A method, apparatus and equipment for obtaining tobacco leaf threshing process parameters

Abstract

This invention relates to the field of tobacco processing technology and discloses a method, apparatus, and equipment for obtaining tobacco threshing process parameters. The method includes: constructing a leaf structure prediction model; inputting tobacco samples into the leaf structure prediction model to obtain predicted values ​​of their leaf structure features; selecting tobacco samples that meet preset leaf structure conditions based on the predicted values; training a neural network regression model using tobacco appearance features as input and threshing process parameters as output using the tobacco samples that meet the preset leaf structure conditions; and inputting the leaf appearance features of the tobacco to be processed into the neural network regression model to obtain tobacco threshing process parameters that meet the leaf structure requirements. This invention automatically outputs threshing process parameters that meet the requirements based on tobacco appearance features and preset leaf structure conditions, avoiding errors and deviations that may occur during manual setting, and improving the accuracy and efficiency of threshing process parameter setting.

Description

Technical Field

[0001] This invention relates to the field of tobacco processing technology, specifically to a method, apparatus, and equipment for obtaining tobacco leaf trimming process parameters. Background Technology

[0002] In recent years, the tobacco industry has placed higher demands on the quality and structure of tobacco flakes during the threshing and re-drying process to meet consumers' taste and health needs. Threshing and re-drying is a process that involves crushing, humidifying, compressing, heating, and cooling tobacco leaves to form flakes. The quality and structure of the flakes mainly depend on the settings of the threshing process parameters, such as threshing temperature, pressure, time, and moisture content.

[0003] Currently, the setting of leaf-threshing process parameters mainly relies on manual adjustment based on experience. This involves manually adjusting parameters such as temperature, pressure, and time of the leaf-threshing machine based on factors such as the variety, grade, moisture content, and blend of the tobacco leaves, as well as target indicators for the tobacco sheets, such as the percentage of large and medium-sized leaves, stem content, and nicotine content. However, this manual parameter setting method is highly subjective and unstable, leading to significant differences between different batches, shifts, and even individual producers using the same blend. This affects the consistency of tobacco sheet quality and structure, resulting in low precision and efficiency. Summary of the Invention

[0004] To overcome the shortcomings of low precision and low efficiency in existing tobacco leaf-cutting process parameter settings, this invention proposes the following technical solution:

[0005] In the first aspect, the present invention proposes a method for obtaining tobacco leaf-cutting process parameters, comprising:

[0006] S1: Construct a blade structure prediction model;

[0007] S2: Input the tobacco sample into the leaf structure prediction model to obtain its predicted leaf structure features;

[0008] S3: Based on the predicted values ​​of leaf structure characteristics, select tobacco samples that meet the preset leaf structure conditions;

[0009] S4: Train a neural network regression model with tobacco appearance features as input and leaf-cutting process parameters as output using tobacco samples that meet the preset leaf structure conditions.

[0010] S5: Input the leaf appearance characteristics of the tobacco to be processed into the neural network regression model to obtain the tobacco leaf pruning process parameters that meet the leaf structure requirements.

[0011] Secondly, the present invention also proposes a device for obtaining tobacco leaf threshing process parameters, comprising:

[0012] The blade structure prediction model building module is used to build blade structure prediction models.

[0013] The leaf structure feature prediction module is used to input tobacco samples into the leaf structure prediction model to obtain predicted values ​​of their leaf structure features.

[0014] The screening module is used to select tobacco samples that meet the preset leaf structure conditions based on the predicted values ​​of leaf structure features.

[0015] The training module is used to train a neural network regression model with tobacco appearance features as input and leaf-cutting process parameters as output using tobacco samples that meet preset leaf structure conditions.

[0016] The leaf-cutting process parameter acquisition module is used to input the leaf appearance characteristics of the tobacco to be processed into a neural network regression model to obtain tobacco leaf-cutting process parameters that meet the leaf structure requirements.

[0017] Thirdly, the present invention also proposes an electronic device, which includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to perform the operations performed by the tobacco leaf-picking process parameter acquisition method of any of the schemes in the first aspect.

