Tobacco planting quality prediction model construction method and system based on multi-source data fusion

Abstract

The invention relates to the technical field of smart tobacco planting, in particular to a tobacco planting quality prediction model construction method and system based on multi-source data fusion. The tobacco leaf planting quality prediction model construction method based on multi-source data fusion comprises the following steps that multi-source data of a tobacco leaf planting area are collected in real time, and the multi-source data comprise soil humidity data Sw, soil nitrogen phosphorus and potassium content data Npk, illumination intensity data Li, environment temperature data Te and tobacco leaf growth image data Im; and acquiring the acquired data, calculating according to a formula to obtain a comprehensive quality index CI, alpha, beta and gamma are weight coefficients, analyzing the tobacco growth image data Im at the same time, and extracting the leaf color saturation Sc and the disease spot number Nd. According to the method, the soil, weather and image data are integrated, the comprehensive quality index is constructed and compared with the preset threshold value, and meanwhile, the threshold value self-adaptive mechanism is combined, so that accurate prediction of the tobacco leaf quality and planting parameter optimization can be achieved.

Description

Technical Field

[0001] The present invention relates to the field of smart tobacco planting technology, and specifically to a method and system for constructing a tobacco planting quality prediction model based on multi-source data fusion. Background Art

[0002] Tobacco leaves are the core raw material for products such as cigarettes, cigars, and pipe tobacco. Their cultivation and management require specific agricultural techniques and environmental conditions. Tobacco cultivation refers to the process of artificially cultivating tobacco plants, harvesting their leaves, and processing them into tobacco products. Accurately predicting tobacco leaf quality is key to improving the economic benefits of tobacco agriculture.

[0003] Existing technologies primarily rely on a single environmental parameter, such as soil moisture sensors, to predict tobacco leaf quality. This is done by combining data from historical experience with a fixed threshold. For example, if soil moisture falls below 40%, the system prompts irrigation; if it rises above 60%, drainage is indicated.

[0004] However, tobacco leaf growth is influenced by multiple factors, and relying solely on a single parameter can lead to one-sided predictions. For example, in high-temperature, high-light environments, tobacco leaves can still suffer from light stress burns even when soil moisture is appropriate. Existing systems are unable to capture these complex influences, resulting in an error rate of up to 30%. Furthermore, fixed thresholds cannot adapt to regional climate variations, further reducing the universality of the prediction model. Summary of the Invention

[0005] The purpose of the present invention is to provide a method and system for constructing a tobacco planting quality prediction model based on multi-source data fusion, so as to solve the problem raised in the above background technology that tobacco growth is affected by the synergistic influence of multiple factors, and relying only on a single parameter will lead to one-sided prediction results.

[0006] To achieve the above object, the present invention provides a method for constructing a tobacco planting quality prediction model based on multi-source data fusion, comprising the following steps:

[0007] S1, real-time collection of multi-source data in the tobacco planting area, wherein the multi-source data includes soil moisture data Sw, soil nitrogen, phosphorus and potassium content data Npk, light intensity data Li, ambient temperature data Te and tobacco growth image data Im;

[0008] S2. Get the collected data and calculate it according to the formula The comprehensive quality index CI is calculated, where α, β, γ and is the weight coefficient and satisfies At the same time, the tobacco leaf growth image data Im is analyzed to extract the leaf color saturation Sc and the number of disease spots Nd. If Nd>0, the calculation formula of the comprehensive quality index CI is updated;

[0009] S3. Preset a threshold value Th, and dynamically modify the threshold value Th based on historical data. Compare the calculated comprehensive quality index CI with the preset threshold value Th. If CI ≥ Th, the tobacco leaf quality is judged to be excellent. If CI < Th, it is determined that the planting parameters need to be adjusted.

[0010] S4. Generate a planting parameter optimization plan based on the determination result and feed it back to the user terminal.

