Tobacco leaf curing data acquisition system, yield prediction system and operation method

Abstract

The invention provides a tobacco leaf curing data acquisition system, a yield prediction system and an operation method, and the method comprises the steps: inputting tobacco leaf initial data through a personal terminal computer device, collecting curing process data through a first acquisition module, collecting packaged tobacco leaf data during warehousing through a second acquisition module, and transmitting the data to a cloud computing device; and the cloud computing equipment processes the collected tobacco initial data, the baking process data and the packaged tobacco data during warehousing, generates tobacco baking result data, and issues the result to the personal terminal computer equipment. The baking result data comprises the tobacco leaf quality after baking, the baking loss and the yellow tobacco rate of the tobacco leaves; and judging whether the tobacco leaves moved out of the tobacco curing barn are consistent with the packaged tobacco leaves stored in the tobacco curing barn or not. According to the method, tobacco leaf curing data collection and yield prediction are carried out in the whole process, the workload of tobacco farmers or tobacco curing workers is reduced, and industrial production is facilitated.

Description

Technical Field

[0001] The invention relates to a data collection and yield prediction system for tobacco leaf baking using data processing technology, in particular to a tobacco leaf baking data collection system, a yield prediction system and an operation method. Background Art

[0002] Tobacco leaf baking is a crucial part of the tobacco leaf production process. It refers to the process of slowly drying the fresh tobacco leaves after harvest by controlling the temperature, humidity, time and ventilation conditions, so that the color changes from green to yellow or reddish brown, and a series of physical and chemical changes occur, ultimately forming high-quality tobacco leaves with specific aroma, taste and burning properties. The technical level of the baking process directly affects the quality and availability of tobacco leaves, and is a key link in determining the final economic value of tobacco leaves.

[0003] The existing tobacco leaf baking process lacks a data collection and yield prediction system. Data entry relies on manual work, resulting in a large workload and inaccurate yield prediction, which is not conducive to industrial production. Therefore, it is necessary to design a tobacco leaf baking data collection and yield prediction system with a high degree of automation. At the same time, due to various complex reasons, the tobacco leaves may be damaged, dropped, or replaced during transportation after baking and storage, resulting in the high-quality tobacco leaves after baking becoming low-quality tobacco leaves or insufficient quality when they are stored. Therefore, it is necessary to solve this problem through technical means. Summary of the invention

[0004] In order to solve this technical problem, the present invention provides a tobacco baking data collection system, a yield prediction system, and an operation method.

[0005] Among them, the tobacco leaf baking data collection and processing system includes a personal terminal computer device with output and display functions, a cloud computing device that has data connection with the personal terminal computer device and processes data, a database of the cloud computing device, and a data collection device.

[0006] Among them, the personal terminal computer device is used to input the initial data of tobacco leaves and display the tobacco leaf baking result data sent by the cloud computing device.

[0007] The data acquisition device includes a first acquisition module for collecting baking process data, performing data preprocessing, and sending the data to a cloud computing device; and a second acquisition module for collecting tobacco leaf packaging data when entering storage and sending the data to a cloud computing device.

[0008] The cloud computing device is used to process the collected tobacco leaf initial data and baking process data as well as the tobacco leaf packaging data when entering the warehouse, generate tobacco leaf baking result data, and send part of the result data to the personal terminal computer device.

[0009] The tobacco leaf baking data collection and processing system as described above, wherein the initial tobacco leaf data at least includes tobacco leaf planting geographical information, tobacco leaf type information, grower information, tobacco flue-curing room coding information, baking type information, tobacco leaf rack number, and identification code on the tobacco leaf rack.

[0010] The baking process data includes: the baking time of the tobacco-curing room, the temperature and humidity information during baking, and the real-time quality information of each tobacco leaf rack collected by the first collection module during baking.

[0011] After the baking is completed, when the tobacco leaves are moved out of the tobacco curing room, the first acquisition module collects the photo of the tobacco rack including the tobacco leaves and the identification code, which is the first photo; the second acquisition module collects the quality of the packaged tobacco leaves and the photo of the packaged tobacco leaves including the identification code, which is the second photo.

[0012] The collected tobacco leaf initial data, baking process data and storage tobacco leaf data are processed to generate tobacco leaf baking result data, specifically: (1) Baking result data include tobacco leaf quality and baking loss after baking; (2) Compare the first photo with the second photo, compare the quality of the tobacco leaves after curing with the quality of the packaged tobacco leaves, and combine the identification code to determine whether the tobacco leaves when they are removed from the tobacco curing room are consistent with the packaged tobacco leaves when they are put into storage; (3) Based on the first and second photos, the yellowing rate of tobacco leaves is obtained.

[0013] The tobacco leaf baking data collection and processing system as described above, wherein the real-time quality information is obtained by a weighing sensor arranged on a tobacco leaf rack; the temperature and humidity information during baking is obtained by a temperature sensor and a humidity sensor in a tobacco baking room, and is automatically uploaded to a cloud computing device in real time under the control of a first acquisition module control program; after baking is completed, the cloud computing device collects complete data on the entire tobacco baking process.

[0014] The tobacco leaf baking data collection and processing system as described above, wherein the identification code is one of a bar code and a QR code; the identification code is detachably arranged on the tobacco leaf rack, and after the baking is completed, the tobacco leaves are removed from the tobacco leaf rack and packaged, and the identification code is transferred to the packaged tobacco leaves; A color calibration card is provided on the identification code.

