A method for controlling the initial baking of tobacco leaves based on an image inference model

Abstract

The present invention provides a tobacco leaf initial baking control method based on an image reasoning model. By establishing a yellow ratio and dry ratio prediction model, the prediction model is written into software and installed in an intelligent control system of an intensive baking room, the actual yellow ratio and dry ratio in the current baking process are obtained through an intelligent control system with an on-site image acquisition function, and compared with the yellow ratio and dry ratio required by the baking curve. If the yellow ratio and the dry ratio are the same, the target is reached in advance. If the yellow ratio and the dry ratio are different, the fire extension time is estimated according to the yellow baking characteristics and the dry baking characteristics of the tobacco leaves, the fire is automatically extended, the dry bulb temperature and the wet bulb temperature are adjusted, and baking is continued. When the fire extension time reaches the target, the next baking stage can be entered. If the fire extension time does not meet the standard after burning, the next baking stage is automatically jumped. The present invention does not rely on manual monitoring by experts. Through real-time image acquisition and comparison, the baking stage can be accurately evaluated, so that the flue-cured tobacco can achieve the best baking effect.

Description

Technical Field

[0001] The present invention belongs to the technical field of tobacco leaf baking, and in particular relates to a tobacco leaf primary baking control method based on an image inference model. Background Art

[0002] The specific processing of flue-cured tobacco is called flue-curing, or simply curing, or "baking tobacco." It's not simply a process of dehydration and drying, but rather a series of unique and complex tobacco conditioning processes. Curing is divided into primary curing and re-curing. The first curing of tobacco leaves harvested in the field (called fresh tobacco) is called primary curing, typically carried out in separate tobacco-growing areas. The resulting tobacco leaves are called primary-cured tobacco or raw tobacco (commonly known as dry tobacco), and are the raw material for human consumption. The raw tobacco is then subjected to a second curing process called re-curing, often performed in centralized re-curing plants. The raw tobacco after curing is called re-cured tobacco.

[0003] Primary tobacco leaf curing is the first step in transforming cultivated tobacco leaves into cigarette raw material. Farmers place fresh tobacco leaves harvested from the fields in a curing barn and heat and condition them, turning them into the "raw tobacco" used as cigarette raw material. Different curing temperatures are set during the yellowing, color-fixing, and tendon-drying stages of the leaves. This ensures that the leaves reach a certain percentage of yellowing, a soft, collapsed structure, and a softened main vein. The leaves then develop yellow flakes, yellow tendons, and curled edges, ultimately resulting in a dried-main vein raw tobacco. This entire process requires not only real-time observation of the tobacco leaves' changing state to flexibly control the temperature and humidity in the curing barn, but also adjustments to the barn environment based on changes in ambient temperature and humidity.

[0004] Traditional techniques often rely on manual judgment for the initial curing of tobacco leaves, which is time-consuming and labor-intensive. Traditionally, experts observe the color and shrinkage of the leaves through observation windows to determine when to adjust the temperature in the curing barn. This method is labor-intensive, time-consuming, and difficult to accurately measure. Current intensive curing barn technology uses intelligent controllers to regulate temperature and humidity in the barn. Specifically, a curing curve is pre-set and then imported into the control system for curing. However, this method fails to account for the actual changes in tobacco leaves during the curing process. Specifically, even if the leaves meet the target, they will still be cured according to the curing curve, resulting in overcooking. Curing will also be stopped even if the optimal curing state is not reached. Without accurate control of the curing stages, curing results are difficult to guarantee. Summary of the Invention

[0005] The present invention is designed to solve the problem that the current intelligent controller of the intensive curing room does not take into account the actual changes of the tobacco leaves during the curing process, resulting in inaccurate baking. The purpose is to provide a tobacco initial baking control method based on an image reasoning model that can accurately control the baking stage through an intelligent controller, obtain the actual changes of the tobacco leaves, and ensure the baking effect.

