Modularized pepper planting agriculture and matched model parameter association method

Abstract

The invention discloses a method for associating modularized pepper planting agriculture and matched model parameters, and belongs to the field of agricultural informatization. The method comprises the steps that firstly, four agricultural modules of seedling raising, transplanting, plant protection and harvesting are divided according to the growth period of processed peppers, agricultural parameters of all the modules are collected through a sensor array, corresponding agricultural machine parameters and threshold values are input in a standardized mode, and two types of parameter sub-libraries are constructed; constructing a correlation model based on a Bayesian network, introducing an environmental factor as a hidden variable, reversely deducing an optimal agricultural machine parameter in combination with a random forest model, and generating an agricultural technology, environment, machine type and probability correlation table; and finally, correcting the model by using measured data through a gradient descent method in each quarter to realize database iteration. Through modularization, machine learning self-optimization and closed-loop iteration, the limitation of a traditional static database is broken through, cross-regional accurate adaptation is achieved, the agricultural machinery efficiency and the pepper quality are improved, and practicability and expansibility are high.

Description

Technical Field

[0001] This invention relates to the field of agricultural planting technology, and in particular to a method for linking modular chili pepper planting agronomy with the parameters of supporting machinery. Background Technology

[0002] As an important economic crop in my country, the scale of chili processing and its processing industry have been expanding year by year. The compatibility of agronomic plans and agricultural machinery directly affects the yield, quality and planting benefits of chili. At present, in the field of chili processing, the formulation of agronomic plans and the selection of agricultural machinery rely heavily on the experience and judgment of growers or technicians, lacking a systematic and data-driven support system. This leads to frequent problems such as low efficiency of agricultural machinery operation and large fluctuations in chili quality.

[0003] While some databases exist for storing agronomic or agricultural machinery-related data, these databases generally suffer from static and singular limitations. Most databases only store agricultural machinery parameters or agronomic indicators in isolation, failing to establish the correlation between environmental parameters, agronomic requirements, and machine parameters, and even less able to achieve dynamic adaptation among the three. In actual planting scenarios, the environmental requirements for processing chili peppers vary significantly at different growth stages, and environmental factors such as soil moisture, light intensity, and rainfall are constantly changing. Traditional static databases cannot capture these real-time changes and adjust agricultural machinery parameter recommendation strategies, resulting in a disconnect between agricultural machinery operation parameters and actual agronomic requirements and environmental conditions.

[0004] Meanwhile, my country's chili processing production areas are widely distributed, and there are significant differences in soil texture, climate conditions, and main varieties in different production areas. Traditional databases lack adaptive adjustment capabilities, and existing technologies have not introduced effective machine learning and self-updating mechanisms, making it impossible to optimize parameter matching logic based on measured data from different production areas, resulting in extremely poor adaptability when applied across regions.

[0005] Furthermore, the cultivation of processing chili peppers is a complex process involving multiple stages, including seedling raising, transplanting, plant protection, and harvesting. The agronomic requirements of each stage are interconnected and may change in combination as planting goals are adjusted. Traditional databases adopt an integrated data storage model without modularizing the agronomic stages or designing corresponding parameter sub-databases and linkage interfaces. When the combination of agronomic stages is adjusted, it is difficult to quickly retrieve and match the corresponding agricultural machinery parameters. Not only is the parameter matching efficiency low, but parameter mismatch between stages is also very likely to occur, which cannot meet the needs of flexible and ever-changing planting scenarios.

[0006] Finally, once traditional databases are built, they are in a relatively fixed state and lack a mechanism for continuous iteration and optimization. With the research and development and promotion of new agricultural machinery and the upgrading and updating of agronomic technologies, existing databases cannot incorporate new machine parameters and agronomic indicators in a timely manner, resulting in a gradual narrowing of their compatibility and a limited usage cycle, making it difficult to achieve long-term compatibility and value continuation of technical solutions.

