Tobacco leaf curing state discrimination method and system

Abstract

The invention relates to the technical field of tobacco leaf curing, and discloses a tobacco leaf curing state discrimination method and system, and the method comprises the steps: obtaining curing state images of different tobacco leaves, carrying out the segmentation fusion of first target regions of each tobacco leaf image based on a preset fusion strategy, and obtaining at least one target tobacco leaf image, generating an initial tobacco image data set; key tobacco leaf pictures in the initial tobacco leaf image data set are extracted, the key tobacco leaf pictures are subjected to automatic degree grade identification, and a target tobacco leaf image data set is generated; constructing a tobacco leaf baking state discrimination model based on the historical tobacco leaf image data set; the real-time tobacco leaf image is input into a tobacco leaf curing state judgment model for judgment, and the tobacco leaf curing degree is recognized; and generating a tobacco leaf curing degree report according to the tobacco leaf curing degree. The system comprises a tobacco leaf curing state obtaining module, a tobacco leaf state judgment model module and a curing state judgment module. According to the method, the tobacco leaf curing state can be accurately judged.

Description

Technical Field

[0001] The present invention relates to the technical field of tobacco leaf baking, and in particular to a method and system for distinguishing the baking state of tobacco leaves. Background Art

[0002] Image recognition technology has limitations, preventing it from accurately determining the final maturity of tobacco leaves. Determining the degree of yellowing, wilting, and tipping of tobacco leaves largely relies on the experience of tobacco farmers and tobacco curers, resulting in significant discrepancies in results. Manual observation and adjustment are time-consuming and labor-intensive, and are easily affected by environmental factors such as light and temperature. The lack of precise control makes it difficult to ensure tobacco leaf quality during curing, leading to uneven curing and over-curing. With the advancement of technology, modern tobacco curing has gradually incorporated intelligent and automated methods. High-definition cameras capture real-time images of tobacco leaves, combining image processing with machine learning algorithms to analyze characteristics such as color and texture, enabling accurate identification of the curing state. Near-infrared spectroscopy analyzes the chemical composition of tobacco leaves (such as moisture and sugar content), combined with artificial intelligence algorithms (such as random forests), to rapidly determine the maturity and curing state of tobacco leaves. Temperature and humidity sensors, gas sensors, and other devices monitor curing environmental parameters in real time, and combined with PLC control systems, automatically adjust curing conditions to improve the stability and efficiency of the curing process.

[0003] Prior art one, Chinese patent, patent number: 202411780984.X discloses a tobacco leaf baking system and its control method. By setting baking process curve parameters, wherein the baking process curve parameters include a first duration of the yellowing period and a second duration of the color fixing period, the yellowing period includes an early yellowing period and a late yellowing period, during the tobacco leaf baking process, a carbon dioxide detection device is used to detect the carbon dioxide concentration value in the tobacco loading chamber in real time, in the late yellowing period, the first baking time corresponding to the inflection point where the carbon dioxide concentration value begins to rise is obtained, and the first duration is adjusted according to the first baking time, and in the color fixing period, the second baking time corresponding to when the carbon dioxide concentration value reaches a preset concentration value is obtained, and the second duration is adjusted according to the second baking time. Although it can improve the control accuracy of baking parameters, improve the baking quality of tobacco leaves, and save baking costs; however, the application of image recognition technology has limitations and cannot accurately judge the final maturity of tobacco leaves.

[0004] Prior art 2, a Chinese patent with the patent number 202411771697.2, belongs to the field of model prediction technology and specifically relates to a method and system for predicting moisture and pigment content in flue-cured tobacco based on a chained integrated model. The method comprises collecting raw near-infrared spectral data of tobacco leaves at different curing stages as spectral data, applying multiple preprocessing methods to each of these data, dividing the data into training and test sets using a stratified sampling method, combining the data obtained from the different preprocessing methods with different regression methods to construct full-band prediction models, and screening the preprocessing methods. The screened preprocessing methods are combined with different characteristic band screening methods to construct characteristic band prediction models, and the optimal sub-models for moisture and pigment content prediction are determined. Multiple identical optimal sub-models are combined in a cascade manner to form a chained integrated model, which is continuously iterated and optimized to output the final prediction results. Although the method can simultaneously predict the moisture and pigment content of tobacco leaves during the curing process using multiple indicators, the application of image recognition technology has limitations and cannot accurately determine the final maturity of the tobacco leaves.

[0005] Prior art three, Chinese patent, patent number: 202411762653.3 relates to the field of tobacco leaf baking regulation technology, specifically disclosing a tobacco leaf intelligent baking platform, which is equipped with a tobacco leaf qualification judgment subsystem, a placement status judgment subsystem, a baking scheme comparison subsystem, a baking status qualification judgment subsystem, and a cooling status qualification judgment subsystem. The present invention solves the problems of traditional tobacco leaf intelligent baking in that the comprehensive management capabilities of tobacco leaf pretreatment, baking, cooling, etc. may be insufficient, the optimization capabilities of the entire process are insufficient, the integration capabilities are insufficient, and the adaptability to different tobacco leaves is insufficient. It greatly improves the overall quality of tobacco leaf products, reduces the production of unqualified products, and the mutual linkage between each subsystem optimizes the operation process and improves production efficiency. Although personalized processing of different tobacco leaf characteristics is achieved, thereby improving the final flavor and quality of the tobacco leaves, it can track and correct possible problems in the production process in a timely manner, thereby improving the response speed of production; however, the application of image recognition technology has limitations and cannot accurately judge the final maturity of the tobacco leaves.

[0006] Currently, the existing technologies 1, 2, and 3 have limitations in the application of image recognition technology and are unable to accurately determine the final maturity of tobacco leaves. To solve the above problem, the present invention provides a method and system for determining the baking state of tobacco leaves. Summary of the Invention

[0007] The main purpose of the present invention is to provide a method and system for determining the baking state of tobacco leaves, so as to solve the problem that the application of image recognition technology in the prior art is limited and cannot accurately determine the final maturity of tobacco leaves.

[0008] To achieve the above object, the present invention provides the following technical solutions:

[0009] A method for determining the curing state of tobacco leaves, comprising:

[0010] Acquire images of different tobacco leaves in a baking state, and fuse the first target region of each tobacco leaf image in segments based on a preset fusion strategy to obtain at least one target tobacco leaf image and generate an initial tobacco leaf image dataset;

[0011] Extract key tobacco leaf images from the initial tobacco leaf image dataset, automatically identify the key tobacco leaf images and grade them to generate a target tobacco leaf image dataset; build a tobacco leaf baking state discrimination model based on the historical tobacco leaf image dataset;

[0012] The real-time tobacco leaf image is input into the tobacco leaf baking state discrimination model for discrimination to identify the baking degree of the tobacco leaves; the tobacco leaf baking degree is generated into a tobacco leaf baking degree report, which is visually output and the tobacco leaf processing progress is controlled through the terminal.

