Fruit planting water and fertilizer management and control system and method based on deep learning

Abstract

The invention relates to the technical field of intelligent agriculture monitoring management, and discloses a fruit tree planting water and fertilizer management and control system and method based on deep learning, and the system comprises a data collection module, a data preprocessing module, a water and fertilizer state monitoring module, a control module, and a water and fertilizer device. Wherein the data acquisition module acquires original data such as weather, soil and fruit tree canopy images in real time or periodically; the data preprocessing module standardizes, corrects and fuses the original data through a preset algorithm, and extracts indexes representing weather, soil conditions and fruit tree growth vigor; the water and fertilizer state monitoring module constructs a soil-fruit tree-climate matching relation model according to a preset algorithm based on the indexes, and outputs an intelligent irrigation and fertilization scheme; the control module controls the water and fertilizer device to achieve precise irrigation and fertilization operation. Through cooperation of the modules, intelligent and precise water and fertilizer management is realized by means of deep learning, and the planting efficiency is improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of intelligent agricultural monitoring and management, in particular to a fruit tree planting water and fertilizer management and control system and method based on deep learning. BACKGROUND

[0002] In the process of fruit tree planting, water and fertilizer management is the core link affecting yield, quality and resource utilization efficiency. Traditional water and fertilizer management highly depends on the personal experience of farmers, which has strong subjectivity and blindness, and is easy to lead to over-fertilization and irrigation or insufficient fertilization, which not only causes a large amount of waste of water resources and fertilizers, and the rise of agricultural production cost, but also may cause soil degradation, environmental problems such as non-point source pollution.

[0003] Some existing intelligent agricultural management systems attempt to integrate sensor data and remote sensing images to build yield prediction, pest warning or fertilization recommendation models. However, most of these systems still have the following limitations: first, the data fusion level is shallow, and the dynamic and quantitative matching relationship model between soil, fruit trees and climate cannot be built, and the decision basis is not sufficient; second, the model is mostly targeted at a single task such as only predicting yield or only diagnosing nutrient deficiency, and lacks a comprehensive intelligent decision support platform integrating irrigation, fertilization, growth monitoring and pest warning; finally, the system lacks closed-loop control capability and cannot form an automatic control and execution closed loop with field execution equipment such as water and fertilizer devices, and still relies on manual intervention, making it difficult to achieve truly intelligent control.

[0004] Therefore, it is necessary to provide a fruit planting water and fertilizer management and control system and method based on deep learning to deeply integrate multi-source data, accurately build the matching relationship between field fertilizer and water key indicators and the soil-fruit tree-climate system, and have an integrated intelligent decision and precise execution capability of intelligent management and control system to overcome the shortcomings of the prior art. SUMMARY

[0005] Therefore, the present application provides a fruit planting water and fertilizer management and control system and method based on deep learning, which aims to solve the problems of not deeply integrating multi-source data, not building the matching relationship between fruit planting fertilizer and water key indicators and the soil-fruit tree-climate system, and no execution device in the prior art.

[0006] The application provides a fruit planting water and fertilizer management and control system based on deep learning, which comprises a data acquisition module, a data preprocessing module, a water and fertilizer state monitoring module, a control module and a water and fertilizer device, wherein the data acquisition module comprises a meteorological sensor network arranged in a field, a soil multi-parameter sensor network and a drone loaded with a hyperspectral camera, is used for acquiring original data in real time or periodically, and sends the original data to the data preprocessing module; the original data comprises field meteorological data, soil environmental data and fruit tree canopy image data; the data preprocessing module receives the original data, and sequentially performs standardization, correction and fusion processing on the acquired original data based on a multi-sensor fusion algorithm, and extracts indexes for representing the growth of fruit trees, soil conditions and meteorological characteristics; the water and fertilizer state monitoring module fuses the indexes for representing the growth of fruit trees, soil conditions and meteorological characteristics based on a machine learning algorithm, constructs a soil-plant-meteorological matching relationship model, and outputs an intelligent irrigation scheme and an intelligent fertilization scheme; and the control module is used for controlling the water and fertilizer device to perform precise irrigation and fertilization operation according to a decision prescription output by the intelligent decision model. The water and fertilizer device comprises a water and fertilizer barrel, a plurality of spray pipes are circumferentially arranged on the side of the water and fertilizer barrel, and an adjusting assembly is arranged between the water and fertilizer barrel and the spray pipes; the adjusting assembly adjusts the displacement and opening and closing of the spray pipes according to the control module, so as to realize precise irrigation and fertilization.