[0007] Therefore, developing a technical solution that can achieve dynamic correlation between environment, agronomy, and machine type, and has cross-regional adaptability, flexible response capability, and long-term iteration capability has become a key requirement for solving the current problem of agronomy and agricultural machinery adaptation in the cultivation of processed chili peppers. Summary of the Invention

[0008] One objective of this invention is to propose a method for linking modular chili pepper cultivation agronomy with the parameters of supporting machinery. This invention can achieve dynamic and precise matching of environmental, agronomic, and machinery parameters in chili pepper cultivation, construct a three-element dynamic association database of environment, agronomy, and machinery to adapt to complex field environments, improve cross-regional adaptability through machine learning and self-updating modules, improve parameter matching efficiency by relying on modular decomposition and sub-database linkage, and achieve long-term expansion and optimization of technical solutions through a closed-loop iterative system.

[0009] A method for associating modular chili pepper cultivation agronomy with matching machine parameters according to an embodiment of the present invention includes the following steps: S1. Modular acquisition of agronomic parameters: The growth period of processed chili peppers is divided into four core agronomic modules: seedling raising, transplanting, plant protection, and harvesting. Each module collects the corresponding core agronomic parameters through a sensor array. S2. Standardized input of machine parameters: For the agricultural machinery types corresponding to the four agronomic modules, input the key parameters of each agricultural machinery and mark the effective threshold of the parameters to form a sub-library of machine parameters; S3. Dynamic Association Model Construction: Based on Bayesian network, an association model is constructed, with the agronomic parameters collected in step S1 as input nodes and the machine parameters entered in step S2 as output nodes. Soil pH and precipitation are used as latent variables. The model is trained through the EM algorithm to generate a probability association table of agronomic parameters, environmental factors and machine parameters. S4. Database self-updating: Regularly collect actual operation data, and correct the parameters of the correlation model through gradient descent to achieve dynamic iteration of the correlation between agronomic and machine parameters.

[0010] Furthermore, the sensor array in step S1 includes a soil moisture sensor, a light intensity sensor, a temperature and humidity sensor, and a plant growth sensor. The core agronomic parameters are as follows: germination temperature of 15-30℃ and light duration of 8-12h / d for the seedling module; planting depth and plant spacing for the transplanting module; pesticide dosage and spraying range for the plant protection module; and fruit coloring rate of ≥90% maturity and fruit stalk toughness of ≥5N for the harvesting module.

[0011] Furthermore, the types of agricultural machinery in step S2 include seedling machines, transplanters, plant protection machines, and harvesters; The key parameters and effective thresholds are specifically: the specifications of the 50-hole / 72-hole seedling trays for the seedling machine; the plant spacing adjustment range of 20-40cm for the transplanter; and the optimal rotation speed range of 200-300r / min for the harvester.

[0012] Furthermore, the rule for generating the probability association table of the Bayesian network in step S3 is as follows: When a specific combination of environmental factors and agronomic parameters is input, the output will show the model parameters with the highest compatibility probability. Specifically, when the soil pH is 6.5-7.0 and the pepper stalk toughness is 5-8 N, the compatibility probability of a harvester speed of 250 r / min is no less than 92%.

[0013] Furthermore, the actual operational data in step S4 includes quality indicators such as the chili pepper damage rate, spiciness, and moisture content after agricultural machinery operation; The periodic data collection cycle is every quarter, and the model parameters are corrected in reverse using the quality indicators to adapt to the environmental differences in different production areas.

[0014] Furthermore, in steps S1 and S2, an agronomic parameter sub-library and a machine model parameter sub-library are formed respectively, and the two sub-libraries are dynamically called through a preset module interface protocol. The interface protocol standardizes the format, calling method and error handling mechanism for data interaction between the sub-libraries.

[0015] Furthermore, in step S3, a random forest machine learning model is introduced, using the quality indicators of the processed chili pepper variety as the target variable and the planting agronomic parameters as the input features to infer the optimal agricultural machinery parameters. The association model has a built-in self-learning module that can iteratively update the model weights based on measured data from different production areas.