[0013] As a further improvement of the present invention, the process of generating an initial tobacco leaf image dataset includes the following steps:

[0014] Obtain images of different tobacco leaves in different baking states, perform image enhancement, noise removal, and normalization preprocessing on the original tobacco leaf images; segment the preprocessed tobacco leaf images and identify the key areas of the tobacco leaves as the first target area;

[0015] Extracting feature vectors from different regions of the same tobacco leaf and the same type of regions of different tobacco leaves for the first target region; horizontally splicing the vector features from different sources in the feature dimension to obtain at least one target tobacco leaf image;

[0016] At least one target tobacco leaf image is subjected to redundant features removal, and a subset of key features is screened and integrated to form an initial tobacco leaf image dataset.

[0017] As a further improvement of the present invention, the process of generating a target tobacco leaf image dataset includes the following steps:

[0018] Based on the fuzzy relationship between yellow and green in natural environments, the color attributes of the tobacco leaf pixels in the initial tobacco leaf image dataset are inferred. By calculating the difference between green and red in the tobacco leaf image, the corresponding relationship between different maturity categories and color values is established.

[0019] The yellow and green areas are distinguished by hue and saturation parameters, and the area proportion of the yellow area is calculated; the yellow area proportion is mapped to the membership probability of different maturity levels; and the maturity of the tobacco leaves is determined according to the preset maturity categories;

[0020] A target tobacco leaf image dataset is formed according to the maturity of tobacco leaves; the target tobacco leaf image dataset is divided into a training set, a test set, and a validation set; and a tobacco leaf baking state discrimination model is constructed based on the historical tobacco leaf image dataset.

[0021] As a further improvement of the present invention, the process of establishing a correspondence between different maturity categories and color values includes the following steps:

[0022] The color features are quantified by calculating the difference between the green and red channels in the tobacco leaf image. If the difference between the green and red channels is greater than a preset value, it indicates low maturity. If the difference between the green and red channels decreases, it indicates increased maturity.

[0023] Analyze the color data of tobacco leaf samples at different maturity levels and calculate the difference between the green channel and the red channel corresponding to each maturity category. The maturity categories are underripe, moderately ripe, moderately ripe, and overripe.

[0024] A piecewise function is defined to map the difference between the continuous green and red channels to a discrete probability distribution of maturity levels to eliminate color bias. The correlation between the difference between the green and red channels and maturity is verified through experiments, and the threshold range is adjusted.

[0025] As a further improvement of the present invention, the process of determining the maturity of tobacco leaves includes the following steps:

[0026] receiving an input area ratio of a yellow area, dividing the yellow area into a plurality of area segments, each area segment including a first plurality of yellow image pixels, and dividing the plurality of area segments into a plurality of sub-area segments, each sub-area segment including a second plurality of yellow image pixels;

[0027] estimating grayscale values of the yellow image corresponding to the first plurality of yellow image pixels and the second plurality of yellow image pixels; calculating a standard deviation based on the grayscale values; and responding to yellow image pixels in which the standard deviation of the first plurality of yellow image pixels and the second plurality of yellow image pixels is lower than a set threshold and is not lower than the set threshold;

[0028] The yellow area proportion corresponding to the yellow image pixels that are not less than the set threshold is mapped to the membership probability of different maturity levels.

[0029] As a further improvement of the present invention, the process of constructing a tobacco leaf baking state discrimination model based on a historical tobacco leaf image dataset includes the following steps:

[0030] A target tobacco leaf image dataset is formed based on the maturity of the tobacco leaves; the target tobacco leaf image dataset is divided into a training set, a test set, and a validation set; the color and texture features of the tobacco leaves are extracted; the texture eigenvalues are obtained through the angular second moment, entropy, and contrast index of the gray-level co-occurrence matrix; and the texture eigenvalues are subjected to variable cluster analysis and correlation analysis;

[0031] A set of parameter combinations is randomly generated as the initial solution; the classification accuracy is used as the fitness indicator to evaluate the performance of each parameter set; the parameters are iteratively optimized through genetic operations to ultimately select the optimal tobacco leaf curing parameter combination;

[0032] The optimized optimal parameter combination was used to train the tobacco leaf baking state model on the training set data, and the complexity of the tobacco leaf baking state model was adjusted through the validation set, and the generalization ability of the tobacco leaf baking state model was evaluated using the test set.

[0033] As a further improvement of the present invention, the process of performing variable cluster analysis and correlation analysis on texture feature values includes the following steps:

[0034] Calculate the internal correlation coefficient matrix for color, contour and texture features respectively, extract high correlation indicators to reduce the dimension; group the extracted multiple texture features according to similarity; set the distance threshold for clustering distance and classify the texture features;

[0035] For each cluster, measure the correlation with the target variable. Analyze the relationship between features and select the features with the strongest correlation with the baking stage in each cluster. Retain the features with the highest correlation with the target variable in each cluster and eliminate redundant features.

[0036] The data with redundant features removed is classified and evaluated to eventually form a tobacco leaf feature subset, which is used as a dataset for training the model.

[0037] As a further improvement of the present invention, the process of ultimately screening out the optimal tobacco leaf curing parameter combination includes the following steps:

[0038] A set of tobacco leaf curing parameter combinations is randomly generated as the initial population; wherein the tobacco leaf curing combination includes key control variables of temperature, humidity and curing time;

[0039] Using classification accuracy as an indicator of adaptability, the performance of each set of tobacco leaf curing parameters was quantitatively evaluated. Tobacco leaf parameter combinations were screened based on fitness values, and individuals with performance that met the standards were retained.

[0040] Perform a genetic crossover operation on the selected parent parameters to generate new offspring parameter combinations; randomly perturb the offspring parameters with a preset probability to increase population diversity; repeat the selection, crossover, and mutation process until the preset number of iterations is reached; finally, select the parameter combination with the highest classification accuracy from the last generation population as the optimal solution.

[0041] As a further improvement of the present invention, the process of evaluating the generalization ability of the tobacco leaf curing state model using a test set includes the following steps:

[0042] The optimized optimal tobacco leaf curing parameter combination is used to train the tobacco leaf curing state model on the training set data; hyperparameter tuning, cross-validation, and complexity control are performed to adjust the tobacco leaf curing state model;

[0043] The tobacco curing stages were classified using comprehensive accuracy, recall, and F1 metrics. The optimal data for identifying key curing stages was selected. The adaptability of the tobacco curing state model to unknown data was verified using an independent test set to obtain the optimal tobacco curing state model.

[0044] The tobacco leaf data collected in real time is input into the tobacco leaf baking status model for evaluation and judgment; when an anomaly is detected, an early warning is issued; and the final result is output visually.

