Field tobacco leaf curing characteristic real-time discrimination method and system

Abstract

The invention belongs to the technical field of tobacco leaf intelligent production, and particularly relates to a field tobacco leaf curing characteristic real-time discrimination method and system. The method comprises the following steps: acquiring the spectral reflectance of fresh tobacco leaves in a field growth period, extracting a wave band sensitive to a baking characteristic index from the spectral reflectance of the fresh tobacco leaves according to the baking characteristic index of the fresh tobacco leaves capable of representing the quality of the baked tobacco leaves, and inputting the wave band sensitive to the baking characteristic index into a fresh tobacco leaf baking characteristic index prediction model, obtaining a prediction result of the field fresh tobacco leaf baking characteristic index data; the fresh tobacco leaf baking characteristic index prediction model is obtained by training wavebands sensitive to each baking characteristic index in fresh tobacco leaves and corresponding actually measured baking characteristic index data. According to the method, real-time and in-situ judgment of the baking characteristics of the fresh tobacco leaves is achieved, the problem that the baking loss rate is increased due to the fact that the quality deviation between the tobacco leaves and the tobacco leaves suitable for baking is too large is solved, and a foundation is laid for scientific tobacco loading in a baking room, optimal baking process matching and baking quality improvement.

Description

Technical Field

[0001] The present invention belongs to the technical field of intelligent tobacco production, and in particular relates to a method and system for real-time identification of baking characteristics of field tobacco leaves. Background Art

[0002] The tobacco production process generally places the tobacco leaves harvested from the field in a curing room, and controls the temperature, humidity and ventilation conditions to bake the tobacco leaves to become tobacco products with certain quality, style and grade standards. Therefore, the quality of the tobacco leaves after baking is predicted, and the baking process or the tobacco leaves sent for baking can be adjusted according to the prediction results to obtain more tobacco products that meet the quality requirements. In the prior art, the tobacco leaves being baked are generally sampled during the baking stage, and the quality of the tobacco leaves after baking is predicted based on the tobacco leaf indicators during the baking process. This method predicts the indicators of tobacco leaves in the baking stage, at which time the tobacco leaves have been harvested. Since the quality of fresh tobacco leaves at the time of harvest (the quality of fresh tobacco leaves is generally reflected in physical and chemical characteristics, dry matter accumulation and enzyme activity) is also an important factor affecting the baking characteristics and the quality of the baked tobacco leaves, when the quality of the harvested tobacco leaves is too large compared to that of the tobacco leaves suitable for baking, the quality of the baked tobacco leaves predicted using the existing prediction method may be greatly deviated. Even if the baking parameters are adjusted, the baking loss rate will increase due to the large deviation. Summary of the invention

[0003] The purpose of the present invention is to provide a method and system for real-time identification of field tobacco leaf baking characteristics, so as to solve the problem that the existing prediction method may lead to an increase in the tobacco leaf baking loss rate.

[0004] In order to solve the above technical problems, the present invention provides a real-time discrimination method for baking characteristics of field tobacco leaves, comprising: obtaining the spectral reflectance of fresh tobacco leaves in the field growth period; extracting the bands sensitive to the baking characteristic index from the spectral reflectance of the fresh tobacco leaves according to the baking characteristic index of the fresh tobacco leaves that can characterize the quality of the baked tobacco leaves; inputting the bands sensitive to the baking characteristic index into a baking characteristic index prediction model for fresh tobacco leaves to obtain the prediction results of the baking characteristic index data of field fresh tobacco leaves; the baking characteristic index prediction model for fresh tobacco leaves is trained by using the bands in the fresh tobacco leaves that are sensitive to each baking characteristic index and the corresponding baking characteristic index data actually measured.

[0005] Furthermore, the fresh tobacco leaf baking characteristic indicators include total nitrogen and starch.

[0006] Furthermore, the fresh tobacco leaf baking characteristic index is obtained by screening, and the screening method is: based on the physical and chemical characteristics of fresh tobacco leaves before baking and the quality of fresh tobacco leaves after baking, a fresh tobacco leaf baking characteristic index is established. The fresh tobacco leaf baking characteristic index can characterize the quality of fresh tobacco leaves after baking.

