An apple yield prediction method and system based on anvil-slip interaction coefficient

Abstract

This invention discloses a method and system for predicting apple yield based on rootstock-scion interaction coefficients, belonging to the field of fruit tree yield prediction technology. The method includes the following steps: Step S4: Based on a Gaussian-exponential coupling model, apple growth is divided into a rapid growth period and a slow ripening and expansion period. Root hydraulic conductivity, branch density, and rootstock-scion interaction coefficients and measured values ​​are substituted into the model to calculate a first predicted apple yield value. Step S6: Based on a gradient-enhancing tree model, the rootstock-scion interaction coefficients and measured values ​​are substituted into the model to calculate a second predicted apple yield value. This invention uses a Gaussian-exponential coupling model and a gradient-enhancing tree model, and uses root hydraulic conductivity, branch density, and fertilizer / water values ​​as the basis for predicting apple yield, respectively, to obtain two predicted values. These two predicted values ​​are then compared with the average apple yield of previous years, and the predicted value that best matches the predicted apple yield is selected, thus improving the accuracy of apple yield prediction.

Description

Technical Field

[0001] This invention relates to the field of fruit tree yield prediction technology, and in particular to a method and system for predicting apple yield based on rootstock-scion interaction coefficient. Background Technology

[0002] Fruit yield forecasting is a complex and important field that integrates agricultural science, data analysis, and modern technology. It is crucial for orchard management, market planning, supply chain optimization, and agricultural policy formulation. Accurate yield forecasts can help fruit growers develop sales strategies, negotiate prices, and arrange warehousing and logistics in advance. Processing plants and retailers can also plan their production and market launches in advance based on the forecasted yields.

[0003] In apple cultivation, common yield forecasting methods rely on growers' years of observation experience, making rough estimates based on tree vigor and flowering conditions, or using drones or ground equipment to collect high-resolution images of the trees during the flowering or fruiting season. However, most existing apple trees are cultivated using rootstock-scion grafting. Since the rootstock-scion interaction coefficient is an important indicator of the tree's growth status, a positive rootstock-scion interaction coefficient will improve tree growth and yield, while a negative rootstock-scion interaction coefficient will affect tree growth and even reduce yield. Common yield forecasting methods struggle to monitor the impact of the rootstock-scion interaction coefficient, easily overlooking its influence on yield and reducing the accuracy of yield forecasting. Summary of the Invention

[0004] The purpose of this invention is to provide a method and system for predicting apple yield based on rootstock-scion interaction coefficient, so as to solve the problems mentioned in the background art.

[0005] To achieve the above objectives, the present invention provides the following technical solution: a method for predicting apple yield based on rootstock-scion interaction coefficient, comprising the following steps: Step S1: Establish a database based on the rootstock-scion interaction coefficients of different rootstocks, and store the rootstock-scion interaction coefficients, environmental data, and yield data of fruit trees from previous years into the database; Step S2: Extract the root hydraulic conductivity, branch density, and fertilizer-water interaction coefficient of the corresponding rootstock from the database according to the rootstock type. At the same time, match the spatial data of the measured fruit tree plots based on GIS technology to ensure that they are consistent in location, shape, and boundary. Step S3: Use UAV multispectral technology to monitor the planting area, branches and canopy area of ​​fruit trees, use sensors to collect data on soil moisture, pH value and nutrients, and use the pressure chamber method to determine root hydraulic conductivity. Step S4: Based on the Gaussian-exponential coupling model, apple growth is divided into a rapid growth period and a slow ripening and expansion period. The root hydraulic conductivity, branch density, and fertilizer-water interaction coefficient between rootstock and scion, along with the measured values ​​from step S3, are substituted into the model to calculate the first predicted value of apple yield. Step S5: Use sensors to monitor sunlight, rainfall, and diurnal temperature range during the fruit-bearing season, and monitor the pests and diseases affecting the fruit trees. Step S6: Based on the gradient boosting tree model, substitute the rootstock-scion interaction coefficient, the measured values ​​in Step S3 and Step S5 into the model to establish a yield prediction model and calculate the second predicted value of apple yield. Step S7: Calculate the average value of apple production data from previous years in the database, compare the average value with the first and second predicted values, calculate the confidence interval of the predicted values, and calculate the NPV and IRR values. The model is updated monthly.