[0045] To achieve the above object, the present invention also provides the following technical solutions:

[0046] A tobacco leaf baking state determination system is applied to the tobacco leaf baking state determination method. The tobacco leaf baking state determination system comprises:

[0047] A tobacco leaf baking state acquisition module is used to acquire baking state images of different tobacco leaves, and based on a preset fusion strategy, segmentally fuse the first target area of each tobacco leaf image to obtain at least one target tobacco leaf image and generate an initial tobacco leaf image dataset;

[0048] The tobacco leaf state discrimination model module is used to extract key tobacco leaf images from the initial tobacco leaf image dataset, automatically identify the key tobacco leaf images and grade them to generate a target tobacco leaf image dataset; and build a tobacco leaf baking state discrimination model based on the historical tobacco leaf image dataset;

[0049] The baking state discrimination module is used to input the real-time tobacco leaf image into the tobacco leaf baking state discrimination model for discrimination and identify the baking degree of the tobacco leaves; generate a tobacco leaf baking degree report based on the tobacco leaf baking degree, output it visually, and control the tobacco leaf processing progress through the terminal.

[0050] The present invention uses image processing and machine learning technologies to accurately identify the baking status of tobacco leaves, reduce the subjectivity and errors of manual judgment, and improve the scientific nature and applicability of the baking process; based on the identification results, the baking parameters (such as temperature, humidity, etc.) are adjusted in real time to ensure the consistency and stability of the tobacco leaf baking process, thereby improving the quality of tobacco leaves; through automation and intelligent technology, the need for manual monitoring and operation is reduced, labor intensity is reduced, and production efficiency is improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] Figure 1 This is a schematic flow chart of steps of an embodiment of a method for determining the baking state of tobacco leaves according to the present invention;

[0052] Figure 2 A schematic flow chart of steps for generating an initial tobacco leaf image dataset according to an embodiment of a tobacco leaf baking state discrimination method of the present invention;

[0053] Figure 3 A schematic flow chart of steps for generating a target tobacco leaf image dataset according to an embodiment of a tobacco leaf baking state discrimination method of the present invention;

[0054] Figure 4 This is a flow chart of steps for controlling tobacco leaf processing progress through a terminal according to one embodiment of a tobacco leaf baking state determination method of the present invention;

[0055] Figure 5 This is a functional module diagram of an embodiment of a tobacco baking state discrimination system of the present invention;

[0056] Figure 6 This is a schematic structural diagram of an embodiment of an electronic device of the present invention;

[0057] Figure 7 This is a schematic structural diagram of an embodiment of a storage medium of the present invention. DETAILED DESCRIPTION

[0058] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. All other embodiments obtained by ordinary technicians in this field based on the embodiments of the present invention without making any creative efforts shall fall within the scope of protection of the present invention.

[0059] The terms "first", "second" and "third" in the present invention are used for descriptive purposes only and should not be understood as indicating or implying relative importance or implicitly indicating the number of the indicated technical features. Therefore, the features defined as "first", "second" and "third" may explicitly or implicitly include at least one of the features. In the description of the present invention, the meaning of "multiple" is at least two, for example, two, three, etc., unless otherwise clearly and specifically defined. All directional indications in the embodiments of the present invention (such as up, down, left, right, front, back...) are only used to explain the relative position relationship, movement, etc. between the components under a certain specific posture (as shown in the accompanying drawings). If the specific posture changes, the directional indication will also change accordingly. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions. For example, a process, method, system, product or device that includes a series of steps or units is not limited to the listed steps or units, but may optionally include steps or units that are not listed, or may optionally include other steps or units that are inherent to these processes, methods, products or devices.

[0060] References herein to "embodiments" mean that a particular feature, structure, or characteristic described in connection with the embodiments may be included in at least one embodiment of the present invention. The appearance of this phrase in various places in the specification does not necessarily refer to the same embodiment, nor does it constitute a separate or alternative embodiment that is mutually exclusive of other embodiments. It is understood, both explicitly and implicitly, by those skilled in the art that the embodiments described herein may be combined with other embodiments.

[0061] like Figure 1 As shown, this embodiment provides an embodiment of a method for determining the baking state of tobacco leaves. In this embodiment, the method for determining the baking state of tobacco leaves specifically includes the following steps:

[0062] Step S1: acquiring images of different tobacco leaves in a baking state, and fusing the first target region of each tobacco leaf image in sections based on a preset fusion strategy to obtain at least one target tobacco leaf image and generate an initial tobacco leaf image dataset;

[0063] Step S2: extracting key tobacco leaf images from the initial tobacco leaf image dataset, and automatically identifying the key tobacco leaf images to generate a target tobacco leaf image dataset; building a tobacco leaf baking state discrimination model based on the historical tobacco leaf image dataset;

[0064] Step S3: Input the real-time tobacco leaf image into the tobacco leaf curing state discrimination model for discrimination to identify the tobacco leaf curing degree; generate a tobacco leaf curing degree report based on the tobacco leaf curing degree, output it visually, and control the tobacco leaf processing progress through the terminal.

[0065] Preferably, this embodiment obtains the baking state images of different tobacco leaves, and performs segmented fusion on the first target area based on a preset fusion strategy to generate an initial tobacco leaf image dataset; extracts key images from the initial tobacco leaf image dataset, and uses automatic recognition technology to grade the baking degree of the tobacco leaves to generate a target tobacco leaf image dataset; constructs a tobacco leaf baking state discrimination model based on the historical tobacco leaf image dataset, and inputs real-time tobacco leaf images into the model for discrimination; generates a report on the baking degree of the tobacco leaves and outputs it visually, and regulates the tobacco leaf processing progress through the terminal. This embodiment uses image processing and machine learning technology to achieve accurate discrimination of the baking state of tobacco leaves, reduces the subjectivity and error of manual judgment, and improves the scientific nature and applicability of the baking process; based on the discrimination results, adjusts the baking parameters (such as temperature, humidity, etc.) in real time to ensure the consistency and stability of the tobacco leaf baking process, thereby improving the quality of tobacco leaves; through automation and intelligent technology, reduces the need for manual monitoring and operation, reduces labor intensity, and improves production efficiency.

[0066] Furthermore, if Figure 2 As shown, the process of generating the initial tobacco leaf image dataset in step S1 specifically includes the following steps:

[0067] Step S11: acquiring images of different tobacco leaves in a baking state, performing image enhancement, noise removal, and normalization preprocessing on the original tobacco leaf images; segmenting the preprocessed tobacco leaf images, and identifying key areas of the tobacco leaves as first target areas;

[0068] Step S12: extracting feature vectors from different regions of the same tobacco leaf and the same type of regions of different tobacco leaves for the first target region; horizontally splicing the vector features from different sources in the feature dimension to obtain at least one target tobacco leaf image;

[0069] Among them, horizontal splicing includes sequentially connecting the color feature vector and texture feature vector of the same tobacco leaf to form a multi-dimensional coincident feature vector; splicing the feature vectors of the same area of different tobacco leaves;

[0070] Step S13: removing redundant features from at least one target tobacco leaf-removed image, screening a subset of key features, and integrating them to form an initial tobacco leaf image dataset.