[0016] Furthermore, the environmental factors in step S3 also include soil moisture and light intensity; When the processed chili peppers are in the harvesting period and the soil moisture is >60%, the harvester speed output by the correlation model is reduced to 200 r / min compared to the normal humidity, so that the fruit damage rate is less than 5%.

[0017] Furthermore, the data collected by the sensor is transmitted to the edge computing node via wireless communication technology. The node performs preprocessing on the data, including cleaning, format conversion, and outlier removal, before using it to construct the agronomic parameter sub-library.

[0018] Furthermore, it also includes a database maintenance step, in which the sensors are calibrated and maintained regularly, the database is backed up and disaster recovery is performed, and the data transmission process is encrypted to prevent leakage and tampering.

[0019] The beneficial effects of this invention are: 1. This invention is based on the working principle of IoT sensing technology and edge computing. By deploying multiple types of sensors in the planting field, it collects environmental parameters, agronomic requirements, and agricultural machinery operation parameters at different growth stages in real time. These parameters are then dynamically mapped in spatiotemporal dimensions through edge computing nodes to construct a ternary association database of environment, agronomy, and machinery type. Simultaneously, relying on a Bayesian network model, environmental factors are incorporated as latent variables into the association analysis to generate a probabilistic association table, which can output optimal machinery parameters based on real-time changes in the environment and agronomy.

[0020] 2. In this invention, a random forest machine learning model and a self-updating module are introduced. Its core working principle is to use the quality index of processed chili peppers as the target variable and the agronomic parameters of planting as the input features. The model is trained through a large amount of historical data to infer the optimal agricultural machinery parameters. At the same time, the database collects measured data every quarter and corrects the model weights through the gradient descent method. This effectively solves the problem of poor cross-regional adaptability of traditional technologies and keeps the deviation of agricultural machinery operation efficiency in different production areas within a low range.

[0021] 3. Based on the modular decomposition principle of agronomic links, this invention breaks down the planting process into four core modules: seedling raising, transplanting, plant protection, and harvesting. Each module is designed with an independent sub-library of agronomic parameters and a sub-library of machine parameters. The sub-libraries are dynamically called through a unified interface protocol. When the combination of agronomic links changes, the system can quickly retrieve the data of the corresponding module's sub-library through the interface protocol and rebuild the relationship. This greatly improves the matching efficiency compared to traditional databases and avoids the parameter mismatch problem caused by changes in the combination of links.

[0022] 4. This invention utilizes a closed-loop working principle of sensor acquisition, model association, actual measurement feedback, and parameter correction to form a continuously iterative technical system. Sensors collect data in real time to update the basic sub-library, the association model optimizes the mapping relationship based on the new data, actual measurement data corrects the model parameters in reverse, and the database is synchronously iterated and upgraded. This mechanism enables the database to adapt to 10-15 more types of machine models each year, and the coverage of agronomic parameters is wider, achieving long-term adaptability and value continuation of the technical solution. Attached Figure Description

[0023] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used in conjunction with embodiments of the invention to explain the invention and do not constitute a limitation thereof. In the drawings: Figure 1 This is a schematic diagram of the overall framework of a method for linking modular chili planting agronomy with the parameters of supporting machinery proposed in this invention. Figure 2 This is a flowchart illustrating the steps of a method for linking modular chili planting agronomy with the parameters of supporting machinery proposed in this invention. Detailed Implementation

[0024] To make the technical means and objectives and effects of the present invention easier to understand, the embodiments of the present invention will be described in detail below with reference to specific illustrations.

