A precision peanut seeding control system and method

Abstract

The application discloses a kind of peanut precision seeding regulation systems and methods, it is related to agricultural machinery technical field.The system includes seeding execution state acquisition module, seed flow state perception module, ground running state identification module, data pre-processing module, seeding state discrimination module, seeding state prediction module, collaborative control module and dynamic sampling module;The method includes parameter acquisition, two-stage filtering pre-processing, state discrimination, fusion model prediction, dynamic weight collaborative control and adaptive sampling mode switching.Through multidimensional parameter coupling acquisition, filtering noise reduction, fusion prediction and dynamic regulation, the problem of existing technology regulation lag, single dimension is solved, the precision control of peanut seeding amount and rhythm is realized, and the stability and adaptability of seeding under complex field environment are improved.

Description

Technical Field

[0001] This invention relates to the field of agricultural machinery technology, specifically to a peanut precision seeding control system and method. Background Technology

[0002] With the development of agricultural mechanization and precision agriculture technologies, peanut planting operations have been transformed towards mechanization and intelligence. Precision planting technology has become the core technology for peanut planting because it can improve seed utilization and enhance the uniformity of seedling emergence in the field. The planting effect of existing peanut planters is affected by various factors such as the operating status of the seed metering mechanism, seed flow characteristics, planter speed, ground undulation, and soil resistance. Under complex field conditions and frequent speed changes, problems such as unstable planting rhythm and fluctuating planting rate are prone to occur.

[0003] In existing technologies, some solutions rely on post-planting adjustments based on monitoring results, which struggle to respond in real-time to dynamic changes during the planting process, resulting in lag in adjustments. Other solutions depend on a single data dimension or fixed control logic, failing to fully consider the coupling relationship between planting execution, seed flow, and ground conditions, thus hindering the effective prediction of planting status evolution trends and leading to insufficient stability and adaptability in regulation. Therefore, there is an urgent need for a technical solution that can comprehensively sense multi-dimensional states, identify and predict trends in real time, and achieve coordinated regulation to meet the needs of precision peanut planting. Summary of the Invention

[0004] The purpose of this invention is to provide a peanut precision seeding control system and method to solve the problem that existing technologies cannot respond to dynamic changes in the seeding process in real time.

[0005] The technical solution of the present invention to solve the above-mentioned technical problems is as follows:

[0006] A peanut precision seeding control system, comprising:

[0007] Sowing execution status acquisition module: used to acquire the execution status parameters of the sowing mechanism and the physical characteristic parameters of the seeds. The execution status parameters include at least the rotation parameters of the seed metering mechanism, the load parameters of the driving component, the vibration characteristic parameters of the sowing component, and the sowing depth parameters.

[0008] Seed flow state sensing module: Acquires relevant data on seed flow through the sowing channel through redundant sensing devices, and generates seed flow state parameters that characterize the continuity, uniformity and rhythm stability of seed flow.

[0009] Ground operation status recognition module: used to collect information on the movement status of the seeder and soil characteristic parameters, and to identify ground operation status parameters formed by ground undulation, movement speed fluctuation, soil resistance changes and soil environment differences;

[0010] Data preprocessing module: Uses filtering algorithms to remove noise and optimize the data of the various parameters collected above;

[0011] Sowing status discrimination module: Based on the combined characteristics of preprocessed sowing execution status parameters, seed flow status parameters, and ground operation status parameters, the current sowing status is discerned;

[0012] Sowing Status Prediction Module: Based on a preset sowing status evolution model and combined with historical status change information, it predicts the status change trend of subsequent sowing intervals.

[0013] The collaborative control module generates feedforward control parameters based on the sowing status prediction results, and generates feedback correction parameters by combining the real-time collected sowing execution status parameters, thereby collaboratively adjusting the operating parameters of the sowing mechanism.

[0014] Dynamic sampling module: Based on the priority of the current sowing status, it adaptively adjusts the sampling frequency of each acquisition module.

[0015] The primary data preprocessing uses the Kalman filter algorithm, as shown in the following formula:

[0016]

[0017] The second stage uses a moving average filtering algorithm, as shown in the following formula:

[0018] ;

[0019]

[0020]

[0021]

[0022]

[0023]

[0024]

[0025]

[0026]

[0027]

[0028]

[0029]

[0030]

[0031]

[0032]

[0033]

[0034]

[0035] ;

[0036] Soil resistance correction uses the following formula:

[0037] ;

[0038] Each module interacts with other data through a standardized bus interface, and the sensor adopts a modular design to support plug-and-play functionality.

