Agricultural seedling raising control method and platform for deep learning teaching

Abstract

The invention relates to an artificial intelligence-based agricultural management platform control method and platform for teaching, and the method comprises the steps: collecting the seedling environment data and controller parameters of the platform, combining with a deep learning model to precisely predict the germination rate and the seedling growth state, generating recommended target environment parameters, and according to the target environment parameters, carrying out the calculation of the germination rate and the seedling growth state. The operation information of the controller and the attributes of the static controller are combined, the operation parameters of the controller are optimized through a PID control algorithm, and the seedling raising environment is adjusted to be optimized in real time; students control the controller through manual programming, the platform records control parameters of the students and operation results of the controller, control results of the students are compared with parameters recommended by the deep learning model, difference indexes are evaluated, and the students adjust control logic and improve equipment operation strategies by checking comparison results and system suggestions. Students not only can manually control equipment through programming, but also can intuitively understand the optimization ability and principle of deep learning through comparison with a recommendation result of the deep learning model, so that the ability of the students to solve complex agricultural problems is improved, and the cultivation of intelligent agricultural talents is promoted.

Description

Technical Field

[0001] The present invention relates to the technical field of artificial intelligence training platforms, and particularly to a control method and platform for an agricultural management platform for teaching based on artificial intelligence. Background Art

[0002] In the field of real intelligent agriculture, with the development of Internet of Things technology, the level of agricultural automation has been significantly improved. Environmental monitoring and equipment control achieved through sensors and controllers have greatly improved the production efficiency of crops and the resource utilization rate. However, existing agricultural teaching training platforms are generally only limited to the basic applications of Internet of Things technology and cannot effectively combine artificial intelligence technology, especially deep learning technology, making it difficult for students to understand and apply these cutting-edge technologies during the learning process.

[0003] Most existing teaching platforms stay at the level of manually operating equipment, lacking intelligent automatic control and prediction capabilities, and unable to compare the differences between students' manual operations and intelligent optimization methods, thus limiting the teaching effect. Summary of the Invention

[0004] In order to overcome the above deficiencies of the prior art, a control method for agricultural seedling cultivation for deep learning teaching is provided, including the following steps: S1. Real-time collect seedling cultivation environment data through sensors of the agricultural seedling cultivation platform. The data includes soil humidity, temperature, and light intensity. Collect images of seed germination and seedling growth through a camera to generate seedling cultivation process data; collect controller operation information, including controller status parameters such as the operation status, current, power, light intensity of supplementary lighting, and drip irrigation flow rate of the fan; collect controller static information, including controller attributes such as the rated power and maximum rotation speed of the fan. S2. Perform data preprocessing on the collected data. The data preprocessing includes the following steps: fill in missing values in the data, remove outliers, and perform data smoothing and normalization processing; remove the background, enhance the image, and extract features from the image data to form a standardized input data set; segment the time series data using a sliding window method to generate a historical data set and real-time input data. S3. Predict the germination rate, including extracting the features of the seed germination image based on a convolutional neural network, analyzing the seedling cultivation environment data and controller status data in combination with a time series model, predicting the germination rate of the seeds, and generating a germination rate time curve from the prediction result as the input for subsequent precise regulation. S4. Use a deep learning image recognition model to identify and analyze the seedling growth image, evaluate the growth parameters of the seedlings, including the number of leaves, leaf color, stem thickness, and height; combine the real-time collected environmental data and controller operation status to generate precise target environmental parameters for subsequent controller adjustment. S5. According to the generated target environment parameters in S4, combined with the controller operation information and static controller attributes, the PID control algorithm optimizes the controller operation parameters and adjusts the seedling raising environment in real time, including: adjusting the running time and speed of the fan to achieve temperature and humidity adjustment; controlling the light intensity and duration of the supplementary light to meet the light requirements of the crops; adjusting the output flow of the humidifier to keep the air and soil humidity stable; adjusting the drip irrigation flow to ensure the optimization of soil moisture.