[0025] Please see Figure 1-2 This embodiment is applied to a 100-mu (approximately 6.7 hectares) chili pepper planting base. The base is functionally divided into a seedling area, a planting area, a plant protection operation area, and a harvesting area. The soil type is sandy loam, with an average annual rainfall of 300 mm and a stable soil pH between 6.8 and 7.2. The objectives are to achieve a chili pepper harvesting damage rate of ≤5%, agricultural machinery operation efficiency of ≥0.5 mu / h, and spiciness stability fluctuation of ≤10%. The specific implementation steps are as follows: S1. Based on the growth cycle rhythm of processed chili peppers, the entire planting process is divided into four core agronomic modules: seedling raising, transplanting, plant protection, and harvesting. The sensor deployment, parameter acquisition, and data processing flow for each module are as follows: Sensors were selected based on the agronomic requirements of each module. In the seedling area, temperature and humidity sensors with an accuracy of ±0.5℃ and light sensors with a measurement range of 0-20000 lux were used, with one set deployed per 10㎡ of seedling bed, totaling 100 sets. In the transplanting area, soil moisture sensors with a measurement range of 0-100%RH and laser rangefinders with an accuracy of ±1cm were used, deployed in a 5m×5m grid, totaling 2800 sets. In the plant protection area, pesticide flow sensors with an accuracy of ±0.1L / min and ultrasonic rangefinders with a measurement radius of 0-10m were used, integrated into the plant protection machine body, totaling 10 sets. In the harvesting area, color recognition sensors and tension sensors with a tensile strength accuracy of ±0.1N were installed at the harvester's picking head, totaling 5 sets. All sensors have an IP68 waterproof rating, adaptable to field dust and rain environments.

[0026] Each sensor collects data at preset intervals. The seedling module collects germination temperature and light duration every 30 minutes, with germination temperature warning thresholds set at <15℃ or >30℃ and light duration warning thresholds at <8h / d or >12h / d. When the data exceeds the thresholds, the seedling greenhouse temperature control system and supplemental lighting system are triggered. The transplanting module collects planting depth and plant spacing every 15 minutes, with a standard planting depth of 8cm (±2cm) and a standard plant spacing of 30cm (±5cm). The plant protection module collects pesticide dosage and spraying range in real time during operation. The standard pesticide dosage for a single plant protection operation is 20L / acre, and the spraying range coverage must be ≥95%. The harvesting module collects fruit maturity and stalk toughness every 5 minutes, with maturity determined by a coloring rate ≥90% and stalk toughness ≥5N. Fruits that do not meet the standards are not included in the harvesting range.

[0027] Data collected by sensors is transmitted to the base's edge computing nodes via LoRa wireless communication technology. The edge computing nodes have pre-defined data processing rules: outliers are removed using the 3σ criterion, format conversion is performed using the Python Pandas library, and missing data is filled in using linear interpolation. The preprocessed data is written to the agronomic parameter sub-database in real time, with a data write latency of ≤1 second.

[0028] S2. For the agricultural machinery types corresponding to the four agronomic modules, establish a standardized parameter input process and construct a structured sub-library of machine model parameters. The specific implementation is as follows: The types of agricultural machinery and key parameters compatible with each module are clearly defined. The seedling module uses the 2ZQ-800 seedling machine, and the key parameters to be entered include tray size, sowing accuracy, and seedling efficiency. The transplanting module uses a transplanter, and the key parameters include plant spacing adjustment range, planting depth adjustment range, and operating efficiency. The plant protection module uses a plant protection machine, and the key parameters include tank volume, spraying pressure, and number of nozzles. The harvesting module uses a chili harvester, and the key parameters include picking head speed, conveyor belt inclination angle, damage rate, and operating efficiency.

[0029] A standardized input interface is used for parameter entry. Operators need to upload a scanned copy of the agricultural machinery's factory certificate of conformity. The parameter values ​​must be consistent with the certificate of conformity. After the entry is completed, the system will automatically verify the format.

[0030] Valid thresholds are marked for key operating parameters, such as the optimal threshold of 200-300 r / min for the harvester's picking head speed and the effective threshold of 20-40 cm for adjusting the plant spacing of the transplanter. The entered parameters are stored in the machine model parameter sub-database.

[0031] Define a module interface protocol based on RESTful API, with the protocol version being V1.0. Specify that the data interaction format is JSON, and the request methods include GET, POST, and PUT.