[0039] A method for precision seeding control of peanuts includes the following steps:

[0040] Step 1: Collect the execution status parameters of the sowing mechanism, the physical characteristic parameters of the seeds, and at the same time collect the travel status information of the sowing machine, soil characteristic parameters, and relevant data on the flow of seeds through the sowing channel;

[0041] Step 2: Use a two-stage filtering algorithm to preprocess the various parameters collected in Step 1, remove noise and optimize data quality;

[0042] Step 3: Based on the preprocessed parameters, extract the combined features of the sowing execution status parameters, seed flow status parameters, and ground operation status parameters to determine the current sowing status;

[0043] Step 4: Based on the preset fusion-type sowing state evolution model and combined with historical state change information, predict the state change trend of subsequent sowing intervals and generate sowing state prediction results.

[0044] Step 5: Generate feedforward control parameters based on the sowing status prediction results, and generate feedback correction parameters by combining the real-time collected sowing execution status parameters. Use a dynamic weight allocation strategy to coordinately adjust the operating parameters of the sowing mechanism.

[0045] Step 6: Based on the current sowing status priority and parameter fluctuation threshold, adaptively switch the sampling mode and adjust the collection frequency of various parameters.

[0046] The two-stage filtering algorithm includes Kalman filtering and moving average filtering. The fusion-type seeding state evolution model adopts a fusion architecture of a sliding time window model and a machine learning model. Dynamic weight allocation is achieved through a formula. accomplish.

[0047] The present invention has the following beneficial effects:

[0048] This invention overcomes the limitations of existing technologies that rely on single-dimensional analysis by coupling multi-dimensional data collection of sowing execution, seed flow, ground operation, and soil characteristics. Two-stage filtering preprocessing effectively eliminates noise interference from the field environment, improving data accuracy. A fusion-based prediction model combines short-term trend and long-term correlation analysis to achieve accurate prediction of sowing status. Dynamic weight control and adaptive sampling modes enable the system to dynamically adjust its response strategy based on field conditions, avoiding regulatory lag. The entire technical solution achieves end-to-end optimization from parameter acquisition, data processing, status identification, trend prediction to collaborative control, improving the stability and adaptability of precision peanut sowing and laying the foundation for subsequent field management and yield improvement. Attached Figure Description

[0049] Figure 1 For the invention system block diagram;

[0050] Figure 2 The diagram illustrates the specific steps of the method of the present invention. Detailed Implementation

[0051] The following detailed description of the specific embodiments of the present invention, in conjunction with the technical solutions and logical connections, ensures that each technical feature strictly corresponds to the claims. Through the technical transformation of "generalization-refinement, definition-implementation", the abstract technical solutions are made to have clear implementability.

[0052] System Detailed Implementation

[0053] 1. Seeding execution status acquisition module

[0054] Detailed technical features: This module adopts a hardware configuration scheme of "core parameter sensors + auxiliary characteristic monitoring units," and the deployment and selection of each sensor are designed around the parameter types defined in the claims.

[0055] Rotation parameters of the seed metering mechanism: An incremental encoder (model: E6B2-CWZ6C, resolution 1024 lines) is selected and fixed to the end of the seed metering shaft through a coupling. The rotation speed of the seed metering disc is collected in real time, and the signal output is in the form of A / B phase pulse. The rotation speed value is obtained after counting and processing by the main controller.

[0056] Drive component load parameters: A torque sensor (model: JN338, measurement range 0-50N・m) is used, connected in series with the connection section between the seeding shaft and the drive motor to collect the driving torque fluctuation parameters of the seeding shaft. The signal is transmitted to the main controller after being conditioned by an amplification circuit (gain 100 times).

[0057] Vibration characteristic parameters of the sowing component: A piezoelectric vibration sensor (model: YD-105, sensitivity 50mV / g) is selected and fixed to the sowing furrow opener bracket by anti-vibration bolts. The sampling direction is consistent with the sowing direction to obtain vibration amplitude and frequency parameters. The sampling frequency is 25Hz by default.

[0058] Sowing depth parameters: An ultrasonic ranging sensor (model: US-100, measuring range 2-450cm, protection level IP67) is used. It is installed next to the sowing furrow opener via an adjustable bracket, with the sensor probe pointing vertically towards the ground surface. The sensor measures the vertical distance between the lowest point of the sowing furrow opener and the ground surface, with a data output resolution of 0.1cm.