[0005] S6. Compare the operation status after monitoring and controlling by the controller, the feedback data of the seedling raising environment and the prediction results, and calculate the error indicators, including: the prediction error of the germination rate; the deviation between the seedling growth parameters and the target values; the error between the controller operation status and the set target; Take the error data as the feedback input, and continuously optimize the germination rate prediction model and the controller environment control parameters until the set environmental conditions and growth goals are met.

[0006] Furthermore, the data preprocessing method in step S2 includes constructing a smooth curve to repair missing values through piecewise polynomial fitting interpolation: , where n is the order of the polynomial, is the interpolation interval; Eliminate outliers through the interquartile range IQR to ensure the rationality of data distribution; Perform mean smoothing through the sliding window method: where w is the size of the sliding window; Normalize the data ranges of different dimensions of data: .

[0007] Furthermore, the prediction of the germination rate in step S3 includes the steps: S3.1 Obtain the germination image data of the seeds during the seedling raising process, and perform standardization, size unification and data enhancement processing on the images. The data enhancement includes random rotation, cropping, flipping and color jitter; S3.2 Collect time series data, record the dynamic parameters of the seedling raising environment, including temperature, humidity, light intensity, and preprocess the parameters. The preprocessing includes missing value filling, data smoothing processing, and data normalization; S3.3 Build an image feature extraction model based on a convolutional neural network: Use randomly initialized weights to build a convolutional neural network model. The model includes multiple convolutional layers, pooling layers and fully connected layers; The convolution operation calculation formula is: ; Generate a fixed-length image feature vector after dimensionality reduction through pooling ; S3.4 Merge the image feature vector with the preprocessed time series environmental data to form a comprehensive feature vector for input to the subsequent time series prediction model; S3.5 Construct a time series prediction model through Transformer, introduce positional encoding based on sine and cosine functions for the time series data to embed the position information of time points, and enhance the model's ability to model the time order; generate weighted feature representations through the self-attention mechanism, calculate time dependencies through the self-attention mechanism, and extract features of the short-term and long-term dependencies of the time series data by stacking multiple layers of Transformer encoders to generate time series feature representations, and map them to the germination rate prediction results through a linear layer ; S3.6 Generate the germination rate prediction value Rt at time point t based on the germination rate prediction results, and draw the germination rate time curve: , to obtain the curve data points {(t, )}, which are used as the reference input for subsequent precise control.

[0008] Further, the steps for generating precise target environmental parameters in step S4 include: Perform convolution operations on the seedling image using a deep learning image recognition model to extract the feature vector of the growth parameters , extract the following growth parameters from the feature vector , at least including: the number of leaves; the leaf color characteristics; the stem thickness; the height; preprocess the real-time collected environmental data and extract the environmental feature vector , fuse the feature vector and the environmental feature vector to form a comprehensive feature vector F, where the environmental data includes: environmental temperature, environmental humidity, and light intensity; convert the comprehensive feature vector F into target environmental parameters for adjusting the operation of the controller through a mapping function.

[0009] Further, the step S5 includes the following specific processes: comparing the target environmental parameters with the real-time collected environmental data, and calculating the error between the current environmental conditions and the target value; using the PID control algorithm or other control methods to adjust the operating parameters of the controller in real time according to the error value: adjusting the running time and speed of the fan to narrow the deviation between the current environmental temperature and humidity and the target value; controlling the light intensity and duration of the supplementary light to ensure that the light intensity is close to the target value; adjusting the output flow of the humidifier to keep the air humidity within the target range; adjusting the drip irrigation flow to make the soil humidity meet the optimized target value; dynamically adjusting the weight coefficients of the gain parameters Kp, Ki, and Kd of the PID controller according to the controller operating status information and the static attributes of the controller to improve the control accuracy and response speed; continuously monitoring the controller operating status and environmental feedback data, updating the error value and looping to execute the control strategy until the environmental parameters reach the target range.

