value and the environment comprehensive deviation value of the cultivation environment data and the ideal cultivation environment data; the decision generation module 03 is used to obtain a state warning signal according to the single parameter relative deviation value and the parameter control priority, to identify the leaf health grade, the plant growth health index and the plant health state grade according to the convolutional neural network image classification model, the random forest algorithm and the plant physiological data, to combine the environment comprehensive deviation value and the plant growth health index, to obtain a candidate control strategy according to the reinforcement learning strategy optimization algorithm, and to determine the optimal control strategy according to the multi-objective optimization algorithm verification of the candidate control strategy; the decision execution module 04 is used to coordinate the irrigation and fertilization control equipment, the light and temperature and humidity control equipment, and the ventilation and gas regulation equipment according to the optimal control strategy, and to monitor and adjust the running state of the execution equipment in real time.

[0025] In the embodiments of the present application, the data acquisition module 01 includes a soil data acquisition unit, a meteorological data acquisition unit and a microorganism data acquisition unit. The soil data acquisition unit is used to acquire soil temperature data and soil humidity data according to a capacitive soil temperature and humidity sensor, remove the influence of soil thermal conductivity according to a temperature compensation algorithm, acquire soil microbial activity data and rhizosphere microenvironment data through a bionic soil probe, acquire soil fertility data and soil electrochemical data through a spectral soil fertility sensor and a pH-conductivity composite sensor, and generate a soil nutrient spatial distribution map through a space-time interpolation algorithm. Specifically, the soil temperature data and the soil humidity data are collected in a layered manner, a capacitive soil temperature and humidity sensor with a precision of ±2% relative humidity and ±0.5℃ temperature is used, the sensor is buried at depths of 10cm, 20cm and 30cm in the vertical direction of the planting bed, 3 groups of sensors are arranged per row, the distance between each group of sensors is 1.5 meters, the sensor captures the dynamic changes of the temperature and humidity of the root development layer in real time, acquires the soil temperature data and the soil humidity data, and the sensor is built-in with a temperature compensation algorithm, which removes the influence of soil thermal conductivity difference on temperature measurement according to the temperature compensation algorithm, so as to ensure the stability of the sensing accuracy in different seasons. The main body of the bionic soil probe is made of food-grade material, the outer layer is covered with root micropore structure with a simulated pore size of 20-50μm, the surface is loaded with low-concentration root exudate simulators (such as L-glutamic acid and glucose), which induce the directional aggregation of rhizosphere microorganisms, and the bionic soil probe is built-in with a micro electrochemical sensor array, which detects the oxidation-reduction current generated by microbial metabolism in real time through the alternating current impedance spectroscopy technology, establishes the mapping relationship between the abundance of dominant flora such as actinomycetes and mycorrhizal fungi and the impedance signal, realizes the in-situ monitoring of the rhizosphere microbial community dynamics (such as the change rate of actinomycetes abundance), and compared with the traditional method which needs 48-72 hours, the response time of the method provided by the embodiments of the present application is controlled within 10 minutes, the time cost is greatly reduced, the soil temperature and conductivity sensors are integrated in the bionic soil probe, the physical and chemical parameters of the rhizosphere microenvironment are synchronously acquired, the soil microbial activity data and the rhizosphere microenvironment data are finally acquired, and the construction of the "microbial activity-soil water-ion concentration" coupling model is realized. The nitrogen, phosphorus and potassium detection accuracy of the spectral soil fertility sensor is ±3%, the pH detection accuracy of the pH-conductivity composite sensor is ±0.1, the conductivity detection accuracy is ±5mS / cm, one group is arranged every 20 square meters, the four-electrode method is used to reduce the polarization effect, the soil available nitrogen, available phosphorus, potassium ion concentration and acid-base balance state are monitored in real time, the soil fertility data and the soil electrochemical data are acquired, and the data is used to generate a soil nutrient spatial distribution map through a space-time interpolation algorithm, which provides a basis for precision fertilization.

[0026] The meteorological data acquisition unit is used to collect air flow data according to an ultrasonic wind speed and direction sensor, to identify carbon dioxide concentration data through a dual-beam infrared carbon dioxide sensor, to collect leaf surface water film thickness through a leaf humidity sensor, to collect plant spectral data through a multi-spectral imaging sensor, and to collect ultraviolet radiation data through an ultraviolet intensity probe. Specifically, a cluster of micro weather stations is deployed in the cultivation area, and the micro weather stations are deployed at a density of 50 m 2 / per station. The micro weather stations are integrated with an ultrasonic wind speed and direction sensor, a dual-beam infrared carbon dioxide sensor, and a leaf humidity sensor. The ultrasonic wind speed and direction sensor has an accuracy of ±0.1 m / s and monitors air flow data in the greenhouse in real time to optimize the start-stop strategy of the ventilation equipment. The dual-beam infrared carbon dioxide sensor has an accuracy of ±5 ppm and identifies carbon dioxide concentration fluctuation data during the vigorous photosynthesis period, such as the midday trough phenomenon, which can trigger the automatic operation of the air supplementing device. The leaf humidity sensor has a resolution of 0.1 μm and detects and collects leaf surface water film thickness data, which can be combined with temperature and humidity data to warn of high disease risk. The multi-spectral imaging sensor is a four-channel (450 nm, 550 nm, 650 nm, 850 nm) line array camera installed 2 meters above the planting area. It scans the plant spectrum every hour to collect plant spectral data, calculates the normalized vegetation index and photochemical reflectance index, quantifies the leaf chlorophyll content and photosystem II activity, and establishes a correlation model between spectral characteristics and medicinal ingredients. The ultraviolet intensity probe senses the ultraviolet radiation intensity in the wavelength range of 280-400 nm in real time and collects ultraviolet radiation data, focusing on analyzing the wavelength range of 280-320 nm to induce the synthesis of stillingia root flavonoids, which can be used to dynamically adjust the shading or light supplementing strategy.