[0015] In the tobacco leaf baking data collection and processing system as described above, the personal terminal computer device at least has: System settings module, used by users to adjust various system settings; User interface module, including the interface for inputting data and displaying prediction results; Reporting and visualization module: Generate reports of forecast results, including charts and data visualization; Cloud computing equipment must have at least: The data module is used to collect relevant data in the tobacco leaf baking process, process and analyze the data, and build a tobacco leaf baking yield prediction model.

[0016] The tobacco leaf baking yield prediction system provided by the present invention obtains the tobacco leaf initial data, tobacco leaf baking process data, and tobacco leaf baking result data through the above tobacco leaf baking data collection and processing system, and obtains the tobacco leaf baking yield prediction model by automatically identifying and training the data; after obtaining the tobacco leaf initial data, the tobacco leaf baking yield prediction model can be used to obtain the quality of the baked tobacco leaves, baking loss, and yellow tobacco rate.

[0017] In the tobacco baking yield prediction system described above, the prediction model is constructed by the following steps: S11: Select key characteristic data that affect tobacco leaf baking yield from tobacco leaf initial data, tobacco leaf baking process data, and tobacco leaf baking result data, and mainly perform the following operations: (1) Use interpolation methods to fill missing values, delete and fill some data; then use statistical methods and anomaly detection algorithms to identify and process outliers. (2) Based on domain knowledge and practical experience, new feature data is constructed to form original feature data; then the original feature data is transformed, including standardization and normalization of continuous features and one-hot encoding of discrete features; (3) Use statistical methods, Bayesian, and deep learning methods to select the most influential features for tobacco yield prediction; S12: Establish a machine learning model, select a machine learning algorithm to build a prediction model, take the feature variables as input, take the tobacco leaf baking yield as output, start training the model and perform verification and evaluation, as follows: (1) The selection of a machine learning algorithm to construct a prediction model specifically refers to the selection of a large language model, and further refers to the selection of a Transformer model and a Multi-head Attention mechanism; When constructing the prediction model, the initial tobacco leaf data and tobacco leaf baking process data collected from a large number of different types of tobacco flue-cured houses are used as model input, and the tobacco leaf baking result data is used as model output to train the tobacco leaf baking yield prediction model; The initial tobacco leaf data includes tobacco leaf planting geographical area information, tobacco leaf quality, and tobacco leaf type data; the tobacco leaf baking process data includes baking type, baking temperature, and baking humidity; the tobacco leaf baking result data includes tobacco leaf quality after baking, baking loss, and yellow tobacco rate information; (2) The model training process is to divide the data set into a training set and a test set, using 70% of the data as the training set and 30% of the data as the test set; at the same time, the training set is used to train the selected regression algorithm to learn the relationship between the model input and the tobacco leaf baking result data; (3) Use the test set to verify the trained model and observe the performance of the model on new data; use cross entropy, mean square error, and root mean square error indicators to evaluate the model's fit and prediction accuracy; (4) Tune the model’s hyperparameters, including regularization parameters and kernel function parameters; based on the model’s feature importance feedback, further optimize the feature engineering process to improve the model’s predictive ability.

[0018] In the tobacco baking yield prediction system described above, the statistical method is the 3σ principle; the anomaly detection algorithm is Isolation Forest; The statistical method is one of variance analysis and correlation analysis; The deep learning method is a Bayesian probabilistic deep learning model.

[0019] The operating method of the tobacco leaf baking yield prediction system of the present invention is executed on the tobacco leaf baking data acquisition and processing system; the method comprises the following steps: S21: The personal terminal computer device inputs the initial data of tobacco leaves, the identification code information on the tobacco leaf rack, and the tobacco leaf rack number, and sends the information to the cloud computing device; S22: During the baking process, the first collection module collects the baking time, the temperature and humidity information during baking, and the real-time quality information of each tobacco leaf rack, and sends the information to the cloud computing device; S23: When the tobacco leaves are removed from the curing room along with the tobacco leaf rack after curing, the first acquisition module acquires a photo of the tobacco leaf rack including the tobacco leaves, the identification code, and the color correction card as the first photo, and sends the photo to the cloud computing device; S24: After the tobacco leaves are packaged, the identification code and the color calibration card are transferred to the packaged tobacco leaves; S25: When the packaged tobacco leaves are put into storage, the second collection module collects the quality information of the packaged tobacco leaves, and collects the photos of the packaged tobacco leaves including the identification code and the color correction card as the second photos; and sends them to the cloud computing device; S26: The cloud computing device determines whether the tobacco leaves when being removed from the curing room with the tobacco leaf racks are consistent with the packaged tobacco leaves when being put into storage; the determination method is: The first photo is compared with the second photo to determine whether the color information is consistent, thereby obtaining a first set of comparison results; the quality information of the stored packaged tobacco leaves is compared with the quality information of the baked tobacco leaves to determine whether the quality information is consistent, thereby obtaining a second set of comparison results; when the first set of comparison results and the second set of comparison results are consistent, the tobacco leaves are considered to be the same; otherwise, the tobacco leaves are considered to be damaged in transportation; Wherein: before the second acquisition module acquires the photos of the packaged tobacco leaves, one of the following information is restricted to be sent to the personal terminal computer device and restricted to be displayed: (1) a photo of the tobacco leaf rack including the tobacco leaves and the identification code; (2) the quality of the tobacco leaves after baking; when one type of information is restricted to be sent to the personal terminal computer device and restricted to be displayed, the remaining information will no longer be restricted to be sent to the personal terminal computer device and displayed; S27: After the second acquisition module collects the photos of the packaged tobacco leaves, at least the quality of the tobacco leaves after baking, baking losses, yellow tobacco rate, the judgment results of whether the baked tobacco leaves are consistent with the packaged tobacco leaves when entering the warehouse, the photos of the tobacco leaf hanging racks, and the photos of the packaged tobacco leaves are sent to the personal terminal computer device and displayed.