[0059] Seed physical characteristics: A miniature image sensor (model: OV7670, 640×480 pixels) and a humidity sensor (model: SHT20, measurement range 0-100%RH) are deployed at the seed box outlet (5cm from the seed metering mechanism inlet). The image sensor identifies the seed size (measurement error ≤0.5mm) and plumpness level (level 1: plumpness ≥90%, level 2: plumpness 60%-90%, level 3: plumpness <60%) through grayscale threshold segmentation and morphological processing algorithms. The humidity sensor directly outputs the seed surface humidity value with a response time ≤8ms.

[0060] 2. Seed flow status sensing module

[0061] Detailed technical features: Redundant sensing devices are deployed "inside the seeding channel + at the exit," and two types of sensors based on different principles are used for cross-verification to ensure data reliability.

[0062] Sensor deployment: The first set of infrared beam sensors (model: E3Z-LS63, detection distance 5-30cm) is installed in the middle of the sowing channel (10cm from the channel entrance, channel inner diameter 3cm), and the second set of miniature weighing sensors (model: FSR402, measurement range 0-100g) is installed at the exit of the sowing channel (5cm from the exit, sensor surface flush with the inner wall of the channel to avoid obstructing seed flow).

[0063] Data acquisition logic: The infrared beam sensor detects the occlusion signal when the seed passes by and generates a sequence of time intervals for the seed to pass by (time resolution 1ms). The miniature weighing sensor synchronously collects the instantaneous weight of a single seed (measurement error ≤0.01g). The two sets of data are synchronized through the synchronous trigger interface of the main controller to achieve timestamp alignment.

[0064] State parameter calculation:

[0065] Seed flow continuity coefficient Cc: calculated based on the time interval sequence, the formula is as follows.

[0066] ;

[0067] in The duration of a single continuous monitoring session (default setting is 10 seconds). The cumulative time for seeds to pass through an interval that exceeds the preset normal interval threshold (calculated based on the target sowing distance and travel speed; for example, when the target distance is 15cm and the travel speed is 1m / s, the normal interval threshold is set to 0.15s).

[0068] Seed flow uniformity coefficient According to the formula Calculation, where The standard deviation of the weight of a single seed during the monitoring period is ( ) represents the average weight of a single seed (calibrated by statistically analyzing the average weight of 100 seeds).

[0069] Anomaly detection logic: Based on the anomaly type, a logical detection model is constructed:

[0070] Seed adhesion: When "the seed passage time interval is greater than twice the normal interval threshold + the weight signal is within the standard weight range ( ±10% (within) Triggered when ≥0.6”, the judgment is based on the fact that the blocking time of the sticky seeds is prolonged when they pass through, but the total weight is within the standard range of a single seed;

[0071] Missed seeding: When "seed transit time > 3 times the normal interval threshold + no weight signal trigger + Triggered when <0.6”, the determination criteria are that there are no seed passing signals for a continuous period of time and the continuity coefficient is lower than the stable threshold;

[0072] Excessive variation in seed size: When "the time interval between seed passages is within the normal threshold range + the weight fluctuation of a single seed is greater than 20%" The condition is triggered when the value is greater than 0.1, and the determination is based on the fact that the rhythm is normal but the weight distribution dispersion exceeds the stable range.

[0073] 3. Ground Operation Status Identification Module

[0074] Technical feature refinement: Based on parameter type, coupled data acquisition of soil properties and travel status is achieved.

[0075] Soil property parameter acquisition: A soil sensor integrated unit (including a soil moisture sensor and a soil hardness sensor) is deployed next to the front wheel of the seeder (30cm horizontally from the center of the front wheel). The soil penetration depth is controlled by an electric lifting mechanism (default setting is 5cm to avoid distortion of surface soil moisture and hardness); among which, soil volumetric water content... Soil moisture data is collected using a soil moisture sensor (model: TDR-300, measurement range 10%-40%, error ≤±1%), and soil hardness data (Sh) is collected using a soil hardness sensor (model: TYD-2, measurement range 0.5-5MPa, error ≤±0.1MPa). Both sets of parameters are synchronously transmitted to the main controller via an I2C bus.

[0076] Travel status information acquisition: A fusion acquisition scheme of "GPS + wheel speed sensor + three-axis gyroscope" is adopted. The GPS module (model: UbloxNEO-6M, update frequency 10Hz, positioning accuracy ±2m) and the wheel speed sensor (model: Hall sensor A1104, installed on the driven wheel axle of the seeder) are fused through Kalman filtering to output the travel speed (measurement error ≤ ±0.02m / s). The three-axis gyroscope (model: MPU6050, sampling frequency 50Hz) collects the pitch angle (representation of longitudinal undulation) and roll angle (representation of lateral tilt) of the machine body, with a measurement range of ±10° and an error of ≤ ±0.1°.