[0027] The microbial data acquisition unit is used to analyze the abundance of dominant flora according to a portable fluorescent quantitative polymerase chain reaction instrument and to identify pest species and quantity through a sex attractant-image recognition composite sensor based on the YOLOv5 algorithm. Specifically, the portable fluorescent quantitative polymerase chain reaction instrument has a detection period of 30 minutes and automatically collects rhizosphere soil samples with a depth of 15 cm. It quantitatively analyzes the abundance of dominant flora through specific primers (such as actinomycete 16S rRNA gene primers) and constructs a dynamic coupling model of microbial community succession and environmental factors based on the data collected by the biomimetic soil probe to guide the precise application of biological agents (such as Bacillus subtilis). The sex attractant-image recognition composite sensor is deployed for major pests such as aphids and thrips. It attracts adult insects through a sex attractant trap and carries an artificial intelligence vision module with a resolution of 1920×1080. It identifies pest species and quantity in real time through the YOLOv5 algorithm. When the pest situation reaches level III (i.e., the density is greater than 10 per square meter), it automatically triggers physical control (such as increased insect-proof net) or biological control (such as releasing ladybugs) instructions.

[0028] In the embodiment of the present application, the data processing module 02 includes a data cleaning unit, an environment matching unit and a bias analysis unit. The data cleaning unit is used for identifying abnormal data in continuous data and periodic abnormal data in time series data by using a statistical algorithm for the original data of the cultivation environment, identifying abnormal data in high-dimensional data by using a machine learning algorithm, repairing single-point missing abnormal data by using a forward-backward filling method, repairing continuous missing abnormal data based on Kalman filtering method, repairing spatial inconsistent abnormal data by using Kriging interpolation method weighted calculation of missing values, removing noise data in high-frequency fluctuation data by using sliding average filtering method, removing noise data in low-frequency slowly varying data by using median filtering method, and obtaining the cultivation environment data. Specifically, for continuous data such as soil temperature and humidity, the quartile range method with a threshold coefficient of 1.5 is used to identify outliers, and the 3σ rule is used to filter sensor drift noise; for time series data such as 24-hour light curve, seasonal decomposition is performed to identify periodic abnormal data (such as sudden increase of midnight light); for high-dimensional data such as multi-spectral image features, isolated forest is used to isolate abnormal points by constructing binary tree, and forward-backward filling method is used to repair single-point missing abnormal data, such as soil pH value missing, the average value of the previous and next 10 minutes is used for filling; for continuous missing abnormal data, a dynamic prediction model is constructed based on Kalman filtering method, and the trend of abnormal data missing for more than 30 minutes is extrapolated and repaired, with an error control of ±1.2%; for spatial inconsistent abnormal data, Kriging interpolation method is used to calculate the missing values by weighting the data of the neighboring sensors within a radius of 5 meters and to repair them, and the spatial data consistency after repair can be improved to 98%.

[0029] Further, adaptive filtering technology is used to remove noise data, sliding average filtering method is used to reduce the influence of short-term weather disturbance and remove noise data for high-frequency fluctuation data such as light intensity and leaf humidity, median filtering method is used to remove impulse noise while retaining data trend for low-frequency slowly varying data such as soil conductivity value and microbial abundance, to remove noise data. In an optional embodiment, soil humidity data is converted into volumetric water content, light intensity data is converted into photosynthetically active photon flux density, data is scaled to the interval [0, 1] through data normalization, to unify the sensor output format and eliminate dimension differences, and finally the cultivation environment data is obtained.

[0030] In the embodiment of the present application, the environment matching unit is used for performing elastic matching on the cultivation environment data and the ideal cultivation environment data by dynamic time warping, calculating the bending path distance, constructing a multi-parameter dependency graph by Bayesian network inference, and inferring the environment suitability by conditional probability table, and finally obtaining the matching degree. Specifically, the database of ideal cultivation environment data contains 3 dimensions, which are divided according to growth stages. In one embodiment, the ideal cultivation environment data table is shown in Table 1.

[0031] Table 1 (ideal data table of cultivation environment) The dynamic time warping is used for flexible matching of the cultivation environment data and the ideal data of cultivation environment, a time axis offset range of ±2 hours is allowed, a curved path distance is calculated, a matching degree of the current environment and the ideal environment is determined, a matching degree greater than 90% is qualified, a multi-parameter dependency graph (for example, "soil humidity low + air temperature high"→"transpiration stress risk") is constructed through Bayesian network reasoning, an environment suitability degree is inferred through a conditional probability table, a risk value in an interval of [0, 1] is output (a risk value greater than 0.7 triggers a warning), and finally the matching degree is obtained.

[0032] In the embodiment of the present application, the bias analysis unit is configured to calculate a single-parameter absolute bias value of the cultivation environment data and the ideal data of cultivation environment, obtain a single-parameter relative bias value according to the single-parameter absolute bias value, and calculate the environment comprehensive bias value according to the single-parameter relative bias value and a cultivation environment parameter weight. Specifically, for each parameter, a single-parameter absolute bias value between the cultivation environment data and the ideal data of cultivation environment is calculated, and the single-parameter absolute bias value calculation formula is represented by the following formula (1): Formula (1); In the formula, represents the cultivation environment data, represents the ideal data of cultivation environment, and represents the single-parameter absolute bias value. For example, the soil temperature data is 14℃, the ideal soil temperature data is 16℃, and the single-parameter absolute bias value is . The single-parameter relative bias value is obtained according to the single-parameter absolute bias value, so as to more accurately reflect the relative degree of bias, and the single-parameter relative bias value calculation formula is represented by the following formula (2): Formula (2); In the formula, represents the single-parameter relative bias value, for example, the soil temperature data is 14℃, the ideal soil temperature data is 16℃, the single-parameter absolute bias value is , and the single-parameter relative bias value is The single-parameter relative bias value can eliminate the dimension influence, and is convenient for unified evaluation and comparison of the bias of different parameters.

[0033] Further, the environment comprehensive deviation value is calculated according to the single-parameter relative deviation value and the cultivation environment parameter weight. According to the sensitivity of the growth of Yuyang root to different environmental factors and the importance of each environmental parameter at the current growth stage, a corresponding weight is set for each parameter. For example, during the flowering period of Yuyang root, light intensity and temperature have a greater influence on flower bud differentiation and flower development, and their weights can be set to 0.3 and 0.25, respectively. The weight of soil moisture content is 0.2, and the weight of soil fertility is 0.15. The setting of the weight is based on the results of plant physiology research, the experience of agricultural experts, and the statistical analysis results of historical planting data. The environment comprehensive deviation value of the cultivation environment data and the ideal cultivation environment data is calculated according to the single-parameter relative deviation value and the cultivation environment parameter weight by using the weighted summation method. The calculation formula of the environment comprehensive deviation value is represented by the following formula (3): Formula (3); In the formula, represents the single-parameter relative deviation value of the i-th environmental parameter; represents the corresponding weight; represents the environment comprehensive deviation value. For example, at a certain time, the single-parameter relative deviation value of soil temperature is , the weight is ; the single-parameter relative deviation value of light intensity is , the weight is ; the single-parameter relative deviation value of soil moisture content is , the weight is ; the single-parameter relative deviation value of soil fertility is , the weight is ; the single-parameter relative deviation value of air humidity is , the weight is , and the comprehensive deviation value is . The larger the environment comprehensive deviation value D is, the more significant the difference between the current cultivation environment and the ideal cultivation environment is, and the more important the regulation and control is.