[0006] The present invention adopts the following technical solutions:

[0007] A tobacco leaf primary baking control method based on an image inference model, the method comprising the following steps:

[0008] S1, establish a prediction model for the yellow ratio and dry ratio during baking, write the prediction model into the software, and install it in the intelligent control system of the intensive baking room;

[0009] S2, obtains the actual yellow ratio and dry ratio of the current baking process through an intelligent control system with on-site image acquisition function;

[0010] S3, compare the actual yellow ratio and dry ratio during the baking process with the yellow ratio and dry ratio required by the baking curve at any time. If the yellow ratio and dry ratio are the same, the target is reached in advance and the next baking stage can be entered. If the yellow ratio and dry ratio still do not reach the target when the time required by the baking curve is reached, the fire extension time is estimated according to the yellow baking characteristics and dry baking characteristics of the tobacco leaves, the fire is automatically extended, the dry bulb temperature and wet bulb temperature are adjusted, and baking is continued. When the fire extension time reaches the target, the next baking stage can be entered. If the fire extension time does not meet the target, the next baking stage will be automatically jumped to.

[0011] In the tobacco leaf primary baking control method based on the image inference model, the steps for establishing the yellow ratio and dry ratio prediction model in S1 are as follows:

[0012] S11, experts scored and evaluated the yellowing ratio and dryness ratio of tobacco leaves during the curing process;

[0013] S12, collecting an image of the flue-cured tobacco sample to be tested, and reading the image as a sample image;

[0014] S13, a polynomial regression model was established with the yellow ratio and dry ratio evaluated by experts as dependent variables and the flue-cured tobacco sample images, dry-bulb temperature, wet-bulb temperature, and curing time as independent variables.

[0015] In the tobacco leaf primary baking control method based on an image inference model, the image of the tobacco sample to be tested in S12 is collected by multiple cameras in the baking room.

[0016] In the tobacco leaf primary baking control method based on an image inference model, multiple cameras in the baking room in S12 move back and forth in real time to capture images.

[0017] In the tobacco leaf primary baking control method based on the image inference model, when a sample image is read in S12, the sample image is converted into HSV and Lab color spaces respectively to obtain the HSV and Lab channel pixel values of the image.

[0018] In the tobacco leaf primary baking control method using an image reasoning model, the yellow ratio and dry ratio prediction polynomial regression model established in S13 is:

[0019]

[0020] Wherein, x1 represents the dry-bulb temperature; x2 represents the wet-bulb temperature; x3 represents the pixel value of the H channel; x4 represents the pixel value of the S channel; x5 represents the pixel value of the V channel; x6 represents the pixel value of the L channel; x7 represents the pixel value of the a channel; x8 represents the pixel value of the b channel; x9 represents the baking time; w0 represents the bias term; w1 represents the regression coefficient of the dry-bulb temperature; w2 represents the regression coefficient of the wet-bulb temperature; w3 represents the regression coefficient of the product of the dry-bulb temperature and the wet-bulb temperature; w4 represents the regression coefficient of the square of the pixel value of the H channel; w5 represents the regression coefficient of the square of the pixel value of the S channel; w6 represents the regression coefficient of the pixel value of the V channel; w7 represents the regression coefficient of the pixel value of the L channel; w8 represents the regression coefficient of the product of the pixel value of the V channel and the pixel value of the L channel; w9 represents the regression coefficient of the square of the pixel value of the a channel; w 10 Represents the regression coefficient of the square of the pixel value of channel b; w 11 represents the regression coefficient of baking time; y represents the predicted value of the dependent variables yellow ratio and dry ratio.

[0021] In the tobacco leaf primary baking control method based on the image inference model, the steps of establishing the yellow baking characteristics and the dry baking characteristics in S3 are as follows:

[0022] S31, obtaining the current yellow ratio and dry ratio according to the yellow ratio and dry ratio prediction model, and subtracting the initial yellow ratio and dry ratio obtained by the yellow ratio and dry ratio measurement model to obtain the changed yellow ratio and changed dry ratio values;

[0023] S32, subtracting the current baking time from the start baking time to obtain the cumulative baking time of the current baking stage;

[0024] S33, dividing the yellow change and the dry change by the cumulative baking time, the dry-bulb temperature, and the wet-bulb temperature to obtain the yellow baking characteristics and the dry baking characteristics of the tobacco leaves.