[0006] Preferably, the calculation of the rootstock-scion interaction coefficient in step S1 includes the following steps: Step S11: Select the rootstock and scion types, as well as the traits to be measured, in the rootstock-scion interaction coefficient; Step S12: Measure the phenotypic data of the selected grafting combination and rootstock selected in step S44 using sensors and measuring instruments to obtain the phenotypic measurement values ​​of the grafting combination and the rootstock. Step S13: By calculating the percentage values ​​of the grafted combination trait measurement value and the rootstock trait measurement value in step S12, the rootstock-scion interaction coefficient of this trait is obtained.

[0007] Preferably, the GIS-based land parcel alignment in step S2 includes the following steps: Step S21: Select UAV orthophotos, high-resolution satellite imagery, and RTK field measurement data as reference data; Step S22: First, manually select a set of corresponding control points on the data layer to be adjusted and the base data layer. Then, use the eraser transformation algorithm to non-uniformly distort and deform the entire layer so that the control points are aligned. Step S23: Based on computer vision, the boundary line is regarded as an image feature, and the feature matching algorithm is used to register the data layer to be adjusted and the reference data.

[0008] Preferably, the trait data collection in step S3 includes the following steps: Step S31: Collect image information on the planting area, branches, and canopy of fruit trees using UAV remote sensing technology, and calculate the ratio of branch pixels to canopy pixels using multispectral images to obtain the branch density; Step S32: Collect data on the moisture and nutrient content of the soil used for planting fruit trees using sensors, calculate the ratio of moisture to nutrient content, and obtain fertilizer and water values. Step S33: The root water conductivity of the rootstock is determined using the pressure chamber method and a flow velocity sensor.

[0009] Preferably, the Gaussian-exponential coupling model in step S4 includes the following steps: Step S41: Apples produce a pulse effect in the early stage of growth. This effect rapidly reaches its peak and then decays over time or space. The rapid growth period of apples is described by a Gaussian function. Step S42: After the initial pulse, the apple growth enters a long-term, slowly decaying tail effect, which is described by an exponential function for the apple ripening and expansion period. Step S43: The advantages of the rapid growth period and the ripening and expansion period of apples are combined by using a Gaussian-exponential coupling model to describe the dynamics of the entire apple growth process and realize the monitoring of the apple production process.

[0010] Preferably, the parameter acquisition in step S5 includes the following steps: Step S51: By setting up sensors in the fruit tree planting area, calculate the light duration and light intensity during the fruit tree fruiting season, detect the rainfall during the apple growth process, and monitor the difference between day and night growth temperatures. Step S52: Detect the number of fruit trees affected by pests and diseases using monitoring equipment, and calculate the percentage of fruit trees affected by pests and diseases.

[0011] Preferably, the establishment of the production prediction model in step S6 includes the following steps: Step S61: Preprocess the data collected in steps S3 and S5 and extract the feature values ​​of the data; Step S62: Select XGBoost as the gradient boosting tree model, and substitute the previous years' rootstock-scion interaction coefficients, environmental data and yield data from the database in Step S1 into the model for training, so that the model can learn the impact of rootstock-scion interaction on yield under different environments. Step S63: By inputting the feature quantities extracted in step S61 into the model in step S62, the predicted yield value and its confidence interval are calculated.

[0012] Preferably, the comparison of production data in step S7 includes the following steps: Step S71: Obtain the average annual apple production by extracting the historical production data from the database in Step S1 and performing a weighted average. Step S72: Compare the confidence intervals of the first predicted value and the second predicted value with the average value in step S71 to determine the accuracy of the first predicted value and the second predicted value. Step S73: When the average value is within the common interval of the confidence intervals of the first predicted value and the second predicted value, take the average of the first predicted value and the second predicted value; otherwise, take the predicted value whose confidence interval is closest to the average value.

[0013] Preferably, the NPV and IRR numerical calculations in step S7 include: NPV calculation is performed by subtracting the initial investment cost from the future cash flow. The future cash flow is the product of the expected net cash inflow over the period and the discount rate. The net cash inflow is the product of apple production and net revenue per unit price. When NPV is greater than 0, it means that the project is feasible and profitable. When NPV is equal to 0, it means that the project is feasible but has no excess revenue. When NPV is less than 0, it means that the project is not feasible. IRR calculation is based on the discount rate at which the net present value is 0, which yields the internal rate of return. When the IRR is greater than the investment cost, the project is considered feasible. When the IRR equals the investment cost, the project is considered to be at break-even. When the IRR is less than the investment cost, the project is considered infeasible.