[0034] In optional embodiments, short-term trend analysis is performed on the change trend of the environmental comprehensive deviation value and the deviation of each main cultivation environment data in units of hours or days, and a deviation time series curve is drawn to observe whether the deviation is gradually decreasing, increasing, or remaining stable. For example, if the environmental comprehensive deviation value shows a gradually increasing trend over the past three days, it indicates that the current control measures have not effectively improved the planting environment, and the control strategy needs to be re-evaluated and adjusted. If the deviation shows a gradually decreasing trend, it indicates that the control measures are working, and the execution can continue to be adjusted according to the trend. From the perspective of weeks, months, and growth cycles, analyzing the long-term change trend of the deviation helps to understand the evolution law of the cultivation environment in the macro time scale, so as to evaluate the overall effect of environmental control in the entire growth cycle. For example, in a growth cycle of Huyanggeng, observe the change of the environmental comprehensive deviation value in different growth stages. If the deviation is large in the early stage but gradually tends to be stable and decreases, it indicates that the control system is continuously adapting and optimizing. If there is a large deviation for a long time and no obvious improvement trend, the entire cultivation system needs to be checked and optimized, including sensor accuracy, control equipment performance, database parameter rationality, etc.

[0035] In optional embodiments, the decision generation module 03 includes a state warning unit, a health index generation unit, and a strategy generation unit. The state warning unit is used to determine the parameter control priority according to the sensitivity and importance of the cultivation environment parameters in the current growth stage, determine the state warning signal level according to the relative deviation value of the single parameter, and determine the state warning signal according to the parameter control priority and the state warning signal level. Specifically, the parameter control priority is defined according to the sensitivity of Huyanggeng growth to different environmental factors and the importance of the cultivation environment parameters in the current growth stage, and the order of processing is "humidity > temperature > light > soil parameters > CO2 gas content", to avoid system shock caused by simultaneous control of multiple parameters. When multiple parameters deviate from the ideal value at the same time, the parameter with high priority is processed first. If the deviation value of a single parameter is very high, only the parameter is precisely controlled, and other low deviation value parameters are temporarily not intervened, to avoid system shock caused by over-adjustment of a single parameter. For example, if the humidity deviation value is greater than 20% and the temperature deviation value is less than 15%, the dehumidification or humidification equipment is started first. If the temperature deviation value is greater than 25% and other parameters are normal, the temperature control system is triggered immediately. The state warning signal level is determined according to the relative deviation value of the single parameter. The state warning signal level is yellow when the deviation value is within 10%-20%, and the state warning signal level is red when the deviation value is greater than 20%. Finally, the state warning signal is determined according to the parameter control priority and the state warning signal level. The state warning signal includes the parameter abnormal type and the recommended measures, such as heating or cooling, irrigation or ventilation, light supplement or shading.

[0036] In an optional embodiment, the health index generating unit is configured to identify the leaf health grade through transfer learning according to the convolutional neural network image classification model and the leaf image recognition result, normalize the six environmental indexes and the six soil nutrient indexes, assign weights to the twelve indexes through a random forest algorithm, linearly combine the normalized index values and the weight values, obtain the plant growth health index, and obtain the plant health state grade according to the leaf health grade and the plant growth health index. Specifically, the leaf image recognition result includes leaf color RGB value, disease spot area ratio, and stem thickness pixel value, the convolutional neural network image classification model is based on a ResNet-18 pre-training model, and the leaf health grade is identified through transfer learning according to the leaf image recognition result. The leaf health grade is divided into 1-5 levels, wherein level 1 is full wilting, and level 5 is fresh green. In addition, an RGB image of 512x512 pixels is input, and after data enhancement (i.e., rotation ±15°, brightness ±20%), training can accurately determine leaf diseases such as yellowing and wilting. The six environmental indexes and the six soil nutrient indexes are temperature, humidity, light, pH, conductivity, carbon dioxide, leaf chlorophyll relative content value, stem thickness, disease spot rate, root volume, nitrogen content, and potassium content. 300 wild Gynostemma pentaphyllum data are used as samples for model training, wherein the proportions of healthy samples, sub-healthy samples, and diseased samples are 5:3:2, and the 5-fold cross-validation accuracy rate reaches 96.2%. The twelve indexes are normalized, the weights of the twelve indexes are assigned through a random forest algorithm, the normalized index values and the weight values are linearly combined to obtain the plant growth health index, and the plant growth health index ranges from 0 to 100. Finally, the plant health state grade is obtained according to the leaf health grade and the plant growth health index. The plant health state grade is divided into three levels of excellent, medium, and poor. When the plant health state grade is excellent, the plant growth health index is greater than or equal to 85, the leaf color is fresh green, the relative content of leaf chlorophyll is greater than or equal to 45, and the root volume is greater than 150 cm 3 , there is no disease spot, and no intervention is required. When the plant health state grade is medium, the plant growth health index is greater than or equal to 60 and less than 85, the leaf color is light green, the relative content of leaf chlorophyll is between 35 and 45, the root volume is between 80 and 150 cm 3 , the disease spot area is less than 5%, and mild intervention (such as foliar fertilization and increased ventilation) is triggered. When the plant health state grade is poor, the plant growth health index is less than 60, the leaf is yellow or wilted, the relative content of leaf chlorophyll is less than 35, the root is brown, the disease spot area is greater than or equal to 5%, and emergency intervention such as disease and pest control and root rejuvenation treatment is started.