[0025] In the tobacco leaf primary baking control method using an image inference model, the calculation formulas for the yellow baking characteristics and the dry baking characteristics in S33 are:

[0026] Yellow baking characteristics = yellow change / (accumulated baking time * dry bulb temperature * wet bulb temperature)

[0027] Dry-bake characteristics = change in dryness / (accumulated baking time * dry-bulb temperature * wet-bulb temperature)

[0028] In the tobacco leaf primary baking control method based on the image inference model, the fire extension time is 2 hours.

[0029] The tobacco leaf primary baking control method based on the image inference model, wherein the on-site image acquisition refers to acquisition by multiple cameras in the baking room, and the multiple cameras in the baking room move back and forth in real time to capture images.

[0030] The beneficial effects of the present invention are as follows: the present invention solves the problem that the current intelligent controller of the intensive flue-curing barn does not consider the actual changes of tobacco leaves during the curing process, resulting in inaccurate baking. By establishing a yellow ratio and dry ratio prediction model, the prediction model is written into the software, and installed in an intelligent control system of the intensive flue-curing barn with a field image acquisition function, the actual yellow ratio and dry ratio in the baking process are predicted and compared with the yellow ratio and dry ratio required by the baking curve. If the yellow ratio and the dry ratio are the same, the target is reached in advance. If the yellow ratio and the dry ratio are different, the fire extension time is estimated according to the yellow baking characteristics and the dry baking characteristics, the fire is automatically extended, the dry bulb temperature and the wet bulb temperature are adjusted, and baking is continued. When the fire extension time reaches the target, the next baking stage can be entered. If the fire extension time does not meet the standard, the next baking stage is automatically jumped. Without relying on manual monitoring by experts, the baking stage can be accurately evaluated through real-time image acquisition and comparison to prevent over-baking and substandard baking.

[0031] By taking pictures with multiple cameras moving back and forth in real time, it is possible to accurately obtain images of tobacco leaves at any time, calculate the yellow and dry ratios of the tobacco leaves in real time, compare them with the yellow and dry ratios of the baking curve, and accurately control the baking stage.

[0032] The yellow baking characteristics and dry baking characteristics can be used to calculate the change in yellow and wet bulb temperature required to achieve the target yellow and wet bulb ratios when the yellow and wet bulb ratios do not meet the standards. This allows the dry bulb temperature and wet bulb temperature to be adjusted during the automatic fire extension period to achieve the target yellow and wet bulb ratios and ensure baking results. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] Figure 1 The figure is a flow chart of a tobacco leaf initial baking control method based on an image reasoning model. DETAILED DESCRIPTION

[0034] The technical content of the present invention is further described below in conjunction with specific embodiments and drawings, but this does not limit the scope of protection of the present invention.

[0035] See also Figure 1 A tobacco leaf primary baking control method based on an image inference model comprises the following steps:

[0036] S1, establish a prediction model for the yellow ratio and dry ratio during the baking process, write the prediction model into the software, and install it in the intelligent control system of the intensive baking room.

[0037] The intelligent control system controls the working temperature and humidity in the flue-curing barn and the curing time of tobacco leaves through temperature and humidity sensors, multiple cameras with variable shooting angles that move back and forth in the flue-curing barn, single-chip computer technology and fuzzy logic software, thereby achieving intelligent and digital control of existing flue-curing barns, making the entire flue-curing process programmed and digital, changing the unstable quality of flue-cured tobacco and improving the efficiency of flue-curing.

[0038] The steps for establishing the yellow ratio and dry ratio prediction model are as follows:

[0039] S11, experts scored and evaluated the yellowing ratio and dryness ratio of tobacco leaves during the curing process;

[0040] S12, collecting an image of the flue-cured tobacco sample to be tested, and reading the image as a sample image;

[0041] S13, a polynomial regression model was established with the yellow ratio and dry ratio evaluated by experts as dependent variables and the flue-cured tobacco sample images, dry-bulb temperature, wet-bulb temperature, and curing time as independent variables.

[0042] Among them, the image acquisition of the flue-cured tobacco sample to be tested in S12 is carried out through multiple cameras in the flue-curing room. The multiple cameras in the flue-curing room in S12 move back and forth in real time to capture images. When reading in the images, the images are converted into HSV and Lab color spaces respectively to obtain the HSV and Lab channel pixel values of the images.