[0071] Preferably, in this embodiment, the quality of tobacco leaf images is improved through operations such as image enhancement, noise removal, and normalization, providing clear input for processing; the key areas of tobacco leaves are identified by image segmentation technology as the first target area; multidimensional feature vectors such as color and texture are extracted from the target area, and feature vectors from different sources are integrated into multidimensional conforming feature vectors by horizontal splicing; an initial tobacco leaf image dataset is formed by removing redundant features and screening key feature subsets. Image enhancement and noise removal improve image quality, making segmentation and feature extraction more accurate; the splicing method of multidimensional feature vectors can capture multiple characteristics of tobacco leaves (such as color, texture, etc.), thereby improving the model's ability to recognize tobacco leaf status; redundant information is removed through feature screening and integration, forming a high-quality initial dataset, which provides a reliable foundation for classification or recognition tasks; the overall process significantly improves the accuracy and robustness of tobacco leaf status classification or recognition through multi-step optimization.

[0072] Furthermore, if Figure 3 As shown, the process of generating the target tobacco leaf image dataset in step S2 specifically includes the following steps:

[0073] Step S21: Based on the fuzzy relationship between yellow and green in a natural environment, the color attributes of the pixels in the tobacco leaf area in the initial tobacco leaf image dataset are inferred; by calculating the difference between green and red in the tobacco leaf image, the corresponding relationship between different maturity categories and color values is established;

[0074] Step S22: Differentiating yellow and green areas by hue and saturation parameters, and calculating the area ratio of the yellow area; mapping the yellow area ratio to the membership probability of different maturity levels; and determining the maturity of the tobacco leaves according to the preset maturity category;

[0075] Step S23: forming a target tobacco leaf image dataset according to the maturity of the tobacco leaves; dividing the target tobacco leaf image dataset into a training set, a test set, and a validation set; and constructing a tobacco leaf baking state discrimination model based on the historical tobacco leaf image dataset.

[0076] Preferably, in this embodiment, the color attributes of pixels in tobacco leaf images are inferred through the fuzzy relationship between yellow and green in a natural environment, and a corresponding relationship between color values and maturity categories is established; the yellow and green areas are distinguished by hue and saturation parameters, and the area proportion of the yellow area is counted; the yellow area proportion is mapped to the membership probability of different maturity levels, further optimizing the accuracy of maturity judgment; the target image data set is divided according to the maturity of tobacco leaves, and is divided into a training set, a test set, and a validation set, and a tobacco leaf baking state discrimination model is constructed based on historical data. This embodiment eliminates the influence of natural light on tobacco leaf color information through fuzzy lighting processing and color feature extraction, significantly improving the accuracy of maturity recognition; it can adapt to tobacco leaf images under different lighting conditions, reduce the interference of environmental factors on the recognition results, and enhance the generalization ability of the model; through the combination of image processing technology and machine learning models, the automatic discrimination of tobacco leaf maturity is realized, which reduces manual intervention and improves efficiency; it provides a scientific basis for tobacco leaf harvesting and baking, helps to optimize the production process, and improve tobacco leaf quality.

[0077] Furthermore, the process of establishing the correspondence between different maturity categories and color values in step S21 specifically includes the following steps:

[0078] Step S211: Calculating the difference between the green channel and the red channel in the tobacco leaf image to quantify the color features; wherein, when the difference between the green channel and the red channel is greater than a preset value, it indicates low maturity; and if the difference between the green channel and the red channel decreases, it indicates increased maturity;

[0079] Step S212: Analyze the color data of tobacco leaf samples of different maturity levels, and calculate the difference range between the green channel and the red channel corresponding to each maturity level; wherein each maturity level is classified into underripe, moderately ripe, moderately ripe, and overripe;

[0080] Step S213: Define a piecewise function to map the difference between the continuous green channel and the red channel to a discrete probability distribution of maturity levels to eliminate color deviation; verify the correlation between the difference between the green channel and the red channel and maturity through experiments, and adjust the threshold range.

[0081] Preferably, this embodiment quantifies color features by calculating the difference between the green channel and the red channel in the tobacco leaf image, and uses color changes to reflect the maturity of the tobacco leaves; maps the difference between the continuous green channel and the red channel to a discrete maturity level probability distribution, adjusts the threshold range through experiments, eliminates color deviation, and improves discrimination accuracy; by analyzing the color data of tobacco leaf samples with different maturity levels, statistics the difference range between the green channel and the red channel corresponding to each maturity category, and defines a piecewise function for mapping to ensure the scientific nature and practicality of the model; verifies the correlation between the difference between the green channel and the red channel and maturity through experiments, and adjusts the threshold range according to the experimental results to ensure the accuracy and robustness of the model. This embodiment, through the quantification of color features and the application of piecewise functions, can more accurately judge the maturity of tobacco leaves, reduce subjective errors in manual judgment, and can quickly and automatically complete the judgment of tobacco leaf maturity, reducing the time and cost of manual operation and improving production efficiency. This embodiment provides technical support for the automation and intelligence of tobacco leaf maturity judgment, which helps to promote the technological upgrading and sustainable development of the tobacco industry. By accurately judging the maturity of tobacco leaves, a scientific basis can be provided for tobacco leaf grading and harvesting, thereby improving tobacco leaf quality and economic benefits.

[0082] Furthermore, the process of determining the maturity of tobacco leaves in step S22 specifically includes the following steps:

[0083] Step S221: receiving an input area ratio of a yellow area, dividing the yellow area into a plurality of area segments, each area segment including a first plurality of yellow image pixels, and dividing the plurality of area segments into a plurality of sub-area segments, each sub-area segment including a second plurality of yellow image pixels;

[0084] Step S222: estimating grayscale values of the yellow image corresponding to the first plurality of yellow image pixels and the second plurality of yellow image pixels; calculating a standard deviation based on the grayscale values; and responding to yellow image pixels in which the standard deviation of the first plurality of yellow image pixels and the second plurality of yellow image pixels is lower than a set threshold and is not lower than the set threshold;

[0085] Step S223: mapping the yellow area proportions corresponding to the yellow image pixels that are not less than the set threshold to membership probabilities of different maturity levels.