[0007] Further, the data preprocessing module sequentially performs standardization, correction and fusion processing on the original data by using an extended Kalman filtering algorithm, and the specific steps are as follows: the meteorological data and the soil data are dimensionally unified, the parameters such as air temperature, air humidity and light intensity in the meteorological data are converted to the interval [0, 1], the parameters such as soil temperature and soil humidity in the soil data are converted to the interval [0, 1], and the fruit tree canopy image data is normalized by pixel value to the range [0, 255]; the fusion result of the standardized multi-source data is taken as a system state quantity, a state equation is established to describe the dynamic change of the data with time, and a state transition matrix is obtained based on the time sequence characteristics of historical data; the measured values of various sensors are taken as observation values, an observation equation is established, and an observation matrix is configured according to the accuracy of the sensors; filter parameters are initialized, including the initial value of the state, the covariance matrix and the process noise and observation noise variance; the state quantity and the covariance matrix at the current time are predicted based on the state equation, the Kalman gain is calculated combined with the observation equation, the predicted value is corrected, and the corrected fusion data is obtained; the indexes for representing the meteorological data, the soil conditions and the growth of fruit trees are extracted from the corrected fusion data, including the humidity index, the light accumulation index calculated based on the meteorological data, the soil moisture index, the nutrient comprehensive index calculated based on the soil data, and the normalized vegetation index, the green waveband normalized vegetation index and the like extracted based on the fruit tree canopy image data.

[0008] Further, the data preprocessing module processes the fruit tree canopy image data, including radiation correction, geometric correction and atmospheric correction, and performs image stitching to generate a complete inter-orchard orthoimage; the stitched image is processed using a semantic segmentation network based on deep learning to separate the fruit tree canopy from the soil background; A plurality of vegetation indices are extracted from the processed fruit tree canopy image, including normalized difference vegetation index, green normalized difference vegetation index, excess green index, visible atmosphere resistance index and modified soil adjusted vegetation index; feature selection is performed on the extracted vegetation indices, and a high correlation vegetation index combination related to the fruit tree nutrient deficiency state is selected based on the Pearson correlation coefficient method to construct a feature subset.

[0009] Further, the water and fertilizer state monitoring module includes a host computer state monitoring and control software and a control unit, wherein: the host computer state monitoring and control software receives and processes the indices and spectral features output by the data preprocessing module, and generates the intelligent irrigation scheme and the intelligent fertilization scheme; the control unit coordinates the operation of each module according to the host computer state monitoring and control software, to ensure the continuity of data transmission and scheme execution.

[0010] Further, the water and fertilizer state monitoring module further includes a knowledge base, which stores the nitrogen, phosphorus, potassium and trace element demand thresholds of different fruit tree varieties at each growth stage, and can be dynamically updated according to field measured data and historical planting records.

[0011] Further, the water and fertilizer state monitoring module predicts future periods of meteorological data based on an LSTM time series prediction model, and combines real-time soil moisture data to output an intelligent irrigation scheme including predicted irrigation frequency, single irrigation water depth and total irrigation water volume; the intelligent fertilization scheme is based on the vegetation index combination after feature selection, establishes a relationship model between the growth of fruit trees and vegetation indices, and further constructs a nutrient deficiency amount model represented by growth difference, and finally fuses the relationship model between the growth of fruit trees and vegetation indices and the nutrient deficiency amount model, and combines the standard nutrient demand in the knowledge base to generate an intelligent fertilization scheme.