[0020] As in the above-mentioned operation method of the tobacco leaf baking yield prediction system, the cloud computing device determines whether the tobacco leaves when being moved out of the baking room with the tobacco leaf racks are consistent with the packaged tobacco leaves when being put into storage; the determination method is: The first photo is compared with the second photo to determine whether the color information is consistent, and a first set of comparison results is obtained, specifically: Adjusting the chroma, saturation and brightness of at least one of the first photo and the second photo so that the chroma, saturation and brightness of the color correction card on the adjusted first photo are the same as those of the color correction card on the second photo; then comparing the chroma and saturation of the tobacco leaves in the first photo and the tobacco leaves in the second photo after the adjustment to see if they are consistent. If they are consistent, the comparison result is consistent; otherwise, it is inconsistent; Compare the quality information of the stored packaged tobacco leaves with the quality information of the baked tobacco leaves to determine whether the quality information is consistent, and obtain a second set of comparison results; When the first set of comparison results and the second set of comparison results are consistent, they are considered to be the same tobacco leaves; otherwise, they are considered to be damaged tobacco leaves.

[0021] Beneficial technical effects: The present invention collects tobacco leaf baking data and predicts yield in the whole process, reduces the workload of tobacco growers or tobacco-curing workers, and is conducive to industrial production.

[0022] The invention can accurately predict the yield according to the initial data of tobacco leaves, which is beneficial to industrial production.

[0023] The present invention can identify whether the tobacco leaves when being moved out of the curing room with the tobacco leaf hanging racks are consistent with the packaged tobacco leaves when being put into storage, so as to technically prevent the situation that the high-quality tobacco leaves after curing become low-quality tobacco leaves or the situation that the quality is insufficient when being put into storage. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] Figure 1 It is a structural diagram of the tobacco baking data acquisition system of the present invention. DETAILED DESCRIPTION

[0025] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0026] Example 1: The tobacco leaf baking data collection and processing system of the present invention comprises: (1) A personal terminal computer device 7 with output and display functions.

[0027] (2) Data acquisition equipment.

[0028] (3) A cloud computing device 6 that has a data connection with the personal terminal computer device and processes data, and a database of the cloud computing device.

[0029] 1. The personal terminal computer device 7 may be a tablet computer or a smart phone with a touch screen and a display screen. This example takes a smart phone as an example.

[0030] Smartphones are used to input initial tobacco leaf data and display tobacco leaf baking result data sent by cloud computing devices.

[0031] The smart phone includes a system setting module, which is used for the user to adjust various system settings.

[0032] The smartphone is used to input the initial data of tobacco leaves and display the tobacco leaf baking result data sent by the cloud computing device. This can be completed by a mini-program developed on WeChat or by a system built through a WeChat public account.

[0033] User interface module on smartphones: provides a friendly user interface, including an interface for inputting data, displaying prediction results, and possible adjustment options; Report and visualization module: generates reports of prediction results, including charts and data visualization, so that users can intuitively understand and analyze the prediction results.

[0034] The initial tobacco leaf data includes tobacco leaf planting geographic information, tobacco leaf type information, grower information, flue-curing room code information, and baking type information. These need to be entered when using the system for the first time. In the subsequent use process, the information can be automatically retrieved from the first input information according to the registered account to avoid repeated input. The initial tobacco leaf data also includes the tobacco leaf rack number and the identification code on the tobacco leaf rack. This is information that needs to be re-entered every time the system is used.

[0035] Tobacco growers upload basic information through smartphones, including tobacco planting geographic information, tobacco type information, grower information, tobacco flue-curing room code information, baking type information, tobacco rack number, and identification code on the tobacco rack. Tobacco growers put fresh tobacco leaves into the baking rack to start baking, and confirm the start and end time of baking on their smartphones.

[0036] 2. Data acquisition equipment.

[0037] The data acquisition device includes a first acquisition module, which is used to collect baking process data, perform data preprocessing, and send the data to a cloud computing device.

[0038] and a second acquisition module for collecting data of packaged tobacco leaves when entering the warehouse and sending the data to a cloud computing device; The baking process data includes: during baking, the baking time of the tobacco-curing room, the temperature and humidity information during baking, and the real-time quality information of each tobacco leaf rack 2 collected by the first collection module; The baking result data include: the quality of tobacco leaves after baking, baking loss; After the baking is completed, the first acquisition module acquires a photo of the tobacco rack including the tobacco leaves and the identification code; the second acquisition module acquires a photo of the packaged tobacco leaves including the identification code; The real-time quality information is obtained through the weighing sensor 4 arranged on the tobacco leaf rack 2; the temperature and humidity information during baking is obtained through the temperature sensor and humidity sensor of the tobacco flue-curing room, and is automatically uploaded to the cloud computing device 6 in real time under the control of the first acquisition module control program; when the baking is completed, the cloud computing device will collect complete data of the entire process of tobacco flue-curing.