[0010] An agricultural seedling raising platform for deep learning teaching, characterized in that it includes a programming interface through which students manually program to control the controller; the platform records the control parameters of the students and the operating results of the controller; a comparison module, which compares the control results of the students with the target environmental parameters of the deep learning model and evaluates the difference indicators, and the difference indicators include: energy consumption, temperature and humidity fluctuation range, germination rate; The device module includes a controller and sensors. The controller includes a fan, a supplementary light, a power supply module, a water pump, and a humidifier. The sensors include: a light sensor, a thermometer and hygrometer, and a camera; The data acquisition module is used to acquire sensor data; The data preprocessing module is used to process the acquired sensor data; removing the background, enhancing the image and extracting features from the image data to form a standardized input data set; segmenting the time series data to generate a historical data set and real-time input data; The prediction module is used to extract the features of the seed germination image through a convolutional neural network and predict the germination rate of the seeds; The image recognition module is used to recognize and analyze the seedling growth image, evaluate the growth parameters of the seedlings; generate accurate target environmental parameters; The device adjustment module is used to optimize the device operating parameters through the PID control algorithm in combination with the target environmental parameters and adjust the seedling raising environment in real time; Further, it also includes a processor, a memory, and a computer program stored on the memory and executable on the processor. When the processor executes the program, it implements the agricultural seedling raising method for deep learning teaching.

[0011] Advantages of the present invention: The present invention provides an agricultural seedling raising platform for teaching. By collecting the seedling raising environment data and controller parameters of the platform, and combining with the accurate prediction of the germination rate and the growth state of seedlings by a deep learning model, target environment parameters are generated. According to the target environment parameters, combined with the controller operation information and static controller attributes, the PID control algorithm optimizes the controller operation parameters to adjust the seedling raising environment to the optimum in real time; students control the controller through manual programming, and the platform records the control parameters of the students and the operation results of the controller, compares the control results of the students with the target environment parameters of the deep learning model, evaluates the difference indicators, and students adjust the control logic and improve the equipment operation strategy by viewing the comparison results and system suggestions. Students can not only control the equipment manually through programming, but also intuitively understand the optimization ability and principle of deep learning by comparing with the recommended results of the deep learning model, so as to improve the students' ability to solve complex agricultural problems and promote the cultivation of intelligent agricultural talents. BRIEF DESCRIPTION OF THE DRAWINGS

[0012] For ease of explanation, the present invention is described in detail by the following preferred embodiments and accompanying drawings.

[0013] Figure 1 It is a schematic structural diagram of a control method and platform for an agricultural management platform for teaching artificial intelligence of the present invention; Figure 2 It is an execution flowchart of a control method and platform for an agricultural management platform for teaching artificial intelligence of the present invention; Figure 3 It is an execution flowchart of a non-sensing intelligent control method for electromechanical equipment of the present invention; Figure 4 It is a communication schematic diagram of an agricultural seedling raising platform for teaching deep learning of the present invention; Figure 5 It is a schematic structural diagram of an agricultural seedling raising platform for teaching deep learning of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0014] To make the implementation objectives, technical solutions and characteristics of the present invention application clearer, the technical solutions implemented in the present invention application will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of the present application. Obviously, the described embodiments are part of the examples of the present invention application, rather than all the implementation cases. Usually, the embodiments of the present invention application described and shown in the accompanying drawings here can be arranged and designed in different configurations.

[0015] Therefore, the detailed description of the embodiments provided in the invention application is not intended to limit the scope claimed by the present application, but only represents the specific implementation cases of the present application. Based on the implementation cases in the present application, all other implementation cases obtained by those of ordinary skill in the art without creative work belong to the scope protected by the present invention application.