[0012] Further, the specific process of the LSTM time series prediction model for future period prediction of meteorological data includes: data input: the preprocessed meteorological data is divided according to time sequence, and meteorological data segments of a continuous preset time length are taken as input samples, and the samples contain time sequence characteristics of air temperature, air humidity, light intensity and wind speed; model training: a sliding window method is used to construct a training set, the window step is set to 1 hour, and through correlation analysis of historical meteorological data and measured soil moisture in the corresponding period, the number of hidden layer neurons, the dropout coefficient and the iteration number of the LSTM model are optimized, so that the prediction error of the model for meteorological parameters in the next 72 hours is controlled within a preset threshold; prediction output: output the prediction values of hourly air temperature, air humidity, light intensity and wind speed in the next 72 hours, and calculate the cumulative reference crop evapotranspiration in the period.

[0013] Further, the water and fertilizer device further comprises a driving assembly and an executing assembly, the driving assembly receives instructions of the control unit and drives the executing assembly, and the executing assembly realizes directional and quantitative irrigation and fertilization of fruit trees at different positions by adjusting the spatial angle, telescopic length and on-off state of the spray pipe.

[0014] Further, the spray pipe of the water and fertilizer device is provided with a flow sensor and a pressure sensor for feeding back the irrigation and fertilization amount and pressure parameters to the control unit in real time, forming a closed loop control to ensure the execution accuracy.

[0015] A fruit planting water and fertilizer management and control method based on deep learning, comprising the following steps: acquiring original data in real time or periodically, the original data including meteorological data, soil data and fruit tree canopy image data; performing standardization, correction and fusion processing on the original data in sequence based on a preset algorithm, and extracting indexes for representing meteorological data, soil conditions and fruit tree growth from the original data; according to a preset algorithm, constructing a soil-fruit-weather matching relationship model based on the indexes of meteorological data, soil conditions and fruit tree growth, and outputting an intelligent irrigation and fertilization scheme for predicting irrigation demand; and controlling a water and fertilizer device to perform precise irrigation and fertilization operation, wherein the water and fertilizer device comprises a water and fertilizer barrel, a plurality of spray pipes are installed on the side of the water and fertilizer barrel in a circumferential direction, and an adjusting assembly is further arranged between the water and fertilizer barrel and the spray pipes, the adjusting assembly adjusts the displacement and opening and closing of the spray pipes according to control instructions to realize precise irrigation and fertilization.

[0016] Compared with the prior art, the fruit planting water and fertilizer management and control system and method based on deep learning provided by the present application have the following beneficial effects: 1. By fusing meteorological, soil multi-parameter and hyperspectral image multi-dimensional data, and constructing a soil-fruit-weather matching relationship model based on deep learning, the limitations of traditional methods of shallow data utilization and single decision basis are overcome, and the scientificity and accuracy of irrigation and fertilization decision are significantly improved; 2. The system integrates growth monitoring, nutrient deficiency diagnosis, irrigation decision-making, fertilization recommendation, and execution control, achieving intelligent control from perception to decision-making to execution, avoiding the problems of relying on manual experience, multiple system fragmentation, and non-closed loop execution in traditional methods; 3. The system generates executable irrigation and fertilization schemes through intelligent algorithms and realizes precise and quantitative application with the help of controllable water and fertilizer devices and sensor feedback, effectively avoiding water and fertilizer excess or deficiency, saving resources, reducing production costs, and reducing the risk of non-point source pollution and soil degradation. BRIEF DESCRIPTION OF DRAWINGS

[0017] Various other advantages and benefits will become apparent to those of ordinary skill in the art upon reading the following detailed description of the preferred embodiments. The accompanying drawings are included to provide a description of the preferred embodiments and are not meant to limit the present application. Moreover, the same reference numerals in the attached drawings indicate the same or similar components. In the drawings: Figure 1 A system block diagram of a deep learning-based fruit planting water and fertilizer management system according to an embodiment of the present application; Figure 2 A flowchart of a deep learning-based fruit planting water and fertilizer management method according to an embodiment of the present application; Figure 3 A water and fertilizer device structure diagram of a deep learning-based fruit planting water and fertilizer management system and method according to an embodiment of the present application.