[0032] The protocol has a built-in error handling mechanism. When the request parameters are incorrect, it returns a 400 status code and a "parameter format error" message. When the server is abnormal, it returns a 500 status code and a "server under maintenance" message, ensuring that the success rate of data calls between sub-databases is ≥99.9%.

[0033] S3. A combined model architecture of Bayesian network and random forest is adopted to achieve deep correlation between agronomic parameters, environmental factors and machine parameters. The specific construction and training process is as follows: Bayesian network model construction: Using 12 agronomic parameters, such as germination temperature, plant spacing, and fruit maturity, as input nodes, and 10 machine parameters, such as picking head speed and planting depth, as output nodes, and environmental factors such as soil pH, precipitation, soil moisture, and light intensity as latent variables, a Bayesian network with 3 layers of nodes is constructed. The connection relationships between nodes are determined based on expert experience and mutual information calculation.

[0034] Historical data from the base over the past three years were collected, including agronomic parameters of each module, environmental monitoring data, agricultural machinery operating parameters, and chili pepper quality data. These data were divided into training and testing sets in a 7:3 ratio, and the initial network parameters were initialized using the maximum likelihood estimation method.

[0035] The EM algorithm is set to iterate 100 times, with a convergence threshold of 0.001. E-step: Calculate the posterior probability of latent variables based on the current model parameters. For example, when the agronomic parameter is "fruit stalk toughness = 6N" and the machine parameter is "speed = 250 r / min", the posterior probability of soil pH = 6.8 is calculated to be 0.85. M-step: Update the model parameters based on the posterior probability, maximizing the log-likelihood function.

[0036] After training, a probability association table is generated. When the soil pH is 6.5-7.0 and the pepper stalk toughness is 5-8N, the fit probability of harvester speed of 250 r / min is 92.3% and the fit probability of speed of 240 r / min is 7.1%. The optimal parameter recommendation rule is to select the value with the highest fit probability.

[0037] Random forest model ensemble: Using the quality index of chili peppers as the target variable and the agronomic parameters of planting as the input features, a random forest model was constructed. The number of decision trees was set to 200, the tree depth was limited to 15 layers, and the minimum number of samples for node splitting was 5. The Gini coefficient was used to evaluate the importance of features.

[0038] The model was trained using the training set, and the parameters were optimized through 5-fold cross-validation. The spiciness prediction error on the validation set was ≤500 SHU, and the moisture content prediction error was ≤2%.

[0039] The output of the trained random forest model is used as an auxiliary input to the Bayesian network model to correct the recommended machine parameters. For example, when the agronomic parameters are "planting density of 4000 plants / acre and harvesting cycle of 15 days", the random forest model back-calculates the optimal harvester speed as 280 r / min. Combined with the environmental adaptability analysis of the Bayesian network, the final output speed parameter is obtained.

[0040] The self-learning module is developed using the Python TensorFlow framework, with the model update cycle set to once a month. An update is triggered when the amount of newly collected measured data reaches 1,000 records.

[0041] The update process includes data filtering, weight calculation, and model fusion to ensure that the model is dynamically optimized as the production area environment changes.

[0042] After continuous rainfall during the harvesting period, the soil moisture increased from the usual 45% to 65%. The model captured the change of this environmental factor in real time through soil moisture sensors and automatically triggered the association rules: the harvester speed was reduced from 250r / min under the usual humidity to 200r / min, and the conveyor belt inclination angle was adjusted from 30° to 28°.

[0043] According to on-site measurements, the fruit breakage rate under this parameter combination was 3.2%, a decrease of 74.4% compared to before the adjustment, which fully meets the preset target.

[0044] S4. Establish a closed-loop self-updating mechanism for data acquisition, deviation calculation, parameter correction, and verification feedback, as follows: At the end of each quarter, a five-day field data collection process is conducted, with a 10-person data collection team. Portable equipment is used to test key indicators: the stem toughness of 100 fruits is randomly tested using a digital tensile tester, the spiciness is measured using a chili pepper meter, and the moisture content is measured using a halogen moisture analyzer. Simultaneously, agricultural machinery operation data and environmental data are recorded. Each collection generates a field data report, including the data source, testing equipment, raw data, and preprocessing results, ensuring data traceability.

