A method and system for tobacco curing management

Abstract

The application discloses a tobacco leaf baking management method and system, and the method comprises the following steps: pre-entering and storing baking stage judgment threshold values corresponding to different regions, varieties, parts and maturity tobacco leaves, and completing threshold value library initialization; in the baking process, multi-source data fusion judgment is realized by synchronously collecting multi-dimensional tobacco leaf state data in combination with a multiple linear regression model, and a judgment result of the current baking stage is output. The application effectively solves the problems of low accuracy and high misjudgment rate of the existing tobacco leaf baking stage judgment scheme, and realizes the effects of improving the judgment accuracy and reducing the baking quality loss.

Description

Technical Field

[0001] This invention relates to a method and system for tobacco curing management, belonging to the field of tobacco curing technology. Background Technology

[0002] Tobacco curing is a core step in the primary processing of tobacco leaves. The precision of the curing process directly determines the final quality and economic value of the tobacco leaves. Accurate determination of the curing stage is a crucial prerequisite for the dynamic adjustment of curing process parameters and a key control point in the entire curing process. The tobacco curing process is typically divided into three main stages: yellowing, color setting, and drying. Each main stage can be further subdivided into multiple sub-process nodes, and the optimal process parameters for different nodes vary significantly.

[0003] Currently, the mainstream methods for determining the curing stage of tobacco leaves in the industry mainly employ two types of technical solutions:

[0004] 1. Manual Judgment Method: This method relies entirely on the practical experience of the tobacco curing personnel. Stage judgments are made by visually observing changes in the color and shape of the tobacco leaves, and by touching them to assess their moisture and hardness. There is no standardized, quantitative testing process. The results of this method are greatly influenced by the experience level, work status, and subjective judgment bias of the personnel, leading to significant differences in judgments among different individuals for the same batch of tobacco leaves.

[0005] 2. Single-detection judgment scheme: This scheme uses a single type of detection technology to collect tobacco leaf parameters, such as using a contact moisture sensor to detect surface moisture or using image recognition technology to identify color changes in the tobacco leaves, and then uses this as the basis for judging the curing stage. However, tobacco curing is a complex physiological and biochemical process involving the coordinated degradation of internal chemical components, changes in external morphology, and overall moisture evaporation. A single detection scheme can only obtain attribute data of one dimension of the tobacco leaf, cannot characterize the coordinated changes of the three types of parameters, cannot truly reflect the actual curing progress, and has a high judgment error.

[0006] Based on the above, the existing tobacco curing stage judgment scheme has low accuracy and high misjudgment rate, which can easily lead to quality problems such as green or blackened tobacco leaves, resulting in direct economic losses. Summary of the Invention

[0007] (a) Purpose of the invention

[0008] Based on the above, the present invention provides a method and system for tobacco curing management, so as to improve the accuracy of tobacco curing stage judgment and reduce curing quality loss caused by misjudgment.

[0009] (II) Technical Solution

[0010] In a first aspect, the present invention provides a method for managing tobacco leaf curing, comprising:

[0011] S1. Pre-enter and store the threshold values ​​for the curing stage of tobacco leaves of different regions, varieties, parts, and maturity levels, and complete the initialization of the threshold library;

[0012] S2. Enter the attribute information of the tobacco leaves to be cured this time, and call the judgment threshold of the corresponding attribute in the threshold library to complete the system configuration;

[0013] S3. During the baking process, near-infrared spectroscopy is used to identify and collect data on the internal moisture and characteristic chemical components of the tobacco leaves, image acquisition is used to identify and collect the yellowness value and wilting grade value of the tobacco leaves, and weight-moisture correlation detection is used to obtain the overall moisture data of the tobacco leaves.

[0014] S4. After preprocessing the three types of detection data collected, input the pre-trained multiple linear regression baking curve model and output the judgment result of the current baking stage.

[0015] S5. Compare the obtained judgment result with the preset baking stage corresponding to the current baking time. If the deviation exceeds the preset threshold, an alarm will be triggered to notify the staff to conduct manual review.

[0016] S6. After the baking is completed, if the manual review confirms that the accuracy of the judgment meets the preset conditions, the parameters will be iteratively updated to the judgment threshold of the corresponding attribute tobacco leaf in the threshold library to complete the adaptation and optimization.