[0007] Furthermore, the tobacco leaf quality prediction model after tobacco leaf curing is established based on the physical and chemical characteristics of fresh tobacco leaves of different maturity before curing and the quality of tobacco leaves after curing.

[0008] Furthermore, the method also includes: after obtaining the baking characteristic index data, inputting the obtained baking characteristic index data into a prediction model for the chemical composition of fresh tobacco leaves after baking to obtain a prediction result of the chemical composition of fresh tobacco leaves after baking; the prediction model for the chemical composition of fresh tobacco leaves after baking is obtained based on the analysis of the baking characteristic index data of fresh tobacco leaves and the corresponding chemical composition measurement data after baking.

[0009] Furthermore, the fresh tobacco leaf baking characteristic index prediction model is a gradient boosting decision tree model.

[0010] Furthermore, the spectral reflectance of fresh tobacco leaves during the field growth period is the reflectance corrected based on the reflectance calibration curve.

[0011] The beneficial effects of the above technical scheme are as follows: the present invention is a pioneering invention, which determines the baking characteristic index of fresh tobacco leaves that can characterize the quality of the tobacco leaves after baking, and uses the bands in the fresh tobacco leaves that are sensitive to each baking characteristic index and the corresponding actually measured baking characteristic index data to establish a fresh tobacco leaf baking characteristic index prediction model; before the tobacco leaves are harvested, the spectral reflectance of the fresh tobacco leaves in the field growth period is directly photographed, and the relevant sensitive bands are extracted from the spectral reflectance of the fresh tobacco leaves according to the baking characteristic index of the fresh tobacco leaves, and the extracted sensitive bands are input into the fresh tobacco leaf baking characteristic index prediction model, so as to obtain the fresh tobacco leaves in situ The baking characteristic index data of leaves can be used to predict the spatial distribution results of the baking characteristic index of fresh tobacco leaves in the field, thereby realizing real-time and in-situ judgment of the baking characteristics of fresh tobacco leaves. According to the predicted baking characteristic index, it can be known whether the fresh tobacco leaves meet the quality requirements after baking, so that the fresh tobacco leaves that meet the quality requirements in the field can be harvested. If they do not meet the quality requirements, the tobacco leaves can be not harvested first, so as to avoid the problem of excessive deviation in quality between tobacco leaves and tobacco leaves suitable for baking, which leads to increased baking loss rate, and lays the foundation for scientific loading of tobacco in the baking room, optimal baking process matching and improvement of baking quality.

[0012] In order to solve the above technical problems, the present invention also provides a real-time discrimination system for field tobacco leaf baking characteristics, comprising a processor and a data acquisition module, wherein the data acquisition module is used to capture spectral reflectance images of fresh tobacco leaves during the field growth period, and send the spectral reflectance of fresh tobacco leaves during the field growth period to the processor; the processor is used to execute stored computer program instructions to implement the above-mentioned real-time discrimination method for field tobacco leaf baking characteristics.

[0013] Furthermore, the system also includes a data display module, which is used to display the predicted baking characteristic index data during the field growth period.

[0014] The beneficial effects of the above technical scheme are as follows: the present invention is a pioneering invention, which determines the baking characteristic index of fresh tobacco leaves that can characterize the quality of the tobacco leaves after baking, and uses the bands in the fresh tobacco leaves that are sensitive to each baking characteristic index and the corresponding actually measured baking characteristic index data to establish a fresh tobacco leaf baking characteristic index prediction model; before the tobacco leaves are harvested, the spectral reflectance of the fresh tobacco leaves in the field growth period is directly photographed, and the relevant sensitive bands are extracted from the spectral reflectance of the fresh tobacco leaves according to the baking characteristic index of the fresh tobacco leaves, and the extracted sensitive bands are input into the fresh tobacco leaf baking characteristic index prediction model, so as to obtain the fresh tobacco leaves in situ The baking characteristic index data of leaves can be used to predict the spatial distribution results of the baking characteristic index of fresh tobacco leaves in the field, thereby realizing real-time and in-situ judgment of the baking characteristics of fresh tobacco leaves. According to the predicted baking characteristic index, it can be known whether the fresh tobacco leaves meet the quality requirements after baking, so that the fresh tobacco leaves that meet the quality requirements in the field can be harvested. If they do not meet the quality requirements, the tobacco leaves can be not harvested first, so as to avoid the problem of excessive deviation in quality between tobacco leaves and tobacco leaves suitable for baking, which leads to increased baking loss rate, and lays the foundation for scientific loading of tobacco in the baking room, optimal baking process matching and improvement of baking quality. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] Figure 1 is a flow chart of real-time identification of field tobacco leaf baking characteristics according to an embodiment of the method of the present invention; Figure 2 It is a block diagram of a baking characteristic real-time determination system according to a system embodiment of the present invention. DETAILED DESCRIPTION