[0043] The established polynomial regression model for predicting yellow ratio and dry ratio is:

[0044]

[0045] Wherein, x1 represents the dry-bulb temperature; x2 represents the wet-bulb temperature; x3 represents the pixel value of the H channel; x4 represents the pixel value of the S channel; x5 represents the pixel value of the V channel; x6 represents the pixel value of the L channel; x7 represents the pixel value of the a channel; x8 represents the pixel value of the b channel; x9 represents the baking time; w0 represents the bias term; w1 represents the regression coefficient of the dry-bulb temperature; w2 represents the regression coefficient of the wet-bulb temperature; w3 represents the regression coefficient of the product of the dry-bulb temperature and the wet-bulb temperature; w4 represents the regression coefficient of the square of the pixel value of the H channel; w5 represents the regression coefficient of the square of the pixel value of the S channel; w6 represents the regression coefficient of the pixel value of the V channel; w7 represents the regression coefficient of the pixel value of the L channel; w8 represents the regression coefficient of the product of the pixel value of the V channel and the pixel value of the L channel; w9 represents the regression coefficient of the square of the pixel value of the a channel; w 10 represents the regression coefficient of the square of the pixel value of channel b; w 11 represents the regression coefficient of baking time; y represents the predicted value of the dependent variables yellow ratio and dry ratio.

[0046] S2, obtains the actual yellow ratio and dry ratio in the current baking process through an intelligent control system with on-site image acquisition function.

[0047] On-site image acquisition is carried out through multiple cameras in the baking room, and the multiple cameras in the baking room move back and forth in real time to capture images.

[0048] S3, compare the actual yellow ratio and dry ratio during the baking process with the yellow ratio and dry ratio required by the baking curve at any time. If the yellow ratio and dry ratio are the same, the target is reached in advance and the next baking stage can be entered. If the yellow ratio and dry ratio still do not reach the target when the time required by the baking curve is reached, the fire extension time is estimated according to the yellow baking characteristics and dry baking characteristics of the tobacco leaves, the fire is automatically extended, the dry bulb temperature and wet bulb temperature are adjusted, and baking is continued. When the fire extension time reaches the target, the next baking stage can be entered. If the fire extension time does not meet the target, the next baking stage will be automatically jumped to.

[0049] The estimated fire extension time based on the yellow-baking characteristics and dry-baking characteristics of tobacco leaves can be 2 hours.

[0050] The steps for establishing yellow baking characteristics and dry baking characteristics are:

[0051] S31, obtaining the current yellow ratio and dry ratio according to the yellow ratio and dry ratio prediction model, and subtracting the initial yellow ratio and dry ratio obtained by the yellow ratio and dry ratio measurement model to obtain the changed yellow ratio and changed dry ratio values;

[0052] S32, subtracting the current baking time from the start baking time to obtain the cumulative baking time of the current baking stage;

[0053] S33, dividing the yellow change and the dry change by the cumulative baking time, the dry-bulb temperature, and the wet-bulb temperature to obtain the yellow baking characteristics and the dry baking characteristics of the tobacco leaves.

[0054] The calculation formulas for yellow baking characteristics and dry baking characteristics are:

[0055] Yellow baking characteristics = yellow change / (accumulated baking time * dry bulb temperature * wet bulb temperature)

[0056] Dry-bake characteristics = change in dryness / (accumulated baking time * dry-bulb temperature * wet-bulb temperature)

[0057] The following is a comparison of the results of yellow-baked and dry-baked characteristics obtained using this method with the experts' evaluation of the yellow-baked and dry-baked characteristics of flue-cured tobacco leaves.

[0058] The initial yellow and dry ratios of the tobacco leaves in the baking process were used as test objects. The tobacco samples were selected from the Chuxiong production area of Yunnan in 2022. The experts' evaluation of the yellow ratio and dry ratio is shown in Table 1. The experts scored the yellow ratio and dry ratio of the tobacco leaves in the current baking stage. The initial yellow ratio and dry ratio were 0. The current yellow ratio and dry ratio were subtracted from the initial yellow ratio and dry ratio to obtain the change yellow and change dry values. The current baking time, dry bulb temperature and wet bulb temperature were obtained from the control terminal sensor. According to the formula: yellow baking characteristics = change yellow / (cumulative baking time * dry bulb temperature * wet bulb temperature); dry baking characteristics = change dry / (cumulative baking time * dry bulb temperature * wet bulb temperature). This table uses the 0.05% integer multiple upper limit counting method to obtain the following data on yellow baking characteristics and dry baking characteristics.