[0086] Among them, the yellow area ratio is calculated as follows:

[0087]

[0088] Where, P yellow Indicates the area ratio of the yellow area; N seg Indicates the total number of region segments divided in the image; M sub Indicates the total number of sub-region segments in each region segment; C yellow(i, j) represents the number of yellow pixels in the i-th region and the j-th sub-region; T total Indicates the total number of pixels in the image;

[0089] Gray value standard deviation calculation:

[0090]

[0091] Where, σ gray represents the standard deviation of the grayscale value of the yellow area; G(i,j) represents the average grayscale value of the yellow pixels in the i-th region segment and the j-th sub-region segment; μ gray Represents the average gray value of all yellow pixels; N seg Indicates the total number of regional segments; M sub Indicates the total number of sub-region segments;

[0092] Maturity membership probability mapping:

[0093]

[0094] Where, P mat represents the probability of maturity membership; H(i,j) represents the yellow area ratio in the i-th region and the j-th sub-region; G(i,j) represents the average grayscale value of the yellow pixels in the i-th region and the j-th sub-region; G th represents the set threshold of the gray value; δ(·) represents the indicator function, which is 1 when the condition is met and 0 otherwise;

[0095] Maturity level classification:

[0096]

[0097] Where, L mat Indicates the maturity level of the final judgment; L mat represents the maturity membership probability; a k ,b k The Sigmoid function parameter representing the kth maturity level; argmax k The above formula fully describes the calculation process of tobacco leaf maturity determination through refined mathematical expression, ensuring the rigor and complexity of the technical characteristics.

[0098] Preferably, this embodiment receives and analyzes the area ratio of the yellow area in the tobacco leaf image, divides it into multiple area segments and sub-area segments, and counts the number of yellow image pixels in each area. By calculating the grayscale values of these pixels, the standard deviation is obtained to evaluate the uniformity and degree of change of the yellow area. When the standard deviation is lower than the set threshold, it indicates that the color change of the yellow area is small and the maturity is relatively consistent; otherwise, it indicates that there is a significant color change in the yellow area and the maturity is not uniform enough; finally, the yellow area ratio corresponding to the yellow image pixels that are not lower than the set threshold is mapped and processed to obtain the membership probability of different maturity levels, thereby achieving quantitative evaluation and classification judgment of tobacco leaf maturity. This embodiment realizes accurate and automated analysis of tobacco leaf maturity, improves the accuracy and reliability of the judgment results, and provides a scientific basis for tobacco leaf baking processing.

[0099] Furthermore, the process of constructing a tobacco leaf baking state discrimination model based on the historical tobacco leaf image dataset in step S23 specifically includes the following steps:

[0100] Step S231: forming a target tobacco leaf image dataset based on the maturity of the tobacco leaves; dividing the target tobacco leaf image dataset into a training set, a test set, and a validation set; extracting color and texture features of the tobacco leaves; obtaining texture eigenvalues using indices such as the angular second moment, entropy, and contrast of the gray-level co-occurrence matrix; and performing variable cluster analysis and correlation analysis on the texture eigenvalues;

[0101] Among them, color includes parameters such as hue, saturation and brightness;

[0102] Step S232: randomly generating a set of parameter combinations as an initial solution; using classification accuracy as a fitness indicator to evaluate the performance of each set of parameters; iteratively optimizing the parameters through genetic operations, and ultimately selecting the optimal tobacco leaf curing parameter combination;

[0103] Step S233: Use the optimized optimal parameter combination to train the tobacco leaf curing state model on the training set data, adjust the complexity of the tobacco leaf curing state model through the validation set, and use the test set to evaluate the generalization ability of the tobacco leaf curing state model.

[0104] Among them, the expression for adjusting the complexity of the tobacco leaf baking state model through the validation set is:

[0105]

[0106]

[0107] Where LRadj represents the adjusted learning rate; LRinit represents the initial learning rate; γ represents the attenuation coefficient, which controls the adjustment amplitude of the learning rate; E val represents the mean square error on the validation set; Nval represents the number of samples in the validation set; y i′ represents the true label of the i′th sample; represents the predicted label of the i′th sample; δ represents the smoothing factor to avoid the denominator being zero.

[0108] Preferably, this embodiment uses image processing technology to divide the tobacco leaf image data set into a training set, a test set, and a validation set, and extracts the color (hue, saturation, brightness) and texture features of the tobacco leaves (such as the angular second moment of the grayscale co-occurrence matrix, entropy, contrast, etc.); randomly generates initial parameter combinations, and uses classification accuracy as the fitness index, iteratively optimizes parameters through genetic operations, and finally screens out the optimal tobacco leaf baking parameter combination; uses the optimized parameter combination to train the tobacco leaf baking state model on the training set data, and adjusts the model complexity through the validation set, and finally uses the test set to evaluate the generalization ability of the model. This embodiment uses image recognition and classification technology, combined with genetic algorithm optimization parameters, to achieve precise control of the tobacco leaf baking process, significantly improving the consistency and scientificity of the baking effect; automated feature extraction and parameter optimization reduce the subjectivity and errors of manual operation, and improve production efficiency and product quality; through cross-validation of the validation set and the test set, the applicability and generalization ability of the model on different data sets are ensured, and the overfitting problem is avoided; this embodiment provides technical support for the intelligent discrimination of the tobacco leaf baking state, and promotes the intelligent and modern development of the tobacco processing industry.

[0109] Furthermore, the process of performing variable cluster analysis and correlation analysis on the texture feature values in step S231 specifically includes the following steps:

[0110] Step S2311: Calculate the internal correlation coefficient matrix for color, contour, and texture features respectively, extract high correlation indicators to reduce the dimension; group the extracted multiple texture features according to similarity; set a distance threshold for the cluster distance and classify the texture features;

[0111] Step S2312: For each feature in each cluster, measure its correlation with the target variable; analyze the relationship between features and select the feature with the strongest correlation with the baking stage in each cluster; retain the feature with the highest correlation with the target variable in each cluster and eliminate redundant features;

[0112] Step S2313: Classify and evaluate the data after removing redundant features to ultimately form a tobacco leaf feature subset; wherein the tobacco leaf feature subset is used as a data set for training the model.

[0113] Preferably, this embodiment calculates the internal correlation coefficient matrix of features such as color, contour and texture, extracts high correlation indicators to reduce the dimension; groups the extracted texture features according to similarity, and classifies the clustering results by setting a distance threshold; classifies and evaluates the data after eliminating redundant features, and finally forms a tobacco leaf feature subset. This embodiment reduces the amount of calculation of redundant data and improves data processing speed and model training efficiency through dimensionality reduction and feature screening; screens out features that are highly correlated with the target variable, reduces the complexity of the model, and improves the prediction accuracy and generalization ability of the model; and through cluster analysis and threshold setting, classifies similar features into one category, further optimizing the quality and stability of the classification results; the tobacco leaf feature subset finally generated provides reliable data support for intelligent classification and discrimination in the baking stage, which helps to realize the automation and intelligence of tobacco leaf grading.