[0017] Preferably, in step S3, the acquisition frequency of various types of data is as follows: near-infrared spectroscopy detection is acquired once every 15 to 60 minutes, image acquisition and recognition is acquired once every 10 to 30 minutes, and weight detection is acquired in real time.

[0018] Preferably, in step S4, the input variables of the multiple linear regression baking curve model include internal moisture value, total sugar content, tobacco yellowness value, wilting grade, and overall moisture conversion value.

[0019] Preferably, the specific calculation formula for the multiple linear regression baking curve model is: Y = a1X1 + a2X2 + a3X3 + a4X4 + a5X5 + b

[0020] Where Y is the output quantitative value of the baking stage, X1 is the internal moisture value, a1 is its corresponding weight coefficient, X2 is the total sugar percentage, a2 is its corresponding weight coefficient, X3 is the yellowness value of the tobacco leaves, a3 is its corresponding weight coefficient, X4 is the wilting level, a4 is its corresponding weight coefficient, X5 is the overall moisture conversion value, a5 is its corresponding weight coefficient, and b is the offset constant.

[0021] Preferably, the default weights for the three types of detection data are: near-infrared spectroscopy detection data weight is 3, image acquisition and recognition data weight is 3, and weight-moisture correlation detection data weight is 2.

[0022] In a second aspect, the present invention provides a tobacco curing management system, comprising:

[0023] Near-infrared spectroscopy detection module: used to collect near-infrared spectral data of tobacco leaves and analyze the content data of internal moisture and characteristic chemical components of tobacco leaves;

[0024] Image acquisition and recognition module: used to acquire real-time images of tobacco leaves and identify the yellowness value and wilting level value of the tobacco leaves;

[0025] Weight detection module: used to collect real-time weight data of the entire rack of tobacco leaves, and to obtain the overall moisture content data of the tobacco leaves by weight-moisture conversion;

[0026] Data processing module: Built-in multiple linear regression curing curve model, used to fuse three types of detection data and output the judgment result of the current tobacco curing stage;

[0027] Threshold optimization module: It is used to store the judgment threshold library for tobacco leaves of different regions, varieties, parts and maturity, and can automatically iterate and update the judgment threshold of tobacco leaves with corresponding attributes when the preset accuracy conditions are met.

[0028] Alarm module: used to trigger a manual review alarm when the deviation between the judgment result and the preset baking stage timing exceeds a preset threshold.

[0029] Preferably, the input variables for the multiple linear regression baking curve model include internal moisture content, total sugar content, yellowness of tobacco leaves, wilting grade, and overall moisture conversion value.

[0030] The preferred calculation formula for the multiple linear regression baking curve model is as follows:

[0031] Y = a1X1+ a2X2+ a3X3+ a4X4+ a5X5+ b

[0032] Where Y is the output quantitative value of the baking stage, X1 is the internal moisture value and a1 is its corresponding weight coefficient, X2 is the total sugar percentage and a2 is its corresponding weight coefficient, X3 is the yellowness value of tobacco leaves and a3 is its corresponding weight coefficient, X4 is the wilting level and a4 is its corresponding weight coefficient, X5 is the overall moisture conversion value and a5 is its corresponding weight coefficient, and b is the offset constant.

[0033] (III) Beneficial Effects

[0034] This invention uses three technologies—near-infrared spectroscopy, image recognition, and weight detection—to simultaneously collect three types of multi-dimensional data: internal physicochemical properties, external morphological characteristics, and overall moisture changes of tobacco leaves. Combined with a multiple linear regression curing curve model, it achieves multi-source data fusion and judgment, which not only eliminates the subjective error of manual judgment but also solves the problem that a single detection dimension is one-sided and cannot reflect the overall state of tobacco leaf curing, thus improving the accuracy of judgment during the curing stage.

[0035] This invention solves the problem of existing general thresholds not taking into account differences in tobacco leaf attributes and having poor adaptability by constructing a dynamically iteratively optimized threshold library that is adapted to tobacco leaves of different regions, varieties, parts, and maturity levels. It can cover various main and specialty tobacco leaf varieties and curing scenarios in different production areas.