[0016] The invention will be described in detail below in conjunction with embodiments and the accompanying drawings. As Figures 1-3 described, a control method for an agricultural management platform for artificial intelligence teaching includes the following steps: S1. Real-time collect seedling raising environment data through sensors of the agricultural seedling raising platform. The data includes soil humidity, temperature, and light intensity. Collect images of seed germination and seedling growth through a camera to generate seedling raising process data; collect controller operation information, including controller status parameters such as fan operation status, current, power, light intensity of supplementary lighting, and drip irrigation flow rate; collect controller static information, including controller attributes such as the rated power and maximum rotation speed of the fan. S2. Perform data preprocessing on the collected data. The data preprocessing includes the following steps: fill in missing values in the data, remove outliers, and perform data smoothing and normalization processing; perform background removal, image enhancement, and feature extraction on the image data to form a standardized input data set; segment the time series data using the sliding window method to generate a historical data set and real-time input data. S3. Predict the germination rate, including extracting the features of the seed germination image based on a convolutional neural network, analyzing the seedling raising environment data and controller status data in combination with a time series model, predicting the germination rate of the seeds, and generating a germination rate time curve for the prediction result as the input for subsequent precise regulation. S4. Use a deep learning image recognition model to identify and analyze the seedling growth images, evaluate the growth parameters of the seedlings, including the number of leaves, leaf color, stem thickness, and height; combine the real-time collected environmental data and controller operation status to generate precise target environmental parameters for subsequent controller adjustment. S5. According to the target environmental parameters generated in S4, combine the controller operation information and static controller attributes, and optimize the controller operation parameters using the PID control algorithm to adjust the seedling raising environment in real time, including: adjusting the fan operation time and rotation speed to achieve temperature and humidity regulation; controlling the light intensity and duration of the supplementary lighting to meet the lighting requirements of the crops; adjusting the output flow rate of the humidifier to keep the air and soil humidity stable; adjusting the drip irrigation flow rate to ensure the optimization of soil moisture.

[0017] S6. Compare the operation status, seedling raising environment feedback data, and prediction results after the monitoring controller is regulated, and calculate error indicators, including: germination rate prediction error; deviation between the seedling growth parameters and the target values; error between the controller operation status and the set target. Use the error data as feedback input to continuously optimize the germination rate prediction model and controller environment control parameters until the set environmental conditions and growth goals are met.

[0018] Further, the data preprocessing method in step S2 includes constructing a smooth curve to repair missing values through piecewise polynomial fitting interpolation: , where n is the order of the polynomial, is the interpolation interval; the optimal polynomial order is determined to be 3 through cross-validation;

[0019] Outlier points are removed through the interquartile range IQR, [Q1 - 1.5⋅IQR, Q3 + 1.5⋅IQR], where IQR = Q3 - Q1, and Q1 and Q3 are the 25th percentile and 75th percentile respectively; to ensure the rationality of the data distribution; for outliers, the method of median replacement is used for repair;

[0020] Mean smoothing is performed through the sliding window method: where w is the size of the sliding window, and w is 3 in the embodiment;

[0021] Data with different dimensions are normalized by unifying the data range: .

[0022] Further, the prediction of the germination rate in step S3 includes the steps: S3.1 Obtain the germination image data of the seeds during the seedling raising process, and perform standardization, size unification, and data augmentation processing on the images. The data augmentation includes random rotation, cropping, flipping, and color jitter; S3.2 Time series data collection, record the dynamic parameters of the seedling raising environment, including temperature, humidity, and light intensity, and preprocess the parameters. The preprocessing includes missing value filling, data smoothing processing, and data normalization; S3.3 Construct an image feature extraction model based on a convolutional neural network: Use randomly initialized weights to construct a convolutional neural network model. The model includes multiple convolutional layers, pooling layers, and fully connected layers; the convolution operation calculation formula is: , where is the output of the convolution kernel k at position (i, j), σ is the activation function, and are the weights and biases of the convolution kernel respectively; the convolutional neural network model uses the ResNet network architecture, contains 50 convolutional layers, the convolution kernel size is 3×3, and the activation function is ReLU; the dimension is reduced through the max pooling or average pooling layer to reduce the size of the feature map. The pooling formula is: ; Advanced image features are extracted through multiple convolutional and pooling operations, and the extracted features are flattened into a one-dimensional vector; a fixed-length feature vector is generated through the fully connected layer ; S3.4 Combine the image feature vector with the preprocessed time series environmental data Merge to form a comprehensive feature vector F, , for input to the subsequent time series prediction model; S3.5 Construct a time series prediction model through Transformer, introduce positional encoding based on sine and cosine functions for time series data to embed the position information of time points, and enhance the model's ability to model time order, , where t is the time step, PE is the positional encoding, i is the feature dimension index, and d is the dimension of the feature vector; generate weighted feature representations through the self-attention mechanism, and calculate time dependence through the self-attention mechanism, , where Q, K, and V are the query, key, and value matrices respectively, is the dimension of the key matrix. By stacking multiple layers of Transformer encoders, extract features from the short-term and long-term dependencies of time series data, generate time series feature representations, and map them to germination rate prediction results through a linear layer , where F is the feature representation output by the Transformer encoder, W is the weight of the linear layer, and b is the bias term; in this embodiment, 6 layers of Transformer encoders are used, and each layer contains 512 hidden units; S3.6 Generate the germination rate prediction value Rt at time point t based on the germination rate prediction result, and draw the germination rate time curve: , to obtain the curve data points {(t, )}, as the reference input for subsequent precise control.