[0037] In an optional embodiment, the policy generation unit is configured to combine the environment comprehensive deviation value and the plant growth and health index, randomly select and execute an initial action from a basic control action combination according to a reinforcement learning policy optimization algorithm, observe the feedback of the plant growth and health index, retain high-reward actions by reward function evaluation, form a stable policy library and screen the candidate control strategies, check the candidate control strategies according to a multi-objective optimization algorithm, and determine the optimal control strategy according to the constraint conditions. Specifically, the input environment comprehensive deviation value, plant growth and health index, seasonal cycle (spring, summer, autumn, winter), past 7-day control records (instruction type, execution time, environment response curve), available resources (water and fertilizer reserves, power load), and manual priority instructions (such as manual intervention in extreme weather) are used to select the optimal control action with high plant growth and health index improvement and low resource consumption from the basic control action combination by the reinforcement learning policy optimization algorithm, and the control action for high-priority parameters (such as humidity and temperature) is preferentially retained to ensure that when multiple parameters deviate, the parameters more sensitive to the growth of Yunchen are preferentially processed. The process is refined into state space mapping, action selection and execution, reward function evaluation, and policy iteration optimization: first, the input multi-dimensional variables are converted into state vectors recognizable by the algorithm (such as “summer + temperature 28°C + humidity 60% + plant growth and health index 80 + water and fertilizer sufficient”), then an initial action (such as “cooling 2°C + humidifying 5% + light supplementing 1 hour”) is randomly selected from the basic control action combination and executed, the feedback of the plant growth and health index improvement and resource consumption is observed, the high-reward actions are retained and the low-reward actions are eliminated by reward function evaluation, and a stable policy library is gradually formed, and finally 1-2 groups of candidate control strategies with “high plant growth and health index improvement (greater than or equal to 5%), high environment compliance rate, and low resource consumption” are screened out. The candidate control strategies are checked according to the multi-objective optimization algorithm to ensure that the constraint conditions are met and the three objectives of “maximizing plant growth and health index growth, minimizing resource consumption, and minimizing environmental fluctuations” are balanced, and finally the optimal control strategy is determined. The two algorithms are applied cooperatively to quickly screen efficient action combinations through the reward mechanism to generate candidate control strategies, and the multi-objective optimization algorithm optimizes the balance through constraint checking and multi-objective balancing, and finally outputs the optimal control strategy with high feasibility and high cost performance, which provides a core basis for precise instruction generation.

[0038] In optional embodiments, the decision execution module 04 is configured to control the irrigation and fertilization control devices, the light and temperature and humidity control devices, and the ventilation and gas regulation devices in coordination according to the optimal control strategy, and to monitor and adjust the operation state of the execution devices in real time. Specifically, the intelligent irrigation device cluster is controlled according to the optimal control strategy, a layered irrigation method is adopted, each drip irrigation unit adopts a pressure-compensating dripper with a flow accuracy of ±5%, is arranged at an interval of 30 cm between planting ridges, and in combination with soil humidity data, precise water replenishment of the root development layer is achieved (single irrigation amount error ≤±3%), each sprinkler unit is equipped with a rotary atomizing nozzle with an atomizing particle size of 50-100 μm, which is used for leaf surface water replenishment and cooling, and is automatically started when the air humidity is <50% or the canopy temperature is >28°C, and the single operation time is 5-15 minutes. In addition, the crop water requirement is calculated based on a soil water potential model, and the irrigation plan is dynamically adjusted in combination with weather forecast data (i.e., 24-hour precipitation probability in the future), which can save more than 30% of water compared with traditional fixed-time irrigation. The intelligent fertilizer machine is controlled according to the optimal control strategy, the nitrogen, phosphorus and potassium three-channel fertilizer pump is integrated in the intelligent fertilizer machine, the flow range is 0.1-5 L / min, the accuracy is ±2%, and water-soluble fertilizers and biological agents (such as actinomycete agents) are supported for quantitative injection. In addition, the fertilization amount is adjusted by using a proportional-integral-derivative control algorithm according to the effective phosphorus and available nitrogen data and the growth stage threshold value, for example, when the effective phosphorus during flowering is lower than 20 mg / kg, a phosphorus fertilizer solution is injected at a flow rate of 0.5 L / min until the standard is met.

[0039] In an optional embodiment, the intelligent light adjustment device is regulated according to the optimal regulation strategy to realize the light supplement-shading linkage regulation, wherein the light emitting diode (LED) light supplement lamp adopts full-spectrum LED with light efficiency ≥ 100 lm / W, the coverage area of a single lamp is 5 m2, and 0-100% power adjustment is supported; when the light intensity during flowering period is insufficient (i.e. less than 2500 Lux), the LED light supplement lamp is automatically turned on and adjusted to the target light intensity (3000-4000 Lux), so as to realize the improvement of light supplement efficiency; when the light intensity is greater than 5000 Lux and the temperature is higher than 28℃, the electric sunshade net is automatically turned on in linkage with the ventilation equipment to avoid strong light from burning the leaves. In addition, according to the photoperiod requirement of different growth stages of the herb (such as 12 hours of light or 12 hours of darkness during the seedling stage), the timer is linked with the light sensor to ensure that the photoperiod accuracy is within 5 minutes, so as to realize the precise photoperiod control. The temperature and humidity regulation equipment is regulated according to the optimal regulation strategy, the intelligent temperature control unit includes a heat pump type cooling and heating machine and an ultrasonic humidifier, the heat pump type cooling and heating machine has a refrigerating capacity of 5-20 kW and a heating capacity of 3-15 kW, supports the precise temperature control in the greenhouse, and the precision is ±0.5℃; when the detected temperature deviates from the target threshold, the output power is adjusted within 10 seconds according to the proportional-integral-derivative control algorithm; the ultrasonic humidifier has an atomization capacity of 1-5 kg / h, and the air humidity sensor is combined to start when the humidity is less than 60%, and the control precision of the target humidity is ±3% relative humidity. At the same time, the paraffin-based phase change material with a melting point of 22℃ is embedded in the wall of the greenhouse to assist in smoothing the day and night temperature difference and reducing the frequency of starting and stopping the equipment.

[0040] In an optional embodiment, the ventilation and gas regulation equipment is regulated according to the optimal regulation strategy, the intelligent ventilation equipment is a variable frequency fan, the variable frequency fan adopts an axial flow fan, the air volume is 500-5000 m 3 / h, the air speed sensor is matched, the speed is automatically adjusted according to the carbon dioxide concentration (target 400-600 ppm) and the air flow rate (target 0.3-0.5 m / s), and the ventilation energy consumption can be reduced. The ventilation is realized through the top window-side window linkage equipment combined with the data of the electric push rod type window opener and the wind direction sensor; when the outdoor temperature is less than 20℃ and the carbon dioxide concentration is greater than 600 ppm, the top window is preferentially opened to exhaust the hot and humid air; when the outdoor temperature is greater than 28℃, the top window and the side window are simultaneously opened to form convection. The carbon dioxide supplement equipment is a steel bottle type carbon dioxide generator matched with a mass flow meter; when the detected carbon dioxide concentration is less than 400 ppm and the light intensity is greater than 2000 Lux (i.e. during the active period of photosynthesis), the carbon dioxide is supplemented at a rate of 500 ppm / h until the target concentration is reached, so as to realize the improvement of the supplement efficiency.