[0011] As a further improvement of this technical solution, the calculation formula for updating the comprehensive quality index CI in step S2 is specifically:

[0012] The tobacco leaf growth image data Im is analyzed by convolutional neural network to extract the leaf color saturation Sc and the number of disease spots Nd, where Nd>0;

[0013] Add Sc and Nd as additional input parameters to the calculation formula of the comprehensive quality index CI, and update its calculation formula to: Among them, δ is the newly added weight coefficient, and satisfies

[0014] As a further improvement of the present technical solution, the threshold value Th is dynamically corrected according to historical data in step S3, specifically:

[0015] Set a time window T and calculate the proportion Py of tobacco leaves judged as excellent in the time window T;

[0016] If Py>70%, Th increases by 5%;

[0017] If Py < 30%, Th decreases by 5%;

[0018] If 30%≤Py≤70%, Th remains unchanged.

[0019] As a further improvement of the present technical solution, in step S4, a planting parameter optimization scheme is generated according to the determination result, specifically:

[0020] Analyze the current comprehensive quality index CI and each sub-parameter value, including soil moisture data Sw, soil nitrogen, phosphorus and potassium content data Npk, light intensity data Li, ambient temperature data Te, leaf color saturation Sc, and number of disease spots Nd;

[0021] If the soil moisture data Sw is lower than the lower limit of soil moisture, it is marked as insufficient irrigation. Then according to the formula ΔQ=(Sw 理想 -Sw 当前 )×η×1000, calculate the amount of water to be replenished ΔQ, where Sw 理想 is the soil moisture data under ideal conditions, Sw 当前is the soil moisture data under the current state, η is the soil permeability coefficient;

[0022] If the NPK content is lower than the standard NPK content, it is marked as needing additional NPK fertilizer, and the type and amount of compound fertilizer are recommended according to the NPK ratio;

[0023] If Nd>5, that is, the number of disease spots exceeds the standard, it is marked as a high risk of pests and diseases. Then, combined with the color abnormality of the leaf color saturation Sc, the pesticide spraying plan and light control strategy are matched.

[0024] The present invention also provides a tobacco planting quality prediction model construction system based on multi-source data fusion, the system comprising a multi-source data acquisition module, a data fusion module, a quality prediction module, a dynamic adjustment module and a user terminal;

[0025] The multi-source data acquisition module includes a soil sensor, a meteorological sensor and an image acquisition unit, which is used to collect soil moisture data Sw, soil nitrogen, phosphorus and potassium content data Npk, light intensity data Li, ambient temperature data Te and tobacco growth image data Im in the tobacco planting area in real time;

[0026] The data fusion module is used to calculate the comprehensive quality index CI according to the formula;

[0027] The quality prediction module is used to compare the calculated comprehensive quality index with a preset threshold and output a corresponding determination result;

[0028] The dynamic adjustment module is used to generate a planting parameter optimization plan based on the determination result and feed it back to the user terminal;

[0029] The user terminal is used to receive the implantation parameter optimization plan and to display the CI real-time change curve and the threshold comparison result.

[0030] As a further improvement of the present technical solution, the image acquisition unit is used to analyze the tobacco leaf growth image data Im through a convolutional neural network, extract the leaf color saturation Sc and the number of disease spots Nd, and add the leaf color saturation Sc and the number of disease spots Nd as additional input parameters to the calculation formula of the comprehensive quality index CI, thereby updating the calculation formula of the comprehensive quality index CI.

[0031] As a further improvement of the present technical solution, the dynamic adjustment module is further configured to dynamically modify the threshold Th according to historical data.

[0032] As a further improvement of the present technical solution, the system further includes a cloud storage module for real-time backup of the multi-source data and model parameters collected by the multi-source data acquisition module, and interacting with the user terminal through an encrypted communication protocol.

[0033] Compared with the prior art, the present invention has the following beneficial effects:

[0034] 1. In the present invention, a comprehensive quality index is constructed by integrating multi-dimensional data such as soil moisture, nitrogen, phosphorus and potassium content, light intensity, ambient temperature and tobacco leaf growth images, and real-time comparison is performed based on preset thresholds. This overcomes the limitation of existing technologies that rely on a single parameter, and can fully reflect the complex environmental effects of tobacco leaf growth, such as the synergistic effects of light damage and soil moisture under high temperature and strong light stress, making the prediction results more in line with the actual growth status, thereby reducing the misjudgment rate and significantly improving the prediction accuracy and reliability.