[0014] An apple yield prediction system based on rootstock-scion interaction coefficient includes: A data storage device is used to store the rootstock-scion interaction coefficients of different rootstocks, as well as the rootstock-scion interaction coefficients of fruit trees in previous years, environmental data, and yield data. The processor is used to process data in the database and collected data, and to calculate the predicted apple yield, confidence interval, average of previous years' yield data, NPV and IRR values ​​by running a Gaussian-exponential coupling model and a gradient boosting tree model. The measuring instrument group consists of sensors, drones, and mobile terminals, and is used to collect environmental parameters and image information of apple growth. A visualization platform is used to display the collected parameters, apple yield forecasts, confidence intervals, average values ​​of historical yield data, NPV, and IRR values ​​in the form of charts. The management platform is used for human-computer interaction, is user-facing, and performs in-depth data analysis and system management.

[0015] The technical effects and advantages of this invention are as follows: This invention uses a Gaussian-exponential coupling model and a gradient boosting tree model to predict apple yield based on root hydraulic conductivity, branch density, and fertilizer and water levels during apple growth. Two predicted values ​​are obtained, and then the two predicted values ​​are compared with the average apple yield of previous years. The predicted value that best matches the apple yield is selected. At the same time, the influence of external factors during apple growth, including environmental parameters and pest and disease conditions, is taken into account to improve the accuracy of apple yield prediction. Attached Figure Description

[0016] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used together with the embodiments of the invention to explain the invention, but do not constitute a limitation thereof. In the drawings: Figure 1 This is a schematic diagram of the production forecasting method of the present invention; Figure 2 This is a schematic diagram illustrating the calculation process of the rootstock-scion interaction coefficient according to the present invention; Figure 3 This is a schematic diagram of the phenotypic data acquisition process of the present invention; Figure 4 This is a schematic diagram of the operation process of the Gaussian-exponential coupling model of the present invention; Figure 5 This is a schematic diagram illustrating the process of establishing the production prediction model of the present invention; Figure 6 This is a schematic diagram of the comparison process for production data in this invention; Figure 7 This is a schematic diagram of the production prediction system framework of the present invention. Detailed Implementation

[0017] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0018] This invention provides, for example Figures 1-6 The apple yield prediction method based on rootstock-scion interaction coefficient shown includes the following steps: Step S1: Establish a database based on the rootstock-scion interaction coefficients of different rootstocks, and store the rootstock-scion interaction coefficients, environmental data, and yield data of fruit trees from previous years into the database; Step S2: Extract the root hydraulic conductivity, branch density, and fertilizer-water interaction coefficients of the corresponding rootstock from the database according to the rootstock type. At the same time, based on GIS plot-level alignment, accurately match the spatial data of plots from different sources, with different precision, or at different times to ensure that they are consistent in location, shape, and boundary, thus ensuring the uniformity of fruit tree planting area parameters. Step S3: Monitor branches and canopy area using UAV multispectral technology, collect data on soil moisture, pH value and nutrient content using sensors, calculate the fertilizer-water ratio in the soil, and measure root hydraulic conductivity using the pressure chamber method to reflect the ability of fruit tree roots to absorb and transport water. Step S4: Based on the Gaussian-exponential coupling model, apple growth is divided into a rapid growth period and a slow ripening and expansion period. The root hydraulic conductivity, branch density, and fertilizer-water interaction coefficient between rootstock and scion, along with the measured values ​​from step S3, are substituted into the model to calculate the first predicted value of apple yield. Step S5: Use sensors to monitor the sunlight, rainfall and diurnal temperature range during the fruit-bearing season of the fruit trees, and monitor the pests and diseases affecting the fruit trees, and calculate the proportion of fruit trees affected by pests and diseases. Step S6: Based on the gradient boosting tree model, substitute the rootstock-scion interaction coefficient, environmental data and yield data of previous years in the database into the model, train and learn the model, establish a yield prediction model, and calculate the second predicted value of apple yield by substituting the root water conductance, branch density and fertilizer-water rootstock-scion interaction coefficient, as well as the measured values ​​in steps S3 and S5 into the model. Step S7: Calculate the average value of apple production data from previous years in the database, compare the average value with the first and second predicted values, calculate the confidence interval of the predicted values, output the predicted yield per acre for the current year and the 95% confidence interval, with an error ≤8%, and simultaneously calculate NPV and IRR for financial credit. The model is updated monthly, and Kalman filtering is used to reduce observation noise, so that the prediction error gradually converges to ≤5%, realizing real-time updates of apple production prediction.