[0018] The beneficial effects of the present invention include at least the following:

[0019] This invention automatically outputs threshing process parameters that meet the requirements based on the appearance characteristics of tobacco and preset leaf structure conditions. This avoids errors and deviations that may occur during manual setting, improves the accuracy of threshing process parameter settings, and allows for rapid acquisition of threshing process parameters for large quantities of tobacco, saving manpower and time costs and increasing efficiency. Furthermore, this invention can dynamically adjust the threshing process parameters according to tobacco formulations in different modules and leaf structure conditions with varying requirements, further improving the quality and effectiveness of tobacco threshing. Attached Figure Description

[0020] Figure 1 This is a flowchart illustrating the method for obtaining tobacco leaf-cutting process parameters provided in an embodiment of this application.

[0021] Figure 2 (a) is a scatter plot of the predicted and actual values ​​of the block rate prediction model in this application embodiment on the test set.

[0022] Figure 2 (b) is a scatter plot of the predicted and actual values ​​of the mid-film rate prediction model in this application embodiment on the test set.

[0023] Figure 2 (c) is a scatter plot of the predicted and actual values ​​of the leaf stalk content prediction model in this application embodiment on the test set.

[0024] Figure 3 This is a structural diagram of the neural network regression model in the embodiments of this application.

[0025] Figure 4 This is a graph showing the training set loss and validation set loss of the neural network regression model in this embodiment of the application with respect to the number of training rounds.

[0026] Figure 5 This is a schematic diagram of the device for obtaining tobacco leaf-cutting process parameters provided in an embodiment of this application.

[0027] Figure 6 This is a schematic diagram of the structure of the electronic device provided in the embodiments of this application. Detailed Implementation

[0028] The embodiments of the present invention will be described below with reference to the accompanying drawings and preferred technical solutions. Those skilled in the art can easily understand other advantages and effects of the present invention from the content disclosed in this specification. The present invention can also be implemented or applied through other different specific embodiments, and various details in this specification can also be modified or changed based on different viewpoints and applications without departing from the spirit of the present invention. It should be understood that the preferred technical solutions are only for illustrating the present invention and not for limiting the scope of protection of the present invention.

[0029] It should be noted that the illustrations provided in the following embodiments are only schematic representations of the basic concept of the present invention. Therefore, the drawings only show the components related to the present invention and are not drawn according to the actual number, shape and size of the components in the actual implementation. In the actual implementation, the form, quantity and proportion of each component can be arbitrarily changed, and the layout of the components may also be more complex.

[0030] In the following description, numerous details are explored to provide a more thorough explanation of embodiments of the invention. However, it will be apparent to those skilled in the art that embodiments of the invention may be practiced without these specific details. In other embodiments, well-known structures and devices are shown in block diagram form rather than in detail to avoid obscuring embodiments of the invention.

[0031] Example 1

[0032] This embodiment proposes a method for obtaining tobacco leaf trimming process parameters.

[0033] Specifically, Figure 1 This is a flowchart illustrating the method for obtaining tobacco leaf-cutting process parameters provided in an embodiment of this application.

[0034] like Figure 1 As shown, the method for obtaining tobacco leaf threshing process parameters includes the following steps:

[0035] S1: Construct a blade structure prediction model.

[0036] S2: Input the tobacco sample into the leaf structure prediction model to obtain its predicted leaf structure features.

[0037] S3: Based on the predicted values ​​of leaf structure characteristics, select tobacco samples that meet the preset leaf structure conditions.

[0038] S4: Train a neural network regression model with tobacco appearance features as input and leaf-cutting process parameters as output using tobacco samples that meet the preset leaf structure conditions.

[0039] S5: Input the leaf appearance characteristics of the tobacco to be processed into the neural network regression model to obtain the tobacco leaf pruning process parameters that meet the leaf structure requirements.

[0040] The tobacco threshing process parameter acquisition method provided in this application automatically outputs threshing process parameters that meet the requirements based on the tobacco's appearance characteristics and preset leaf structure conditions. This avoids errors and deviations that may occur during manual setting, improves the accuracy of threshing process parameter settings, and allows for rapid acquisition of threshing process parameters for large quantities of tobacco, saving manpower and time costs and improving efficiency. Furthermore, by dynamically adjusting the threshing process parameters according to different module formulations and varying leaf structure conditions, the quality and effectiveness of tobacco threshing are improved.