[0114] Furthermore, the process of finally screening out the optimal tobacco leaf curing parameter combination in step S232 specifically includes the following steps:

[0115] Step S2321: randomly generating a set of tobacco leaf curing parameter combinations as an initial population; wherein the tobacco leaf curing combination includes key control variables such as temperature, humidity, and curing time;

[0116] Step S2322: using the classification accuracy as the adaptability index, quantitatively evaluate the performance of each set of tobacco leaf curing parameters; screening tobacco leaf parameter combinations based on the fitness values, and retaining individuals that meet the performance standards;

[0117] Step S2323: Perform a genetic crossover operation on the selected parent generation parameters to generate a new offspring parameter combination; randomly perturb the offspring parameters with a preset probability to increase population diversity; repeat the selection, crossover, and mutation process until the preset number of iterations is reached; finally, select the parameter combination with the highest classification accuracy from the last generation population as the optimal solution.

[0118] Preferably, this embodiment randomly generates a set of tobacco leaf baking parameter combinations (such as temperature, humidity and baking time, etc.) as the initial population to ensure the diversity and randomness of the population; uses classification accuracy as an adaptability indicator to quantitatively evaluate the performance of each set of parameters, and screens out individuals with performance that meets the standards based on the fitness value; generates new offspring parameter combinations through gene crossover and mutation operations to increase population diversity and prevent the algorithm from converging prematurely; repeats the selection, crossover and mutation process until a preset number of iterations is reached, and finally selects the parameter combination with the highest classification accuracy from the last generation population as the optimal solution. This embodiment can significantly improve the classification accuracy and baking quality of tobacco leaves by optimizing the baking parameter combination, ensuring that the appearance, aroma and chemical composition of tobacco leaves reach the optimal state; randomly generating initial populations and genetic operations (such as crossover and mutation) increases the diversity of the population, improves the global search capability of the algorithm, and avoids the trap of local optimal solutions; through the automated optimization process, reduces the need for manual parameter setting, and improves the intelligence level of the baking process; the optimized baking parameters can reduce energy consumption costs while increasing the market value of tobacco leaves, with significant economic benefits.

[0119] Furthermore, the process of using the test set to evaluate the generalization ability of the tobacco leaf curing state model in step 233 specifically includes the following steps:

[0120] Step S2331: using the optimized optimal tobacco leaf curing parameter combination, respectively training the tobacco leaf curing state model on the training set data; performing operations such as hyperparameter tuning, cross validation, and complexity control to adjust the tobacco leaf curing state model;

[0121] Step S2332: using comprehensive accuracy, recall rate, F1 and other indicators to classify tobacco leaf curing stages; screening the data with the best recognition effect on key curing stages; verifying the adaptability of the tobacco leaf curing state model to unknown data through an independent test set, and obtaining the optimal tobacco leaf curing state model;

[0122] Step S2333: input the tobacco leaf data collected in real time into the tobacco leaf baking state model for evaluation and judgment; when an abnormality is detected, issue an early warning; and output the final result in a visual format.

[0123] Preferably, this embodiment uses the optimized tobacco leaf baking parameter combination to perform model training on the training set data, and combines hyperparameter tuning, cross-validation and complexity control methods to improve the performance and generalization ability of the model; uses comprehensive accuracy, recall rate, F1 score and other indicators to classify the tobacco leaf baking stages, and screens out the data with the best recognition effect on the key baking stages; inputs the tobacco leaf data collected in real time into the model for evaluation and judgment, issues an early warning when an anomaly is detected, and outputs the final result through visualization technology. This embodiment can optimize baking process parameters (such as temperature and humidity) through precise model prediction and real-time monitoring, reduce baking losses, and improve the quality and economic benefits of tobacco leaves; the intelligent system reduces dependence on manual experience and reduces labor intensity, while reducing losses caused by improper operation through automated control and early warning mechanisms; and through independent test set verification and multi-indicator evaluation, ensures the adaptability of the model to unknown data, thereby improving the practical application value of the model; by improving the efficiency and quality of tobacco leaf baking, it helps to increase farmers' income and promote the sustainable development of the tobacco industry.

[0124] Furthermore, if Figure 4 As shown, the process of regulating the tobacco leaf processing progress through the terminal in step S3 specifically includes the following steps:

[0125] Step S31: Classify and identify the tobacco leaf state based on the tobacco leaf curing state model and output the curing stage determination result; evaluate the moisture migration state by combining the correlation analysis between the tobacco leaf free water content and the morphological shrinkage rate; and analyze the deviation of the current parameters with reference to the historical curing process execution curve;

[0126] Step S32: Automatically generate a tobacco leaf status report including real-time roasting stage determination, key indicator trend charts, quality assessment parameters, process optimization suggestions, and abnormality warnings; and convert the tobacco leaf status report into a visual interface through the cloud platform;

[0127] The visualization interface includes a heat map showing the baking uniformity of different areas of the tobacco leaf, a three-dimensional model showing the changes in tobacco leaf morphology, and a dynamic process execution curve comparison chart.

[0128] Step S33: associate each report with the original image data, the tobacco leaf baking state model determination result and the sensor calibration data record, and automatically store the tobacco leaf report in the tobacco leaf quality traceability database.

[0129] Preferably, this embodiment classifies and identifies the tobacco leaf status through a tobacco leaf baking state model, combines the correlation analysis between free water content and morphological shrinkage rate to evaluate the moisture migration status, and refers to the historical baking process execution curve to perform deviation analysis on the current parameters; and generates a real-time tobacco leaf status report through the cloud platform, including baking stage judgment, key indicator trend chart, quality assessment parameters, process optimization suggestions and abnormal warnings; and through visualization means such as heat maps, three-dimensional models and dynamic process execution curves, intuitively displays the uniformity, morphological changes and process execution of tobacco leaf baking, thereby improving the operator's understanding and control ability of the baking process; each report is associated with the original image data, the tobacco leaf baking state model judgment results and the sensor calibration data records, and stored in the tobacco leaf quality traceability database, thereby realizing the full traceability of tobacco leaf quality. This embodiment uses image recognition technology and real-time data analysis to accurately judge the baking stage and moisture migration status of tobacco leaves, avoids errors caused by traditional subjective judgment, and significantly improves the scientific nature and applicability of the baking process; combines historical data and real-time monitoring to dynamically adjust baking parameters (such as temperature, humidity, wind speed, etc.), optimize the baking process, reduce resource waste, and improve tobacco quality; this embodiment uses an intelligent system to reduce dependence on manual experience, reduce labor intensity, and reduce tobacco loss caused by improper operation; this embodiment improves the appearance, aroma and chemical composition of tobacco leaves by accurately controlling key parameters in the baking process, thereby improving the overall quality and market competitiveness of tobacco leaves; and through a data traceability system, ensures that the production process and quality information of each batch of tobacco leaves are traceable, provides a reference basis for production, and meets quality supervision needs.