[0019] The calculation of the rootstock-scion interaction coefficient in step S1 includes the following steps: Step S11: Select the rootstock and scion types, as well as the traits to be measured, in the rootstock-scion interaction coefficient; Step S12: Measure the phenotypic data of the selected grafting combination and rootstock selected in step S44 using sensors and measuring instruments to obtain the phenotypic measurement values ​​of the grafting combination and the rootstock. Step S13: By calculating the percentage values ​​of the grafted combination trait measurement value and the rootstock trait measurement value in step S12, the rootstock-scion interaction coefficient of this trait is obtained.

[0020] After the rootstock and scion are combined, they do not grow independently, but form an interactive symbiotic complex. The rootstock-scion interaction coefficient is a numerical indicator used to quantify the degree and effect of mutual influence between the rootstock and scion. The database contains the rootstock-scion interaction coefficients of different traits of different rootstock-scion combinations. By measuring specific traits of grafting combination traits and rootstock traits, the percentage values ​​of grafting combination trait measurements and rootstock trait measurements reflect the rootstock-scion interaction coefficient of specific traits. By quantifying the rootstock-scion interaction coefficient and combining it with environmental data and tree physiological data, high-precision and forward-looking predictions of apple yield can be achieved.

[0021] Step S2, the parcel alignment using GIS technology, includes the following steps: Step S21: Select UAV orthophotos, high-resolution satellite imagery, and RTK field measurement data as reference data; Step S22: First, manually select a set of corresponding control points on the data layer to be adjusted and the base data layer. The selected control points should be located on clear and unchanging feature points. Then, use the eraser transformation algorithm to non-uniformly distort and deform the entire layer so that the control points are aligned. Step S23: Based on computer vision, the boundary line is regarded as an image feature, and the feature matching algorithm is used to register the data layer to be adjusted and the reference data.

[0022] Based on GIS technology, the rubber sheet transformation algorithm is first used to align the data layer to be adjusted and the reference data layer. At the same time, computer vision is used to register the data layer to be adjusted and the reference data. Combining the above two methods of plot alignment, a mutual reference is formed to ensure the accuracy of plot alignment. The rubber sheet transformation algorithm simulates the stretching of a rubber sheet to non-uniformly twist and deform the entire layer, so that the control points are aligned as much as possible, while maintaining the relative topological relationship between the elements within the layer.

[0023] Step S3, trait data collection, includes the following steps: Step S31: Collect image information on fruit tree planting area, branches and canopy using UAV remote sensing technology, calculate the ratio of branch pixels to canopy pixels using multispectral images to obtain branch density, which reflects the density of the canopy or the internal structure of the plant community. Step S32: Collect data on the humidity and nutrient content of the soil for fruit tree planting using a humidity sensor and a photoelectric sensor with colorimetric method, respectively, calculate the ratio of humidity to nutrient content, obtain fertilizer and water values, and accurately reflect the nutritional status, water requirements and growth environment of the fruit trees. Step S33: Using the pressure chamber method, the pressure is measured at 0.5 MPa. The root water conductivity of the rootstock is measured using a flow velocity sensor with an accuracy of ±1%.

[0024] The core principle of the pressure chamber method for determining root hydraulic conductivity is to pressurize water or nutrients through a section of detached rootstock root system under controlled pressure, while using a flow velocity sensor to accurately measure the flow velocity. Then, the hydraulic conductivity is calculated according to Darcy's law. This method can precisely control the pressure and simulate the root system's hydraulic conductivity under different water potential gradients from low to high.

[0025] By correlating root hydraulic conductivity, branch density, and the rootstock-scion interaction coefficient of fertilizer and water with apple yield during the apple growth process, this study reflects the absorption status of water and nutrients by fruit trees during apple growth based on root hydraulic conductivity, pruning index, and fertilizer and water splitting. This provides a data basis for apple yield prediction and improves the accuracy of yield prediction.