[0036] This invention employs an automatic judgment and anomaly alarm control method, which can reduce the high-frequency inspection workload of baking practitioners, while retaining a backup mechanism for manual review to avoid judgment errors in extreme cases, thus balancing automation and reliability of control. Attached Figure Description

[0037] Figure 1 This is a flowchart of the tobacco curing management method. Detailed Implementation

[0038] To make the above-mentioned objects, features, and advantages of the present invention more apparent and understandable, specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings. Many specific details are set forth in the following description to provide a thorough understanding of the present invention. However, the present invention can be practiced in many other ways different from those described herein, and those skilled in the art can make similar modifications without departing from the spirit of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0039] Example 1: A method for managing tobacco curing

[0040] refer to Figure 1 The method specifically includes:

[0041] Step S1: Construct a decision threshold library

[0042] Pre-enter and store the judgment thresholds for different regions, varieties, parts, and maturity levels of tobacco leaves for the different curing stages, and complete the initialization of the threshold library. The purpose of building the judgment threshold library is to reserve judgment benchmarks for tobacco leaves with different attributes in advance and eliminate the adaptation error of general thresholds. In one example, the judgment thresholds stored in the threshold library may include the internal moisture range, total sugar ratio range, yellowness value range, wilting grade threshold, and overall moisture range for each sub-stage of the yellowing stage, color fixing stage, and dry rib stage. The thresholds can be entered based on historical curing experience or experimental data.

[0043] Step S2: Parameter configuration before baking

[0044] Enter the attribute information of the tobacco leaves to be cured, and call the corresponding judgment threshold in the threshold library to complete the system configuration. In one example, the entered tobacco leaf attribute information may include four core parameters: planting region, variety, harvested part, and maturity. The system automatically matches the completely corresponding threshold group in the threshold library as the judgment criterion for this curing.

[0045] Step S3: Synchronous Acquisition of Multi-Dimensional Status Data

[0046] During the curing process, data on the internal moisture and characteristic chemical components of tobacco leaves are simultaneously collected through near-infrared spectroscopy, external color and morphological characteristics are collected through image acquisition, and overall moisture data of the tobacco leaves is obtained through weight-moisture correlation detection. Near-infrared spectroscopy can penetrate the surface of the tobacco leaves to detect core chemical components such as internal moisture, total sugar, and nicotine, reflecting the physiological and biochemical changes within the leaves. Image acquisition can obtain external characteristics such as yellowness, wilting degree, and midrib dryness, reflecting observable morphological changes. Weight-moisture correlation detection calculates the overall moisture evaporation by converting the weight change of the entire tobacco rack, verifying the dehydration progress of the tobacco leaves from a global perspective. The simultaneous collection of these three types of data comprehensively covers the coordinated changes of these three core parameters during the tobacco curing process, eliminating the limitations of single-detection methods.

[0047] In one example, near-infrared spectroscopy is performed every 15 to 60 minutes. Near-infrared spectroscopy is performed using a near-infrared camera with a wavelength range of 400 nm to 1100 nm. Before powering on, calibration is required using standard tobacco leaves with different moisture contents. The calibration process can be performed by selecting three standard tobacco leaf samples with moisture contents of 20%, 40%, and 60% to ensure that the detection error is ≤2%. The internal moisture value and total sugar content of the tobacco leaf are obtained through near-infrared spectroscopy.

[0048] In one example, images are captured every 10-30 minutes to identify the yellowness and wilting levels of tobacco leaves. The image acquisition and recognition are implemented using a CNN image recognition algorithm, the construction method of which is as follows:

[0049] S31 Dataset Construction

[0050] Image samples of tobacco leaves from different regions, varieties, parts, and maturity levels were collected and labeled at various curing stages. The labels included the yellowness value and wilting level of the corresponding tobacco leaves. The samples covered actual curing barn scenes with different lighting and shooting angles.

[0051] S32 network structure configuration

[0052] The lightweight MobileNetV2 is used as the backbone network. The network structure includes an input layer, a 3×3 standard convolutional layer, 17 inverted residual bottleneck modules, a 1×1 convolutional layer, a global average pooling layer, a fully connected layer, and an output layer. The input image size of the input layer is uniformly adjusted to 224×224×3. The output layer outputs two types of recognition results simultaneously: yellowness value (a continuous value from 0 to 100) and wilting level (a discrete level value from 1 to 5).

[0053] S33 Model Training

[0054] The Adam optimizer was used with an initial learning rate of 0.001 and a batch size of 16. The cross-entropy loss function was used for the wilting level and the mean squared error loss function was used for the yellowness value for joint training. The training iterations were no less than 100 rounds until the average accuracy of the model in the two-class feature recognition on the validation set was ≥95%, and the model training was completed.