[0077] Application of the soil resistance correction formula: According to the formula Calculate, where:

[0078] The measured value is that of the traction resistance sensor (model: LC1003, measuring range 0-500N, error ≤±5N), which is installed at the traction connection of the seeder.

[0079] α (soil moisture content influence coefficient) was set to 0.05, and β (soil hardness influence coefficient) was set to 0.03, both within the range of 0.01-0.1. The values ​​were calibrated through field trials of three typical soil types (sand loam, loam, and clay loam) to ensure that the corrected data can truly reflect the actual soil resistance at the time of sowing.

[0080] 4. Data Preprocessing Module

[0081] Detailed technical features: Strictly adhering to a "two-stage filtering" strategy to achieve noise removal and data optimization.

[0082] First-level Kalman filter: For parameters containing high-frequency noise, such as seed metering shaft drive torque, seeding component vibration, and travel speed, the filter parameters are set as follows:

[0083] State transition matrix A: Set as identity matrix I2×2. Since the temporal correlation of the above parameters is weak, it is assumed by default that the current state has no significant change from the previous state.

[0084] Control input matrix B: Set as a zero matrix 02×2, since the filtering process is only optimized for the acquired data and there is no additional external control input;

[0085] The process noise covariance matrix Q is set as diag([1e−4,1e−4]), based on the inherent noise level of the sensor.

[0086] The observation noise covariance matrix R is set to 0.01 and is determined by statistically analyzing the noise variance during static sensor acquisition.

[0087] Observation matrix H: set as the identity matrix The sensor-collected values ​​are used directly as the observation input;

[0088] Second-level moving average filtering: For parameters after Kalman filtering and continuous stable parameters such as sowing depth and seed moisture, the moving window size N=5 (set according to sampling frequency and data stability requirements), the formula is:

[0089] ;

[0090] Achieve data smoothing and avoid misjudgments caused by instantaneous fluctuations;

[0091] Data quality labeling: Automatically labeled by the main controller according to the three-level labeling rules:

[0092] "0 (invalid)" indicates that the sensor has no signal output, the signal amplitude is outside the measurement range, or there is no change for 5 consecutive sampling points.

[0093] Indicates "1 (Valid):" The parameters are within the preset normal range (set based on peanut planting operation requirements), and the error of the filtered data is <3%;

[0094] The label "2 (requires verification)" indicates that the parameter is outside the normal range but has not triggered the abnormal threshold, or the error after filtering is between 3% and 5%, and needs to be cross-verified with data from other modules.

[0095] 5. Sowing Status Determination Module

[0096] Technical feature refinement: Based on the "combined features + three-state" discrimination logic, accurate classification of sowing states is achieved.

[0097] Combined Feature Extraction: Six core combined features were extracted from the preprocessed parameters, including "sowing depth fluctuation value, seed flow continuity coefficient". Seed flow uniformity coefficient Soil resistance correction value Fluctuation values, speed fluctuation values, and pitch / roll angle fluctuation values ​​are all calculated as "maximum value - minimum value within 10 seconds".

[0098] The discrimination threshold was determined through multiple field trials, as follows:

[0099] Sowing depth fluctuation <0.3cm ≥0.8 ≤0.1、 Fluctuation <10%, speed fluctuation <0.1m / s, pitch / roll angle fluctuation <1°;

[0100] State determination logic:

[0101] Steady-state sowing: All combined features meet the above threshold requirements and the duration is ≥30s, indicating that the sowing process is stable and there is no significant interference;

[0102] Transitional seeding state: 1-3 combined features exceed the threshold but do not reach the instability standard, or the duration is <30s, indicating that the seeding process is slightly disturbed and is in a transitional stage between stability and instability;

[0103] Unstable sowing state: ≥4 combined features exceed the threshold, or the fluctuation of a single feature is greater than twice the threshold, indicating that the sowing process is severely disturbed and requires emergency control;

[0104] Instability Mode Subdivision: Based on the implicit classification logic, instability modes are subdivided into three categories:

[0105] Seeding mechanism failure type: Seeding shaft drive torque fluctuation > 2 times the threshold, seeding disc speed fluctuation > 10%, and other characteristics are normal;

[0106] Ground undulation type: Pitch / roll angle fluctuation > 2 times the threshold, soil resistance correction value fluctuation > 20%, and other characteristics are normal;

[0107] Abnormal seed characteristics: Seed flow uniformity coefficient > 5%, seed moisture fluctuation > 5%, and other characteristics are normal.