[0041] In optional embodiments, cultivation facility regulation is achieved through the intelligent planting bed device. The lifting planting bed uses a motorized lead screw to complete lifting, with a lifting stroke of 0-50 cm and an accuracy of ±1 mm. The bed level is automatically adjusted according to the terrain slope to ensure smooth drainage. When the slope is greater than 5°, the tilt compensation is started. The substrate circulation device integrates a substrate mixer and a conveyor belt. When the soil conductivity is greater than 1.8 mS / cm (i.e., the salt content is too high), the top 5 cm of substrate is automatically replaced, combined with the leaching program to reduce salt accumulation. The efficiency is greatly improved compared to manual operation. Drainage and seepage prevention are completed through an intelligent drainage valve. The intelligent drainage valve can be a 50-100 mm electromagnetic drainage valve. It is connected to a soil moisture sensor. When the soil moisture at a depth of 30 cm is greater than 60%, the valve is automatically opened. The drainage rate is dynamically adjusted according to the water depth. The seepage prevention membrane with a thickness of 0.5 mm is laid. The edge is equipped with a pressure sensor. When the water pressure under the membrane is detected to be greater than 5 kPa, the drainage valve is triggered to automatically pump water to prevent root waterlogging.

[0042] In optional embodiments, multi-device coordinated regulation is achieved through a multi-device linkage strategy. For example, in the comprehensive regulation of flowering period, when the data processing module 02 outputs "the comprehensive deviation value of the flowering period environment is 0.7", the decision execution module 04 starts the following linkage: the light supplement lamp is turned on at 80% power, the target light intensity is 3500 Lux, the fan speed is increased to 80%, the air volume is 4000 m 3 / h, the top window is opened at an angle of 60°, the drip irrigation device irrigates at a flow rate of 1.5 L / min for 10 minutes until the target soil moisture is 55%, and the carbon dioxide generator supplements air at a rate of 300 ppm / h to 500 ppm. The entire linkage process is completed within 3 minutes, with an efficiency improvement of 80% compared to manual intervention. In addition, while the devices are coordinated, the device status is monitored in real time. Each execution device is equipped with a current sensor and a travel switch, which report the running status in real time, such as "irrigation pump running", "sunshade net closed", etc. The status update frequency is once every second. Each regulation device follows a fault self-healing mechanism. When it is determined to be a first-level fault, such as motor overload, the device automatically stops and attempts to restart. If it fails more than three times, it reports to the decision generation module 03 to generate a maintenance work order. When it is determined to be a second-level fault, such as valve jamming, it switches to a backup device and triggers a manual maintenance reminder. The manual maintenance reminder can be pushed through the management software.

[0043] In summary, the wild-to-cultivated plant cultivation system of Gynostemma pentaphyllum provided by the embodiments of the present application collects the original data of the cultivation environment of Gynostemma pentaphyllum through the data collection module, the original data of the cultivation environment including soil environment data, meteorological environment data and microbial data, collecting all kinds of environment data in full dimension in real time to provide data support for precise regulation and control and realizing early warning of insect pests; the data processing module repairs and removes abnormal data and noise data in the original data of the cultivation environment, improving the reliability of data repair, obtaining the cultivation environment data, judging the matching degree of the cultivation environment data and ideal cultivation environment data through dynamic time warping and Bayesian network inference, quantifying the dynamic environment matching degree, facilitating accurate identification of environmental deviation, and calculating the single-parameter relative deviation value and the environmental comprehensive deviation value of the cultivation environment data and the ideal cultivation environment data, weighting the environmental comprehensive deviation value according to the parameter sensitivity to clarify the priority of regulation and control and avoid system shock caused by simultaneous adjustment of multiple parameters; in the decision generation module, the state warning signal is obtained according to the single-parameter relative deviation value and the parameter regulation priority, the leaf health grade, the plant growth health index and the plant health state grade are identified according to the convolutional neural network image classification model, the random forest algorithm and the plant physiological data, the dynamic evaluation of the plant growth state is realized, the candidate regulation strategy is obtained according to the reinforcement learning strategy optimization algorithm combined with the environmental comprehensive deviation value and the plant growth health index, the high-cost-effective regulation action is screened through the reward function, the efficiency of artificial decision is improved by 80%, the candidate regulation strategy is verified according to the multi-objective optimization algorithm to ensure the synergy of the three targets of "maximizing the plant growth health index, minimizing the resources and minimizing the environmental fluctuations", and the optimal regulation strategy is determined; in the decision execution module, the irrigation and fertilization control equipment, the light and temperature and humidity regulation equipment and the ventilation and gas regulation equipment are cooperatively regulated according to the optimal regulation strategy, the running state of the execution equipment is monitored and adjusted in real time, millisecond-level linkage response of the equipment is realized, multiple equipment cooperatively completes the regulation within 3 minutes, the efficiency of artificial response is improved by 80%, and accurate resource control is realized to ensure zero waste of resources and strengthen the stability of the cultivation environment.

[0044] The above is an introduction to the system embodiments of the present application. Based on the foregoing embodiments, the method embodiments of the present application are introduced below.

[0045] Please refer to Figure 2 which shows a flowchart of a wild-to-cultivated plant cultivation method of Gynostemma pentaphyllum provided by the embodiments of the present application. The method is applied to the wild-to-cultivated plant cultivation system of Gynostemma pentaphyllum as shown in Figure 1 For details not disclosed in the method embodiments, please refer to the system embodiments. The system includes a data collection module, a data processing module, a decision generation module and a decision execution module, which are connected in sequence. As shown in Figure 2As shown, the method comprises the following steps S210 to S240.

[0046] In step S210, original data of a cultivation environment of the plant is collected, which includes soil environment data, meteorological environment data and microbial data.