[0016] In order to make the purpose, technical solutions and advantages of the present invention more clear, the specific implementation methods of the present invention are further described below in conjunction with the accompanying drawings.

[0017] The present invention determines the baking characteristic index of fresh tobacco leaves that can characterize the quality of tobacco leaves after baking from the physical and chemical characteristics of tobacco leaves, and establishes a baking characteristic index prediction model for fresh tobacco leaves by using the wavebands in fresh tobacco leaves that are sensitive to each baking characteristic index and the corresponding baking characteristic index data actually measured. Before the tobacco leaves are harvested, the spectral reflectance of the fresh tobacco leaves in the field growth period is directly photographed to obtain the baking characteristic index data of the fresh tobacco leaves in situ. According to the predicted baking characteristic index data of the fresh tobacco leaves that can characterize the quality of the tobacco leaves after baking, it can be known whether the fresh tobacco leaves meet the quality requirements after baking, so that the fresh tobacco leaves that meet the quality requirements in the field are harvested. If they do not meet the quality requirements, the tobacco leaves may not be harvested first, so as to avoid the problem of increased baking loss rate due to excessive deviation in quality between tobacco leaves and tobacco leaves suitable for baking. The present invention can realize real-time and in-situ judgment of fresh tobacco leaves, and lay a foundation for real-time judgment of tobacco leaves suitable for baking in the field, timely harvesting of tobacco leaves, scientific loading of tobacco in a baking room, optimal baking process matching, and improvement of baking quality.

[0018] Method Embodiment A method for real-time identification of the baking characteristics of field tobacco leaves of the present invention is as follows: Figure 1 As shown, the specific process is as follows: 1. A quality prediction model for fresh tobacco leaves after baking is established based on the physical and chemical characteristics of fresh tobacco leaves before baking and the quality of fresh tobacco leaves after baking. The quality prediction model for fresh tobacco leaves after baking based on the physical and chemical characteristics of fresh tobacco leaves is analyzed using the stepwise regression method, and the baking characteristic index data of fresh tobacco leaves during the field growth period that can characterize the quality of fresh tobacco leaves after baking are screened. The physical and chemical characteristics before baking include leaf moisture content and chemical composition, and the quality of fresh tobacco leaves after baking is characterized by baking loss rate, appearance quality of fresh tobacco leaves after baking, and chemical composition of fresh tobacco leaves after baking. Furthermore, in order to improve the prediction accuracy, the present invention establishes a quality prediction model for fresh tobacco leaves after baking based on the physical and chemical characteristics of fresh tobacco leaves of different maturity before baking and the quality of fresh tobacco leaves after baking. Specifically, it includes the following steps: 1) Select plots with different nutritional levels in the field, harvest each plot separately, and classify the fresh tobacco leaves according to different maturity levels after harvesting to form differentiated fresh tobacco leaf types. Different nutritional levels include three types: weak development level, medium development level and strong development level, and different maturity levels include three different maturity levels: under-ripe, mature and mature.

[0019] 2) Each type of fresh tobacco leaf was divided into three equal parts, which were used for leaf moisture content measurement, fresh tobacco leaf chemical composition test and baking. After baking, the baking loss rate, appearance quality of the baked tobacco leaves and the chemical composition of the baked tobacco leaves were calculated.