[0035] 2. In the present invention, by combining the threshold adaptive mechanism, the preset threshold can be dynamically corrected according to historical planting data. This mechanism enables the model to adapt to the climate conditions and planting cycle changes in different regions, avoiding the misjudgment problem caused by fixed thresholds. At the same time, the dynamically adjusted threshold is linked with the real-time optimized planting parameters (such as irrigation amount and fertilization strategy), which can accurately guide users to adjust field management measures, achieve tobacco leaf quality improvement and maximize resource utilization efficiency, thereby saving water and fertilizer costs. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] Figure 1 The figure is a flow chart of the steps of the method for constructing a tobacco planting quality prediction model based on multi-source data fusion according to the present invention.

[0037] Figure 2 This is a principle block diagram of the tobacco planting quality prediction model construction system based on multi-source data fusion of the present invention. DETAILED DESCRIPTION

[0038] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0039] In a specific embodiment, Figure 1 As shown in FIG, a method for constructing a tobacco planting quality prediction model based on multi-source data fusion includes the following steps:

[0040] The first step is to collect multi-source data from tobacco-growing areas in real time. This data includes soil moisture data Sw, soil nitrogen, phosphorus, and potassium content data Npk, light intensity data Li, ambient temperature data Te, and tobacco growth image data Im. By integrating this multi-dimensional data, we can fully reflect the complex environmental effects on tobacco growth, improving prediction accuracy and reliability.

[0041] The second step is to obtain the collected data and analyze the tobacco leaf growth image data Im through a convolutional neural network to extract the leaf color saturation Sc and the number of disease spots Nd.

[0042] If Nd=0, according to the formula The comprehensive quality index CI is calculated, where α, β, γ and is the weight coefficient and satisfies

[0043] If Nd>0, Sc and Nd are added as additional input parameters to the calculation formula of the comprehensive quality index CI, and the calculation formula is updated to: Among them, δ is the newly added weight coefficient, and satisfies

[0044] By using convolutional neural networks to analyze image data, extract key information, and update the calculation formula of the comprehensive quality index, the prediction results can be more in line with the actual growth status and the misjudgment rate can be reduced.

[0045] Step 3: Preset the threshold Th and dynamically modify the threshold Th based on historical data. Specifically:

[0046] Set a time window T and calculate the proportion Py of tobacco leaves judged as excellent in the time window T;

[0047] If Py>70%, Th increases by 5%;

[0048] If Py < 30%, Th decreases by 5%;

[0049] If 30%≤Py≤70%, Th remains unchanged;

[0050] Then the calculated comprehensive quality index CI is compared with the preset threshold Th. If CI ≥ Th, the tobacco leaf quality is judged to be excellent. If CI < Th, it is determined that the planting parameters need to be adjusted.

[0051] Combined with the threshold adaptive mechanism, the preset threshold is dynamically corrected according to historical planting data, so that the model can adapt to the climatic conditions and planting cycle changes in different regions, avoiding the misjudgment problem caused by fixed thresholds.

[0052] Step 4: Generate a planting parameter optimization plan based on the judgment results and feed it back to the user terminal. The specific operations for generating the planting parameter optimization plan are as follows:

[0053] Analyze the current comprehensive quality index CI and each sub-parameter value, including soil moisture data Sw, soil nitrogen, phosphorus and potassium content data Npk, light intensity data Li, ambient temperature data Te, leaf color saturation Sc, and number of disease spots Nd;

[0054] If the soil moisture data Sw is lower than the lower limit of soil moisture, it is marked as insufficient irrigation. Then according to the formula ΔQ=(Sw 理想 -Sw 当前 )×η×1000, calculate the amount of water to be replenished ΔQ, where Sw 理想 is the soil moisture data under ideal conditions, Sw 当前 is the soil moisture data under the current state, η is the soil permeability coefficient;

[0055] If the NPK content is lower than the standard NPK content, it is marked as needing additional NPK fertilizer, and the type and amount of compound fertilizer are recommended according to the NPK ratio;

[0056] If Nd>5, that is, the number of disease spots exceeds the standard, it is marked as a high risk of pests and diseases. Then, combined with the color abnormality of the leaf color saturation Sc, the pesticide spraying plan and light control strategy (such as shade net deployment) are matched.