[0039] The data acquisition equipment needs to collect relevant data during the tobacco leaf baking process, including environmental conditions and baking room parameters, and perform preliminary processing; therefore, the data acquisition module needs to have certain data processing capabilities, that is, it needs to have a processor, a memory, and a data transceiver module 5. For example, when pre-processing data, the collected data needs to be cleaned, including removing outliers, filling missing values, and data standardization and normalization operations; of course, these processes can also be directly processed by cloud computing devices 6, which have stronger processing capabilities.

[0040] The first acquisition module is used to collect baking process data. Tobacco growers upload basic information and initial tobacco data through WeChat applet or WeChat public account. Tobacco growers put fresh tobacco leaves into tobacco rack 2 to start baking, and confirm the start and end time on the WeChat client.

[0041] Weighing sensors, temperature sensors and humidity sensors are installed on each tobacco leaf rack in the flue-curing room. Under the control of the control program, the weight data, temperature and humidity data and collection time are automatically uploaded to the cloud computing device. The control end can view the process data of the flue-cured tobacco in real time through the network. When the flue-curing is completed, the cloud computing device will collect complete data on the entire flue-cured tobacco process.

[0042] The working process of the first acquisition module is as follows: the tobacco leaf rack 2 is placed under the weighing sensor 4, the controller reads the weighing sensor readings regularly, and uploads them to the cloud computing device 6. The error of the weighing sensor is 0.1%, so as to collect the weight changes of the tobacco leaves in real time during the baking process. Each tobacco leaf rack is attached with a high-temperature resistant and reusable identification code and color correction card 3, which is used for traceability, correction, and consistency comparison in the tobacco leaf baking and acquisition links. After the tobacco leaves are hung on the rack, the identification code and color correction card 3 are attached to the tobacco leaf rack, and after the baking is completed, the tobacco leaves are removed from the tobacco leaf rack and packaged, and the identification code and color correction card 3 are transferred to the packaged tobacco leaves.

[0043] The identification code can be a barcode or a QR code. The color calibration card 3 and the barcode are printed on the same carrier and are transferred synchronously during transfer.

[0044] At the exit of the curing room, an image collector, that is, a camera, and a fill light with uniform illumination are set. When the worker moves the tobacco leaves out of the curing room with the tobacco leaf rack, the worker points the tobacco leaves with the tobacco leaf rack to the camera and takes several photos, which are the first photos. The first photo has the tobacco leaves on the rack, the identification code, and the color calibration card.

[0045] The identification code and color correction card are high temperature resistant and can be repeatedly torn and pasted. After tobacco curing is completed, image information (the image information includes tobacco leaves 1 on the tobacco rack, the identification code and the color correction card 3), i.e., the first photo, can be collected for each rack of tobacco leaves.

[0046] At the entrance of the flue-cured tobacco warehouse, an image collector, that is, a camera, and a fill light with uniform illumination are set. When the workers move to the warehouse, they must be weighed again and the packaged tobacco leaves must be photographed. The second photo is sent to the cloud computing device. In the acquisition stage (i.e., when entering the warehouse), the flue-cured tobacco image (the image information includes the packaged tobacco leaves, the identification code and the color correction card 3), i.e., the second photo, can be obtained according to the identification code, and matched with the grading information to achieve the traceability of the flue-cured tobacco, and compared with the first photo to determine the consistency of the tobacco leaves on the tobacco rack and the final warehoused tobacco leaves, to prevent transportation damage, falling, or being replaced in the middle, resulting in the high-quality tobacco leaves after baking becoming low-quality tobacco leaves or insufficient quality when entering the warehouse, or to identify the transportation damage, falling, or being replaced in the middle, and identify the high-quality tobacco leaves after baking becoming low-quality tobacco leaves or insufficient quality when entering the warehouse.

[0047] 3. The cloud computing device 6 is used to process data and has a data connection with the personal terminal computer device. The cloud computing device has a large database.

[0048] The cloud computing device is used to process the collected initial tobacco leaf data, the baking process data, and the packaged tobacco leaf data when entering the warehouse, and to generate tobacco leaf baking result data, and send part of the result data to the personal terminal computer device.

[0049] include: (1) Based on the real-time quality information of tobacco leaves 1, the quality and baking loss of the tobacco leaves after baking are generated.

[0050] (2) Compare the photos of the cured tobacco leaves with the photos of the packaged tobacco leaves, compare the quality of the cured tobacco leaves with the quality of the packaged tobacco leaves, and combine the identification code to determine whether the cured tobacco leaves and the packaged tobacco leaves at the time of storage are consistent.

[0051] (3) Obtain the yellowing rate of tobacco leaves based on the photos of the baked tobacco leaves and the packaged tobacco leaves.

[0052] The cloud computing device provides the system operator with a backend data module, including a user interface module that provides a user-friendly interface, including an interface for inputting data, displaying forecast results, and possible adjustment options; a report and visualization module: generates a report of the forecast results, including charts and data visualization, so that users can intuitively understand and analyze the forecast results.

