Fruit tree growth regulation and control method based on multi-source data fusion, terminal and storage medium

Abstract

The invention belongs to the technical field of orchard management, and particularly relates to a fruit tree growth regulation and control method based on multi-source data fusion, a terminal and a storage medium, and the method comprises the steps: collecting the video and image data of a fruit tree canopy through a multispectral camera installed in an orchard, and judging the type of a fruit tree through a pre-constructed fruit tree recognition model; extracting growth characteristic parameters of the fruit tree from the collected video and image data; the method comprises the following steps: detecting weather forecast data at an orchard through a micro weather monitoring station, acquiring soil environment data of a fruit tree root layer through a soil temperature and humidity sensor, and acquiring soil chemical element data of soil in the orchard through a chemical element sensor; multi-dimensional data such as fruit tree canopy images, meteorology, soil environment and chemical elements are comprehensively collected, fruit tree growth state prediction results in several days in the future are predicted according to the multi-dimensional data, the moisture requirements of fruit trees in different growth stages and under different environmental conditions can be accurately reflected, and a scientific basis is provided for accurate irrigation.

Description

Technical Field

[0001] The present invention belongs to the technical field of orchard management, and in particular relates to a fruit tree growth control method based on multi-source data fusion, a terminal and a storage medium. Background Art

[0002] In traditional fruit cultivation, plantation management has long relied on empirical judgment and extensive practices. This model has gradually exposed numerous drawbacks in adapting to the complex and ever-changing natural environment and the growing needs of fruit trees. Empirical management is often limited by the knowledge and practical experience of individual growers. Different growers have significant differences in their understanding of the relationship between fruit tree growth stages, environmental factors, and growth status, leading to a lack of scientific and consistent management decisions. For example, when determining whether fruit trees need irrigation, traditional methods often rely on apparent soil dryness or subjective judgment by the grower. This makes it difficult to accurately grasp the actual water needs of the fruit tree's root zone. This can lead to problems such as over-irrigation, which wastes water resources and nutrient loss, or under-irrigation, which affects the normal physiological activities of the fruit tree. The same is true for fertilization. Fertilization based on empirical evidence cannot accurately match the nutrient needs of fruit trees at different growth stages. This not only results in fertilizer waste and increases production costs, but can also lead to physiological disorders in the fruit trees due to nutrient imbalances, such as nutrient deficiencies or elemental poisoning, which in turn affect fruit quality and yield. This extensive management approach fails to meet the requirements of modern agriculture for refined and efficient production, hindering the sustainable development of the fruit industry.

[0003] With the penetration of information technology in the agricultural field, some orchards have begun to try to introduce a single type of sensor to monitor the growth environment parameters of fruit trees, such as using only soil temperature and humidity sensors to monitor soil moisture and temperature, or relying on weather stations to obtain meteorological data. However, fruit tree growth is a complex dynamic process, affected by a combination of factors. A single data source cannot fully reflect the inherent relationship between the growth status of fruit trees and environmental changes. Although soil temperature and humidity data can reflect the soil moisture and temperature conditions, they cannot directly reflect the actual use of water by fruit trees and the impact of soil nutrients on growth; although meteorological data can provide information such as light, temperature, and precipitation, it is difficult to accurately assess the actual effects of these factors within the fruit tree canopy. Summary of the Invention

[0004] In view of the fact that traditional fruit tree planting in the existing technology has long relied on experience-based judgment and extensive operations, when dealing with the complex and changeable natural environment and the growth needs of fruit trees, experience-based management is often limited by the knowledge reserves and practical accumulation of individual growers, or irrigation and fertilization are adjusted based on the introduction of a single type of sensor to monitor the fruit tree growth environment parameters. However, a single data source cannot fully reflect the intrinsic relationship between the growth status of fruit trees and environmental changes, and is prone to defects such as excessive irrigation causing waste of water resources, loss of soil nutrients, or physiological disorders of fruit trees due to nutrient imbalance. The present invention provides a fruit tree growth regulation method, terminal and storage medium based on multi-source data fusion to solve the above technical problems.

[0005] In a first aspect, the present invention provides a method for regulating fruit tree growth based on multi-source data fusion, comprising the following steps: S1. Collect video and image data of the fruit tree canopy using a multispectral camera installed in the orchard, and use a pre-built fruit tree recognition model to determine the fruit tree type; and extract growth characteristic parameters of the fruit tree from the collected video and image data, including leaf area index, fruit morphology index, and branch density; S2. Meteorological forecast data at the orchard is detected by a micro-meteorological monitoring station. Soil environmental data of the root zone of the fruit trees is obtained by a soil temperature and humidity sensor. The soil environmental data includes soil moisture and soil temperature. Soil chemical element data of the soil in the orchard is obtained by a chemical element sensor, including soil nitrogen content, phosphorus content, and potassium content. S3, inputting the fruit tree type, growth characteristic parameters, weather forecast data, soil environment data, and soil chemical element data into a pre-built fruit tree growth prediction model to generate prediction results of the fruit tree growth status in the next few days; S4. Calculate the water demand index and fertilizer demand index based on the prediction results, and dynamically adjust the infiltration irrigation amount and fertilization plan.