[0023] The steps for generating precise target environmental parameters in step S4 include:

[0024] Extract features through the convolutional layer of the deep learning image recognition model ResNet: X = W * X + b, where X is the input image, W is the convolutional kernel with a size of 3x3, and b is the bias: ;

[0025] Extract the feature vector of the growth parameters , and the regression network extracts the following growth parameters P = W ⋅ from the feature vector + b, where: W is the weight matrix, is the feature vector extracted from ResNet, and b is the bias; the growth parameters at least include: the number of leaves, the leaf color feature, the stem diameter, and the height;

[0026] Preprocess the real-time collected environmental data and extract the environmental feature vector , and the environmental data includes: environmental temperature, environmental humidity, and light intensity;

[0027] The feature vector and the environmental feature vector are fused to form a comprehensive feature vector F, . Combining the real-time collected environmental data, the comprehensive feature vector F is converted into target environmental parameters for adjusting the operation of the controller, such as the target temperature T, the target humidity H, the target light intensity L, and the target soil humidity M through the regression network mapping function, where , and g represents the mapping function of the model.

[0028] The step S5 includes the following specific processes: comparing the target environmental parameters with the real-time collected environmental data, and calculating the error between the current environmental conditions and the target value; using the PID control algorithm or other control methods to adjust the operation parameters of the controller in real time according to the error value: adjusting the running time and speed of the fan to narrow the deviation between the current environmental temperature and humidity and the target value; controlling the light intensity and duration of the supplementary light to ensure that the light intensity is close to the target value; adjusting the output flow of the humidifier to keep the air humidity within the target range; adjusting the drip irrigation flow to make the soil humidity meet the optimized target value; dynamically adjusting the weight coefficients of the gain parameters Kp, Ki, and Kd of the PID controller according to the controller operation status information and the static attributes of the controller to improve the control accuracy and response speed. Generally, the platform is placed indoors and the environment is relatively stable. Due to the relatively stable environment, the gain parameters can be fixed; continuously monitoring the controller operation status and the environmental feedback data, updating the error value and looping to execute the control strategy until the environmental parameters reach the target range; determining the optimal PID parameters through experiments as Kp = 1.2, Ki = 0.5, Kd = 0.3.

[0029] As Figures 4-5 described, the embodiment of this specification also includes an agricultural seedling raising platform for deep learning teaching, including a programming interface through which students can manually program to control the controller; the platform records the control parameters of the students and the operation results of the controller; a comparison module that compares the control results of the students with the target environmental parameters of the deep learning model and evaluates the difference indicators, and the difference indicators include: energy consumption, temperature and humidity fluctuation range, germination rate. The system generates improvement suggestions for students' operations based on the comparative analysis results, including specific guidance on optimizing equipment parameters and control strategies; students adjust the programming logic according to the feedback and gradually improve the operation strategy to make the control effect close to the optimal solution recommended by the model; students can intuitively understand the optimization ability and principle of deep learning, thereby improving the students' ability to solve complex agricultural problems and promoting the cultivation of intelligent agricultural talents

[0030] Device module, including a controller and sensors. The controller includes a fan, a supplementary light, a power supply module, a water pump, and a humidifier. The sensors include: a light sensor, a temperature and humidity meter, and a camera; the temperature and humidity meter uses a high-precision soil humidity sensor, and the sampling frequency of the sensors is once per minute; Data acquisition module, used for acquiring sensor data; Data preprocessing module, used for processing the acquired sensor data; removing the background, enhancing the image, and extracting features from the image data to form a standardized input data set; segmenting the time series data to generate a historical data set and real-time input data; Prediction module, used for extracting the features of the seed germination image through a convolutional neural network and predicting the germination rate of the seeds; Image recognition module, used for recognizing and analyzing the seedling growth image, evaluating the growth parameters of the seedlings; generating accurate target environment parameters; Device adjustment module, used for optimizing the device operation parameters through a PID control algorithm in combination with the target environment parameters and adjusting the seedling raising environment in real time; It further includes a processor, a memory, and a computer program stored on the memory and executable on the processor. Among them, when the processor executes the program, it implements an agricultural seedling raising method for deep learning teaching.