[0047] In the embodiment, soil temperature data and soil humidity data are collected according to a capacitive soil temperature and humidity sensor, soil microbial activity data and rhizosphere microenvironment data are collected through a bionic soil probe, soil fertility data and soil electrochemical data are collected through a spectral soil fertility sensor and a pH-conductivity compound sensor, and a soil nutrient spatial distribution map is generated through a space-time interpolation algorithm; air flow data are collected according to an ultrasonic wind speed and direction sensor, carbon dioxide concentration data are identified through a dual-beam infrared carbon dioxide sensor, leaf surface water film thickness is collected through a leaf humidity sensor, plant spectral data are collected through a multispectral imaging sensor, and ultraviolet radiation data are collected through an ultraviolet intensity probe; the abundance of dominant flora is analyzed according to a portable fluorescent quantitative polymerase chain reaction instrument, and pest species and quantity are identified according to a YOLOv5 algorithm through a sex attractant-image recognition compound sensor.

[0048] In step S220, abnormal data and noise data in the original data of the cultivation environment are repaired and removed to obtain cultivation environment data, the matching degree of the cultivation environment data and ideal cultivation environment data is judged through dynamic time warping and Bayesian network inference, and the single-parameter relative deviation value and the environmental comprehensive deviation value of the cultivation environment data and the ideal cultivation environment data are calculated.

[0049] In the embodiment, for the original data of the cultivation environment, statistical algorithms are used to identify abnormal data in continuous data and periodic abnormal data in time series data, machine learning algorithms are used to identify abnormal data in high-dimensional data, single-point missing abnormal data are repaired through a forward-backward filling method, continuous missing abnormal data are repaired based on a Kalman filter method, missing values are calculated by weighted calculation through a Kriging interpolation method to repair spatial inconsistent abnormal data, noise data in high-frequency fluctuation data are removed through a moving average filter method, noise data in low-frequency slowly varying data are removed through a median filter method, and the cultivation environment data are obtained; for the cultivation environment data and the ideal cultivation environment data, elastic matching is performed through dynamic time warping, the bending path distance is calculated, a multi-parameter dependency graph is constructed through Bayesian network inference, the environmental suitability is inferred through a conditional probability table, and finally the matching degree is obtained; the single-parameter absolute deviation value of the cultivation environment data and the ideal cultivation environment data is calculated, the single-parameter relative deviation value is obtained according to the single-parameter absolute deviation value, and the environmental comprehensive deviation value is calculated according to the single-parameter relative deviation value and the cultivation environment parameter weight.

[0050] Step S230, obtaining a state warning signal according to the single parameter relative deviation value and the parameter regulation priority, identifying a leaf health grade, a plant growth health index and a plant health state grade according to a convolutional neural network image classification model, a random forest algorithm and plant physiological data, combining the environmental comprehensive deviation value and the plant growth health index, obtaining a candidate regulation strategy according to a reinforcement learning strategy optimization algorithm, and checking the candidate regulation strategy according to a multi-objective optimization algorithm to determine an optimal regulation strategy.

[0051] In the embodiments of the present application, the parameter regulation priority is determined according to the sensitivity and importance of the current growth stage cultivation environment parameters, the state warning signal grade is determined according to the single parameter relative deviation value, and the state warning signal is determined according to the parameter regulation priority and the state warning signal grade; the leaf health grade is identified through transfer learning according to a convolutional neural network image classification model and a leaf image identification result, six environmental indicators and six soil nutrient indicators are normalized, a random forest algorithm is used to assign weights to twelve indicators, the normalized indicator values and the weight values are linearly combined to obtain the plant growth health index, and the plant health state grade is obtained according to the leaf health grade and the plant growth health index; the environmental comprehensive deviation value and the plant growth health index are combined, an initial action is randomly selected and executed from a basic regulation action combination according to a reinforcement learning strategy optimization algorithm, feedback of the plant growth health index is observed, high-reward actions are reserved through a reward function evaluation, a stable strategy library is formed and the candidate regulation strategy is screened, the candidate regulation strategy is checked according to a multi-objective optimization algorithm, and the optimal regulation strategy is determined according to a constraint condition.

[0052] Step S240, according to the optimal regulation strategy, the irrigation and fertilization control equipment, the light and temperature and humidity regulation equipment and the ventilation and gas regulation equipment are cooperatively regulated, and the operation state of the execution equipment is monitored and adjusted in real time.

[0053] In the embodiments of the present application, the crop water requirement is calculated based on a soil water potential model, the irrigation plan is dynamically adjusted in combination with weather forecast data, and the fertilization amount is adjusted using a proportional-integral-derivative control algorithm according to soil fertility sensor data and growth stage threshold values.

[0054] In summary, the wild to domestic cultivation method of the Campanumoea javanica provided by the embodiment of the present application collects the original data of the cultivation environment of the Campanumoea javanica, the original data of the cultivation environment including soil environment data, meteorological environment data and microbial data, all kinds of environment data are collected in full dimension and in real time, data support is provided for precise regulation and control, and early warning of insect pests is realized; the abnormal data and noise data in the original data of the cultivation environment are repaired and removed, the reliability of data repair is improved, the data of the cultivation environment are obtained, the matching degree of the data of the cultivation environment and ideal data of the cultivation environment is judged through dynamic time warping and Bayesian network inference, the dynamic environment matching degree is quantified, the environment deviation is accurately identified, the single parameter relative deviation value and the environment comprehensive deviation value of the data of the cultivation environment and the ideal data of the cultivation environment are calculated, the environment comprehensive deviation value is calculated according to the parameter sensitivity weighting, the priority of regulation and control is clear, and system shock caused by simultaneous adjustment of multiple parameters is avoided; the state warning signal is obtained according to the single parameter relative deviation value and the parameter regulation and control priority, the leaf health grade, the plant growth health index and the plant health state grade are identified according to the convolutional neural network image classification model, the random forest algorithm and the plant physiological data, the dynamic evaluation of the plant growth state is realized, the candidate regulation and control strategy is obtained according to the reinforcement learning strategy optimization algorithm combined with the environment comprehensive deviation value and the plant growth health index, the high cost performance regulation and control action is screened through the reward function, the efficiency of artificial decision is improved by 80%, the candidate regulation and control strategy is verified according to the multi-objective optimization algorithm, the three targets of "maximizing the plant growth health index, minimizing the resources and minimizing the environment fluctuation" are coordinated, and the optimal regulation and control strategy is determined; the irrigation and fertilization control equipment, the light and temperature and humidity regulation and control equipment and the ventilation and gas regulation and control equipment are cooperatively regulated and controlled according to the optimal regulation and control strategy, the running state of the execution equipment is monitored and adjusted in real time, millisecond level linkage response of the equipment is realized, multiple equipment cooperates to complete the regulation and control within 3 minutes, the efficiency of artificial response is improved by 80%, accurate resource control is realized, resource waste is avoided, and the stability of the cultivation environment is strengthened.