[0006] A further improvement of this technical solution is to extract growth characteristic parameters of fruit trees from the collected video and image data, and the method specifically includes: S11. For the video data of the fruit tree canopy, a pre-stored background subtraction method is used to extract the dynamic change area between consecutive frames to obtain the dynamic growth information of the fruit tree canopy; S12. For the fruit tree canopy image data, use the pre-stored Mask R-CNN image segmentation algorithm to perform instance segmentation on the fruits in the image, and calculate the fruit morphology index based on the segmentation results; S13. Obtain the red light reflectance and the near-infrared reflectance of the canopy using a multispectral camera, and calculate the leaf area index based on the red light reflectance and the near-infrared reflectance; S14. Use a pre-stored skeleton extraction algorithm to obtain the skeleton structure of the branches from the fruit tree canopy image data, and calculate the branch density based on the pixel ratio of the branch skeleton length and the number of branches in the skeleton structure.

[0007] A further improvement of this technical solution is that the method for calculating the fruit morphology index based on the segmentation results includes: S121, performing a morphological closing operation on the segmented fruit region to fill the holes; S122, calculating the actual projected area of ​​the fruit according to the pixel scale; S123, according to the current date and the fruit tree type output by the fruit tree identification model, matching the target fruit area of ​​the current fruit tree type in the current growth stage from a reference table of target fruit areas of standard growth stages pre-stored in the database; S124, performing Canny edge detection on the segmented fruit binary image, and obtaining the symmetry coefficient of the fruit outline using randomized Hough transform; S125. Calculate the mean symmetry value of all fruits in the orchard based on the symmetry coefficient of the fruit outline; S126. Calculate a fruit morphology index based on the actual projected area of ​​the fruit, the target fruit area of ​​the current fruit tree type at the current growth stage, the symmetry coefficient of the fruit outline, and the mean symmetry of all fruits in the orchard.

[0008] A further improvement of this technical solution is that step S2 includes: S21. Preset multiple sets of soil temperature and humidity sensor arrays in the root distribution area of ​​the fruit trees. The soil temperature and humidity sensor arrays are arranged in layers to cover the root activity layer. Each set of soil temperature and humidity sensors includes a temperature probe and a humidity probe, and transmits data in real time to the edge computing node via the LoRa wireless communication module. S22. Deploy a soil chemical element sensor group based on ion-selective electrodes, which integrates three-channel detection modules for nitrogen, phosphorus, and potassium; S23. Perform dynamic compensation processing on the collected soil chemical element data.

[0009] A further improvement of this technical solution is that step S3 includes: S31, performing interpolation and synchronization processing on the sampling timestamps of the multispectral camera, micro-meteorological monitoring station and sensor based on the fruit tree type; S32, converting the fruit tree type into a one-hot encoding, and normalizing the growth characteristic parameters, weather forecast data, soil environment data, and soil chemical element data; S33, using the ResNet-50 network to extract features from the video and image data of the time-series synchronized fruit tree canopy layer, and outputting a spatial feature vector containing fruit tree type and growth characteristic parameters; S34. Use a two-layer bidirectional LSTM network to perform time series modeling on soil environmental data, weather forecast data, and soil chemical element data, capture the dynamic change patterns of the data, and output a time series feature vector. S35, perform feature fusion on the spatial feature vector and the temporal feature vector through tensor splicing and cross-modal attention to generate multimodal fusion features; S36. Input the multimodal fusion features into the fully connected layer of the pre-built fruit tree growth prediction model for regression prediction, and output the prediction results of the fruit tree growth status in the next few days.

[0010] A further improvement of this technical solution is that in step S4, the water demand index is calculated based on the prediction results. The formula is: ; in, For the future The optimal leaf area index for the day is calculated based on the future The fruit tree growth stage prediction results included in the fruit tree growth state prediction results are matched from the reference table of optimal leaf area of ​​standard growth stages pre-stored in the database; For the future The predicted leaf area index included in the prediction results of the growth status of the Tianguo tree; is the future soil moisture content, which is determined by weather forecast data and future The predicted soil moisture change trend included in the prediction results of the growth status of the Tianguo tree is calculated; The water holding capacity of the orchard soil pre-stored in the database, that is, the maximum amount of water that the soil can hold after gravity drainage; For the future The air temperature at which the photosynthetic efficiency of fruit trees reaches its highest level during the day is determined by historical meteorological data corresponding to the prediction results of the fruit tree type and growth stage; For the future of weather forecasting The average temperature of the day; 、 and is a weight coefficient used to balance the effects of leaf area index, soil moisture content and air humidity on water demand.

[0011] A further improvement of this technical solution is that in step S4, the fertilization demand index is calculated based on the prediction results. The formula is: ; in, For the future Ideal soil nutrient concentration in the future The fruit tree growth stage prediction results included in the fruit tree growth state prediction results are matched from the reference table of ideal soil nutrient concentrations at standard growth stages pre-stored in the database; For the future The prediction results of soil nutrient concentration included in the prediction results of the growth status of the Tianguo tree.

[0012] Further improvements to this technical solution include dynamically adjusting the infiltration irrigation amount and fertilization plan, the method of which includes: Determining the water demand index Is the slope greater than 0.1 / day? If yes, increase the infiltration rate. , infiltration volume The calculation formula is: ;in, is the soil permeability coefficient, in L / m²·day; is the water demand index at the current moment, is the projected area of ​​the fruit tree root system, estimated based on the type and age of the fruit tree, in m²; Determining the water demand index Is the peak value greater than 1.5? If yes, start preventive irrigation, the amount of preventive irrigation The calculation formula is: ; Determine the fertilizer demand index Is it greater than 1.5? If so, it means that the soil nutrient conditions cannot meet the growth needs of fruit trees, triggering the fertilization instruction and generating a corresponding ratio plan based on the nutrient deficiency situation.

[0013] In a second aspect, the present invention provides a terminal, comprising: processor, memory, wherein The memory is used to store computer programs, The processor is used to call and run the computer program from the memory, so that the terminal executes the above-mentioned terminal method.