[0057] The dynamically adjusted thresholds are linked with real-time optimized planting parameters to accurately guide users to adjust field management measures, thereby improving tobacco leaf quality and maximizing resource utilization efficiency, saving water and fertilizer costs.

[0058] In another specific embodiment, Figure 2 As shown, a tobacco planting quality prediction model construction system based on multi-source data fusion includes a multi-source data acquisition module, a data fusion module, a quality prediction module, a dynamic adjustment module, a cloud storage module and a user terminal.

[0059] The multi-source data acquisition module includes soil sensors, meteorological sensors, and an image acquisition unit. It collects real-time data on soil moisture (Sw), soil nitrogen, phosphorus, and potassium (Npk), light intensity (Li), ambient temperature (Te), and tobacco growth image data (Im) from tobacco-growing areas. This real-time data collection provides a comprehensive and accurate foundation for model construction.

[0060] The data fusion module is used to calculate the comprehensive quality index CI according to the formula. Through data fusion, the comprehensive quality index is constructed to fully reflect the growth status of tobacco leaves.

[0061] The image acquisition unit is used to analyze tobacco leaf growth image data Im using a convolutional neural network to extract leaf color saturation Sc and the number of diseased spots Nd. The leaf color saturation Sc and the number of diseased spots Nd are then added as additional input parameters to the calculation formula of the comprehensive quality index CI, thereby updating the calculation formula of the comprehensive quality index CI and improving the accuracy and fit of the prediction results.

[0062] The quality prediction module is used to compare the calculated comprehensive quality index with a preset threshold and output a corresponding judgment result. Based on the preset threshold, the judgment is made and clear planting parameter adjustment suggestions are provided to the user.

[0063] The dynamic adjustment module is used to generate a planting parameter optimization plan based on the determination results and feed it back to the user terminal. The dynamic adjustment module is also used to dynamically modify the threshold Th based on historical data to achieve dynamic adjustment and optimization of planting parameters and improve tobacco leaf quality.

[0064] The cloud storage module is used to back up the multi-source data and model parameters collected by the multi-source data acquisition module in real time, and interacts with the user terminal through an encrypted communication protocol, ensuring the security and reliability of the data and facilitating user access and use at any time.

[0065] The user terminal is used to receive the implant parameter optimization plan and display the CI real-time change curve and threshold comparison results, providing users with an intuitive and convenient interface to facilitate users to view and adjust implant parameters.

[0066] The beneficial effects of the present invention are described below by means of specific embodiments:

[0067] Example 1: Comprehensive quality prediction and irrigation optimization under high temperature and strong light environment.

[0068] A tobacco field is located in a plain area. It has a continuous high temperature in summer (daytime ambient temperature reaches 35℃), the light intensity is as high as 1200Lux, and the soil moisture is monitored at 45% (lower than the ideal value Sw 理想 =55%). Mild burn symptoms appeared on tobacco leaves, with reduced color saturation. The system achieves precise irrigation and light damage warnings through multi-source data fusion and dynamic threshold adjustment.

[0069] The first step is multi-source data collection.

[0070] Soil data: soil moisture data Sw = 45%, soil nitrogen, phosphorus and potassium content data Npk = 1.2% (standard value must be ≥1.5%).

[0071] Meteorological data: light intensity data Li = 1200 Lux, ambient temperature data Te = 35°C.

[0072] Image data: The tobacco leaf growth image shows that the leaf color saturation Sc=0.65 (healthy value ≥0.8) and the number of disease spots Nd=3.

[0073] Step 2: Calculate the comprehensive quality index.

[0074] Weight coefficient: Optimized based on historical data, α = 0.3, β = 0.25, γ = 0.2, δ=0.1。

[0075] Comprehensive quality index CI = 0.3·45% + 0.25·1.2% + 0.2·1200 + 0.15·35 + 0.1·(0.65-3) = 245.135.

[0076] Step 3: Dynamically adjust the threshold and compare.

[0077] Historical data: In the past 30 days, the proportion of high-quality tobacco leaves Py = 25% (<30%), and the trigger threshold was lowered by 5%. The original threshold Th = 260, and the adjusted threshold Th = 247.