[0055] It should be noted that in the embodiments of the present application, if the above-mentioned wild to domestic cultivation method of the Campanumoea javanica based on ecological simulation is realized in the form of a software function module and sold or used as an independent product, it can also be stored in a computer readable storage medium. Based on this understanding, the technical solutions of the embodiments of the present application can be embodied in the form of a software product, which is stored in a storage medium and includes a plurality of instructions for causing an electronic device to execute all or part of the methods described in the embodiments of the present application. The aforementioned storage medium includes a U disk, a mobile hard disk, a read-only memory (ROM), a magnetic disk or an optical disk and various program code storage media. Thus, the embodiments of the present application are not limited to any specific hardware and software combination.

[0056] Correspondingly, the embodiment of the present application provides a computer readable storage medium, which stores a computer program, and the computer program is executed by a processor to implement the steps of any one of the wild to domestic cultivation methods of Gynostemma pentaphyllum based on ecological simulation.

[0057] Based on the same technical concept, the embodiment of the present application provides an electronic device for implementing the wild to domestic cultivation method of Gynostemma pentaphyllum based on ecological simulation. Figure 3 The hardware entity diagram of the electronic device provided by the embodiment of the present application is shown in FIG. 3. Figure 3 As shown in FIG. 3, the electronic device 300 includes a memory 310 and a processor 320, the memory 310 stores a computer program executable on the processor 320, and the processor 320 implements the steps of any one of the wild to domestic cultivation methods of Gynostemma pentaphyllum based on ecological simulation when executing the program.

[0058] The memory 310 is configured to store instructions and applications executable by the processor 320, and can also cache data (for example, image data, audio data, voice communication data and video communication data) to be processed by the processor 320 and each module in the electronic device. The memory 310 can be implemented by a FLASH or a Random Access Memory (RAM).

[0059] The processor 320 implements the steps of any one of the wild to domestic cultivation methods of Gynostemma pentaphyllum based on ecological simulation when executing the program. The processor 320 generally controls the overall operation of the electronic device 300.

[0060] The processor can be at least one of an Application Specific Integrated Circuit (ASIC), a Digital Signal Processor (DSP), a Digital Signal Processing Device (DSPD), a Programmable Logic Device (PLD), a Field Programmable Gate Array (FPGA), a Central Processing Unit (CPU), a controller, a microcontroller, or a microprocessor. It can be understood that the electronic device for implementing the functions of the processor can also be other devices, and the embodiments of the present application are not limited in this regard.

[0061] The computer storage medium / memory can be a Read Only Memory (ROM), a Programmable Read-Only Memory (PROM), an Erasable Programmable Read-Only Memory (EPROM), an Electrically Erasable Programmable Read-Only Memory (EEPROM), a Ferromagnetic Random Access Memory (FRAM), a Flash Memory, a magnetic surface memory, an optical disc, a Compact Disc Read-Only Memory (CD-ROM), or the like. It can also be various electronic devices including one or any combination of the above memories, such as a mobile phone, a computer, a tablet device, a personal digital assistant, and the like.

[0062] It should be noted that the above description of the storage medium and device embodiments is similar to the description of the above method embodiments, and has similar beneficial effects as the method embodiments. For technical details not disclosed in the storage medium and device embodiments of the present application, please refer to the description of the method embodiments for understanding.

[0063] It should be understood that every feature, structure, or characteristic described herein is within a preferred embodiment of the present application. Thus, it is meant that the features, structures, or characteristics can be combined with each other in any manner within a preferred embodiment of the present application. In addition, it is contemplated that each feature, structure, or characteristic can be implemented in hardware, software, or a combination thereof.

[0064] It should be noted that, as used herein, the articles "a", "an", "the", and "at least one" are intended to mean that there is one or more of the elements in the preceding descriptions. The articles "a" (or "an"), as well as the first article "the" and "at least one" are intended to be interpreted as including both the singular and the plural, unless otherwise indicated. Thus, for example, "a" and "the" can mean one or more than one.

[0065] In several embodiments provided in the present application, it should be understood that the disclosed devices and methods can be implemented in other ways. The above-described device embodiments are merely illustrative, for example, the division of the units is only a logical function division, and actual implementation can have another division manner, such as: multiple units or components can be combined, or can be integrated into another system, or some features can be ignored or not executed. In addition, the coupling or direct coupling or communication connection between the displayed or discussed components can be through some interfaces, indirect coupling or communication connection between devices or units, which can be electrical, mechanical or other forms.

[0066] The units described above as separate components can or can not be physically separate, and the components displayed as units can or can not be physical units; they can be located in one place or distributed on multiple network units; and some or all of the units can be selected according to actual needs to achieve the purpose of the embodiments of the present application.

[0067] In addition, each functional unit in each embodiment of the present application can be integrated into one processing unit, or each unit can be a separate unit, or two or more units can be integrated into one unit; the integrated unit can be realized in the form of hardware or hardware plus software functional unit.

[0068] Alternatively, the above-mentioned integrated units of the present application, if implemented in the form of software function modules and sold or used as independent products, can also be stored in a computer readable storage medium. Based on such understanding, the technical solutions of the embodiments of the present application can be embodied in the form of a software product, and the computer software product is stored in a storage medium, and includes several instructions to enable the equipment automatic test line to perform all or part of the methods described in the embodiments of the present application. The foregoing storage medium includes: mobile storage devices, ROM, magnetic disks or optical disks and various media that can store program codes.

[0069] The methods disclosed in the several method embodiments provided by the present application can be combined arbitrarily without conflict to obtain new method embodiments.

[0070] The features disclosed in the several method or device embodiments provided by the present application can be combined arbitrarily without conflict to obtain new method embodiments or device embodiments.

[0071] The above is only an implementation of the present application, but the protection scope of the present application is not limited thereto, and any person skilled in the art can easily think of changes or replacements within the technical scope disclosed by the present application, which should be covered within the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.