[0014] In a third aspect, the present invention provides a computer storage medium, wherein the computer-readable storage medium stores instructions, which, when executed on a computer, enable the computer to execute the methods described in the above aspects.

[0015] The beneficial effects of the present invention are: This method integrates multiple sources of equipment, including multispectral cameras, micro-meteorological monitoring stations, soil temperature and humidity sensors, and chemical element sensors, to comprehensively collect multidimensional data, including fruit tree canopy images, meteorological data, soil environmental data, and chemical element data. This multi-source data acquisition method overcomes the limitations of traditional single-source monitoring and can reflect the inherent relationship between fruit tree growth status and environmental changes from different angles and levels, providing rich and comprehensive data support for subsequent precise analysis and scientific decision-making.

[0016] This method uses a ResNet-50 network to extract features from time-synchronized canopy video and image data, outputting a spatial feature vector containing fruit tree type and growth characteristic parameters. A two-layer, bidirectional LSTM network is then used to perform time-series modeling on soil environmental data, weather forecast data, and soil chemical element data, capturing the dynamic changes in the data and outputting a time-series feature vector. The spatial and time-series feature vectors are then fused using tensor concatenation and cross-modal attention to generate multimodal fusion features. This innovative feature extraction and fusion method fully exploits the inherent correlations between data from different modalities, improving feature representation and providing more comprehensive and accurate feature information for model prediction. The multimodal fusion features are then input into the fully connected layer of a pre-built fruit tree growth prediction model for regression prediction, outputting predictions of fruit tree growth status for several days in the future. By integrating multi-source data, employing a scientific feature extraction and fusion method, and employing advanced model building techniques, this method enables more accurate predictions of fruit tree growth status, providing a reliable basis for subsequent water and fertilizer regulation decisions, effectively overcoming the shortcomings of traditional prediction methods, such as low accuracy and poor reliability.

[0017] The formula for calculating the water demand index based on the prediction results of the present invention comprehensively considers multiple factors such as the future optimal leaf area index, the predicted leaf area index, the future soil moisture content, the orchard soil water holding capacity, the future air temperature, and the average temperature of the weather forecast, and balances the influence of each factor on the water demand through the weight coefficient. This calculation method can accurately reflect the water demand of fruit trees at different growth stages and under different environmental conditions, provides a scientific basis for precision irrigation, avoids the problems of over-irrigation or under-irrigation, and realizes the efficient use of water resources. The calculation formula of the fertilization demand index is based on the future ideal soil nutrient concentration and the predicted results of soil nutrient concentration, and can accurately evaluate the degree of match between soil nutrient status and the growth requirements of fruit trees.

[0018] In addition, the present invention has a reliable design principle, a simple structure and a very broad application prospect. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, for ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.

[0020] Figure 1 A schematic flow chart of a method according to an embodiment of the present invention.

[0021] Figure 2 A schematic diagram of the structure of a terminal provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0022] In order to make the purpose, features, and advantages of the present invention more obvious and easy to understand, the technical solutions of the present invention will be clearly and completely described below in conjunction with the drawings in the specific embodiments. Obviously, the embodiments described below are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments in this application, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of this application.

[0023] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those commonly understood by those skilled in the art of the present invention. The terms used in this specification of the present invention are only for the purpose of describing specific embodiments and are not intended to limit the present invention.

[0024] The fruit tree growth control method based on multi-source data fusion provided by the embodiment of the present invention is executed by a computer device. Figure 1 This is a schematic flow chart of a method according to an embodiment of the present invention. According to different requirements, the order of the steps in the flow chart can be changed, and some steps can be omitted.

[0025] like Figure 1 As shown, the method includes: S1. Collect video and image data of the fruit tree canopy using a multispectral camera installed in the orchard, and use a pre-built fruit tree recognition model to determine the fruit tree type; and extract growth characteristic parameters of the fruit tree from the collected video and image data, including leaf area index, fruit morphology index, and branch density; S2. Meteorological forecast data at the orchard is detected by a micro-meteorological monitoring station. Soil environmental data of the root zone of the fruit trees is obtained by a soil temperature and humidity sensor. The soil environmental data includes soil moisture and soil temperature. Soil chemical element data of the soil in the orchard is obtained by a chemical element sensor, including soil nitrogen content, phosphorus content, and potassium content. S3, inputting the fruit tree type, growth characteristic parameters, weather forecast data, soil environment data, and soil chemical element data into a pre-built fruit tree growth prediction model to generate prediction results of the fruit tree growth status in the next few days; S4. Calculate the water demand index and fertilizer demand index based on the prediction results, and dynamically adjust the infiltration irrigation amount and fertilization plan.

[0026] This invention utilizes multi-source equipment, including multispectral cameras, micro-meteorological monitoring stations, soil temperature and humidity sensors, and chemical element sensors, to comprehensively collect multi-dimensional data on the fruit tree canopy, weather, soil environment, and chemical elements. This overcomes the limitations of traditional single-source monitoring, provides a rich and accurate data foundation for subsequent analysis, and can more accurately reflect the relationship between fruit tree growth status and environmental changes. This multi-source data is input into a pre-built fruit tree growth prediction model, which comprehensively considers various factors and generates a forecast of the fruit tree's growth status for several days in the future. This multi-source data fusion and scientific model construction method significantly improves the accuracy and reliability of predictions, helping to understand fruit tree growth trends in advance and gain time for timely management measures. Based on the predicted results, the water demand index and fertilization demand index are calculated, and the infiltration irrigation rate and fertilization plan are dynamically adjusted. This method can accurately match the water and fertilizer needs of fruit trees at different growth stages, avoiding over- or under-use, improving water and fertilizer utilization, reducing production costs, and mitigating physiological disorders caused by nutrient imbalances, thereby ensuring fruit quality and yield.