[0053] Based on the Internet of Things and cloud computing technologies, the entire baking process of the tobacco flue-curing room is monitored to obtain multi-dimensional tobacco leaf fresh-to-dry ratio data, baking damage ratio data, and average single leaf weight data before and after baking for different regions, different types of flue-curing rooms, different baking methods, different baking processes, different grades, and different parts.

[0054] It includes a data module, which is used to collect relevant data in the tobacco leaf baking process, process and analyze the data, and build a tobacco leaf baking yield prediction model. It includes a system setting module, which is used to adjust various system settings and protect the data security of the model.

[0055] Includes a backend data module that allows users to configure system parameters, such as model update frequency and data source. Includes a user management and permission module: manage system users, set user permissions, and ensure that only authorized users can access and use the system. Includes a system monitoring and log module: implements real-time monitoring of system operation status, records key events and logs, and facilitates tracking and problem solving. Includes a security and privacy protection module: strengthens system security, ensures data confidentiality and integrity, and complies with relevant privacy protection regulations. Includes a system maintenance and update module: provides system maintenance functions, including regular model updates, system problem solving, and performance optimization.

[0056] The cloud computing equipment has built a tobacco baking yield prediction model, including: (1) Feature engineering module: used to select and extract features related to baking yield for subsequent model training.

[0057] (2) Model training module: includes model selection, parameter adjustment and training process, and uses historical data to train the production prediction model.

[0058] (3) Model evaluation module: Evaluate the trained model and use the validation set to check the performance of the model to ensure that it can accurately predict on unseen data.

[0059] (4) Real-time data receiving module: used to receive real-time production-related data to ensure that the system can continuously monitor and predict the current tobacco leaf baking conditions.

[0060] (5) Forecast generation module: Use the trained model to generate production forecast results based on real-time input data.

[0061] (6) Alarm module: Through big data analysis, the flue-curing process model is obtained, and alarms are issued for data omissions and missed data collection when the data violates the model rules.

[0062] Example 2: The work of the cloud computing equipment also includes the construction of a tobacco baking yield prediction model based on machine learning. The construction steps are as follows: Step 1: Feature selection and engineering: Based on the results of data collection, the key features that affect tobacco yield in the planting and curing stages are selected. The main operations are as follows: (1) Data cleaning: Processing missing values: use interpolation methods to fill missing values, or delete or fill data according to specific circumstances; Outlier processing: identify and process outliers, which can be processed by statistical methods or special anomaly detection algorithms; the statistical method is the 3σ principle. The anomaly detection algorithm is Isolation Forest.

[0063] (2) Feature Engineering: Feature construction: construct new features based on domain knowledge and practical experience; Feature transformation: perform data transformation on original features, such as standardizing and normalizing continuous features and performing one-hot encoding on discrete features; Feature selection: Use statistical methods, Bayesian, and deep learning methods to select the most influential features for tobacco yield prediction. The statistical methods are variance analysis and correlation analysis, and the deep learning method is the Bayesian probability deep model.

[0064] Step 2: Establish a machine learning model. According to the actual situation, select the appropriate machine learning algorithm to build a tobacco yield estimation model for the planting stage and the baking stage. Take the characteristic variables as input and the tobacco yield as output. Train the model and perform verification and evaluation. The main process is as follows: (1) Model selection: The Transformer model and Multi-head Attention mechanism widely used in large language models are used to build a baking yield prediction model. The tobacco leaf quality, moisture content, temperature, humidity, baking type, and tobacco leaf type data collected from a large number of different types of tobacco flue-cured rooms are used as model input, and the quality of tobacco leaves after baking, baking loss, and yellow tobacco rate are used as output to train the tobacco leaf baking yield prediction model. (2) Model training: The data set is divided into a training set and a test set. Usually 70% of the data is used as the training set and 30% of the data is used as the test set. At the same time, the training set is used to train the selected regression algorithm to learn the relationship between features and yield. (3) Model verification and evaluation: Model validation: Use the test set to validate the trained model and observe the performance of the model on new data; Evaluation indicators: Cross entropy (CE), mean square error (MSE), and root mean square error (RMSE) indicators are used to evaluate the model's fit and prediction accuracy; (4) Model optimization: Hyperparameter tuning: Tune the model's hyperparameters, such as regularization parameters and kernel function parameters, to improve the model's performance; Feature engineering feedback: Based on the feature importance feedback of the model, the feature engineering process is further optimized to improve the model’s predictive ability; (5) Model application; Apply the trained model to actual scenarios to provide tobacco yield grading prediction and management decision support at the planting, baking and purchasing stages, which is applicable to different decision-making levels; Example 3: Step 1: The personal terminal computer device inputs the initial data of tobacco leaves, the identification code information on the tobacco rack, and the tobacco rack number, and sends the information to the cloud computing device. The cloud computing device associates the tobacco rack number with the identification code information, and later collects the baking time, temperature and humidity information during baking, and the real-time quality information of each tobacco rack, and sends it to the personal terminal computer device through the associated identification code information.

[0065] Step 2: During the baking process, the first acquisition module collects the baking time, temperature and humidity information during baking, and the real-time quality information of each tobacco rack, and sends the information to the cloud computing device.