Claims

1. A wild-to-domestic cultivation system for Panax notoginseng based on ecological simulation, characterized in that: The system comprises: A data acquisition module is used to collect the original data of the cultivation environment of the ginseng, wherein the original data of the cultivation environment includes soil environment data, meteorological environment data and microbial data; a data processing module, configured to repair and remove abnormal data and noise data in the raw cultivation environment data to obtain cultivation environment data, determine the degree of matching between the cultivation environment data and the ideal cultivation environment data through dynamic time warping and Bayesian network reasoning, and calculate a single parameter relative deviation value and an environmental comprehensive deviation value between the cultivation environment data and the ideal cultivation environment data; a decision generation module, configured to obtain a status warning signal based on the single parameter relative deviation value and the parameter control priority, identify the leaf health level, the plant growth health index, and the plant health status level based on a convolutional neural network image classification model, a random forest algorithm, and plant physiological data, obtain a candidate control strategy based on the environmental comprehensive deviation value and the plant growth health index based on a reinforcement learning strategy optimization algorithm, verify the candidate control strategy based on a multi-objective optimization algorithm, and determine the optimal control strategy; The decision-making execution module is used to coordinately control the irrigation and fertilization control equipment, the lighting and temperature and humidity control equipment, and the ventilation and gas regulation equipment according to the optimal control strategy, and monitor and adjust the operating status of the execution equipment in real time.

2. The system according to claim 1, wherein: The data acquisition module includes a soil data acquisition unit, a meteorological data acquisition unit and a microbial data acquisition unit, wherein: The soil data acquisition unit is used to collect soil temperature data and soil moisture data using a capacitive soil temperature and humidity sensor, remove the influence of soil thermal conductivity using a temperature compensation algorithm, collect soil microbial activity data and rhizosphere microenvironment data using a bionic soil probe, collect soil fertility data and soil electrochemical data using a spectral soil fertility sensor and a pH-conductivity composite sensor, and generate a soil nutrient spatial distribution map using a spatiotemporal interpolation algorithm; The meteorological data acquisition unit is used to collect air flow data using an ultrasonic wind speed and direction sensor, identify carbon dioxide concentration data using a dual-beam infrared carbon dioxide sensor, collect water film thickness on leaf surfaces using a leaf humidity sensor, collect plant spectrum data using a multispectral imaging sensor, and collect ultraviolet radiation data using an ultraviolet intensity probe; The microbial data acquisition unit is used to analyze the abundance of dominant bacterial communities based on a portable fluorescent quantitative polymerase chain reaction instrument, and to identify the types and quantities of pests based on the YOLOv5 algorithm using a sex attractant-image recognition composite sensor.

3. The system according to claim 1, wherein: The data processing module includes a data cleaning unit, an environment matching unit and a deviation analysis unit, wherein: The data cleaning unit is used to use a statistical algorithm to identify abnormal data in continuous data and periodic abnormal data in time series data, a machine learning algorithm to identify abnormal data in high-dimensional data, a forward and backward filling method to repair single-point missing abnormal data, a Kalman filter method to repair continuous missing abnormal data, a Kriging interpolation method to weightedly calculate missing values, repair spatially inconsistent abnormal data, a sliding average filter method to remove noise data in high-frequency fluctuation data, and a median filter method to remove noise data in low-frequency slowly varying data, so as to obtain the cultivation environment data; The environment matching unit is used to perform elastic matching between the cultivation environment data and the ideal cultivation environment data through dynamic time warping, calculate the curved path distance, construct a multi-parameter dependency graph through Bayesian network reasoning, and infer the environmental suitability through a conditional probability table to finally obtain the matching degree; The deviation analysis unit is used to calculate the single parameter absolute deviation value between the cultivation environment data and the ideal cultivation environment data, obtain the single parameter relative deviation value based on the single parameter absolute deviation value, and calculate the comprehensive environmental deviation value based on the single parameter relative deviation value and the cultivation environment parameter weight.

4. The system according to claim 1, wherein: The decision generation module includes a status warning unit, a health index generation unit and a strategy generation unit, wherein: The state warning unit is used to determine the parameter control priority according to the sensitivity and importance of the cultivation environment parameter in the current growth stage, determine the state warning signal level according to the single parameter relative deviation value, and determine the state warning signal according to the parameter control priority and the state warning signal level; The health index generating unit is configured to identify the leaf health level through transfer learning based on a convolutional neural network image classification model and leaf image recognition results, normalize six environmental indicators and six soil nutrient indicators, assign weights to the twelve indicators through a random forest algorithm, linearly combine the normalized indicator values ​​with the weight values ​​to obtain the plant growth health index, and obtain the plant health status level based on the leaf health level and the plant growth health index; The strategy generation unit is used to combine the comprehensive environmental deviation value and the plant growth health index, randomly select and execute an initial action from the basic control action combination according to the reinforcement learning strategy optimization algorithm, observe the feedback of the plant growth health index, retain high-reward actions through reward function evaluation, form a stable strategy library and screen the candidate control strategies, verify the candidate control strategies according to the multi-objective optimization algorithm, and determine the optimal control strategy according to the constraint conditions.

5. The system according to claim 1, wherein: The decision execution module is also used to calculate crop water requirements based on the soil water potential model, dynamically adjust irrigation plans based on weather forecast data, and adjust fertilizer application using a proportional-integral-differential control algorithm based on soil fertility sensor data and growth stage thresholds.

6. A method for converting wild ginseng to domestic cultivation based on ecological simulation, characterized in that: The method is applied to the wild-to-domestic cultivation system of Huanyang ginseng, which includes a data acquisition module, a data processing module, a decision generation module, and a decision execution module. The method includes: Collecting original data of the cultivation environment of the ginseng, wherein the original data of the cultivation environment includes soil environment data, meteorological environment data and microbial data; Repairing and removing abnormal data and noise data in the original cultivation environment data to obtain cultivation environment data, determining the matching degree between the cultivation environment data and the ideal cultivation environment data through dynamic time warping and Bayesian network reasoning, and calculating the single parameter relative deviation value and the environmental comprehensive deviation value between the cultivation environment data and the ideal cultivation environment data; Obtaining a status warning signal based on the single parameter relative deviation value and the parameter control priority, identifying the leaf health level, plant growth health index, and plant health status level based on a convolutional neural network image classification model, a random forest algorithm, and plant physiological data, combining the environmental comprehensive deviation value and the plant growth health index, obtaining a candidate control strategy based on a reinforcement learning strategy optimization algorithm, verifying the candidate control strategy based on a multi-objective optimization algorithm, and determining an optimal control strategy; According to the optimal control strategy, the irrigation and fertilization control equipment, the light and temperature and humidity control equipment, and the ventilation and gas regulation equipment are coordinated and controlled, and the operating status of the execution equipment is monitored and adjusted in real time.

7. An electronic device comprising a memory and a processor, wherein the memory stores a computer program that can be run on the processor, characterized in that: When the processor executes the program, the steps in the method according to claim 6 are implemented.

8. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to claim 6 are implemented.

Images