[0041] Example 2

[0042] This embodiment is an improvement on the method for obtaining tobacco leaf-cutting process parameters proposed in Embodiment 1.

[0043] In this embodiment, an image acquisition device integrating an electronic scale and an optical sensor is used to automatically extract information from the RGB, HSV, and Lab color spaces of the tobacco leaf image, including the mean and standard deviation. The leaf appearance features include the leaf's weight, length, width, perimeter, depth, uniformity, thickness, structure, B-symmetric deviation, R-symmetric deviation, S-symmetric deviation, a-mean, a-symmetric deviation, and derived features such as perimeter ratio, thickness-to-weight ratio, area ratio, and GR ratio.

[0044] In this embodiment, the leaf structure indicators include the percentage of large leaves, the percentage of medium leaves, and the percentage of leaves containing stems.

[0045] In this embodiment, the leaf-beating process parameters include the set frequency of the beating roller motor, the set frequency of the grading fan, and the set frequency of the return stem beating roller motor.

[0046] In the specific implementation process, the set frequencies of the roller beaters include those for the first batch of roller beaters (M14, M15, M16, M17, and M18). The set frequencies of the grading fans include those for the first batch of one-grade left fans (M25, M26, M34, M35, M42, M43, M50, M51, M58, M59), the second batch of roller beaters (M71, M72, M79), the third batch of roller beaters (M79), and the fourth batch of roller beaters (M79). The following motors are set to frequencies: M80 (two-beat, seven-tenths left fan), M87 (two-beat, seven-tenths right fan), M88 (two-beat, eight-tenths left fan), M95 (two-beat, eight-tenths right fan), M96 (two-beat, eight-tenths right fan), M107 (three-beat roller motor), M108 (three-beat roller motor), M115 (three-beat, nine-tenths left fan), M116 (three-beat, nine-tenths right fan), M123 (three-beat, ten-tenths left fan), M124 (three-beat, ten-tenths right fan), M131 (four-beat roller motor), M136 (four-beat, eleven-tenths left fan), and M137 (four-beat, eleven-tenths right fan). The return roller motor is set to frequency M144.

[0047] In this embodiment, the specific steps of S1 include:

[0048] S101: Construct a feature dataset S = {a1, a2, ..., a...} using the leaf appearance characteristics, leaf-cutting process parameters, and corresponding leaf structure indices of tobacco samples. K};

[0049] S102: In feature a i Uniform distribution U(a) i,min ,a i,max Randomly select an observation value a from ) i,c As the dividing points, we obtain K dividing points {s1, s2, ..., s...} K}; where a i,min ≤a i,c ≤a i,max , 1≤i≤K, a i,min For feature a i The minimum value of a i,max For feature a i The maximum value;

[0050] S103: Using the split point with the largest current split evaluation score as the root node, the feature dataset is split into left leaf nodes and right leaf nodes;

[0051] In this embodiment, in S103, the method further includes a split evaluation of the segmentation points, the expression of which is as follows:

[0052]

[0053] Among them, S l and S r These are the left and right leaf node subsets, respectively, and var{y|·} represents the variance of the corresponding output value y in different feature datasets.

[0054] S104: Recursively split the left and right leaf nodes until the preset maximum depth of the decision tree is reached or the number of samples in the left and right leaf nodes no longer decreases, thus obtaining the decision tree T. j ;

[0055] S105: Repeat S102 to S104 to obtain the decision tree set. As a leaf structure prediction model; where M is the number of decision trees in the decision tree set, the leaf structure prediction model uses the mean of the outputs of all decision trees as the predicted value of the leaf structure features.

[0056] In the specific implementation process, prediction models for large leaf rate, medium leaf rate, and leaf stem content rate were constructed based on the leaf structure indicators to be predicted. and By substituting each sample in the dataset into the corresponding model and calculating the mean of the model output, the predicted values ​​of large leaf rate, medium leaf rate, and leaf stem content rate are obtained.

[0057] When selecting the modeling set based on leaf structure requirements and leaf structure prediction results, the predicted values ​​of large leaf rate, medium leaf rate, and leaf stem content rate for each sample in the dataset are compared with the required values ​​for large leaf rate, medium leaf rate, and leaf stem content rate, respectively. Specifically: medium film rate Stem content in leaves When the sample has a large proportion of predicted values Predicted rate of medium-length films Stem content in leaves When both conditions are met, the sample is set as a tobacco sample that satisfies the preset leaf structure conditions.