[0055] S34. Real-time recognition

[0056] The collected real-time tobacco leaf images are input into the trained CNN model. After preprocessing such as size normalization, illumination correction, and Gaussian denoising, the model automatically outputs the corresponding yellowness value and wilting level as two types of external feature data.

[0057] In one example, weight is collected in real time. The weight sensor is installed at the connection point between the tobacco hanging beam and the curing barn's load-bearing structure, with a detection accuracy of up to 10g, ensuring that the overall moisture conversion error is ≤1%. The weight-moisture conversion is based on the principle of dry matter conservation, requiring the initial total weight of the entire rack of tobacco leaves loaded into the oven to be collected before curing begins. and initial average moisture content The conversion formula is:

[0058] in, for The overall moisture content of the tobacco leaves at any given time. for The real-time total weight of the entire tobacco rack is collected by the constant weight detection module. This represents the initial total weight of the entire rack of tobacco leaves before curing. The initial average moisture content of the entire tobacco rack before curing.

[0059] Step S4: Multi-source data fusion determination

[0060] After preprocessing the three types of detection data collected, the data is input into a pre-trained multiple linear regression baking curve model, which outputs the judgment result of the current baking stage.

[0061] In one example, preprocessing includes standard data processing operations in this field, such as denoising and normalization. The input variables for the multiple linear regression baking curve model include internal moisture content, total sugar content, tobacco leaf yellowness, wilting grade, and overall moisture conversion value. The specific calculation formula for the multiple linear regression baking curve model is as follows:

[0062] Y = a1X1+ a2X2+ a3X3+ a4X4+ a5X5+ b

[0063] Where Y is the output quantitative value of the baking stage, X1 is the internal moisture value and a1 is its corresponding weight coefficient, X2 is the total sugar percentage and a2 is its corresponding weight coefficient, X3 is the yellowness value of tobacco leaves and a3 is its corresponding weight coefficient, X4 is the wilting level and a4 is its corresponding weight coefficient, X5 is the overall moisture conversion value and a5 is its corresponding weight coefficient, and b is the offset constant.

[0064] The method for constructing a multiple linear regression baking curve model is as follows:

[0065] S41 Sample Set Construction

[0066] Labeled samples of tobacco leaves from different regions, varieties, parts, and maturity levels were collected at each curing stage. Each sample corresponds to a set of input features, including internal moisture value, total sugar content, yellowness value, wilting level, and overall moisture conversion value, as well as manually labeled quantitative values ​​for the curing stage. The samples cover all preset curing sub-stages.

[0067] S42 Variable Normalization

[0068] All input features are subjected to min-max normalization according to their dimensions, and the feature values ​​of each dimension are uniformly mapped to the [0,1] interval to eliminate the influence of the difference in dimensions on the model weights.

[0069] S43 parameter fitting

[0070] The least squares method is used to fit and solve the input features and corresponding labels of the sample set to obtain the initial values ​​of the weight coefficients a1~a5 and the offset constant b for each input feature.

[0071] S44 Model Verification

[0072] 30% of the samples in the sample set are selected as the validation set to verify the accuracy of the fitted initial model. The model is required to have an accuracy of ≥90% in judging the baking stage on the validation set. If this is not met, the sample size is increased and the model is refitted until the accuracy requirement is met, thus completing the model training. Training samples can be obtained from historical baking test data or standard baking experiments. The output judgment results can be refined into 10-15 sub-stages, including the yellowing stage, color fixing stage, and drying stage, to meet the accuracy requirements for adjusting the baking process.

[0073] In one example, the default weights for the three types of detection data are: the weight for near-infrared spectroscopy detection is 3 (i.e., The default sum is 3), and the corresponding data weight for image acquisition and recognition is 3 (i.e., The default sum is 3), and the weight of the corresponding data for weight-moisture correlation detection is 2 (i.e., The default value is 2), and the weight can be adjusted according to the actual scenario.

[0074] Step S5: Deviation Verification and Alarm

[0075] The obtained judgment result is compared with the preset baking stage corresponding to the current baking time. If the deviation exceeds the preset threshold, an alarm is triggered, and staff are notified to conduct a manual review. The purpose of this step is to avoid judgment errors caused by extreme abnormal situations (such as equipment failure or sample abnormality), balancing automation efficiency and judgment reliability.