[0108] 6. Sowing Status Prediction Module

[0109] Technical features refined: A state trend prediction is achieved using a fusion architecture of "sliding time window model + machine learning model".

[0110] Model components:

[0111] Sliding time window model: The window size is set to 10s (based on the seeder's travel speed and field disturbance response time), and the slope of change of 6 combined features within the window is extracted (e.g., The rate of change within 10 seconds (as a short-term trend characteristic) ;

[0112] Machine learning model: A lightweight LSTM model (suitable for embedded device deployment) is selected, employing 3 hidden layers (64 neurons per layer). The input is a preprocessed parameter sequence of approximately 30 seconds (sampling frequency 25Hz, totaling 750 data points), and the output is long-term trend features. ;

[0113] Application of integrated formulas: By formula Calculation, where (Sliding time window model weights) (LSTM model weights), and satisfy The weight values ​​were obtained through training and optimization using experimental data from 10 different field environments to ensure a balance between short-term response and long-term prediction.

[0114] Prediction Output: Generates the seeding state prediction results (steady state / transitional state / unstable state) for the next 20 seconds, and outputs the prediction confidence level. (Value range 0-1), the confidence level is calculated by back-calculating the model prediction error (the smaller the error, the higher the confidence level).

[0115] 7. Coordinated Regulation Module

[0116] Detailed technical features: A dynamic weighted control strategy combining "feedforward" and "feedback" is used to achieve coordinated adjustment of sowing parameters.

[0117] Application of the dynamic weight allocation formula: According to the formula:

[0118] Calculate, where:

[0119] γ = 0.5 (state influence coefficient), used to balance the influence of the current state and the prediction confidence;

[0120] (Sowing state coefficient): 0.7 in steady state, 0.5 in transient state, and 0.3 in unstable state, set according to the stability of the state;

[0121] (Prediction confidence level): Output by the sowing status prediction module, ranging from 0 to 1;

[0122] Control parameters and logic:

[0123] Control parameters include: rotation speed of the seed metering mechanism (adjustable range 50-150 r / min), seeding rhythm (adjustable by controlling the opening and closing time of the seed metering valve, range 0.1-0.5 s / seed), seeding depth (adjustable range 3-8 cm), and seeder travel speed (adjustable range 0.8-1.5 m / s).

[0124] Feedforward control: Parameters are generated based on the prediction results. For example, when the prediction is "evolution from transitional state to unstable state", the seeding speed is reduced by 10% and the travel speed is reduced by 0.2m / s in advance.

[0125] Feedback correction: Parameters are generated based on real-time collected sowing execution status parameters. For example, if the sowing depth is detected to be 0.5cm too shallow, the hydraulic lifting mechanism is immediately controlled to deepen the sowing depth by 0.3-0.5cm.

[0126] Cooperative logic: based on computation and Weighted fusion of feedforward and feedback parameters, for example, in steady state =0.6、 =0.4, mainly driven by predictive feedforward regulation; in unstable state =0.3、 =0.7, mainly based on real-time data-driven feedback correction.

[0127] 8. Dynamic Sampling Module

[0128] Technical features refined: Adaptive sampling is achieved based on the logic of "three sampling modes + priority switching".

[0129] Sampling mode parameters:

[0130] Low power mode: Execution status acquisition frequency 10Hz, seed flow acquisition frequency 15Hz, ground operation acquisition frequency 8Hz, seed image sensor is turned off, only the core sensor is kept working;

[0131] Standard mode: Execution status acquisition frequency 25Hz, seed flow acquisition frequency 30Hz, ground operation acquisition frequency 20Hz, all sensors are working normally;

[0132] High-frequency capture mode: Execution status acquisition frequency 50Hz, seed flow acquisition frequency 60Hz, ground operation acquisition frequency 40Hz, local data caching is enabled (cache the most recent 10s high-frequency data, storage medium is SD card, capacity ≥8GB).

[0133] Triggering conditions and priority:

[0134] Triggering conditions: Low power mode (steady-state seeding state for 30s+ with parameter fluctuation ≤2%), standard mode (transitional seeding state or parameter fluctuation 2%-5%), high frequency capture mode (unstable seeding state or abnormal signal trigger).

[0135] Switching priority: abnormal signal trigger > unstable seeding state > transitional seeding state > steady-state seeding state; switching delay ≤ 100ms; fast switching is achieved through the interrupt triggering mechanism of the main controller.