[0058] like Figure 2 As shown, the model's prediction performance is checked on the test set. Figure 2 (a), (b), and (c) are scatter plots of the predicted values ​​and actual values ​​of the large leaf rate, medium leaf rate, and leaf stem content rate prediction models on the test set, respectively.

[0059] Depend on Figure 2 It is evident that the prediction models for large leaf rate, medium leaf rate, and leaf stem content have good predictive performance on the test set. The correlation coefficients between the predicted and actual values ​​on the test set are all around 0.7, with the corresponding correlation coefficients for the large leaf rate prediction model and the medium leaf rate prediction model reaching over 0.75. This demonstrates that the prediction models have application value.

[0060] The following uses these three blade structure prediction models to predict the blade structure of tobacco leaves on the entire dataset, and selects the data that meets the blade structure requirements.

[0061] Firstly, the leaf structure requirements are: large leaf rate <40.0%; medium leaf rate >38.0%; and stem content in the leaf <1.1%.

[0062] Based on this standard, a dataset is selected from the entire dataset that simultaneously meets the above three requirements: predicted large leaf rate, predicted medium leaf rate, and predicted stem content in leaves. This dataset is then divided into training and testing sets, on which a neural network regression model for predicting leaf removal process parameters can be constructed.

[0063] To simplify the model structure, this embodiment first selects leaf-cutting process parameters with a Pearson correlation coefficient of no more than 0.75 to form a neural network regression model. The selected leaf-cutting process parameters are shown in Table 1.

[0064] Table 1 Selected Leaf Removal Process Parameters

[0065] M14 set frequency of the beater roller motor One hit, one score, left fan M25 set frequency One-to-two split left fan M34 set frequency The set frequency of the second beater roller motor M71 Two-six-point left fan M79 set frequency Two-seven-point left fan M87 set frequency Two-strikes-eight-points left fan M95 set frequency The set frequency of the M107 three-beat roller motor Three-strikes-nine-points left fan M115 set frequency The four-beater roller motor M131 has a set frequency. Four blowers, eleven points, left fan M136 set frequency. M144 back-stalk roller motor set frequency Dimensions of a two-piece frame (in inches)

[0066] In this embodiment, in step S5, before inputting the leaf appearance features of the tobacco to be processed into the neural network regression model, the method further includes:

[0067] The leaf appearance characteristics of the tobacco to be treated are normalized, and the expression is as follows:

[0068]

[0069] Where μ is the mean of the appearance characteristics of the blade, and σ is the standard deviation of the appearance characteristics of the blade.

[0070] In this embodiment, the specific steps of S4 include:

[0071] S401: Set the number of hidden layers in the initial neural network and the number of neurons in each hidden layer, and randomly initialize the network connection weights;

[0072] S402: Based on the appearance characteristics of tobacco x i,外观特征Perform forward propagation on the input vector of the initial neural network and compute the output value of each neuron;

[0073] S403: Calculate a weighted average of the calculated output values ​​of each neuron and apply the ReLU activation function to obtain the initial output value y of the neural network. i,工艺参数 ;

[0074] S404: Calculate the output value y i,工艺参数 The mean squared error (MSE) of the true label is expressed as follows:

[0075]

[0076] S405: Calculate the loss gradient of the initial neural network parameters based on the mean squared error (MSE), as shown in the following expression:

[0077]

[0078] Where L is the loss function, This represents the connection weight between the i-th neuron in layer (l-1) and the j-th neuron in layer l. This is the weighted average of the inputs to the j-th neuron in the l-th layer. This represents the output of the i-th neuron in the (l-1)-th layer. This represents the bias term in the j-th neuron of the l-th layer;

[0079] S406: Update the connection weights of the initial neural network based on the loss gradient of the parameters:

[0080]

[0081] Where η is the learning rate. and This represents the gradient of the loss function with respect to the connection weights and bias terms.

[0082] S407: Repeat S402 to S406 until the initial neural network converges or the set number of iterations is reached to obtain the neural network regression model.