[0045] The measured data are input into the correlation model, and the correction amount is calculated using the gradient descent method.

[0046] Taking the irrigation linkage parameters of the transplanting module as an example, the initial model recommends "plant spacing 30cm → irrigation volume 200ml / plant". The actual test data in the first quarter showed that the root growth rate of peppers reached 15% under this parameter. The irrigation volume correction coefficient was calculated to be 0.75 by the gradient descent method. The corrected parameter is "plant spacing 30cm → irrigation volume 150ml / plant".

[0047] Second-quarter retesting showed that the root elongation rate decreased to 4%, indicating a significant improvement in parameter adaptability. During the correction process, the probability correlation table was updated simultaneously, incorporating the adaptability probability of the new parameter combinations into the calculations to ensure the database dynamically responds to changes in the production area environment.

[0048] After each parameter correction, a one-month verification experiment was conducted, with a control group and an experimental group set up, each with an area of ​​5 mu and other conditions being the same.

[0049] The verification indicators include agricultural machinery operation efficiency and chili pepper quality. When the indicators of the experimental group are improved by ≥10% compared with the control group, the update is confirmed to be effective and the parameters are fixed; if the improvement is less than 5%, the data collection stage is returned for re-optimization.

[0050] In addition, by establishing a full lifecycle maintenance system, the stable operation of sensors, data transmission, and databases can be guaranteed. Specific measures are as follows: Sensor calibration is conducted in the first week of each month. Temperature and humidity sensors are calibrated using a standard constant temperature and humidity chamber, soil moisture sensors are calibrated using the drying method, and tensile sensors are calibrated using standard weights. Calibration records are kept in the sensor maintenance log, and sensors with calibration errors exceeding ±5% are replaced promptly. Sensor cleaning is performed quarterly, using a soft brush to remove surface dust and wiping the detection probe with anhydrous ethanol to ensure data acquisition accuracy.

[0051] Data transmission employs the AES-256 encryption algorithm, with an encryption chip integrated into the LoRa module for end-to-end encryption. The key is automatically updated every 7 days. The database uses a master-slave architecture, with the master database responsible for data writing and the slave database responsible for data querying and backup. A full backup is performed daily at 2 AM, and incremental backups are performed hourly. Backup data is stored on a remote server using an encrypted and compressed format. A fault response mechanism is established; when the database fails, the slave database can automatically switch to the master database within 30 seconds, with a data loss of ≤1 record.

[0052] Deploy a server monitoring system to monitor the CPU utilization and memory usage of edge computing nodes and database servers in real time. Send SMS and email alerts to administrators when the metrics exceed thresholds. Generate a monthly system operation report to analyze the causes of failures and identify optimization directions, ensuring that the system's annual operational stability is ≥99.5%.

[0053] Through the above embodiments, the present invention has achieved precise dynamic correlation between agronomic and machine parameters in chili pepper planting bases in arid Northwest China, increasing agricultural machinery operation efficiency by 26%, reducing chili pepper damage rate to 3.1%, and controlling spiciness stability fluctuations within 8%, fully achieving the preset technical goals. Furthermore, it can be adapted to other processed chili pepper varieties such as sweet peppers and small red chilies by adjusting module parameters, and has broad promotional value.

[0054] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for linking modular chili pepper planting agronomy with the parameters of supporting machinery, characterized in that, Includes the following steps: S1. Modular acquisition of agronomic parameters: The growth period of processed chili peppers is divided into four core agronomic modules: seedling raising, transplanting, plant protection, and harvesting. Each module collects the corresponding core agronomic parameters through a sensor array. S2. Standardized input of machine parameters: For the agricultural machinery types corresponding to the four agronomic modules, input the key parameters of each agricultural machinery and mark the effective threshold of the parameters to form a sub-library of machine parameters; S3. Dynamic Association Model Construction: Based on Bayesian network, an association model is constructed, with the agronomic parameters collected in step S1 as input nodes and the machine parameters entered in step S2 as output nodes. Soil pH and precipitation are used as latent variables. The model is trained through the EM algorithm to generate a probability association table of agronomic parameters, environmental factors and machine parameters. S4. Database self-updating: Regularly collect actual operation data, and correct the parameters of the correlation model through gradient descent to achieve dynamic iteration of the correlation between agronomic and machine parameters.