Claims

1. An agricultural seedling raising control method for deep learning teaching, characterized in that, Including the following steps: S1. Real-time collect the seedling raising environment data through the sensors of the agricultural seedling raising platform. The data includes soil humidity, temperature, and light intensity. Collect the images of seed germination and seedling growth through a camera to generate the seedling raising process data; collect the controller operation information, including the controller status parameters such as the fan operation status, current, power, light intensity of the supplementary light, and drip irrigation flow rate; collect the controller static information; S2. Perform data preprocessing on the collected data. The data preprocessing includes the following steps: fill in the missing values in the data, remove the outliers, and perform data smoothing and normalization processing; perform background removal, image enhancement, and feature extraction on the image data to form a standardized input data set; segment the time series data using the sliding window method to generate a historical data set and real-time input data; S3. Predict the germination rate, including extracting the features of the seed germination image based on the convolutional neural network, analyzing the seedling raising environment data and the controller status data in combination with the time series model, predicting the germination rate of the seeds, and generating a germination rate time curve for the prediction result as the input for subsequent precise control; S4. Use the deep learning image recognition model to identify and analyze the seedling growth images, evaluate the growth parameters of the seedlings, including the number of leaves, leaf color, stem thickness, and height; combine the real-time collected environmental data and the controller operation status to generate precise target environmental parameters for subsequent controller adjustment; S5. According to the target environmental parameters generated in S4, combine the controller operation information and the static controller attributes, and optimize the controller operation parameters using the PID control algorithm to adjust the seedling raising environment in real time, including: adjusting the fan operation time and speed to achieve temperature and humidity adjustment; controlling the light intensity and duration of the supplementary light to meet the lighting requirements of the crops; adjusting the output flow rate of the humidifier to keep the air and soil humidity stable; adjusting the drip irrigation flow rate to ensure the optimization of soil moisture; S6. Compare the operation status, seedling raising environment feedback data, and prediction results after the monitoring controller is adjusted, and calculate the error indicators, including: germination rate prediction error; deviation between the seedling growth parameters and the target values; error between the controller operation status and the set target; Use the error data as the feedback input to continuously optimize the germination rate prediction model and the controller environment control parameters until the set environmental conditions and growth goals are met.

2. The agricultural seedling raising control method for deep learning teaching according to claim 1, characterized in that, The data preprocessing method in the step S2 includes constructing a smooth curve to repair missing values through piecewise polynomial fitting interpolation: , where n is the order of the polynomial, is the interpolation interval; Remove the outliers through the interquartile range IQR to ensure the rationality of the data distribution; Mean smoothing is performed by the sliding window method: where w is the size of the sliding window; Normalize the unified data range for data with different dimensions: .

3. The agricultural seedling raising control method for deep learning teaching according to claim 2, wherein, The prediction of the germination rate in step S3 includes the steps: S3.1 Obtain the germination image data of the seeds during the seedling raising process, and perform standardization, size unification, and data enhancement processing on the images. The data enhancement includes random rotation, cropping, flipping, and color jitter; S3.2 Collect time series data, record the dynamic parameters of the seedling raising environment, including temperature, humidity, and light intensity, and perform preprocessing on the parameters. The preprocessing includes missing value filling, data smoothing processing, and data normalization; S3.3 Construct an image feature extraction model based on a convolutional neural network: Construct a convolutional neural network model using randomly initialized weights. The model includes multiple convolutional layers, pooling layers, and fully connected layers. The calculation formula for the convolution operation is: ; Generate an image feature vector with a fixed length after dimensionality reduction through pooling ; S3.