[0130] like Figure 5 As shown, this embodiment also provides an embodiment of a tobacco baking state discrimination system. In this embodiment, the tobacco baking state discrimination system is applied to the tobacco baking state discrimination method in the above embodiment. The tobacco baking state discrimination system includes:

[0131] The tobacco leaf baking state acquisition module 1 is used to acquire baking state images of different tobacco leaves, and fuse the first target area of each tobacco leaf image in segments based on a preset fusion strategy to obtain at least one target tobacco leaf image and generate an initial tobacco leaf image dataset;

[0132] Tobacco leaf state discrimination model module 2 is used to extract key tobacco leaf images from the initial tobacco leaf image dataset, automatically identify the key tobacco leaf images and generate a target tobacco leaf image dataset; and build a tobacco leaf baking state discrimination model based on the historical tobacco leaf image dataset;

[0133] The baking state judgment module 3 is used to input the real-time tobacco leaf image into the tobacco leaf baking state judgment model for judgment and identify the baking degree of the tobacco leaf; generate a tobacco leaf baking degree report based on the tobacco leaf baking degree, output it visually, and control the tobacco leaf processing progress through the terminal.

[0134] Preferably, this embodiment uses a high-definition camera or other equipment to collect image data of the tobacco leaf baking process in real time, and pre-processes the image (such as denoising, contrast enhancement, etc.) to improve image quality and provide a reliable basis for analysis; uses image segmentation technology (such as convolutional neural network) to extract target areas in the tobacco leaf image, and fuses these areas in segments according to a preset fusion strategy to generate a target tobacco leaf image; extracts key features of the tobacco leaf image (such as color, texture, shape, etc.) through deep learning and machine learning algorithms (such as support vector machines, extreme learning machines, etc.), and constructs a tobacco leaf baking state discrimination model based on historical data; inputs the real-time collected tobacco leaf image into the discrimination model, automatically identifies the baking degree of the tobacco leaf, and generates a visual report, and regulates the tobacco leaf processing progress through the terminal. This embodiment uses intelligent image recognition technology to accurately identify the baking status of tobacco leaves, reduce manual intervention, and improve baking efficiency and quality; based on real-time data and historical data analysis, it dynamically adjusts baking parameters (such as temperature, humidity, etc.) to ensure the consistency and stability of the baking process; and replaces manual operations with automated systems, reducing labor intensity and improving production efficiency; and through accurate identification of tobacco leaf maturity and baking status, effectively improving the industrial availability and quality of tobacco leaves; through data analysis and model optimization, it provides a scientific basis for improving the tobacco leaf baking process, further promoting the intelligent development of the tobacco industry.

[0135] like Figure 6 As shown, this embodiment provides an embodiment of an electronic device. In this embodiment, the electronic device 4 includes a processor 41 and a memory 42 coupled to the processor 41.

[0136] The memory 42 stores program instructions for implementing the tobacco leaf baking state determination method of any of the above embodiments.

[0137] The processor 41 is used to execute the program instructions stored in the memory 42 to determine the baking state of the tobacco leaves.

[0138] The processor 41 may also be referred to as a CPU (Central Processing Unit). The processor 41 may be an integrated circuit chip having signal processing capabilities. The processor 41 may also be a general-purpose processor, a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA), or other programmable logic device, a discrete gate or transistor logic device, or a discrete hardware component. The general-purpose processor may be a microprocessor or any conventional processor.

[0139] Further, Figure 7 This is a schematic diagram of the structure of a storage medium in an embodiment of the present application. The storage medium 5 in the embodiment of the present application stores program instructions 51 that can implement all the above methods, wherein the program instructions 51 can be stored in the above storage medium in the form of a software product, including a number of instructions for causing a computer device (which can be a personal computer, server, or network device, etc.) or a processor to execute all or part of the steps of the method described in each embodiment of the present application. The aforementioned storage medium includes: various media that can store program codes, such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk, or terminal devices such as a computer, server, mobile phone, and tablet.

[0140] In the several embodiments provided by the present invention, it should be understood that the disclosed systems, devices, and methods can be implemented in other ways. For example, the device embodiments described above are merely illustrative. For example, the division of units is merely a logical function division. In actual implementation, there may be other division methods, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. In addition, the mutual coupling or direct coupling or communication connection shown or discussed can be an indirect coupling or communication connection through some interface, device or unit, which can be electrical, mechanical or other forms.

[0141] In addition, the functional units in the various embodiments of the present invention may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated units may be implemented in the form of hardware or in the form of software functional units. The above is only an embodiment of the present invention and does not limit the patent scope of the present invention. Any equivalent structure or equivalent process transformation made using the contents of the present invention specification and drawings, or directly or indirectly applied in other related technical fields, are also included in the patent protection scope of the present invention.

[0142] The above detailed description of the specific embodiments of the invention is intended to be illustrative only, and the present invention is not limited to the specific embodiments described above. For those skilled in the art, any equivalent modifications or substitutions to the invention are also within the scope of the present invention. Therefore, equivalent changes, modifications, and improvements made without departing from the spirit and scope of the present invention are also encompassed within the scope of the present invention.

Claims

1. A method for determining the baking state of tobacco leaves, characterized in that: The method for determining the tobacco leaf baking state comprises: Acquire images of different tobacco leaves in a baking state, and fuse the first target region of each tobacco leaf image in segments based on a preset fusion strategy to obtain at least one target tobacco leaf image and generate an initial tobacco leaf image dataset; Extract key tobacco leaf images from the initial tobacco leaf image dataset, automatically identify the key tobacco leaf images and grade them to generate a target tobacco leaf image dataset; build a tobacco leaf baking state discrimination model based on the historical tobacco leaf image dataset; The real-time tobacco leaf image is input into the tobacco leaf baking state discrimination model for discrimination to identify the baking degree of the tobacco leaves; the tobacco leaf baking degree is generated into a tobacco leaf baking degree report, which is visually output and the tobacco leaf processing progress is controlled through the terminal.

2. The method for distinguishing the baking state of tobacco leaves according to claim 1, wherein: The process of generating the initial tobacco leaf image dataset includes the following steps: Obtain images of different tobacco leaves in different baking states, perform image enhancement, noise removal, and normalization preprocessing on the original tobacco leaf images; segment the preprocessed tobacco leaf images and identify the key areas of the tobacco leaves as the first target area; Extracting feature vectors from different regions of the same tobacco leaf and the same type of regions of different tobacco leaves for the first target region; horizontally splicing the vector features from different sources in the feature dimension to obtain at least one target tobacco leaf image; At least one target tobacco leaf image is subjected to redundant features removal, and a subset of key features is screened and integrated to form an initial tobacco leaf image dataset.