2. The method for correlating modular chili planting agronomy with matching machine parameters according to claim 1, characterized in that, The sensor array in step S1 includes a soil moisture sensor, a light intensity sensor, a temperature and humidity sensor, and a plant growth sensor. The core agronomic parameters are as follows: germination temperature of 15-30℃ and light duration of 8-12h / d for the seedling module; planting depth and plant spacing for the transplanting module; pesticide dosage and spraying range for the plant protection module; and fruit coloring rate of ≥90% maturity and fruit stalk toughness of ≥5N for the harvesting module.

3. The method for correlating modular chili planting agronomy with matching machine parameters according to claim 1, characterized in that, The types of agricultural machinery in step S2 include seedling machines, transplanters, plant protection machines, and harvesters; The key parameters and effective thresholds are specifically: the specifications of the 50-hole / 72-hole seedling trays for the seedling machine; the plant spacing adjustment range of 20-40cm for the transplanter; and the optimal rotation speed range of 200-300r / min for the harvester.

4. The method for correlating modular chili planting agronomy with matching machine parameters according to claim 1, characterized in that, The rule for generating the probability association table of the Bayesian network in step S3 is as follows: When a specific combination of environmental factors and agronomic parameters is input, the output will show the model parameters with the highest compatibility probability. Specifically, when the soil pH is 6.5-7.0 and the pepper stalk toughness is 5-8 N, the compatibility probability of a harvester speed of 250 r / min is no less than 92%.

5. The method for correlating modular chili planting agronomy with matching machine parameters according to claim 1, characterized in that, The actual operational data in step S4 includes quality indicators such as the chili pepper damage rate, spiciness, and moisture content after agricultural machinery operation. The periodic data collection cycle is every quarter, and the model parameters are corrected in reverse using the quality indicators to adapt to the environmental differences in different production areas.

6. The method for correlating modular chili planting agronomy with matching machine parameters according to claim 1, characterized in that, In steps S1 and S2, an agronomic parameter sub-library and a machine model parameter sub-library are formed respectively. The two sub-libraries are dynamically called through a preset module interface protocol. The interface protocol standardizes the format, calling method and error handling mechanism for data interaction between the sub-libraries.

7. The method for correlating modular chili planting agronomy with matching machine parameters according to claim 1, characterized in that, In step S3, a random forest machine learning model is also introduced, which uses the quality indicators of the processed chili pepper variety as the target variable and the planting agronomic parameters as the input features to infer the optimal agricultural machinery parameters. The association model has a built-in self-learning module that can iteratively update the model weights based on measured data from different production areas.

8. The method for correlating modular chili planting agronomy with matching machine parameters according to claim 1, characterized in that, The environmental factors in step S3 also include soil moisture and light intensity; When the processed chili peppers are in the harvesting period and the soil moisture is >60%, the harvester speed output by the correlation model is reduced to 200 r / min compared to the normal humidity, so that the fruit damage rate is less than 5%.

9. The method for correlating modular chili planting agronomy with matching machine parameters according to claim 2, characterized in that, The data collected by the sensor is transmitted to the edge computing node via wireless communication technology. The node performs preprocessing on the data, including cleaning, format conversion, and outlier removal, before using it to construct the agronomic parameter sub-library.

10. The method for correlating modular chili planting agronomy with matching machine parameters according to claim 1, characterized in that, It also includes a database maintenance step, in which the sensors are calibrated and maintained regularly, the database is backed up and disaster recovery is performed, and the data transmission process is encrypted to prevent leakage and tampering.

Images