[0026] The Gaussian-exponential coupling model in step S4 involves the following steps: Step S41: Apples produce a pulse effect in the early stage of growth. This effect rapidly reaches its peak and then decays over time or space. The rapid growth period of apples is described by a Gaussian function. Step S42: After the initial pulse, the apple growth enters a long-term, slowly decaying tail effect, which is described by an exponential function for the apple ripening and expansion period. Step S43: The advantages of the rapid growth period and the ripening and expansion period of apples are combined by using a Gaussian-exponential coupling model to describe the dynamics of the entire apple growth process and realize the monitoring of the apple production process.

[0027] The Gaussian-exponential coupling model, by combining the local concentration properties of the Gaussian function and the long-term decay properties of the exponential function, flexibly describes the common two-stage or two-mechanism behavior in complex systems, and can be expressed as: ;; in Represented as peak amplitude, Indicated as peak position, Represented as the width of the distribution, This represents the root hydraulic conductivity and the rootstock-scion interaction coefficient. Indicates the rootstock-scion interaction coefficient based on branch density. R represents the interaction coefficient of fertilizer, water, rootstock, and scion; P represents the root hydraulic conductivity of Osmanthus fragrans; F represents the branch density of Osmanthus fragrans; and F represents the fertilizer and water splitting number of Osmanthus fragrans.

[0028] Based on the growth trend of apples, apple growth is divided into a rapid growth period and a ripening and expansion period. During the rapid growth period, apples grow rapidly to a peak and then decline, which follows a Gaussian function. During the ripening and expansion period, apple growth slows down, which follows an exponential function. The Gaussian-exponential coupling model combines the advantages of both to more accurately describe the dynamics of the entire apple growth process. By combining the rapid growth period and the ripening and expansion period of apples, the Gaussian-exponential coupling model accurately reflects the growth trend of apples. Furthermore, based on the root water conductance, pruning index, and fertilizer and water status of apple growth stages, it predicts the state of apple growth, reflects the relationship between apple growth volume and weight and yield, and improves the accuracy of apple yield prediction.

[0029] The parameter acquisition in step S5 includes the following steps: Step S51: By setting up sensors in the fruit tree planting area, the parameters of the apple growth environment are monitored. Quantum sensors are used to calculate the light time and light intensity during the fruit tree fruiting season. Weighing rain sensors are used to detect the rainfall during the apple growth process. Temperature sensors are used to monitor the difference between day and night growth temperatures. Step S52: Detect the number of fruit trees affected by pests and diseases using monitoring equipment, calculate the proportion of fruit trees affected by pests and diseases, and reflect the impact of pests and diseases on apple yield.

[0030] Step S6, establishing the production prediction model, includes the following steps: Step S61: Preprocess the data collected in steps S3 and S5 and extract the feature values ​​of the data; Step S62: Select XGBoost as the gradient boosting tree model, and input the rootstock-scion interaction coefficient, environmental data and yield data of previous years in the database of Step S1 into the model for model training. This allows the model to learn the influence of rootstock-scion interaction on yield under different environments. The rootstock-scion interaction coefficients of root hydraulic conductivity, branch density and fertilizer and water are used as the first sequence features to reflect the influence of root hydraulic conductivity, branch density and fertilizer and water on apple yield. Step S63: By inputting the feature quantities extracted in step S61 into the model in step S62, the predicted yield value and its confidence interval are calculated.

[0031] XGBoost is an open-source software library that provides parallel tree boosting functionality. It's an advanced machine learning library for regression, classification, and ranking problems. It constructs an optimal model by minimizing a loss function. The objective function includes a regularization term for the model's prediction error and model complexity. The gradient boosting tree model is trained based on historical rootstock-scion interaction coefficients, environmental data, and yield data from a database. By processing historical data, a model for predicting apple yield is obtained. Root hydraulic conductivity, branch density, rootstock-scion interaction coefficients for fertilizer and water, as well as real-time collected apple growth environmental parameters and the proportion of diseased and pest-infested trees are input into the model to predict apple yield, reflecting the impact of environmental parameters and disease / pest on apple yield and improving the accuracy of yield prediction.

[0032] The comparison of production data in step S7 includes the following steps: Step S71: Obtain the average annual apple production by extracting the historical production data from the database in Step S1 and performing a weighted average. Step S72: Compare the confidence intervals of the first predicted value and the second predicted value with the average value in step S71 to determine the accuracy of the first predicted value and the second predicted value. Step S73: When the average value is within the common interval of the confidence intervals of the first predicted value and the second predicted value, take the average of the first predicted value and the second predicted value; otherwise, take the predicted value whose confidence interval is closest to the average value.