[0066] Step 3: After the baking is finished, when the tobacco leaves are moved out of the baking room with the tobacco leaf racks, the first acquisition module collects photos of the tobacco leaf racks including the tobacco leaves, identification codes, and color correction cards as the first photos, and sends them to the cloud computing device. To this end, an image collector, that is, a camera, is set at the exit of the baking room, and a fill light with uniform illumination is set. When the workers move the tobacco leaves out of the baking room with the tobacco leaf racks, they aim the tobacco leaves with the tobacco leaf racks at the camera and take several photos as the first photos. The first photos include the tobacco leaves on the racks, identification codes, and color correction cards.

[0067] Step 4: When the workers take the tobacco leaves off the tobacco leaf racks and pack them, they will transfer the identification code and color calibration card to the packed tobacco leaves. Then they will be loaded onto trucks and transported to the storage area (that is, the tobacco leaf purchasing point for purchase).

[0068] At this time, the photos of the tobacco racks with tobacco leaves and identification codes, or the quality of the tobacco leaves after baking, or at least one of the two, will be restricted from being sent to personal terminal computer devices, and even if they are sent to personal terminal computer devices, they will be restricted from being displayed. In other words, the transportation link or tobacco growers only know the photos of the tobacco leaves after baking, but not the quality after baking, or only know the quality after baking but not the photos after baking. Prevent the possibility of replacing tobacco leaves through technical means in the transportation link.

[0069] One example is that the photo of the tobacco rack including tobacco leaves and identification codes cannot be sent to the personal terminal computer device, and the display is restricted, because the first photo is used to compare the color with the second photo. If the personal terminal computer device knows the status of the first photo in advance, it is equivalent to providing a reference for the transportation link to replace tobacco leaves or damage tobacco leaves artificially. Therefore, restricting the display can prevent the possibility of replacing tobacco leaves or damaging tobacco leaves artificially through technical means in the transportation link.

[0070] However, there is no restriction on sending and displaying other information except the photos of the tobacco racks to personal terminal computer devices, such as baking time, temperature and humidity information during baking, quality information of each tobacco rack, quality of tobacco after baking, baking loss, and yellow tobacco rate. By letting tobacco growers know this information in advance, tobacco growers can build trust in the accuracy of the system and calculate their own tobacco income in advance.

[0071] Step 5: When the packaged tobacco leaves are put into storage, the second acquisition module collects the quality information of the packaged tobacco leaves and collects the photos of the packaged tobacco leaves including the identification code and color correction card as the second photo; for this purpose, an image collector, that is, a camera, is set at the entrance of the flue-cured tobacco warehouse, and a fill light with uniform illumination is set. When the workers move the packaged tobacco leaves into the warehouse, they must be weighed again and photographed. The second photo is sent to the cloud computing device.

[0072] Step 6: The cloud computing device determines whether the tobacco leaves removed from the drying room with the tobacco leaf racks are consistent with the packaged tobacco leaves put into storage; the determination method is: Compare the first photo with the second photo to determine whether the color information is consistent, and obtain the first set of comparison results; compare the quality information of the stored packaged tobacco leaves with the quality information of the baked tobacco leaves to determine whether the quality information is consistent, and obtain the second set of comparison results; when the first set of comparison results and the second set of comparison results are consistent, they are considered to be the same tobacco leaves; otherwise, they are considered to be damaged tobacco leaves. In case of inconsistency, it may be that the computer recognition is wrong, and the results can be manually reviewed and corrected on site.

[0073] Compare the first photo with the second photo to determine whether the color information is consistent. The specific method is: Adjust the hue, saturation, and brightness of the first or second photo, or adjust both photos at the same time. The color correction card in RGB color mode has at least four colors, namely, a grayscale bar, which displays grayscale 0-100% through multiple grayscale blocks on the grayscale bar, and then a red (R) channel, a green (G) channel, and a blue (B) channel. The processor can first identify the grayscale of multiple grayscale blocks in the first photo, and then adjust the grayscale of multiple grayscale blocks in the second photo to be consistent with the grayscale of multiple grayscale blocks in the first photo by adjusting the brightness and contrast of the second photo; then identify the saturation, hue, and hue of the red (R) channel in the first photo, and then adjust the saturation, hue, and hue of the red (R) channel of the second photo to be consistent with the red (R) channel in the first photo; and then use the same method to adjust the green (G) channel and the blue (B) channel. In short, the color correction card on the adjusted first photo is made to have the same hue, saturation, and brightness as the color correction card on the second photo. Similarly, the color calibration card can also be a CMYK composite channel, cyan (C) channel, magenta (M) channel, yellow (Y) channel, and black (K) channel.

[0074] Then, the hue and saturation of the tobacco leaves in the first photo and the second photo are compared to see if they are consistent. The comparison method is to extract the images of the tobacco leaves in the first photo and the second photo, and extract the RGB value of the image of the tobacco leaves. When the RGB value of the tobacco leaves in the second photo is the same as the RGB value of the tobacco leaves in the first photo, the comparison result is consistent; the same here should be regarded as setting a threshold interval, and the RGB value of the tobacco leaves in the second photo is the same as the RGB value of the tobacco leaves in the first photo within a threshold interval. When the color information is judged to be consistent, the comparison result is consistent; otherwise, it is inconsistent.

[0075] After the second collection module collects the photos of the packaged tobacco leaves, the judgment result of whether the tobacco leaves that have been cured are consistent with the packaged tobacco leaves when they are put into storage, the photos of the tobacco leaf racks, and the photos of the packaged tobacco leaves can be sent to the personal terminal computer device and displayed.