[0076] Step S6: Threshold Dynamic Iterative Optimization

[0077] After the roasting process is completed, if manual verification confirms that the accuracy rate meets preset conditions, such as an accuracy rate ≥ 95%, the parameters are iteratively updated to the corresponding threshold values ​​for tobacco leaves in the threshold library, completing the adaptation and optimization. Continuously optimizing the adaptability of the threshold library through actual roasting data can improve the accuracy rate of subsequent judgments for tobacco leaves of the same attribute.

[0078] Taking the curing of fully mature tobacco leaves of the Yunyan 87 variety, the main tobacco variety cultivated in Liupanshui area, as an example, the above method is described in detail:

[0079] Pre-configuration stage: The thresholds for judging the curing of fully mature tobacco leaves of Yunyan 87 variety have been pre-entered into the threshold library, including the internal moisture range, yellowness value range, and overall moisture range corresponding to each sub-stage of the yellowing stage, color fixing stage, and dry rib stage.

[0080] Pre-baking configuration: The tobacco leaves to be baked this time are from Liupanshui, the variety is Yunyan 87, the part is the middle leaf, and the maturity is fully mature. The system calls the corresponding threshold to complete the configuration. The judgment deviation alarm threshold is set to 20%. The collection frequency of the three types of detection data is as follows: near-infrared is collected once every 30 minutes, image acquisition and recognition is collected once every 15 minutes, and weight detection is collected in real time.

[0081] Baking Process: After 48 hours of baking, the system collected three types of data: internal moisture 35%, total sugar content 28%, yellowness value 82, wilting level 4, and overall moisture 38%. The result of inputting this data into a multiple linear regression model was determined to be the early stage of color fixation, while the preset 48-hour stage corresponds to the late stage of yellowing, a deviation of 25%, exceeding the alarm threshold. The system simultaneously triggered on-site audible and visual alarms and sent an alarm SMS to the administrator's mobile phone, notifying the administrator to review the data. After on-site review, the administrator confirmed that the current stage was indeed the early stage of color fixation, and the deviation was caused by an unreasonable preset timing. The administrator adjusted the stage threshold corresponding to this attribute of the tobacco leaves at 48 hours, and the accuracy rate of the judgment throughout the baking process was 100%.

[0082] After baking, the threshold optimization module automatically updates the parameters of this iteration into the threshold library of mature tobacco leaves in the middle of Liupanshui Yunyan 87, thus completing the threshold optimization.

[0083] Through testing, it was found that after adopting the method of this embodiment, the proportion of high-grade tobacco leaves in this batch was improved compared with the traditional manual judgment method, and no quality problems such as green or blackened tobacco were found.

[0084] Example 2: A tobacco curing management system

[0085] The system includes the following modules:

[0086] Near-infrared spectroscopy detection module: used to collect near-infrared spectral data of tobacco leaves and analyze the content data of internal moisture and characteristic chemical components of tobacco leaves;

[0087] Image acquisition and recognition module: used to acquire real-time images of tobacco leaves and identify the yellowness value and wilting level value of the tobacco leaves;

[0088] Weight detection module: used to collect real-time weight data of the entire rack of tobacco leaves, and to obtain the overall moisture content data of the tobacco leaves by weight-moisture conversion;

[0089] Data processing module: Built-in multiple linear regression curing curve model, used to fuse three types of detection data and output the judgment result of the current tobacco curing stage;

[0090] Threshold optimization module: It is used to store the judgment threshold library for tobacco leaves of different regions, varieties, parts and maturity, and can automatically iterate and update the judgment threshold of tobacco leaves with corresponding attributes when the preset accuracy conditions are met.

[0091] Alarm module: used to trigger a manual review alarm when the deviation between the judgment result and the preset baking stage timing exceeds a preset threshold.

[0092] The processing method of the relevant modules is the same as in Example 1, and will not be repeated here.

[0093] The embodiments described above are merely illustrative of several implementations of the present invention, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of the present invention. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of the present invention, and these modifications and improvements all fall within the scope of protection of the present invention. Therefore, the scope of protection of this patent should be determined by the appended claims.