3. The method for distinguishing the baking state of tobacco leaves according to claim 1, wherein: The process of generating the target tobacco leaf image dataset includes the following steps: Based on the fuzzy relationship between yellow and green in natural environments, the color attributes of the tobacco leaf pixels in the initial tobacco leaf image dataset are inferred. By calculating the difference between green and red in the tobacco leaf image, the corresponding relationship between different maturity categories and color values is established. The yellow and green areas are distinguished by hue and saturation parameters, and the area proportion of the yellow area is calculated; the yellow area proportion is mapped to the membership probability of different maturity levels; and the maturity of the tobacco leaves is determined according to the preset maturity categories; A target tobacco leaf image dataset is formed according to the maturity of tobacco leaves; the target tobacco leaf image dataset is divided into a training set, a test set, and a validation set; and a tobacco leaf baking state discrimination model is constructed based on the historical tobacco leaf image dataset.

4. The method for determining the tobacco leaf baking state according to claim 3, wherein: The process of establishing the correspondence between different maturity categories and color values includes the following steps: The color features are quantified by calculating the difference between the green and red channels in the tobacco leaf image. If the difference between the green and red channels is greater than a preset value, it indicates low maturity. If the difference between the green and red channels decreases, it indicates increased maturity. Analyze the color data of tobacco leaf samples at different maturity levels and calculate the difference between the green channel and the red channel corresponding to each maturity category. The maturity categories are underripe, moderately ripe, moderately ripe, and overripe. A piecewise function is defined to map the difference between the continuous green and red channels to a discrete probability distribution of maturity levels to eliminate color bias. The correlation between the difference between the green and red channels and maturity is verified through experiments, and the threshold range is adjusted.

5. The method for distinguishing the baking state of tobacco leaves according to claim 3, wherein: The process of determining the maturity of tobacco leaves includes the following steps: receiving an input area ratio of a yellow area, dividing the yellow area into a plurality of area segments, each area segment including a first plurality of yellow image pixels, and dividing the plurality of area segments into a plurality of sub-area segments, each sub-area segment including a second plurality of yellow image pixels; estimating grayscale values of the yellow image corresponding to the first plurality of yellow image pixels and the second plurality of yellow image pixels; calculating a standard deviation based on the grayscale values; and responding to yellow image pixels in which the standard deviation of the first plurality of yellow image pixels and the second plurality of yellow image pixels is lower than a set threshold and is not lower than the set threshold; The yellow area proportion corresponding to the yellow image pixels that are not less than the set threshold is mapped to the membership probability of different maturity levels.

6. The method for determining the tobacco leaf baking state according to claim 3, wherein: The process of building a tobacco leaf curing state discrimination model based on the historical tobacco leaf image dataset includes the following steps: A target tobacco leaf image dataset is formed based on the maturity of the tobacco leaves; the target tobacco leaf image dataset is divided into a training set, a test set, and a validation set; the color and texture features of the tobacco leaves are extracted; the texture eigenvalues are obtained through the angular second moment, entropy, and contrast index of the gray-level co-occurrence matrix; and the texture eigenvalues are subjected to variable cluster analysis and correlation analysis; A set of parameter combinations is randomly generated as the initial solution; the classification accuracy is used as the fitness indicator to evaluate the performance of each parameter set; the parameters are iteratively optimized through genetic operations to ultimately select the optimal tobacco leaf curing parameter combination; The optimized optimal parameter combination was used to train the tobacco leaf baking state model on the training set data, and the complexity of the tobacco leaf baking state model was adjusted through the validation set, and the generalization ability of the tobacco leaf baking state model was evaluated using the test set.

7. The method for determining the tobacco leaf baking state according to claim 6, wherein: The process of performing variable cluster analysis and correlation analysis on texture feature values includes the following steps: Calculate the internal correlation coefficient matrix for color, contour and texture features respectively, extract high correlation indicators to reduce the dimension; group the extracted multiple texture features according to similarity; set the distance threshold for clustering distance and classify the texture features; For each cluster, measure the correlation with the target variable. Analyze the relationship between features and select the features with the strongest correlation with the baking stage in each cluster. Retain the features with the highest correlation with the target variable in each cluster and eliminate redundant features. The data with redundant features removed is classified and evaluated to eventually form a tobacco leaf feature subset, which is used as a dataset for training the model.

8. The method for determining the baking state of tobacco leaves according to claim 6, wherein: The process of ultimately selecting the optimal tobacco leaf curing parameter combination includes the following steps: A set of tobacco leaf curing parameter combinations is randomly generated as the initial population; wherein the tobacco leaf curing combination includes key control variables of temperature, humidity and curing time; Using classification accuracy as an indicator of adaptability, the performance of each set of tobacco leaf curing parameters was quantitatively evaluated. Tobacco leaf parameter combinations were screened based on fitness values, and individuals with performance that met the standards were retained. Perform a genetic crossover operation on the selected parent parameters to generate new offspring parameter combinations; randomly perturb the offspring parameters with a preset probability to increase population diversity; repeat the selection, crossover, and mutation process until the preset number of iterations is reached; finally, select the parameter combination with the highest classification accuracy from the last generation population as the optimal solution.

9. The method for determining the tobacco leaf baking state according to claim 6, wherein: The process of evaluating the generalization ability of the tobacco leaf curing state model using the test set includes the following steps: The optimized optimal tobacco leaf curing parameter combination is used to train the tobacco leaf curing state model on the training set data; hyperparameter tuning, cross-validation, and complexity control operations are performed to adjust the tobacco leaf curing state model; The tobacco curing stages were classified using comprehensive accuracy, recall, and F1 metrics. The optimal data for identifying key curing stages was selected. The adaptability of the tobacco curing state model to unknown data was verified using an independent test set to obtain the optimal tobacco curing state model. The tobacco leaf data collected in real time is input into the tobacco leaf baking status model for evaluation and judgment; when an anomaly is detected, an early warning is issued; and the final result is output visually.

10. A tobacco leaf baking state identification system, applied to the tobacco leaf baking state identification method according to any one of claims 1 to 9, characterized in that: The tobacco leaf baking state discrimination system comprises: A tobacco leaf baking state acquisition module is used to acquire baking state images of different tobacco leaves, and based on a preset fusion strategy, segmentally fuse the first target area of each tobacco leaf image to obtain at least one target tobacco leaf image and generate an initial tobacco leaf image dataset; The tobacco leaf state discrimination model module is used to extract key tobacco leaf images from the initial tobacco leaf image dataset, automatically identify the key tobacco leaf images and grade them to generate a target tobacco leaf image dataset; and build a tobacco leaf baking state discrimination model based on the historical tobacco leaf image dataset; The baking state discrimination module is used to input the real-time tobacco leaf image into the tobacco leaf baking state discrimination model for discrimination and identify the baking degree of the tobacco leaves; generate a tobacco leaf baking degree report based on the tobacco leaf baking degree, output it visually, and control the tobacco leaf processing progress through the terminal.

Images