[0033] By using a Gaussian-exponential coupling model and a gradient boosting tree model, and based on root hydraulic conductivity, branch density, and fertilizer and water values ​​during the apple growth process, apple yield was predicted. Two predicted values ​​were obtained, and then compared with the average apple yield of previous years. The predicted value that best matches the apple yield was selected. At the same time, the influence of external factors during the apple growth process, including environmental parameters and pest and disease conditions, was taken into account to improve the accuracy of apple yield prediction.

[0034] Step S7 involves calculating NPV and IRR values. The net present value (NPV) is obtained by subtracting the initial investment cost from the future cash flow. The future cash flow is the product of the expected net cash inflow over the specified number of years and the discount rate. The net cash inflow is the product of apple production and net revenue per unit price. When NPV is greater than 0, the project is feasible and profitable; when NPV equals 0, the project is feasible but without excess returns; and when NPV is less than 0, the project is not feasible. The internal rate of return (IRR) is obtained by calculating the discount rate when the NPV is 0. When IRR is greater than the investment cost, the project is feasible; when IRR equals the investment cost, the project is at break-even; and when IRR is less than the investment cost, the project is not feasible. By calculating the NPV and IRR values ​​for apple production, NPV is theoretically a superior indicator, directly measuring the absolute value of value creation, and its reinvestment assumptions are more conservative and realistic. IRR is used for rapid screening first, followed by NPV for the final decision, to predict the profitability of apple cultivation and provide a data foundation for investment in apple cultivation.

[0035] An apple yield prediction system based on rootstock-scion interaction coefficient, such as Figure 7As shown, the system includes a data storage device, a processor, a set of measuring instruments, a visualization platform, and a management platform. The data storage device stores the rootstock-scion interaction coefficients for different rootstocks, as well as historical rootstock-scion interaction coefficients, environmental data, and yield data, providing a data foundation for apple yield prediction. The processor processes data from the database and collected data, and calculates the predicted apple yield, confidence interval, average historical yield data, NPV, and IRR by running a Gaussian-exponential coupling model and a gradient boosting tree model. The processor is configured with the Gaussian-exponential coupling model and gradient boosting tree model, which, combined with data from the database, predict apple yield. The measuring instrument set consists of sensors, drones, and mobile terminals, used to collect environmental parameters and image information related to apple growth. Sensors include temperature sensors, humidity sensors, flow rate sensors, weighing-type rain sensors, photoelectric sensors, and quantum sensors. The mobile terminal allows farmers to manually input phenological information and agricultural records. Sensors and drones are used to collect data on root hydraulic conductivity, branch density, fertilizer and water data, and environmental parameters during apple growth, providing a data foundation for the prediction system and supporting apple yield forecasting. The platform ensures the accuracy of yield forecasts. The visualization platform displays collected parameters, apple yield forecasts, confidence intervals, average yield data from previous years, NPV, and IRR values ​​in chart form, facilitating real-time data visualization within the forecasting system. By placing data within a context on the display screen, visualization helps decision-makers comprehensively assess the situation, compare different options, and make forecasts based on data trends, leading to more reliable and faster decisions. Charts highlight the forecast results for complete visualization. The management platform facilitates human-computer interaction, providing users with in-depth data analysis and system management. It transforms complex and abstract backend data, logic, and functions into intuitive, easy-to-operate, and effective user interfaces and tools. Complex backend commands and scripts are encapsulated into simple buttons, forms, and menus. Multi-step, dependent, complex tasks are designed as clear, wizard-driven processes, guiding users step-by-step. A centralized interface is provided for managing and adjusting system, software, or service settings, rules, and strategies. Users can control and manage the apple yield system through the management platform.