[0083] In the specific implementation process, such as Figure 3The diagram shown illustrates the structure of the neural network regression model in this embodiment. The model uses a 30-neuron Dense layer as the input layer, employing the computationally efficient ReLU activation function. The second layer expands the features, using a 60-neuron Dense layer with the same ReLU activation function. The output Dense layer has the same number of neurons as the recommended leaf-piercing process parameters (13 in this embodiment) and requires no activation function. Since the network depth is not very deep, there is no need to worry about the vanishing gradient problem; therefore, the BatchNormalization layer is omitted to improve training and inference efficiency.

[0084] In the compilation of the neural network regression model, the loss calculation used is the mean squared error commonly used in regression problems, and the optimizer is Adam with an adaptive learning rate (the initial learning rate is set to 0.003, and beta_1 = 0.9, beta_2 = 0.999).

[0085] The neural network regression model was trained with 150 epochs and a batch size of 16. 10% of the data was randomly selected as the validation set to observe whether the model was overfitting. The curves showing the training and validation losses of the neural network regression model with respect to the number of training epochs are as follows: Figure 4 As shown. From Figure 4 As can be seen from the loss curves of the training and validation sets, as the number of training rounds increases, the training loss and validation loss of the model decrease to the same level simultaneously, indicating that the model has not overfitted.

[0086] Finally, the prediction effect of the neural network regression model was checked on the test set: the appearance features of tobacco leaves in the test set were standardized and then input into the neural network model. The threshing process parameters output by the model are shown in Table 2 below (some threshing process parameter names are replaced with numbers).

[0087] Table 2. Leaf-picking process parameters output by the neural network regression model

[0088]

[0089] The leaf trimming process parameters recommended by the neural network regression model are compared with the actual leaf trimming process parameters that meet the blade structure. The difference between the leaf trimming process parameters is shown in Table 3 below.

[0090] Table 3. Differences in Leaf Removal Process Parameters

[0091]

[0092]

[0093] As shown in Table 3, the leaf trimming process parameters recommended by the neural network regression model differ slightly from the actual leaf trimming process parameters that meet the blade structure requirements in only a few parameters. The recommended values ​​for the remaining leaf trimming process parameters accurately reflect the actual leaf trimming process parameter values ​​that meet the blade structure requirements.

[0094] Example 3

[0095] See Figure 5 This embodiment proposes a device for obtaining tobacco threshing process parameters, applied to the tobacco threshing process parameter obtaining method described in any of the above embodiments, including:

[0096] The blade structure prediction model building module is used to build blade structure prediction models.

[0097] The leaf structure feature prediction module is used to input tobacco samples into the leaf structure prediction model to obtain predicted values ​​of their leaf structure features.

[0098] The screening module is used to select tobacco samples that meet the preset leaf structure conditions based on the predicted values ​​of leaf structure features.

[0099] The training module is used to train a neural network regression model with tobacco appearance features as input and leaf-cutting process parameters as output using tobacco samples that meet preset leaf structure conditions.

[0100] The leaf-cutting process parameter acquisition module is used to input the leaf appearance characteristics of the tobacco to be processed into a neural network regression model to obtain tobacco leaf-cutting process parameters that meet the leaf structure requirements.

[0101] The tobacco threshing process parameter acquisition device proposed in this application automatically outputs the required threshing process parameters based on the tobacco's appearance characteristics and preset leaf structure conditions. This avoids errors and deviations that may occur during manual setting and allows for rapid acquisition of threshing process parameters for large quantities of tobacco, saving manpower and time costs. Furthermore, the threshing process parameters can be dynamically adjusted according to different module formulations of tobacco and varying leaf structure conditions, improving the quality and effectiveness of tobacco threshing.

[0102] Example 4

[0103] This application provides an electronic device. Figure 6 This is a schematic diagram of the structure of an electronic device 100 provided in an embodiment of this application. The electronic device 100 includes a memory 101, a processor 102, and a computer program stored in the memory 101 and executable on the processor 102. When the processor 102 executes the program, it implements the method for obtaining tobacco leaf-cutting process parameters provided in the above embodiment.

[0104] Furthermore, the electronic device 100 also includes a communication interface 103 for communication between the memory 101 and the processor 102.