[0027] To facilitate understanding of the present invention, the following is a further description of the fruit tree growth regulation method based on multi-source data fusion provided by the present invention based on the principle of the fruit tree growth regulation method based on multi-source data fusion of the present invention, combined with the process of regulating fruit tree growth based on multi-source data fusion in the embodiment.

[0028] Specifically, the method for extracting growth characteristic parameters of fruit trees from the collected video and image data includes: S11. For the video data of the fruit tree canopy, a pre-stored background subtraction method is used to extract the dynamic change area between consecutive frames to obtain the dynamic growth information of the fruit tree canopy; S12. For the fruit tree canopy image data, use the pre-stored Mask R-CNN image segmentation algorithm to perform instance segmentation on the fruits in the image, and calculate the fruit morphology index based on the segmentation results; S13. Obtain the red light reflectance and the near-infrared reflectance of the canopy using a multispectral camera, and calculate the leaf area index based on the red light reflectance and the near-infrared reflectance; S14. Use a pre-stored skeleton extraction algorithm to obtain the skeleton structure of the branches from the fruit tree canopy image data, and calculate the branch density based on the pixel ratio of the branch skeleton length and the number of branches in the skeleton structure.

[0029] Collect a large amount of canopy image data containing fruit at different growth stages and varieties to build a fruit image dataset. Annotate the images in the dataset, using annotation tools (such as LabelImg) to mark the location and category of each fruit and generate the corresponding annotation file (e.g., in XML format). During the annotation process, ensure the accuracy of the annotations to avoid missing or mislabeled items. Divide the annotated dataset into training, validation, and test sets, typically in a ratio of 7:1.5:1.5. The training set is used for model training, the validation set is used to adjust model hyperparameters and prevent overfitting, and the test set is used to evaluate the final model performance.

[0030] Select a deep learning framework (such as TensorFlow or PyTorch) to build the Mask R-CNN model. Mask R-CNN is an instance segmentation algorithm based on convolutional neural networks. It detects target objects and generates accurate segmentation masks for each object. Train the Mask R-CNN model using the training set, setting appropriate training parameters such as the learning rate, batch size, and number of training rounds. During training, the model weights are continuously adjusted using the backpropagation algorithm and the stochastic gradient descent optimizer to ensure accurate fruit recognition and segmentation. Evaluate the trained model using the validation set, adjusting model hyperparameters such as anchor box size and scale based on evaluation metrics such as average precision (AP) to improve performance. Repeated rounds of training and fine-tuning are performed until the model achieves satisfactory performance on the validation set. Input the canopy image to be processed into the trained Mask R-CNN model, which outputs instance segmentation results for each fruit, including the fruit's bounding box and segmentation mask.

[0031] Furthermore, the method for calculating the fruit morphology index based on the segmentation result includes: S121, performing a morphological closing operation on the segmented fruit region to fill the holes; S122, calculating the actual projected area of ​​the fruit according to a pixel scale (obtained by pre-measuring the number of pixels corresponding to an object of known length in the image); S123, according to the current date and the fruit tree type output by the fruit tree identification model, matching the target fruit area of ​​the current fruit tree type in the current growth stage from a reference table of target fruit areas of standard growth stages pre-stored in the database; S124, performing Canny edge detection on the segmented fruit binary image, and obtaining the symmetry coefficient of the fruit outline using randomized Hough transform; S125. Calculate the mean symmetry value of all fruits in the orchard based on the symmetry coefficient of the fruit outline; S126. Calculate a fruit morphology index based on the actual projected area of ​​the fruit, the target fruit area of ​​the current fruit tree type at the current growth stage, the symmetry coefficient of the fruit outline, and the mean symmetry of all fruits in the orchard.

[0032] As shown in Table 1, the reference table of target fruit area for standard growth stages is designed to provide standard reference values ​​for fruit area at different growth stages for different fruit tree types, allowing for quick matching of target fruit area based on the current date and tree type. This reference table, constructed based on extensive field measurements and data analysis, covers a variety of common fruit tree types and their key growth stages. Four common fruit tree types—apple, pear, peach, and grape—are used as examples. Each fruit tree type is further subdivided based on its varietal characteristics and growth patterns. (This example is simplified, presenting only broad categories of fruit trees; in practice, specific varietals can be subdivided.) Growth stage classification varies slightly across different fruit tree types, but can generally be divided into the following key stages: young fruit (when the fruit has just formed, is small, and undergoes rapid initial growth); expansion (when the fruit enters a period of rapid growth, significantly increasing in size and gradually filling out); veraison (when the fruit begins to change color, such as apples changing from green to red or yellow, and growth slows); and maturity (when the fruit reaches peak edible condition, meeting maturity standards for size, color, and taste).

[0033] Table 1: Reference table of fruit target area at standard growth stage (unit: )

[0034] The formula for calculating the fruit morphology index FMI based on the segmentation results is: ; in, is the actual projected area of ​​the i-th fruit; The target fruit area for the current fruit tree type at the current growth stage; is the symmetry coefficient of the fruit outline; The fruit morphology index (FMI) is calculated as the mean symmetry value of all fruits in the orchard; n is the number of fruits involved in the calculation of the fruit morphology index (FMI).

[0035] According to the red light reflectivity Reflectivity in the near-infrared band The formula for calculating leaf area index is: ; Among them, NDVI is the normalized difference vegetation index; SAVI is the soil adjusted vegetation index; is the canopy structure correction factor, for example, 0.65 for apple trees and 0.72 for citrus trees.