[0076] The quality of the tobacco leaves after baking is the last real-time quality information of each tobacco leaf rack collected by the first collection module, which is the quality of the tobacco leaves after baking on each tobacco leaf rack.

[0077] The first real-time quality information of each tobacco leaf rack collected by the first collection module is the quality of tobacco leaves before baking on each tobacco leaf rack; the baking loss is the difference between the quality of tobacco leaves before baking and the quality of tobacco leaves after baking.

[0078] The yellow tobacco rate is the ratio of the mass of tobacco leaves after baking to the mass of tobacco leaves before baking.

[0079] The above is an exemplary description and does not represent the protection scope of the present invention; the above examples can be combined to generate new examples.

Claims

1. A tobacco leaf baking data collection and processing system, comprising a personal terminal computer device with output and display functions, a cloud computing device that has a data connection with the personal terminal computer device and processes data, a database of the cloud computing device, and a data collection device; characterized in that: The personal terminal computer device is used to input the initial data of tobacco leaves and display the tobacco leaf baking result data sent by the cloud computing device; The data acquisition device includes a first acquisition module for collecting baking process data, performing data preprocessing, and sending the data to a cloud computing device; and a second acquisition module for collecting data of packaged tobacco leaves when entering the warehouse and sending the data to a cloud computing device; The cloud computing device is used to process the collected tobacco leaf initial data and baking process data as well as the tobacco leaf packaging data when entering the warehouse, generate tobacco leaf baking result data, and send part of the result data to the personal terminal computer device.

2. The tobacco leaf baking data collection and processing system according to claim 1, characterized in that: The initial tobacco leaf data at least includes tobacco leaf planting geographical information, tobacco leaf type information, grower information, tobacco flue-curing room coding information, curing type information, tobacco leaf rack number, and identification code on the tobacco leaf rack; The baking process data includes: during baking, the baking time of the tobacco room, the temperature and humidity information during baking, and the real-time quality information of each tobacco leaf rack collected by the first collection module; After the baking is completed, when the tobacco leaves are moved out of the tobacco-curing room, the first acquisition module acquires a photo of the tobacco rack including the tobacco leaves and the identification code, which is the first photo; the second acquisition module acquires the quality of the packaged tobacco leaves and the photo of the packaged tobacco leaves including the identification code, which is the second photo; The collected tobacco leaf initial data, baking process data and storage tobacco leaf data are processed to generate tobacco leaf baking result data, specifically: (1) Baking result data include tobacco leaf quality and baking loss after baking; (2) Compare the first photo with the second photo, compare the quality of the tobacco leaves after curing with the quality of the packaged tobacco leaves, and combine the identification code to determine whether the tobacco leaves when they are removed from the tobacco curing room are consistent with the packaged tobacco leaves when they are put into storage; (3) Based on the first and second photos, the yellowing rate of tobacco leaves is obtained.

3. The tobacco leaf baking data collection and processing system according to claim 2, characterized in that: The real-time quality information is obtained through a weighing sensor arranged on a tobacco leaf rack; the temperature and humidity information during baking is obtained through a temperature sensor and a humidity sensor in a tobacco flue-curing room, and is automatically uploaded to a cloud computing device in real time under the control of a first acquisition module control program; after baking is completed, the cloud computing device collects complete data on the entire process of tobacco flue-curing.

4. The tobacco leaf baking data collection and processing system according to claim 2, characterized in that: The identification code is one of a barcode and a QR code; the identification code is detachably arranged on the tobacco leaf rack, and after the tobacco leaves are removed from the tobacco leaf rack and packaged after baking, the identification code is transferred to the packaged tobacco leaves; A color calibration card is provided on the identification code.

5. The tobacco leaf baking data collection and processing system according to claim 1, characterized in that: The personal terminal computer equipment shall have at least: System settings module, used by users to adjust various system settings; User interface module, including the interface for inputting data and displaying prediction results; Reporting and visualization module: Generate reports of forecast results, including charts and data visualization; Cloud computing equipment must have at least: The data module is used to collect relevant data in the tobacco leaf baking process, process and analyze the data, and build a tobacco leaf baking yield prediction model.

6. The tobacco leaf baking yield prediction system is characterized by: The initial tobacco data, tobacco baking process data, and tobacco baking result data obtained by the tobacco baking data collection and processing system in any one of claims 1 to 5 are automatically recognized and trained to obtain a tobacco baking yield prediction model; after obtaining the initial tobacco data, the quality of the baked tobacco, baking loss, and yellow tobacco rate can be obtained through the tobacco baking yield prediction model.