[0059] Table 1 Expert evaluation of first-cured tobacco leaf samples

[0060]

[0061] The first-cured tobacco leaves from Chuxiong, Yunnan in 2022 were selected as samples. The following tobacco leaf first-curing control method based on the image inference model was used to obtain the yellow-curing characteristics and dry-curing characteristics of this sample:

[0062] First, based on the correlation between dry-bulb and wet-bulb temperatures, HSV and Lab channel pixel values acquired in real time by a camera moving back and forth in the baking room, baking time, and the yellow ratio and dry ratio evaluated by experts, a prediction model for the yellow ratio and dry ratio was established. The yellow ratio and dry ratio evaluated by experts were used as dependent variables, and the dry-bulb and wet-bulb temperatures, HSV and Lab channel pixel values, and baking time were read as independent variables. The independent and dependent variables were used to build a polynomial regression model as follows to predict the yellow ratio and dry ratio:

[0063]

[0064]

[0065] y represents the predicted value of the dependent variable yellow ratio; y2 represents the predicted value of the dependent variable dry ratio; x1 represents the dry-bulb temperature; x2 represents the wet-bulb temperature; x3 represents the H channel pixel value; x4 represents the S channel pixel value; x5 represents the V channel pixel value; x6 represents the L channel pixel value; x7 represents the a channel pixel value; x8 represents the b channel pixel value; x9 represents the baking time.

[0066] Next, the yellow roast characteristics and dry roast characteristics are calculated.

[0067] (1) The model is used to predict the yellow ratio and dry ratio values at the current time. The prediction results are shown in Table 2.

[0068] Table 2 Predicted values of yellow and dry ratios of first-cured tobacco leaf samples

[0069]

[0070] (2) Subtract the obtained yellow ratio and dry ratio from the yellow ratio and dry ratio at the starting time to obtain the change yellow and change dry values.

[0071] (3) Subtract the current baking time from the starting baking time to obtain the cumulative baking time of this baking stage.

[0072] (4) Divide the yellow change and dry change by the product of the cumulative baking time and the average dry-bulb temperature and the average wet-bulb temperature during this baking period: yellow baking characteristic = yellow change / (cumulative baking time * dry-bulb temperature * wet-bulb temperature); dry baking characteristic = dry change / (cumulative baking time * dry-bulb temperature * wet-bulb temperature) to obtain the yellow baking characteristic and dry baking characteristic values respectively. Take the average values of the yellow baking characteristic and dry baking characteristic of baking stage 6, baking stage 7, and baking stage 8, see Table 3.

[0073] Table 3 Mean values of yellow baking characteristics and dry baking characteristics

[0074] Current baking stage Yellow baking characteristics Dry baking characteristics 6 0.07％ 0.02％ 7 0.05％ 0.02％ 8 0.04％ 0.01％

[0075] The yellow-baked and dry-baked characteristics of this method were compared with the expert evaluation of flue-cured tobacco leaves. The comparison results are shown in Table 4.

[0076] Table 4 Comparison results of yellow baking characteristics and dry baking characteristics

[0077]

[0078] As can be seen from Table 4, based on the comparison of the experts' evaluation of the degree of yellow baking characteristics and dry baking characteristics of tobacco leaves during the baking process and the results of yellow baking characteristics and dry baking characteristics obtained by applying this method, it was found that the proportion of consistent evaluation results between the two was 100%, with a high accuracy rate.

[0079] It can be seen that the application of the baking control method based on the image reasoning model can accurately evaluate the baking effect of the initial baking stage of tobacco leaves and whether the baking meets the standards. The tobacco leaf initial baking control method based on the image reasoning model of the present invention can accurately control the baking stage through an intelligent controller to ensure the baking effect and prevent over-baking and baking that does not meet the standards. It provides a research basis for the future application in digital baking, the inclusion of baking yellow baking characteristics and dry baking characteristics, and the improvement of the efficiency of digital baking of tobacco leaves.