[0036] Calculate branch density based on the pixel ratio of branch skeleton length and number of branches in the skeleton structure The formula is: ; in, is the pixel ratio of the branch skeleton length, , is the total length of the branch skeleton (traversing the skeleton image and counting the sum of the Euclidean distances between all adjacent skeleton pixels), is the total number of pixels in the image; is the pixel ratio of the number of branches, , is the total number of branches; and separately is the weight coefficient of the pixel ratio of branch skeleton length and the pixel ratio of branch number.

[0037] This invention integrates video, image, and multispectral data to extract characteristic parameters of fruit tree growth from multiple dimensions. Background subtraction captures dynamic canopy growth information, enabling timely detection of changes in fruit tree growth. Mask R-CNN instances segment fruit and calculate morphological indices, accurately depicting features such as fruit size and symmetry. Multispectral data is used to calculate the leaf area index, reflecting leaf growth. A skeleton extraction algorithm calculates branch density, revealing branch growth trends and providing rich data support for a comprehensive understanding of fruit tree growth. By establishing a fruit image dataset and training a Mask R-CNN model, accurate fruit segmentation and recognition are achieved, laying the foundation for the calculation of the fruit morphological index. The morphological index comprehensively considers the actual projected area, target area, symmetry coefficient, and other factors, enabling scientific assessment of fruit growth status and quality. This helps to promptly detect abnormally growing fruit, guide precise fertilization, fruit thinning, and other agricultural operations, and improve fruit yield and quality.

[0038] In addition, step S2 includes: S21. Preset multiple sets of soil temperature and humidity sensor arrays in the root distribution area of ​​the fruit trees. The soil temperature and humidity sensor arrays are arranged in layers to cover the root activity layer. Each set of soil temperature and humidity sensors includes a temperature probe and a humidity probe, and transmits data in real time to the edge computing node via the LoRa wireless communication module. S22. Deploy a soil chemical element sensor group based on ion-selective electrodes, which integrates three-channel detection modules for nitrogen, phosphorus, and potassium; S23. Perform dynamic compensation processing on the collected soil chemical element data.

[0039] Soil temperature and humidity sensors are buried in the main root area of ​​fruit trees in a triangular layout (depth of 20-50cm). Take the weighted average of the three point measurements: ;in, is the main root zone measurement; 、 It is the side auxiliary measuring point value.

[0040] This invention deploys multiple soil temperature and humidity sensor arrays in a triangular configuration across the main root zone (20-50 cm deep). This arrangement accurately captures the temperature and humidity dynamics of the active root layer in all directions and at multiple levels. This arrangement avoids the limitations of single-location measurements and accurately reflects the spatial variations in soil temperature and humidity across the root distribution area. By calculating humidity values ​​using a weighted average, using measurements from the main root zone as a key reference and combining them with values ​​from auxiliary lateral measurement points, this not only highlights key information about the main root zone but also comprehensively considers the influence of the surrounding environment, making the humidity data more realistic. Real-time data transmission to edge computing nodes via a LoRa wireless communication module allows growers to instantly monitor soil temperature and humidity changes and adjust irrigation strategies promptly, preventing root rot caused by excessive water or the impact of insufficient water on fruit tree growth, thus achieving scientific and precise irrigation. A soil chemical element sensor array integrating three-channel detection modules for nitrogen, phosphorus, and potassium can simultaneously and accurately measure the levels of these three key nutrients in the soil. These elements are crucial for fruit tree physiological processes such as growth, flowering, and fruiting. By real-time monitoring of its content, we can timely grasp the soil nutrient status, provide a reliable basis for precise fertilization, and avoid resource waste and environmental pollution caused by blind fertilization.

[0041] Furthermore, step S3 includes: S31, performing interpolation and synchronization processing on the sampling timestamps of the multispectral camera, micro-meteorological monitoring station and sensor based on the fruit tree type; S32, converting the fruit tree type into a one-hot encoding, and normalizing the growth characteristic parameters, weather forecast data, soil environment data, and soil chemical element data; S33, using the ResNet-50 network to extract features from the video and image data of the time-series synchronized fruit tree canopy layer, and outputting a spatial feature vector containing fruit tree type and growth characteristic parameters; S34. Use a two-layer bidirectional LSTM network to perform time series modeling on soil environmental data, weather forecast data, and soil chemical element data, capture the dynamic change patterns of the data, and output a time series feature vector. S35, perform feature fusion on the spatial feature vector and the temporal feature vector through tensor splicing and cross-modal attention to generate multimodal fusion features; S36. Input the multimodal fusion features into the fully connected layer of the pre-built fruit tree growth prediction model for regression prediction, and output the prediction results of the fruit tree growth status in the next few days.

[0042] In addition, in step S4, the water demand index is calculated based on the prediction results. The formula is: ; in, For the future The optimal leaf area index for the day is calculated based on the future The fruit tree growth stage prediction results included in the fruit tree growth state prediction results are matched from the reference table of optimal leaf area of ​​standard growth stages pre-stored in the database; For the future The predicted leaf area index included in the prediction results of the growth status of the Tianguo tree; is the future soil moisture content, which is determined by weather forecast data and future The predicted soil moisture change trend included in the prediction results of the growth status of the Tianguo tree is calculated; The water holding capacity of the orchard soil pre-stored in the database, that is, the maximum amount of water that the soil can hold after gravity drainage; For the future The air temperature at which the photosynthetic efficiency of fruit trees reaches its highest level during the day is determined by historical meteorological data corresponding to the prediction results of the fruit tree type and growth stage; For the future of weather forecasting The average temperature of the day; 、 and is a weight coefficient used to balance the effects of leaf area index, soil moisture content and air humidity on water demand.