[0136] Energy consumption optimization: In low power mode, the power supply to non-core sensors is cut off (achieved through the power management module), reducing energy consumption by more than 35% compared to the standard mode, and supporting continuous operation of the seeder for ≥8 hours.

[0137] 9. Hardware integration module

[0138] Detailed technical features: Designed according to "modular + standardized interfaces" to achieve system integration:

[0139] Modular design: All sensors are packaged as independent modules, equipped with standardized interfaces (power interface DC5V / 12V, communication interface CAN bus), supporting plug and play, and modules can be flexibly added or removed according to different seeder models;

[0140] Communication protocol: CAN bus communication (transmission rate 500kbps) is adopted, which supports synchronous communication of multiple devices, has strong anti-interference ability, and is suitable for complex electromagnetic environment in the field.

[0141] Power Management: A DC-DC converter (model: LM2596) is used to convert the vehicle's 12V power supply to 5V (for sensors) and 3.3V (for the main controller), and overvoltage and overcurrent protection circuits are provided to ensure power supply stability;

[0142] Main controller: The STM32H743 microcontroller (480MHz) is selected, which has powerful data processing capabilities and supports real-time filtering, model calculation and control command output.

[0143] Detailed implementation of this method

[0144] Step 1: Parameter Acquisition

[0145] After the seeder starts, the main controller sends a synchronization trigger command via the CAN bus, and each acquisition module works synchronously according to the preset initial mode (standard mode):

[0146] Sowing execution status acquisition module: Real-time acquisition of seed metering disc rotation speed, seed metering shaft drive torque, sowing component vibration amplitude / frequency, sowing depth, seed particle size, seed moisture, and seed plumpness level, with a data transmission cycle of 25ms;

[0147] Seed flow status sensing module: Infrared beam sensor and miniature weighing sensor synchronously collect seed passage time interval sequence and single seed weight, data transmission cycle 10ms;

[0148] Ground operation status identification module: Soil sensor integrated unit collects soil volumetric water content Sm and soil hardness Sh; GPS module, wheel speed sensor and three-axis gyroscope collect travel speed, body pitch angle / roll angle and measured traction resistance Fmeas; data transmission cycle 40ms.

[0149] All collected data carries a timestamp (accuracy 1ms) and is transmitted to the main controller via the CAN bus. The main controller stores the data according to module type.

[0150] Step 2: Two-stage filtering preprocessing

[0151] After receiving the raw data, the main controller performs preprocessing in the order of "Kalman filtering first, then moving average filtering":

[0152] Kalman filtering: Filters parameters containing high-frequency noise, such as seed metering shaft drive torque, seeding component vibration, and travel speed, to remove noise caused by engine vibration and field electromagnetic interference, and outputs the filtered parameter values;

[0153] Moving average filtering: Smooths the parameters after Kalman filtering and continuous parameters such as sowing depth and seed moisture. The sliding window size is N=5, and the optimized parameter data is output.

[0154] Data quality labeling: The main controller automatically adds a quality label field (0 / 1 / 2) to each parameter, filters out data labeled "invalid" and discards it, and temporarily stores data labeled "needs verification" for subsequent cross-validation.

[0155] Step 3: State determination

[0156] Based on the preprocessed valid parameters, the execution status is determined according to the following logic:

[0157] Combined feature extraction: Calculate the sowing depth fluctuation value and seed flow continuity coefficient within 10 seconds. Uniformity coefficient Soil resistance correction value Fluctuation values, speed fluctuation values, and pitch / roll angle fluctuation values;

[0158] Threshold comparison: The extracted combined features are compared one by one with the preset discrimination threshold, and the number of features exceeding the threshold is counted.

[0159] Status output: Based on the number of features exceeding the threshold and the duration, determine whether the current sowing state is steady, transitional, or unstable. If it is unstable, further subdivide it into the corresponding unstable mode. The determination result is transmitted to the sowing state prediction module in real time.

[0160] Step 4: Fusion Model Prediction

[0161] Invoke the preset fusion-type seeding state evolution model and perform predictions according to the following process:

[0162] Short-term trend extraction: The sliding time window model extracts the slope of change of 6 combined features within the last 10 seconds and outputs the short-term trend. ;

[0163] Long-term trend mining: The LSTM model takes the preprocessed parameter sequence of the most recent 30 seconds as input and outputs the long-term trend through model calculation. ;

[0164] Fusion computing: according to the formula Calculate the fusion prediction results and output the sowing status and prediction confidence for the next 20 seconds. ;

[0165] Results transmission: The prediction results and confidence levels are transmitted synchronously to the collaborative control module as the basis for feedforward control.