[0105] The memory 101 may include high-speed RAM (Random Access Memory) memory, and may also include non-volatile memory, such as at least one disk storage.

[0106] If the memory 101, processor 102, and communication interface 103 are implemented independently, then the communication interface 103, memory 101, and processor 102 can be interconnected via a bus to complete communication between them. The bus can be an ISA (Industry Standard Architecture) bus, a PCI (Peripheral Component Interconnect) bus, or an EISA (Extended Industry Standard Architecture) bus, etc. The bus can be divided into address bus, data bus, control bus, etc. For ease of representation, Figure 6 The bus is represented by a single thick line, but this does not mean that there is only one bus or one type of bus.

[0107] Optionally, in a specific implementation, if the memory 101, processor 102, and communication interface 103 are integrated on a single chip, then the memory 101, processor 102, and communication interface 103 can communicate with each other through an internal interface.

[0108] The processor 102 may be a CPU (Central Processing Unit), an ASIC (Application Specific Integrated Circuit), or one or more integrated circuits configured to implement the embodiments of this application.

[0109] In the description of this specification, the references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of this application. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples. Moreover, without contradiction, those skilled in the art can combine and integrate the different embodiments or examples described in this specification, as well as the features of different embodiments or examples.

[0110] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include at least one of that feature. In the description of this application, "N" means at least two, such as two, three, etc., unless otherwise explicitly specified.

[0111] Any process or method described in the flowchart or otherwise herein can be understood as representing a module, segment, or portion of code comprising one or more N executable instructions for implementing custom logic functions or processes, and the scope of the preferred embodiments of this application includes additional implementations in which functions may be performed not in the order shown or discussed, including substantially simultaneously or in reverse order depending on the functions involved, as should be understood by those skilled in the art to which embodiments of this application pertain.

[0112] It should be understood that the various parts of this application can be implemented using hardware, software, firmware, or a combination thereof. In the above embodiments, the N steps or methods can be implemented using software or firmware stored in memory and executed by a suitable instruction execution system. For example, if implemented in hardware, as in another embodiment, it can be implemented using any one or a combination of the following techniques known in the art: discrete logic circuits having logic gates for implementing logical functions on data signals, application-specific integrated circuits (ASICs) having suitable combinational logic gates, programmable gate arrays (FPGAs), field-programmable gate arrays (FPGAs), etc.

[0113] Those skilled in the art will understand that all or part of the steps of the methods in the above embodiments can be implemented by a program instructing related hardware. The program can be stored in a computer-readable storage medium, and when executed, the program includes one or a combination of the steps of the method embodiments.

[0114] Obviously, the above embodiments of the present invention are merely examples for clearly illustrating the present invention, and are not intended to limit the implementation of the present invention. Those skilled in the art can make other variations or modifications based on the above description. It is neither necessary nor possible to exhaustively describe all embodiments here. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the scope of protection of the claims of the present invention.

Claims

1. A method for obtaining tobacco leaf threshing process parameters, characterized in that, include: S1: Construct a blade structure prediction model, including: S101: Construct a feature dataset S = {a1, a2, ..., a...} using the leaf appearance characteristics, leaf-cutting process parameters, and corresponding leaf structure indices of tobacco samples. K }; S102: In feature a i Uniform distribution U(a) i,min ,a i,max Randomly select an observation value a from ) i,c As the dividing points, we obtain K dividing points {s1, s2, ..., s...} K }; where a i,min ≤a i,c ≤a i,max , 1≤i≤K, a i,min For feature a i The minimum value of a i,max For feature a i The maximum value; S103: Using the split point with the largest current split evaluation score as the root node, the feature dataset is split into left leaf nodes and right leaf nodes; S104: Recursively split the left and right leaf nodes until the preset maximum depth of the decision tree is reached or the number of samples in the left and right leaf nodes no longer decreases, thus obtaining the decision tree T. j ; S105: Repeat S102 to S104 to obtain the decision tree set. As a leaf structure prediction model; where M is the number of decision trees in the decision tree set, the leaf structure prediction model uses the mean of the outputs of all decision trees as the predicted value of the leaf structure features. S2: Input the tobacco sample into the leaf structure prediction model to obtain its predicted leaf structure features; S3: Based on the predicted values ​​of leaf structure characteristics, select tobacco samples that meet the preset leaf structure conditions; S4: Train a neural network regression model with tobacco appearance features as input and leaf-cutting process parameters as output using tobacco samples that meet the preset leaf structure conditions. S5: Input the leaf appearance characteristics of the tobacco to be processed into the neural network regression model to obtain the tobacco leaf pruning process parameters that meet the leaf structure requirements.