4. Combine the image feature vector with the preprocessed time series environmental data to form a comprehensive feature vector for input to the subsequent time series prediction model; S3.5 constructs a time series prediction model through Transformer, introduces positional encoding based on sine and cosine functions for time series data to embed the position information of time points, and enhances the model's ability to model time order; generates weighted feature representations through the self-attention mechanism, calculates time dependence through the self-attention mechanism, extracts features of short-term and long-term dependencies of time series data by stacking multiple layers of Transformer encoders, generates time series feature representations, and maps them to germination rate prediction results through a linear layer ; S3.6 Generate the predicted germination rate value Rt at time point t based on the predicted germination rate results, and plot the germination rate time curve: , and obtain the curve data points , which serve as the reference input for subsequent precise regulation.

4. The agricultural seedling raising control method for deep learning teaching according to claim 3, characterized in that, The steps for generating accurate target environmental parameters in step S4 include: Perform a convolution operation on the seedling image using a deep learning image recognition model to extract the feature vector of the growth parameters , extract the growth parameters from the feature vector . The growth parameters at least include: the number of leaves, the leaf color feature, the stem diameter, and the height; preprocess the real-time collected environmental data and extract the environmental feature vector , fuse the feature vector and the environmental feature vector to form a comprehensive feature vector F. The environmental data includes: environmental temperature, environmental humidity, and light intensity; convert the comprehensive feature vector F into the target environmental parameters for adjusting the operation of the controller through the regression network mapping function.

5. According to an agricultural seedling raising control method for deep learning teaching as claimed in claim 4, step S5 includes the following specific processes: Compare the target environmental parameters with the environmental data collected in real time, and calculate the error between the current environmental conditions and the target value; Use the PID control algorithm or other control methods to adjust the operating parameters of the controller in real time according to the error value: Adjust the running time and speed of the fan to reduce the deviation between the current environmental temperature and humidity and the target value; Control the light intensity and duration of the supplementary light to ensure that the light intensity is close to the target value; Adjust the output flow of the humidifier to keep the air humidity within the target range; Adjust the drip irrigation flow to make the soil humidity meet the optimized target value; According to the operating status information of the controller and the static attributes of the controller, dynamically adjust the weight coefficients of the gain parameters Kp, Ki, and Kd of the PID controller to improve the control accuracy and response speed; Continuously monitor the operating status of the controller and the environmental feedback data, update the error value and loop to execute the control strategy until the environmental parameters reach the target range.

6. An agricultural seedling raising platform for deep learning teaching, characterized in that, It includes a programming interface through which students can manually program to control the controller, at least including setting the fan speed, supplementary light intensity, humidifier flow, and drip irrigation flow; The platform records the control parameters of the students and the operating results of the controller; It includes a comparison module that compares the control results of the students with the target environmental parameters of the deep learning model and evaluates the difference indicators, and the difference indicators include: energy consumption, temperature and humidity fluctuation range, germination rate; It also includes a device module, including a controller and sensors, the controller includes a fan, a supplementary light, a power supply module, a water pump, and a humidifier, and the sensors include: a light sensor, a thermometer and hygrometer, and a camera; It also includes a data acquisition module for acquiring sensor data; It also includes a data preprocessing module for processing the acquired sensor data; Remove the background, enhance the image, and extract features from the image data to form a standardized input data set; Segment the time series data to generate a historical data set and real-time input data; It also includes a prediction module for extracting the features of the seed germination image through a convolutional neural network to predict the germination rate of the seeds; It also includes an image recognition module for recognizing and analyzing the seedling growth image and evaluating the growth parameters of the seedlings; Generate accurate target environmental parameters; It also includes a controller adjustment module for optimizing the operating parameters of the controller through the PID control algorithm in combination with the target environmental parameters and adjusting the seedling raising environment in real time.

7. An agricultural seedling raising platform for deep learning teaching according to claim 6, characterized in that, It further includes a processor, a memory, and a computer program stored on the memory and executable on the processor. When the processor executes the program, it implements the agricultural seedling raising method for deep learning teaching as claimed in any one of claims 1-5.

Images