Claims

1. A method for managing tobacco leaf curing, characterized in that, include: S1. Pre-enter and store the threshold values ​​for the curing stage of tobacco leaves of different regions, varieties, parts, and maturity levels, and complete the initialization of the threshold library; S2. Enter the attribute information of the tobacco leaves to be cured this time, and call the judgment threshold of the corresponding attribute in the threshold library to complete the system configuration; S3. During the baking process, near-infrared spectroscopy is used to identify and collect data on the internal moisture and characteristic chemical components of the tobacco leaves, image acquisition is used to identify and collect the yellowness value and wilting grade value of the tobacco leaves, and weight-moisture correlation detection is used to obtain the overall moisture data of the tobacco leaves. S4. After preprocessing the three types of detection data collected, input the pre-trained multiple linear regression baking curve model and output the judgment result of the current baking stage. S5. Compare the obtained judgment result with the preset baking stage corresponding to the current baking time. If the deviation exceeds the preset threshold, an alarm will be triggered to notify the staff to conduct manual review. S6. After the baking is completed, if the manual review confirms that the accuracy of the judgment meets the preset conditions, the parameters will be iteratively updated to the judgment threshold of the corresponding attribute tobacco leaf in the threshold library to complete the adaptation and optimization.

2. The tobacco leaf curing management method according to claim 1, characterized in that, In step S3, the data acquisition frequencies for each type of data are as follows: near-infrared spectroscopy is acquired every 15-60 minutes, image acquisition and recognition is acquired every 10-30 minutes, and weight detection is acquired in real time.

3. The tobacco leaf curing management method according to claim 1, characterized in that, In step S4, the input variables for the multiple linear regression baking curve model include internal moisture value, total sugar content, tobacco yellowness value, wilting level, and overall moisture conversion value.

4. The tobacco leaf curing management method according to claim 3, characterized in that, The specific calculation formula for the multiple linear regression baking curve model is: Y = a1X1 + a2X2 + a3X3 + a4X4 + a5X5 + b Where Y is the output quantitative value of the baking stage, X1 is the internal moisture value and a1 is its corresponding weight coefficient, X2 is the total sugar percentage and a2 is its corresponding weight coefficient, X3 is the yellowness value of tobacco leaves and a3 is its corresponding weight coefficient, X4 is the wilting level and a4 is its corresponding weight coefficient, X5 is the overall moisture conversion value and a5 is its corresponding weight coefficient, and b is the offset constant.

5. The tobacco leaf curing management method according to claim 3, characterized in that, The default weights for the three types of detection data are as follows: near-infrared spectroscopy detection data has a weight of 3, image acquisition and recognition data has a weight of 3, and weight-moisture correlation detection data has a weight of 2.

6. A tobacco leaf curing management system, characterized in that, include: Near-infrared spectroscopy detection module: used to collect near-infrared spectral data of tobacco leaves and analyze the content data of internal moisture and characteristic chemical components of tobacco leaves; Image acquisition and recognition module: used to acquire real-time images of tobacco leaves and identify the yellowness value and wilting level value of the tobacco leaves; Weight detection module: used to collect real-time weight data of the entire rack of tobacco leaves, and to obtain the overall moisture content data of the tobacco leaves by weight-moisture conversion; Data processing module: Built-in multiple linear regression curing curve model, used to fuse three types of detection data and output the judgment result of the current tobacco curing stage; Threshold optimization module: It is used to store the judgment threshold library for tobacco leaves of different regions, varieties, parts and maturity, and can automatically iterate and update the judgment threshold of tobacco leaves with corresponding attributes when the preset accuracy conditions are met. Alarm module: used to trigger a manual review alarm when the deviation between the judgment result and the preset baking stage timing exceeds a preset threshold.

7. The tobacco curing management system according to claim 6, characterized in that, The input variables for the multiple linear regression baking curve model include internal moisture content, total sugar content, yellowness of tobacco leaves, wilting grade, and overall moisture conversion value.

8. The tobacco curing management system according to claim 7, characterized in that, The specific calculation formula for the multiple linear regression baking curve model is: Y = a1X1 + a2X2 + a3X3 + a4X4 + a5X5 + b Where Y is the output quantitative value of the baking stage, X1 is the internal moisture value and a1 is its corresponding weight coefficient, X2 is the total sugar percentage and a2 is its corresponding weight coefficient, X3 is the yellowness value of tobacco leaves and a3 is its corresponding weight coefficient, X4 is the wilting level and a4 is its corresponding weight coefficient, X5 is the overall moisture conversion value and a5 is its corresponding weight coefficient, and b is the offset constant.

Images