2. The method for obtaining tobacco leaf-cutting process parameters according to claim 1, characterized in that, The specific steps of S4 include: S401: Set the number of hidden layers in the initial neural network and the number of neurons in each hidden layer, and randomly initialize the network connection weights; S402: Based on the appearance characteristics of tobacco x i,外观特征 Perform forward propagation on the input vector of the initial neural network and compute the output value of each neuron; S403: Calculate a weighted average of the calculated output values ​​of each neuron and apply the ReLU activation function to obtain the initial output value y of the neural network. i,工艺参数 ; S404: Calculate the output value y i,工艺参数 The mean squared error (MSE) of the true label is expressed as follows: S405: Calculate the loss gradient of the initial neural network parameters based on the mean squared error (MSE), as shown in the following expression: Where L is the loss function, This represents the connection weight between the i-th neuron in layer (l-1) and the j-th neuron in layer l. This is the weighted average of the inputs to the j-th neuron in the l-th layer. This represents the output of the i-th neuron in the (l-1)-th layer. This represents the bias term in the j-th neuron of the l-th layer; S406: Update the connection weights of the initial neural network based on the loss gradient of the parameters; S407: Repeat S402 to S406 until the initial neural network converges or the set number of iterations is reached to obtain the neural network regression model.

3. The method for obtaining tobacco leaf-cutting process parameters according to claim 1, characterized in that, In S103, the method also includes a split evaluation of the segmentation points, the expression of which is shown below: Among them, S l and S r These are the left and right leaf node subsets, respectively, and var{y|·} represents the variance of the corresponding output value y in different feature datasets.

4. The method for obtaining tobacco leaf-cutting process parameters according to claim 1, characterized in that, The blade appearance characteristics include the blade's weight, length, width, circumference, depth, uniformity, thickness, structure, B-axis deviation, R-axis deviation, S-axis deviation, a-mean value, a-axis deviation, and derived characteristics such as circumference ratio, thickness-to-weight ratio, area ratio, and GR ratio.

5. The method for obtaining tobacco leaf-cutting process parameters according to claim 1, characterized in that, The leaf structure indicators include the percentage of large leaves, the percentage of medium leaves, and the percentage of stems in the leaves.

6. The method for obtaining tobacco leaf-cutting process parameters according to claim 1, characterized in that, The leaf-beating process parameters include the set frequency of the beating roller motor, the set frequency of the grading fan, and the set frequency of the return stem beating roller motor.

7. The method for obtaining tobacco leaf threshing process parameters according to claim 1, characterized in that, In step S5, before inputting the leaf appearance features of the tobacco to be processed into the neural network regression model, the method further includes: The leaf appearance characteristics of the tobacco to be treated are normalized, and the expression is as follows: Where μ is the mean of the appearance characteristics of the blade, and σ is the standard deviation of the appearance characteristics of the blade.

8. A device for obtaining tobacco threshing process parameters, applied to the method for obtaining tobacco threshing process parameters as described in any one of claims 1 to 7, characterized in that, include: The blade structure prediction model building module is used to build blade structure prediction models. The leaf structure feature prediction module is used to input tobacco samples into the leaf structure prediction model to obtain predicted values ​​of their leaf structure features. The screening module is used to select tobacco samples that meet the preset leaf structure conditions based on the predicted values ​​of leaf structure features. The training module is used to train a neural network regression model with tobacco appearance features as input and leaf-cutting process parameters as output using tobacco samples that meet preset leaf structure conditions. The leaf-cutting process parameter acquisition module is used to input the leaf appearance characteristics of the tobacco to be processed into a neural network regression model to obtain tobacco leaf-cutting process parameters that meet the leaf structure requirements.

9. An electronic device, characterized in that, The electronic device includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to perform the operations performed by the method for obtaining tobacco leaf-cutting process parameters as described in any one of claims 1 to 7.

Images