[0166] Step 5: Dynamic weight coordination and regulation

[0167] After receiving the forecast results, the coordinated regulation module executes regulation according to the following logic:

[0168] Feedforward parameter generation: Feedforward control parameters are generated based on the predicted sowing state. For example, if the prediction is "steady state maintenance", the current parameters are maintained; if the prediction is "transitional state to unstable state evolution", parameters such as sowing speed and travel speed are adjusted in advance.

[0169] Feedback parameter generation: Based on the deviation between the real-time collected sowing execution status parameters (such as sowing depth and seed metering speed) and the target parameters, feedback correction parameters are generated;

[0170] Weight calculation: according to the formula Calculate feedforward control weights and feedback correction weights ;

[0171] Coordinated adjustment: The feedforward and feedback parameters are integrated according to weight to generate the final control command, which is sent to the actuators of the seeder (such as seed metering motor, hydraulic lifting device, and walking motor) via CAN bus to achieve precise adjustment of the seeding parameters.

[0172] Step 6: Adaptive sampling mode switching

[0173] The main controller monitors the current sowing status and parameter fluctuation thresholds in real time, and switches the sampling mode according to the following logic:

[0174] Status monitoring: Continuously track the seeding status judgment results and parameter fluctuations to determine whether the mode switching conditions are met;

[0175] Mode switching: Trigger the corresponding sampling mode (low power / standard / high frequency capture) according to the priority of "abnormal signal trigger > unstable seeding state > transitional seeding state > steady state seeding state".

[0176] Parameter adjustment: When switching modes, the main controller sends frequency adjustment commands to each acquisition module to adjust the sampling frequency synchronously. In high-frequency mode, the data buffering function is activated.

[0177] Looping execution: During the sowing operation, the above switching logic is continuously repeated to achieve adaptive adjustment of the sampling mode.

[0178] 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 peanut precision seeding control system, characterized in that, include: The sowing execution status acquisition module is used to acquire the execution status parameters of the sowing mechanism and the physical characteristic parameters of the seeds. The execution status parameters include at least the rotation parameters of the seed metering mechanism, the load parameters of the driving component, the vibration characteristic parameters of the sowing component, and the sowing depth parameters. The seed flow state sensing module acquires relevant data on seed flow through the sowing channel through redundant sensing devices, and generates seed flow state parameters that characterize the continuity, uniformity and rhythm stability of the seed flow. The ground operation status recognition module is used to collect information on the movement status of the seeder and soil characteristic parameters, and to identify ground operation status parameters formed by ground undulation, movement speed fluctuation, soil resistance changes and soil environment differences. The data preprocessing module uses filtering algorithms to remove noise and optimize the data of the various parameters collected above; The sowing status discrimination module determines the current sowing status based on the combined characteristics of preprocessed sowing execution status parameters, seed flow status parameters, and ground operation status parameters. The sowing status prediction module, based on a preset sowing status evolution model and combined with historical status change information, predicts the status change trend of subsequent sowing intervals. The collaborative control module generates feedforward control parameters based on the sowing status prediction results and generates feedback correction parameters by combining the real-time collected sowing execution status parameters, thereby collaboratively adjusting the operating parameters of the sowing mechanism. The dynamic sampling module adaptively adjusts the sampling frequency of each acquisition module based on the priority of the current sowing status; the data preprocessing adopts the Kalman filtering algorithm.

2. The peanut precision seeding control system and method according to claim 1, characterized in that, The seed physical characteristic parameters collected by the sowing execution status acquisition module include at least seed size, seed moisture, and seed plumpness level; the sowing depth parameter is the vertical distance between the sowing furrow opener and the ground surface; the data preprocessing module adopts a two-stage filtering strategy, the first stage being the Kalman filter algorithm, with the following formula: The second stage uses a moving average filtering algorithm, as shown in the following formula: 。 3. The peanut precision seeding control system according to claim 1, characterized in that, The redundant sensing device includes at least two sets of sensors of different types, respectively installed inside the sowing channel and at the outlet of the sowing channel; the seed flow state parameters include the seed flow continuity coefficient. and seed flow uniformity coefficient The formula for calculating the uniformity coefficient is as follows: The sowing state discrimination module constructs an anomaly discrimination model based on seed flow state parameters. Based on the continuity coefficient, uniformity coefficient, and observation data from redundant sensing devices, it identifies the types of seed flow anomalies. These anomaly types include at least seed adhesion, missed sowing, and excessively large differences in seed particle size. The standard deviation of the weight of a single seed. This represents the average weight of a single seed.