[0020] 3) A quality prediction model for fresh tobacco leaves after curing was established based on the physical and chemical characteristics of fresh tobacco leaves of different maturity before curing and the quality of fresh tobacco leaves after curing. The correlation between the independent variables (moisture content of tobacco leaves and chemical composition of fresh tobacco leaves) and the dependent variables (baking loss rate, appearance quality of fresh tobacco leaves and chemical composition of fresh tobacco leaves) was analyzed. The stepwise regression method was used to determine the independent variables and dependent variables with good correlation, and the independent variables with good correlation were used as the baking characteristic indicators of fresh tobacco leaves.

[0021] In this embodiment, the most relevant baking characteristic indicators obtained by stepwise regression are total nitrogen and starch, and the chemical components after baking are total nitrogen, reducing sugar and sugar-alkali ratio. Based on the total nitrogen and starch data of fresh tobacco leaves and the corresponding measurement data of chemical components after baking, a prediction model of chemical components after baking of fresh tobacco leaves is obtained, and the prediction results of total nitrogen, reducing sugar and sugar-alkali ratio of baked tobacco leaves are obtained.

[0022] 2. Obtain the spectral reflectance of fresh tobacco leaves during the field growth period, extract the bands sensitive to the baking characteristic index from the spectral reflectance of fresh tobacco leaves, input the bands sensitive to the baking characteristic index into the prediction model of fresh tobacco leaf baking characteristic index, and obtain the prediction results of the baking characteristic index data of fresh tobacco leaves in the field. The baking characteristic index data of fresh tobacco leaves during the field growth period is obtained based on the real-time identification system of baking characteristics, such as Figure 2As shown, the device includes a data acquisition module, a data processing module and a data display module. The data acquisition module includes a drone and a hyperspectral camera. The hyperspectral camera is mounted on the drone and is used to collect spectral reflectance images of fresh tobacco leaves during the field growth period. The drone can adopt a series of drone platforms such as quad-rotor, hexa-rotor, and octa-rotor. The data processing module includes a gray plate, a reflectance correction module, a sensitive band extraction module, an index acquisition module, a storage module and a file transfer module. The data display module communicates with the data processing module in real time, and displays the predicted values ​​of the baking characteristic indicators of tobacco leaves during the field growth period in real time.

[0023] The reflectivity correction module corrects the real-time collected spectral reflectivity image based on the gray plate and the pre-acquired reflectivity calibration curve. In this embodiment, the reflectivity calibration curve is a 50% reflectivity calibration curve.

[0024] The sensitive band extraction module extracts the bands of the corrected spectral reflectance to obtain the bands sensitive to the baking characteristic index; among them, the sensitive bands to the total nitrogen content of fresh tobacco leaves include 406nm, 408nm, 417nm, 481nm, 689nm, 706nm, 715nm, 727nm, 763nm and 774nm. The sensitive bands to the starch content of fresh tobacco leaves include 415nm, 487nm, 757nm, 759nm, 647nm, 834nm, 881nm, 891nm, 913nm and 949nm. The sensitive band extraction module extracts the sensitive bands of baking characteristic index based on the SHAP value theory, feature importance or optimal narrow band vegetation index selection method.

[0025] The index acquisition module has a built-in fresh tobacco leaf baking characteristic index prediction model, and the extracted bands sensitive to the baking characteristic index are input into the fresh tobacco leaf baking characteristic index prediction model to estimate the baking characteristic index data, so as to distinguish the field distribution of the baking characteristic index of fresh tobacco leaves in the field in real time and in situ. Among them, the baking characteristic index estimation model is obtained by training the relevant model using the bands in the fresh tobacco leaves that are sensitive to each baking characteristic index and the corresponding baking characteristic index data actually measured. The model can adopt a partial least squares regression model (PLSR), a random forest regression model (RF), a gradient boosting decision tree model (GBDT) or a support vector machine regression model (SVR). Preferably, the baking characteristic index estimation model adopts a gradient boosting decision tree model. The storage module is used to store the estimation results in the storage module; the file transfer module transmits the storage results to the data display module in real time. The data display module and the data processing module communicate with each other in real time, and display the predicted baking characteristic index prediction value of tobacco leaves in the field growth period in real time.

[0026] 3. Input the predicted baking characteristic index data into the prediction model of the chemical composition of fresh tobacco leaves after baking to obtain the prediction results of the chemical composition of the baked tobacco leaves, so as to in situ determine the predicted results of the chemical composition of fresh tobacco leaves after baking in the field, which can guide the harvesting period of tobacco leaves and provide a basis for real-time determination of tobacco leaves suitable for baking in the field, timely harvesting of tobacco leaves and baking of tobacco leaves.