[0043] For three common fruit trees, apple trees, pear trees, and grape trees, we focused on the four key growth stages of young fruit, expansion, color change, and maturity. Based on actual agricultural observations, research data, and the physiological characteristics of fruit trees, we compiled a reference table for the optimal leaf area in standard growth stages, as shown in Table 2.

[0044] Table 2: Reference table for optimal leaf area during standard growth stages

[0045] Step S31 of the present invention performs interpolation and synchronization processing on multi-source data such as multispectral cameras, micro-meteorological monitoring stations, and sensors, ensuring the consistency of data collected by different devices in the time dimension, providing an accurate time reference for subsequent analysis, avoiding analysis errors caused by data time misalignment, and enabling the model to be trained and predicted based on a complete data set under a unified time framework, greatly improving the validity and reliability of the data. S32 converts the fruit tree type into a unique hot encoding and normalizes the various data, eliminating the differences in dimensions and value ranges between different data types, allowing all types of data to participate in model calculations under the same standard, avoiding the problem of imbalanced weights of certain features in the model due to different data scales, enhancing the stability and generalization ability of the model, and laying a solid data foundation for subsequent feature extraction and model training. S33 uses a ResNet-50 network to extract features from video and image data, effectively capturing the spatial characteristics of the fruit tree canopy. Combined with tree type and growth parameters, it generates a rich spatial feature vector. S34 uses a two-layer bidirectional LSTM network to model multiple time series data, accurately capturing the dynamic patterns of the data and outputting a time series feature vector. These two feature extraction methods comprehensively characterize fruit tree growth from both spatial and temporal dimensions, respectively, enabling the model to capture more detailed and accurate feature information. S35 fuses spatial and time series feature vectors through tensor concatenation and a cross-modal attention mechanism, leveraging the complementary strengths of different modal features. The resulting multimodal features more comprehensively reflect the complex state of fruit tree growth, improving the model's understanding and representation of growth status, and thus enhancing the accuracy and reliability of prediction results. S36 feeds the multimodal fusion features into a fully connected layer for regression prediction, outputting predictions of fruit tree growth status for several days in the future. This provides orchard managers with advanced insights into fruit tree growth trends and enables refined orchard management.

[0046] Furthermore, in step S4, the fertilization demand index is calculated based on the prediction results. The formula is: ; in, For the future Ideal soil nutrient concentration in the future The fruit tree growth stage prediction results included in the fruit tree growth status prediction results are matched from the reference table of ideal soil nutrient concentrations for standard growth stages stored in the database (Table 3), where y = 1, 2, 3, representing the three chemical elements nitrogen, phosphorus, and potassium involved in the calculation of the fertilization requirement index; For the future The prediction results of soil nutrient concentration included in the prediction results of the growth status of the Tianguo tree.

[0047] Table 3: Reference table of ideal soil nutrient concentrations during standard growth stages

[0048] Among them, the dynamic adjustment of infiltration irrigation volume and fertilization plan includes the following implementation methods: Determining the water demand index Is the slope greater than 0.1 / day? If yes, increase the infiltration rate. , infiltration volume The calculation formula is: ;in, is the soil permeability coefficient, in L / m²·day; is the water demand index at the current moment, is the projected area of ​​the fruit tree root system, estimated based on the type and age of the fruit tree, in m²; Determining the water demand index Is the peak value greater than 1.5? If yes, start preventive irrigation, the amount of preventive irrigation The calculation formula is: ; Determine the fertilizer demand index Is it greater than 1.5? If so, it means that the soil nutrient conditions cannot meet the growth needs of fruit trees, triggering the fertilization instruction and generating a corresponding ratio plan based on the nutrient deficiency situation.

[0049] Generate corresponding rationing schemes according to nutrient deficiency, including: Calculate nitrogen, phosphorus and potassium deficiency, taking nitrogen N as an example: ; in, is the nitrogen deficiency; is the ideal content of chemical elements in the soil; is the actual current content of chemical elements in the soil; the calculation method for the deficiency of phosphorus P and potassium K is the same; Generate fertilizer ratio according to nitrogen, phosphorus and potassium deficiency: ; According to soil volume Calculate the amount of fertilizer applied per time : ; in, For the soil The difference between the ideal content and the current actual content of a chemical element; is the soil density; For nutrient absorption efficiency.

[0050] The calculated single fertilizer application amount , select the appropriate fertilizer type and fertilization method, and evenly apply the fertilizer to the root distribution area of ​​the fruit tree to meet the nutrient needs of the fruit tree and promote its healthy growth. During the fertilization process, the fertilization time and frequency can be adjusted based on the fruit tree growth status prediction results in step S3. For example, the number of fertilizations can be appropriately increased during the fruit tree's peak growth period.

[0051] This invention calculates a fertilizer requirement index based on predicted fruit tree growth status and generates specific fertilizer ratios and dosages based on the deficiencies of three key chemical elements: nitrogen, phosphorus, and potassium. This allows for precise fertilization tailored to soil nutrient status and the growth needs of fruit trees. This approach avoids the problems of nutrient waste, soil pollution, and nutrient imbalance in fruit trees caused by blind fertilization in traditional fertilization methods, improving fertilizer utilization and reducing production costs. It also ensures that fruit trees receive an adequate and balanced nutrient supply, promoting healthy growth and high-quality, high-yield fruit trees. The water requirement index slope and peak value are used to dynamically adjust the infiltration irrigation rate, achieving precise irrigation. A high slope indicates a rapidly increasing water requirement, and timely increases in the infiltration irrigation rate can meet the tree's water needs. When the peak value of the water requirement index is too high, preventive infiltration irrigation is initiated to prevent the tree from experiencing water stress. This dynamic regulation method rationally utilizes water resources, improves water resource utilization efficiency, prevents the adverse effects of over- or under-irrigation on fruit tree growth, and ensures normal physiological function.