[0036] Principle of this invention: Based on apple growth trends, apple growth is divided into a rapid growth phase and a ripening and expansion phase. During the rapid growth phase, apples grow rapidly to their peak and then decline. During the ripening and expansion phase, apple growth slows down. A Gaussian-exponential coupling model combines the rapid growth and ripening / expansion phases to accurately reflect the apple growth trend. Furthermore, based on root hydraulic conductivity, pruning index, and fertilizer and water status at each growth stage, the model predicts the apple's growth status, reflecting the relationship between apple volume and weight and yield. Simultaneously, by processing historical data, a model for predicting apple yield is obtained. Finally, the model incorporates root hydraulic conductivity, branch density, rootstock-scion interaction coefficients, and real-time data collection... The model inputs apple growth environment parameters and the proportion of diseased and pest-infested fruit trees into the model to predict apple yield, reflecting the impact of environmental parameters and disease / pests on apple yield. Two predicted values ​​are obtained, and then the two predicted values ​​are compared with the average apple yield of previous years. The predicted value that best matches the apple yield is selected. At the same time, the influence of external factors during apple growth, including environmental parameters and disease / pest conditions, is taken into account to improve the accuracy of apple yield prediction. Based on the predicted apple yield value, NPV and IRR values ​​are calculated. IRR is first used for quick screening, and then NPV is used for final decision-making to predict the profitability of apple planting and provide a data foundation for planting investment.

[0037] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A method for predicting apple yield based on rootstock-scion interaction coefficient, characterized in that, Includes the following steps: Step S1: Establish a database based on the rootstock-scion interaction coefficients of different rootstocks, and store the rootstock-scion interaction coefficients, environmental data, and yield data of fruit trees from previous years into the database; Step S2: Extract root hydraulic conductivity, branch density, and fertilization-water interaction coefficient of rootstock from the database according to rootstock type, and match the spatial data of the measured fruit tree plots based on GIS technology. Step S3: Monitor the planting area, branches, and canopy area of ​​fruit trees using UAV multispectral technology, collect soil data using sensors, and determine root hydraulic conductivity using the pressure chamber method; Step S4: Based on the Gaussian-exponential coupling model, apple growth is divided into a rapid growth period and a slow ripening and expansion period. The rootstock-scion interaction coefficient and the measured values ​​in step S3 are substituted into the model to calculate the first predicted value of apple yield. Step S5: Use sensors to monitor sunlight, rainfall, and diurnal temperature range during the fruit-bearing season, and monitor the pests and diseases affecting the fruit trees. Step S6: Based on the gradient boosting tree model, substitute the rootstock-scion interaction coefficient, the measured values ​​in Step S3 and Step S5 into the model to establish a yield prediction model and calculate the second predicted value of apple yield. Step S7: Calculate the average value of apple production data from previous years in the database, compare the average value with the first and second predicted values, calculate the confidence interval of the predicted values, and calculate the NPV and IRR values. The model is updated monthly.

2. The apple yield prediction method based on rootstock-scion interaction coefficient according to claim 1, characterized in that, The calculation of the rootstock-scion interaction coefficient in step S1 includes the following steps: Step S11: Select the rootstock and scion types, as well as the traits to be measured, in the rootstock-scion interaction coefficient; Step S12: Measure the phenotypic data of the selected grafting combination and rootstock selected in step S44 using sensors and measuring instruments to obtain the phenotypic measurement values ​​of the grafting combination and the rootstock. Step S13: By calculating the percentage values ​​of the grafted combination trait measurement value and the rootstock trait measurement value in step S12, the rootstock-scion interaction coefficient of this trait is obtained.

3. The apple yield prediction method based on rootstock-scion interaction coefficient according to claim 2, characterized in that, The land parcel alignment using GIS technology in step S2 includes the following steps: Step S21: Select UAV orthophotos, high-resolution satellite imagery, and RTK field measurement data as reference data; Step S22: First, manually select a set of corresponding control points on the data layer to be adjusted and the base data layer. Then, use the eraser transformation algorithm to non-uniformly distort and deform the entire layer so that the control points are aligned. Step S23: Based on computer vision, the boundary line is regarded as an image feature, and the feature matching algorithm is used to register the data layer to be adjusted and the reference data.

4. The apple yield prediction method based on rootstock-scion interaction coefficient according to claim 3, characterized in that, The trait data collection in step S3 includes the following steps: Step S31: Collect image information on the planting area, branches, and canopy of fruit trees using UAV remote sensing technology, and calculate the ratio of branch pixels to canopy pixels using multispectral images to obtain the branch density; Step S32: Collect data on the moisture and nutrient content of the soil used for planting fruit trees using sensors, calculate the ratio of moisture to nutrient content, and obtain fertilizer and water values. Step S33: The root water conductivity of the rootstock is determined using the pressure chamber method and a flow velocity sensor.