4. The peanut precision seeding control system according to claim 1, characterized in that, In the ground operation status identification module, the soil resistance correction uses the following formula: ； The soil characteristic parameters collected by the ground operation status identification module include at least soil volumetric water content. and soil hardness ; The collected travel status information includes at least travel speed, aircraft pitch angle, and roll angle; in the soil resistance correction formula, This is a correction value for soil resistance. For the actual traction resistance, α is the influence coefficient of soil moisture content, β is the influence coefficient of soil hardness, and α and β are preset empirical coefficients, with values ​​ranging from 0.01 to 0.

1.

5. The peanut precision seeding control system according to claim 1, characterized in that, The sampling modes of the dynamic sampling module include at least a low-power mode, a standard mode, and a high-frequency capture mode. The sampling mode is switched based on the seeding state judgment result and the parameter fluctuation threshold, with the priority being: abnormal signal trigger > unstable seeding state > transitional seeding state > steady-state seeding state. The sampling frequency is lowest in the low-power mode and highest in the high-frequency capture mode. The data caching function is also triggered synchronously when the high-frequency capture mode is started.

6. The peanut precision seeding control system according to claim 1, characterized in that, The seeding state evolution model adopts a fusion architecture of a sliding time window model and a machine learning model, and the fusion formula for the prediction results is as follows: The sowing state discrimination module classifies the sowing state into steady-state sowing state, transitional sowing state, and unstable sowing state. The unstable sowing state is further subdivided into at least three different types of instability modes; wherein... The prediction result after fusion at time k is... These are the weighting coefficients for the prediction results of the sliding time window model. These are the weight coefficients for the prediction results of the machine learning model, and + =1, These are the predicted values ​​from the sliding time window model. These are the predicted values ​​from the machine learning model.

7. The peanut precision seeding control system according to claim 1, characterized in that, The collaborative control module adopts a dynamic weight allocation strategy. The feedforward control weight and the feedback correction weight are adaptively adjusted according to the current sowing status and prediction confidence. The weight allocation formula is as follows: The collaborative control module coordinates the rotational speed of the seed metering mechanism, the seeding rhythm, the seeding depth, and the seeder's travel speed based on the aforementioned weights; wherein, For feedforward control weights, The feedback adjustment weights are used, and γ is the state influence coefficient. The sowing state coefficient (values ​​are 0.6-0.8 in steady state, 0.4-0.6 in transition state, and 0.2-0.4 in unstable state). The prediction confidence level is set to a value between 0 and 1.

8. The peanut precision seeding control system according to claim 1, characterized in that, The system also includes a hardware integration module. The acquisition sensors in each module adopt a modular design, support plug-and-play, and communicate with the main system through a standardized bus interface. The data output by the data preprocessing module carries a data quality identification field, which includes at least three levels: invalid, valid, and requiring verification.

9. A method for precision seeding control of peanuts, characterized in that, Includes the following steps: Step 1: Collect the execution status parameters of the sowing mechanism, the physical characteristic parameters of the seeds, and at the same time collect the travel status information of the sowing machine, soil characteristic parameters, and relevant data on the flow of seeds through the sowing channel; Step 2: Use a two-stage filtering algorithm to preprocess the various parameters collected in Step 1, remove noise and optimize data quality; Step 3: Based on the preprocessed parameters, extract the combined features of the sowing execution status parameters, seed flow status parameters, and ground operation status parameters to determine the current sowing status; Step 4: Based on the preset fusion-type sowing state evolution model and combined with historical state change information, predict the state change trend of subsequent sowing intervals and generate sowing state prediction results. Step 5: Generate feedforward control parameters based on the sowing status prediction results, and generate feedback correction parameters by combining the real-time collected sowing execution status parameters. Use a dynamic weight allocation strategy to coordinately adjust the operating parameters of the sowing mechanism. Step 6: Based on the current sowing status priority and parameter fluctuation threshold, adaptively switch the sampling mode and adjust the collection frequency of various parameters.

10. The peanut precision sowing control method according to claim 9, characterized in that, The execution status parameters mentioned in step 1 include at least the rotation parameters of the seed metering mechanism, the load parameters of the drive component, the vibration characteristic parameters of the seeding component, and the seeding depth parameters; the seed physical characteristic parameters include at least the seed particle size, seed moisture content, and seed plumpness level; and the travel status information includes at least the travel speed, the body pitch angle, and the roll angle.

Images