[0052] Intelligent decision support: Adjust the time and frequency of fertilization based on the predicted results of fruit tree growth status. For example, appropriately increase the number of fertilizations during the peak growth period of fruit trees, making fertilization and irrigation management more scientific and reasonable.

[0053] Figure 2 This is a structural diagram of a terminal 200 provided in an embodiment of the present invention. The terminal 200 can be used to execute the fruit tree growth control method based on multi-source data fusion provided in an embodiment of the present invention.

[0054] The terminal 200 may include a processor 210, a memory 220, and a communication module 230. These components communicate via one or more buses. Those skilled in the art will appreciate that the server structure shown in the figure does not limit the present invention. The server structure may be a bus structure or a star structure, and may include more or fewer components than shown, or may combine certain components or arrange the components differently.

[0055] The memory 220 can be used to store execution instructions of the processor 210. The memory 220 can be implemented by any type of volatile or non-volatile storage terminal, or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic memory, flash memory, magnetic disk, or optical disk. When the execution instructions in the memory 220 are executed by the processor 210, the terminal 200 can perform some or all of the steps in the above-described method embodiments.

[0056] The processor 210 is the control center of the storage terminal. It uses various interfaces and lines to connect various parts of the entire electronic terminal. It executes various functions of the electronic terminal and / or processes data by running or executing software programs and / or modules stored in the memory 220, and calling data stored in the memory. The processor can be composed of an integrated circuit (IC), for example, it can be composed of a single packaged IC, or it can be composed of multiple packaged ICs with the same or different functions. For example, the processor 210 can only include a central processing unit (CPU). In the embodiment of the present invention, the CPU can be a single computing core or multiple computing cores.

[0057] The communication module 230 is used to establish a communication channel so that the storage terminal can communicate with other terminals, receive user data sent by other terminals, or send user data to other terminals.

[0058] The present invention also provides a computer storage medium, wherein the computer storage medium may store a program that, when executed, may include some or all of the steps of each embodiment provided herein. The storage medium may be a magnetic disk, an optical disk, a read-only memory (ROM), or a random access memory (RAM).

[0059] Those skilled in the art will clearly understand that the techniques in the embodiments of the present invention can be implemented using software plus a necessary general-purpose hardware platform. Based on this understanding, the technical solutions in the embodiments of the present invention, or the portion that contributes to the prior art, can be embodied in the form of a software product. This computer software product is stored in a storage medium such as a USB flash drive, a mobile hard drive, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk, among other media capable of storing program code, and includes instructions for causing a computer terminal (which can be a personal computer, a server, or a second terminal, a network terminal, etc.) to execute all or part of the steps of the methods described in various embodiments of the present invention.

[0060] Although the present invention has been described in detail with reference to the accompanying drawings and in conjunction with preferred embodiments, the present invention is not limited thereto. Without departing from the spirit and essence of the present invention, persons of ordinary skill in the art may make various equivalent modifications or substitutions to the embodiments of the present invention, and such modifications or substitutions shall be within the scope of the present invention. Any changes or substitutions that can be easily conceived by persons skilled in the art within the technical scope disclosed in the present invention shall be within the scope of protection of the present invention.

Claims

1. A fruit tree growth control method based on multi-source data fusion, characterized in that: The following steps are involved: S1. Collect video and image data of the fruit tree canopy using a multispectral camera installed in the orchard, and use a pre-built fruit tree recognition model to determine the fruit tree type; and extract growth characteristic parameters of the fruit tree from the collected video and image data, including leaf area index, fruit morphology index, and branch density; S2. Meteorological forecast data at the orchard is detected by a micro-meteorological monitoring station. Soil environmental data of the root zone of the fruit trees is obtained by a soil temperature and humidity sensor. The soil environmental data includes soil moisture and soil temperature. Soil chemical element data of the soil in the orchard is obtained by a chemical element sensor, including soil nitrogen content, phosphorus content, and potassium content. S3, inputting the fruit tree type, growth characteristic parameters, weather forecast data, soil environment data, and soil chemical element data into a pre-built fruit tree growth prediction model to generate prediction results of the fruit tree growth status in the next few days; S4. Calculate the water demand index and fertilizer demand index based on the prediction results, and dynamically adjust the infiltration irrigation amount and fertilization plan.

2. The fruit tree growth control method based on multi-source data fusion according to claim 1, characterized in that: Extracting growth characteristic parameters of fruit trees from collected video and image data, the method specifically includes: S11. For the video data of the fruit tree canopy, a pre-stored background subtraction method is used to extract the dynamic change area between consecutive frames to obtain the dynamic growth information of the fruit tree canopy; S12. For the fruit tree canopy image data, use the pre-stored Mask R-CNN image segmentation algorithm to perform instance segmentation on the fruits in the image, and calculate the fruit morphology index based on the segmentation results; S13. Obtain the red light reflectance and the near-infrared reflectance of the canopy using a multispectral camera, and calculate the leaf area index based on the red light reflectance and the near-infrared reflectance; S14. Use a pre-stored skeleton extraction algorithm to obtain the skeleton structure of the branches from the fruit tree canopy image data, and calculate the branch density based on the pixel ratio of the branch skeleton length and the number of branches in the skeleton structure.