5. The apple yield prediction method based on rootstock-scion interaction coefficient according to claim 4, characterized in that, The Gaussian-exponential coupling model in step S4 includes the following steps: Step S41: Apples produce a pulse effect in the early stage of growth. This effect rapidly reaches its peak and then decays over time or space. The rapid growth period of apples is described by a Gaussian function. Step S42: After the initial pulse, the apple growth enters a long-term, slowly decaying tail effect, which is described by an exponential function for the apple ripening and expansion period. Step S43: The advantages of the rapid growth period and the ripening and expansion period of apples are combined by using a Gaussian-exponential coupling model to describe the dynamics of the entire apple growth process and realize the monitoring of the apple production process.

6. The apple yield prediction method based on rootstock-scion interaction coefficient according to claim 5, characterized in that, The parameter acquisition in step S5 includes the following steps: Step S51: By setting up sensors in the fruit tree planting area, calculate the light duration and light intensity during the fruit tree fruiting season, detect the rainfall during the apple growth process, and monitor the difference between day and night growth temperatures. Step S52: Detect the number of fruit trees affected by pests and diseases using monitoring equipment, and calculate the percentage of fruit trees affected by pests and diseases.

7. The apple yield prediction method based on rootstock-scion interaction coefficient according to claim 6, characterized in that, The establishment of the production prediction model in step S6 includes the following steps: Step S61: Preprocess the data collected in steps S3 and S5 and extract the feature values ​​of the data; Step S62: Select XGBoost as the gradient boosting tree model, and substitute the previous years' rootstock-scion interaction coefficients, environmental data and yield data from the database in Step S1 into the model for training, so that the model can learn the impact of rootstock-scion interaction on yield under different environments. Step S63: By inputting the feature quantities extracted in step S61 into the model in step S62, the predicted yield value and its confidence interval are calculated.

8. The apple yield prediction method based on rootstock-scion interaction coefficient according to claim 7, characterized in that, The comparison of production data in step S7 includes the following steps: Step S71: Obtain the average annual apple production by extracting the historical production data from the database in Step S1 and performing a weighted average. Step S72: Compare the confidence intervals of the first predicted value and the second predicted value with the average value in step S71 to determine the accuracy of the first predicted value and the second predicted value. Step S73: When the average value is within the common interval of the confidence intervals of the first predicted value and the second predicted value, take the average of the first predicted value and the second predicted value; otherwise, take the predicted value whose confidence interval is closest to the average value.

9. The apple yield prediction method based on rootstock-scion interaction coefficient according to claim 8, characterized in that, The calculation of NPV and IRR values ​​in step S7 includes: NPV calculation is performed by subtracting the initial investment cost from the future cash flow. The future cash flow is the product of the expected net cash inflow over the period and the discount rate. The net cash inflow is the product of apple production and net revenue per unit price. When NPV is greater than 0, it means that the project is feasible and profitable. When NPV is equal to 0, it means that the project is feasible but has no excess revenue. When NPV is less than 0, it means that the project is not feasible. IRR calculation is based on the discount rate at which the net present value is 0, which yields the internal rate of return. When the IRR is greater than the investment cost, the project is considered feasible. When the IRR equals the investment cost, the project is considered to be at break-even. When the IRR is less than the investment cost, the project is considered infeasible.

10. An apple yield prediction system based on rootstock-scion interaction coefficient, characterized in that, The apple yield prediction method based on rootstock-scion interaction coefficient as described in any one of claims 1-9 includes: A data storage device is used to store the rootstock-scion interaction coefficients of different rootstocks, as well as the rootstock-scion interaction coefficients of fruit trees in previous years, environmental data, and yield data. The processor is used to process data in the database and collected data, and to calculate the predicted apple yield, confidence interval, average of previous years' yield data, NPV and IRR values ​​by running a Gaussian-exponential coupling model and a gradient boosting tree model. The measuring instrument group consists of sensors, drones, and mobile terminals, and is used to collect environmental parameters and image information of apple growth. A visualization platform is used to display the collected parameters, apple yield forecasts, confidence intervals, average values ​​of historical yield data, NPV, and IRR values ​​in the form of charts. The management platform is used for human-computer interaction, is user-facing, and performs in-depth data analysis and system management.

Images