[0078] Judgment result: CI=245<Th=247, marking that the implantation parameters need to be adjusted.

[0079] Step 4: Output the planting parameter optimization plan.

[0080] Soil moisture data Sw = 45 is lower than the soil moisture lower limit of 55%. According to the formula ΔQ = (Sw 理想 -Sw 当前 )×η×1000, calculate the amount of water to be added ΔQ=(55%-45%)×0.8×1000=8, then 8m3 of water is added per mu of land. 3 The amount of water.

[0081] The soil nitrogen, phosphorus and potassium content data Npk = 1.2%, which is lower than the standard value of 1.5%, indicating that nitrogen, phosphorus and potassium fertilizers need to be supplemented. It is recommended to apply nitrogen, phosphorus and potassium compound fertilizers at a dosage of 20kg / mu.

[0082] The number of disease spots Nd=3<5, and the number of disease spots has not exceeded the standard for the time being.

[0083] Example 2: Linkage of multiple concurrent disease scenarios.

[0084] A tobacco field has a high incidence of disease (number of disease spots Nd = 8) due to continuous rainfall (soil moisture = 75%). At the same time, the soil nitrogen, phosphorus, and potassium content data Npk = 1.0%, which is low. The system updates the comprehensive quality index through image analysis.

[0085] The number of disease spots Nd = 8>5 is marked as a high risk of pests and diseases. Combined with the color abnormality of the leaf color saturation Sc, the pesticide spraying plan and light control strategy are matched. It is recommended to spray 50% carbendazim wettable powder (dilution ratio 1:800) and increase ventilation to reduce humidity. At the same time, since the soil nitrogen, phosphorus and potassium content data Npk = 1.0% is low, it is necessary to apply 10 kg / mu of nitrogen, phosphorus and potassium compound fertilizer to increase the nitrogen, phosphorus and potassium content.

[0086] In summary, the present invention integrates multi-dimensional data such as soil moisture, nitrogen, phosphorus and potassium content, light intensity, ambient temperature and tobacco growth images to construct a comprehensive quality index, and performs real-time comparison based on preset thresholds, which overcomes the limitations of existing technologies that rely on a single parameter. It can fully reflect the complex environmental effects of tobacco growth, such as the synergistic effects of light damage and soil moisture under high temperature and strong light stress, so that the prediction results are more in line with the actual growth state, thereby reducing the misjudgment rate and significantly improving the prediction accuracy and reliability. At the same time, the present invention combines a threshold adaptive mechanism to dynamically correct the preset threshold according to historical planting data. This mechanism enables the model to adapt to different regional climate conditions and planting cycle changes, avoiding the misjudgment problem caused by fixed thresholds. At the same time, the dynamically adjusted threshold is linked with real-time optimized planting parameters (such as irrigation amount and fertilization strategy), which can accurately guide users to adjust field management measures to achieve tobacco quality improvement and maximize resource utilization efficiency, thereby saving water and fertilizer costs.

[0087] The above shows and describes the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The above embodiments and descriptions are merely preferred examples of the present invention and are not intended to limit the present invention. Various changes and improvements may be made to the present invention without departing from the spirit and scope of the present invention. Such changes and improvements fall within the scope of the present invention. The scope of protection claimed in the present invention is defined by the appended claims and their equivalents.

Claims

1. A method for constructing a tobacco planting quality prediction model based on multi-source data fusion, characterized in that: The following steps are involved: S1, real-time collection of multi-source data in the tobacco planting area, wherein the multi-source data includes soil moisture data Sw, soil nitrogen, phosphorus and potassium content data Npk, light intensity data Li, ambient temperature data Te and tobacco growth image data Im; S2. Get the collected data and calculate it according to the formula The comprehensive quality index CI is calculated, where α, β, γ and is the weight coefficient and satisfies At the same time, the tobacco leaf growth image data Im is analyzed to extract the leaf color saturation Sc and the number of disease spots Nd. If Nd>0, the calculation formula of the comprehensive quality index CI is updated; S3. Preset a threshold value Th, and dynamically modify the threshold value Th based on historical data. Compare the calculated comprehensive quality index CI with the preset threshold value Th. If CI ≥ Th, the tobacco leaf quality is judged to be excellent. If CI < Th, it is determined that the planting parameters need to be adjusted. S4. Generate a planting parameter optimization plan based on the determination result and feed it back to the user terminal.