[0027] System Example The present invention provides a real-time identification system for the baking characteristics of field tobacco leaves, including a processor and a data acquisition module, the data acquisition module is used to shoot the spectral reflectance image of fresh tobacco leaves during the field growth period, and send the spectral reflectance of fresh tobacco leaves during the field growth period to the processor, the processor is used to execute the stored computer program instructions to implement a real-time identification method for the baking characteristics of field tobacco leaves, the method is the real-time identification method for the baking characteristics of field tobacco leaves described in the above method embodiment, and will not be described in detail here. Further, the system also includes a data display module, the data display module is used to display the predicted baking characteristic index data of the field growth period.

Claims

1. A method for real-time identification of the curing characteristics of field tobacco leaves, characterized in that: include: Obtain the spectral reflectance of fresh tobacco leaves during the field growth period; According to the fresh tobacco leaf baking characteristic index which can characterize the quality of the tobacco leaf after baking, the band sensitive to the baking characteristic index is extracted from the spectral reflectance of the fresh tobacco leaf; The bands sensitive to the baking characteristic indexes are input into the baking characteristic index prediction model of fresh tobacco leaves to obtain the prediction results of the baking characteristic index data of field fresh tobacco leaves; the baking characteristic index prediction model of fresh tobacco leaves is trained using the bands in fresh tobacco leaves that are sensitive to each baking characteristic index and the corresponding actually measured baking characteristic index data.

2. The method for real-time identification of the curing characteristics of field tobacco leaves according to claim 1, characterized in that: The fresh tobacco leaf baking characteristic indicators include total nitrogen and starch.

3. The method for real-time identification of the curing characteristics of field tobacco leaves according to claim 1 or 2, characterized in that: The fresh tobacco leaf baking characteristic index is obtained by screening, and the screening method is: based on the physical and chemical characteristics of fresh tobacco leaves before baking and the quality of fresh tobacco leaves after baking, a fresh tobacco leaf baking quality prediction model is established, and the fresh tobacco leaf baking characteristic index that can characterize the quality of fresh tobacco leaves after baking is analyzed by using a stepwise regression method.

4. The method for real-time identification of the curing characteristics of field tobacco leaves according to claim 3, characterized in that: The tobacco leaf quality prediction model after flue-curing is established based on the physical and chemical characteristics of fresh tobacco leaves of different maturity before flue-curing and the quality of the flue-cured tobacco leaves.

5. The method for real-time identification of the curing characteristics of field tobacco leaves according to any one of claims 1 to 4, characterized in that: The method also includes: after obtaining the baking characteristic index data, inputting the obtained baking characteristic index data into a prediction model for the chemical composition of fresh tobacco leaves after baking to obtain a prediction result of the chemical composition of the fresh tobacco leaves after baking; the prediction model for the chemical composition of fresh tobacco leaves after baking is obtained based on the analysis of the baking characteristic index data of fresh tobacco leaves and the corresponding measurement data of the chemical composition after baking.

6. The method for real-time identification of the curing characteristics of field tobacco leaves according to claim 1, characterized in that: The fresh tobacco leaf baking characteristic index prediction model is a gradient boosting decision tree model.

7. The method for real-time identification of the curing characteristics of field tobacco leaves according to claim 1, characterized in that: The spectral reflectance of fresh tobacco leaves during the field growth period is the reflectance corrected based on the reflectance calibration curve.

8. A real-time identification system for field tobacco leaf baking characteristics, comprising a processor and a data acquisition module, characterized in that: The data acquisition module is used to capture the spectral reflectance images of fresh tobacco leaves during the field growth period, and send the spectral reflectance of fresh tobacco leaves during the field growth period to the processor; the processor is used to execute stored computer program instructions to implement a real-time discrimination method for field tobacco leaf baking characteristics as described in any one of claims 1-7.

9. The real-time identification system of field tobacco leaf curing characteristics according to claim 8, characterized in that: The system also includes a data display module, which is used to display the predicted baking characteristic index data during the field growth period.

Images