7. The tobacco leaf baking yield prediction system according to claim 6, characterized in that: The prediction model is constructed through the following steps: S11: Select key characteristic data that affect tobacco leaf baking yield from tobacco leaf initial data, tobacco leaf baking process data, and tobacco leaf baking result data, and mainly perform the following operations: (1) Use interpolation methods to fill missing values, delete and fill some data; then use statistical methods and anomaly detection algorithms to identify and process outliers. (2) Based on domain knowledge and practical experience, new feature data is constructed to form original feature data; then the original feature data is transformed, including standardization and normalization of continuous features and one-hot encoding of discrete features; (3) Use statistical methods, Bayesian, and deep learning methods to select the most influential features for tobacco yield prediction; S12: Establish a machine learning model, select a machine learning algorithm to build a prediction model, take the feature variables as input, take the tobacco leaf baking yield as output, start training the model and perform verification and evaluation, as follows: (1) The selection of a machine learning algorithm to construct a prediction model specifically refers to the selection of a large language model, and further refers to the selection of a Transformer model and a Multi-head Attention mechanism; When constructing the prediction model, the initial tobacco leaf data and tobacco leaf baking process data collected from a large number of different types of tobacco flue-cured houses are used as model input, and the tobacco leaf baking result data is used as model output to train the tobacco leaf baking yield prediction model; The initial tobacco leaf data includes tobacco leaf planting geographical area information, tobacco leaf quality, and tobacco leaf type data; the tobacco leaf baking process data includes baking type, baking temperature, and baking humidity; the tobacco leaf baking result data includes tobacco leaf quality after baking, baking loss, and yellow tobacco rate information; (2) The model training process is to divide the data set into a training set and a test set, using 70% of the data as the training set and 30% of the data as the test set; at the same time, the training set is used to train the selected regression algorithm to learn the relationship between the model input and the tobacco leaf baking result data; (3) Use the test set to verify the trained model and observe the performance of the model on new data; use cross entropy, mean square error, and root mean square error indicators to evaluate the model's fit and prediction accuracy; (4) Tune the model’s hyperparameters, including regularization parameters and kernel function parameters; based on the model’s feature importance feedback, further optimize the feature engineering process to improve the model’s predictive ability.

8. The tobacco leaf baking yield prediction system according to claim 7, characterized in that: The statistical method is the 3σ principle; the anomaly detection algorithm is Isolation Forest; The statistical method is one of variance analysis and correlation analysis; The deep learning method is a Bayesian probabilistic deep learning model.

9. The operation method of the tobacco leaf baking yield prediction system is characterized in that: The method is performed on the tobacco leaf baking data collection and processing system according to any one of claims 1 to 6; the method comprises the following steps: S21: The personal terminal computer device inputs the initial data of tobacco leaves, the identification code information on the tobacco leaf rack, and the tobacco leaf rack number, and sends the information to the cloud computing device; S22: During the baking process, the first collection module collects the baking time, the temperature and humidity information during baking, and the real-time quality information of each tobacco leaf rack, and sends the information to the cloud computing device; S23: When the tobacco leaves are removed from the curing room along with the tobacco leaf rack after curing, the first acquisition module acquires a photo of the tobacco leaf rack including the tobacco leaves, the identification code, and the color correction card as the first photo, and sends the photo to the cloud computing device; S24: After the tobacco leaves are packaged, the identification code and the color calibration card are transferred to the packaged tobacco leaves; S25: When the packaged tobacco leaves are put into storage, the second collection module collects the quality information of the packaged tobacco leaves, and collects the photos of the packaged tobacco leaves including the identification code and the color correction card as the second photos; and sends them to the cloud computing device; S26: The cloud computing device determines whether the tobacco leaves when being removed from the curing room with the tobacco leaf racks are consistent with the packaged tobacco leaves when being put into storage; the determination method is: Compare the first photo with the second photo to determine whether the color information is consistent, and obtain a first set of comparison results; Compare the quality information of the stored packaged tobacco leaves with the quality information of the baked tobacco leaves to determine whether the quality information is consistent, and obtain a second set of comparison results; When the first set of comparison results and the second set of comparison results are consistent, they are considered to be the same tobacco leaves; otherwise, they are considered to be damaged tobacco leaves; Wherein: before the second acquisition module acquires the photos of the packaged tobacco leaves, one of the following information is restricted to be sent to the personal terminal computer device and restricted to be displayed: (1) a photo of the tobacco leaf rack including the tobacco leaves and the identification code; (2) the quality of the tobacco leaves after baking; when one type of information is restricted to be sent to the personal terminal computer device and restricted to be displayed, the remaining information will no longer be restricted to be sent to the personal terminal computer device and displayed; S27: After the second acquisition module collects the photos of the packaged tobacco leaves, at least the quality of the tobacco leaves after baking, baking losses, yellow tobacco rate, the judgment results of whether the baked tobacco leaves are consistent with the packaged tobacco leaves when entering the warehouse, the photos of the tobacco leaf hanging racks, and the photos of the packaged tobacco leaves are sent to the personal terminal computer device and displayed.

10. The operating method of the tobacco leaf baking yield prediction system according to claim 1, characterized in that: The cloud computing device determines whether the tobacco leaves when being removed from the curing room with the tobacco leaf racks are consistent with the packaged tobacco leaves when being put into storage; the determination method is: The first photo is compared with the second photo to determine whether the color information is consistent, and a first set of comparison results is obtained, specifically: Adjusting the chroma, saturation, and brightness of at least one of the first photo and the second photo so that the chroma, saturation, and brightness of the color correction card on the adjusted first photo are the same as those of the color correction card on the second photo; then comparing the chroma and saturation of the tobacco leaves in the first photo and the tobacco leaves in the second photo after the adjustment to see if they are consistent, and if so, the comparison result is consistent; Otherwise it is inconsistent; Compare the quality information of the stored packaged tobacco leaves with the quality information of the baked tobacco leaves to determine whether the quality information is consistent, and obtain a second set of comparison results; When the first set of comparison results and the second set of comparison results are consistent, they are considered to be the same tobacco leaves; otherwise, they are considered to be damaged tobacco leaves.

Images