2. The method for constructing a tobacco planting quality prediction model based on multi-source data fusion according to claim 1, characterized in that: The calculation formula for updating the comprehensive quality index CI in step S2 is specifically: The tobacco leaf growth image data Im is analyzed by convolutional neural network to extract the leaf color saturation Sc and the number of disease spots Nd, where Nd>0; Add Sc and Nd as additional input parameters to the calculation formula of the comprehensive quality index CI, and update its calculation formula to: Among them, δ is the newly added weight coefficient, and satisfies 3. The method for constructing a tobacco planting quality prediction model based on multi-source data fusion according to claim 1, characterized in that: In step S3, the threshold value Th is dynamically modified according to historical data, specifically: Set a time window T and calculate the proportion Py of tobacco leaves judged as excellent in the time window T; If Py>70%, Th increases by 5%; If Py < 30%, Th decreases by 5%; If 30%≤Py≤70%, Th remains unchanged.

4. The method for constructing a tobacco planting quality prediction model based on multi-source data fusion according to claim 1, characterized in that: In step S4, a planting parameter optimization scheme is generated according to the determination result, specifically: Analyze the current comprehensive quality index CI and each sub-parameter value, including soil moisture data Sw, soil nitrogen, phosphorus and potassium content data Npk, light intensity data Li, ambient temperature data Te, leaf color saturation Sc, and number of disease spots Nd; If the soil moisture data Sw is lower than the lower limit of soil moisture, it is marked as insufficient irrigation. Then according to the formula ΔQ=(Sw 理想 -Sw 当前 )×η×1000, calculate the amount of water to be replenished ΔQ, where Sw 理想 is the soil moisture data under ideal conditions, Sw 当前 is the soil moisture data under the current state, η is the soil permeability coefficient; If the NPK content is lower than the standard NPK content, it is marked as needing additional NPK fertilizer, and the type and amount of compound fertilizer are recommended according to the NPK ratio; If Nd>5, that is, the number of disease spots exceeds the standard, it is marked as a high risk of pests and diseases. Then, combined with the color abnormality of the leaf color saturation Sc, the pesticide spraying plan and light control strategy are matched.

5. A tobacco planting quality prediction model construction system based on multi-source data fusion, characterized by: The system includes a multi-source data acquisition module, a data fusion module, a quality prediction module, a dynamic adjustment module and a user terminal; The multi-source data acquisition module includes a soil sensor, a meteorological sensor and an image acquisition unit, which is used to collect soil moisture data Sw, soil nitrogen, phosphorus and potassium content data Npk, light intensity data Li, ambient temperature data Te and tobacco growth image data Im in the tobacco planting area in real time; The data fusion module is used to calculate the comprehensive quality index CI according to the formula; The quality prediction module is used to compare the calculated comprehensive quality index with a preset threshold and output a corresponding determination result; The dynamic adjustment module is used to generate a planting parameter optimization plan based on the determination result and feed it back to the user terminal; The user terminal is used to receive the implantation parameter optimization plan and to display the CI real-time change curve and the threshold comparison result.

6. The method for constructing a tobacco planting quality prediction model based on multi-source data fusion according to claim 5, characterized in that: The image acquisition unit is used to analyze the tobacco leaf growth image data Im through a convolutional neural network, extract the leaf color saturation Sc and the number of disease spots Nd, and add the leaf color saturation Sc and the number of disease spots Nd as additional input parameters to the calculation formula of the comprehensive quality index CI, thereby updating the calculation formula of the comprehensive quality index CI.

7. The method for constructing a tobacco planting quality prediction model based on multi-source data fusion according to claim 5, characterized in that: The dynamic adjustment module is further configured to dynamically modify the threshold Th according to historical data.

8. The tobacco planting quality prediction model construction system based on multi-source data fusion according to claim 5 is characterized in that: The system also includes a cloud storage module for real-time backup of multi-source data and model parameters collected by the multi-source data acquisition module, and interacting with the user terminal through an encrypted communication protocol.

Images