3. The fruit tree growth control method based on multi-source data fusion according to claim 2, characterized in that: Methods for calculating fruit morphological index based on segmentation results include: S121, performing a morphological closing operation on the segmented fruit region to fill the holes; S122, calculating the actual projected area of ​​the fruit according to the pixel scale; S123, according to the current date and the fruit tree type output by the fruit tree identification model, matching the target fruit area of ​​the current fruit tree type in the current growth stage from a reference table of target fruit areas of standard growth stages pre-stored in the database; S124, performing Canny edge detection on the segmented fruit binary image, and obtaining the symmetry coefficient of the fruit outline using randomized Hough transform; S125. Calculate the mean symmetry value of all fruits in the orchard based on the symmetry coefficient of the fruit outline; S126. Calculate a fruit morphology index based on the actual projected area of ​​the fruit, the target fruit area of ​​the current fruit tree type at the current growth stage, the symmetry coefficient of the fruit outline, and the mean symmetry of all fruits in the orchard.

4. The fruit tree growth control method based on multi-source data fusion according to claim 1, characterized in that: Step S2 includes: S21. Preset multiple sets of soil temperature and humidity sensor arrays in the root distribution area of ​​the fruit trees. The soil temperature and humidity sensor arrays are arranged in layers to cover the root activity layer. Each set of soil temperature and humidity sensors includes a temperature probe and a humidity probe, and transmits data in real time to the edge computing node via the LoRa wireless communication module. S22. Deploy a soil chemical element sensor group based on ion-selective electrodes, which integrates three-channel detection modules for nitrogen, phosphorus, and potassium; S23. Perform dynamic compensation processing on the collected soil chemical element data.

5. The fruit tree growth control method based on multi-source data fusion according to claim 1, characterized in that: Step S3 includes: S31, performing interpolation and synchronization processing on the sampling timestamps of the multispectral camera, micro-meteorological monitoring station and sensor based on the fruit tree type; S32, converting the fruit tree type into a one-hot encoding, and normalizing the growth characteristic parameters, weather forecast data, soil environment data, and soil chemical element data; S33, using the ResNet-50 network to extract features from the video and image data of the time-series synchronized fruit tree canopy layer, and outputting a spatial feature vector containing fruit tree type and growth characteristic parameters; S34. Use a two-layer bidirectional LSTM network to perform time series modeling on soil environmental data, weather forecast data, and soil chemical element data, capture the dynamic change patterns of the data, and output a time series feature vector. S35, perform feature fusion on the spatial feature vector and the temporal feature vector through tensor splicing and cross-modal attention to generate multimodal fusion features; S36. Input the multimodal fusion features into the fully connected layer of the pre-built fruit tree growth prediction model for regression prediction, and output the prediction results of the fruit tree growth status in the next few days.

6. The fruit tree growth control method based on multi-source data fusion according to claim 5, characterized in that: In step S4, the water demand index is calculated based on the prediction results. The formula is: ； in, For the future The optimal leaf area index for the day is calculated based on the future The fruit tree growth stage prediction results included in the fruit tree growth state prediction results are matched from the reference table of optimal leaf area of ​​standard growth stages pre-stored in the database; For the future The predicted leaf area index included in the prediction results of the growth status of the Tianguo tree; is the future soil moisture content, which is determined by weather forecast data and future The predicted soil moisture change trend included in the prediction results of the growth status of the Tianguo tree is calculated; The water holding capacity of the orchard soil pre-stored in the database, that is, the maximum amount of water that the soil can hold after gravity drainage; For the future The air temperature at which the photosynthetic efficiency of fruit trees reaches its highest level during the day is determined by historical meteorological data corresponding to the prediction results of the fruit tree type and growth stage; For the future of weather forecasting The average temperature of the day; 、 and is a weight coefficient used to balance the effects of leaf area index, soil moisture content and air humidity on water demand.

7. The fruit tree growth control method based on multi-source data fusion according to claim 6, characterized in that: In step S4, the fertilizer demand index is calculated based on the prediction results. The formula is: ； in, For the future Ideal soil nutrient concentration in the future The fruit tree growth stage prediction results included in the fruit tree growth state prediction results are matched from the reference table of ideal soil nutrient concentrations at standard growth stages pre-stored in the database; For the future The prediction results of soil nutrient concentration included in the prediction results of the growth status of the Tianguo tree.

8. The fruit tree growth control method based on multi-source data fusion according to claim 7, characterized in that: Dynamically adjust the infiltration irrigation volume and fertilization plan, including: Determining the water demand index Is the slope greater than 0.1 / day? If yes, increase the infiltration rate. , infiltration volume The calculation formula is: ;in, is the soil permeability coefficient, in L / m²·day; is the water demand index at the current moment, is the projected area of ​​the fruit tree root system, estimated based on the type and age of the fruit tree, in m²; Determining the water demand index Is the peak value greater than 1.5? If yes, start preventive irrigation, the amount of preventive irrigation The calculation formula is: ; Determine the fertilizer demand index Is it greater than 1.5? If so, it means that the soil nutrient conditions cannot meet the growth needs of fruit trees, triggering the fertilization instruction and generating a corresponding ratio plan based on the nutrient deficiency situation.

9. A terminal, characterized in that: include: processor; a memory for storing execution instructions of the processor; The processor is configured to execute the method according to any one of claims 1 to 8.

10. A computer-readable storage medium storing a computer program, characterized in that: When the program is executed by a processor, the method according to any one of claims 1 to 8 is implemented.

Images