Claims

1. A tobacco leaf primary baking control method based on an image inference model, comprising the following steps: S1, establish a prediction model for the yellow ratio and dry ratio during baking, including: S11, experts scored and evaluated the yellowing ratio and dryness ratio of tobacco leaves during the curing process; S12, performing image acquisition on the flue-cured tobacco sample to be tested, reading in the image as a sample image, and converting the sample image into HSV and Lab color spaces respectively, to obtain HSV and Lab channel pixel values of the image; In S13, a polynomial regression model was established with the yellow ratio and dry ratio evaluated by experts as dependent variables, and the flue-cured tobacco sample images, dry-bulb temperature, wet-bulb temperature, and curing time as independent variables: Wherein, x1 represents dry-bulb temperature; x2 represents wet-bulb temperature; x3 represents H channel pixel value; x4 represents S channel pixel value; x5 represents V channel pixel value; x6 represents L channel pixel value; x7 represents a channel pixel value; x8 represents b channel pixel value; x9 represents baking time; w0 represents bias term; w1 represents the regression coefficient of dry-bulb temperature; w2 represents the regression coefficient of wet-bulb temperature; w3 represents the regression coefficient of the product of dry-bulb temperature and wet-bulb temperature; w4 represents the regression coefficient of the square of H channel pixel value; w5 represents the regression coefficient of the square of S channel pixel value; w6 represents the regression coefficient of V channel pixel value; w7 represents the regression coefficient of L channel pixel value; w8 represents the regression coefficient of the product of V channel pixel value and L channel pixel value; w9 represents the regression coefficient of a channel pixel value square; w10 represents the regression coefficient of b channel pixel value square; w11 represents the regression coefficient of baking time; y represents the predicted value of the dependent variable yellow ratio and dry ratio; Write the prediction model into the software and install it in the intelligent control system of the intensive baking room; S2, obtains the actual yellow ratio and dry ratio of the current baking process through an intelligent control system with on-site image acquisition function; S3, compare the actual yellow ratio and dry ratio during the baking process with the yellow ratio and dry ratio required by the baking curve at any time. If the yellow ratio and dry ratio are the same, the target is achieved in advance and the next baking stage is entered. If the yellow ratio and dry ratio still do not reach the target after the time required by the baking curve, the fire extension time is estimated according to the yellow baking characteristics and dry baking characteristics of the tobacco leaves, the fire is automatically extended, the dry bulb temperature and wet bulb temperature are adjusted, and baking is continued. When the fire extension time reaches the target, the next baking stage is entered. If the fire extension time does not meet the target after the burning, the next baking stage is automatically skipped. The steps for establishing the yellow baking characteristics and the dry baking characteristics are as follows: S31, obtaining a current yellow ratio and dry ratio according to a yellow ratio and dry ratio prediction model, and subtracting an initial yellow ratio and dry ratio obtained by the yellow ratio and dry ratio prediction model to obtain a changed yellow ratio and a changed dry ratio value; S32, subtracting the current baking time from the start baking time to obtain the cumulative baking time of the current baking stage; S33, dividing the yellow change and dry change values by the cumulative baking time, dry bulb temperature, and wet bulb temperature to obtain yellow baking characteristics and dry baking characteristics of the tobacco leaves.

2. The tobacco leaf initial baking control method based on the image reasoning model according to claim 1 is characterized in that: The image of the flue-cured tobacco sample to be tested in S12 is collected by multiple cameras in the flue-curing room.

3. The tobacco leaf initial baking control method based on the image reasoning model according to claim 2 is characterized in that: In the S12 , the multiple cameras in the baking room move back and forth in real time to capture images.

4. The tobacco leaf initial baking control method based on image inference model according to claim 1 is characterized in that: The calculation formulas for the yellow baking characteristics and the dry baking characteristics in S33 are: Yellow baking characteristics = yellow change / (accumulated baking time * dry bulb temperature * wet bulb temperature) Dry bake characteristics = change in dryness / (accumulated bake time * dry bulb temperature * wet bulb temperature).

5. A tobacco leaf primary baking control method based on an image reasoning model according to any one of claims 1 to 4, characterized in that: The fire delay time is 2 hours.

Images