[0018] In the figure: 01 - fertilizer storage tank, 02 - fertilizer outlet, 03 - rotating assembly, 04 - lifting assembly, 05 - water and fertilizer outlet pipe, 06 - water inlet pipe, 07 - water and fertilizer tank, 08 - fertilizer outlet pipe, 09 - maintenance opening. DETAILED DESCRIPTION

[0019] Exemplary embodiments of the present disclosure will be described in greater detail below with reference to the accompanying drawings. Although exemplary embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure can be implemented in various forms and should not be limited by the embodiments described herein. Rather, these embodiments are provided to enable a more thorough understanding of the present disclosure and to fully convey the scope of the present disclosure to those skilled in the art. It should be noted that the embodiments in the present application and the features in the embodiments can be combined with each other without conflict. The present application will be described in detail below with reference to the accompanying drawings and in conjunction with the embodiments.

[0020] The application discloses a fruit planting water and fertilizer management and control system based on deep learning, which comprises a data acquisition module, a data preprocessing module, a water and fertilizer state monitoring module, a control module and a water and fertilizer device, wherein the data acquisition module is used for acquiring original data in real time or periodically, and sending the original data to the data preprocessing module, wherein the original data comprises meteorological data, soil data and fruit tree canopy image data; the data preprocessing module receives the original data and sequentially carries out standardization, correction and fusion processing on the acquired original data based on a preset algorithm, and extracts indexes for representing meteorological data, soil conditions and fruit tree growth from the original data; the water and fertilizer state monitoring module constructs a soil-fruit tree-climate matching relationship model based on the indexes of the meteorological data, soil conditions and fruit tree growth according to a preset algorithm, and outputs an intelligent irrigation and fertilization scheme for predicting irrigation demand; the control module receives and controls the water and fertilizer device to carry out precise irrigation and fertilization operation according to the intelligent irrigation and fertilization scheme; and the water and fertilizer device comprises a water and fertilizer barrel, a plurality of spray pipes are circumferentially installed on the top of the water and fertilizer barrel, and an adjusting assembly is further arranged between the water and fertilizer barrel and the spray pipes, the adjusting assembly adjusts the displacement and opening and closing of the spray pipes according to the control module, so as to realize precise irrigation and fertilization.

[0021] The working process of the system is as shown in the figure: Figure 2 The system acquires original data in real time or periodically through a meteorological sensor network, a soil multi-parameter sensor network and a high-spectrum camera-carrying unmanned aerial vehicle which is regularly patrolled in the field, wherein the original data comprises air temperature, humidity, illumination, wind speed, soil temperature and humidity, pH value, nitrogen, phosphorus and potassium content and fruit tree canopy multispectral image.

[0022] The collected original data is sent to the data preprocessing module for deep processing. In this stage, the meteorological and soil data are subjected to standardization processing, while the unmanned aerial vehicle image needs to be subjected to radiation correction, geometric correction and atmospheric correction, and then spliced to generate a complete field orthographic image. Subsequently, the system automatically identifies and separates the fruit tree canopy and the soil background by using a deep learning-based semantic segmentation network, accurately extracts a plurality of vegetation indexes therefrom, and further screens out a vegetation index combination most related to the nutrition state of the fruit tree by using a Pearson correlation coefficient method, to form a feature subset for decision-making.

[0023] Next, the water and fertilizer state monitoring module receives various indexes after preprocessing, firstly uses an LSTM time sequence prediction model to predict future meteorological conditions, combines real-time soil humidity data to generate a predictive intelligent irrigation scheme, and specifically comprises irrigation frequency, single water volume and total water volume. Meanwhile, it constructs a fruit tree growth and nutrition condition relationship model based on the processed vegetation index combination, and generates an intelligent fertilization scheme by fusing a knowledge base storing different fruit tree variety and growth period fertilizer requirement thresholds.

[0024] The whole decision-making process is completed by the host computer software, and the control unit coordinates the coherent operation of each module. Finally, the control unit converts the generated irrigation and fertilization scheme into specific control instructions and issues them to the water and fertilizer device. The drive assembly and execution assembly of the water and fertilizer device receive the instructions, accurately adjust the displacement, spatial angle and opening and closing state of the circumferentially arranged spray pipes on the top, and realize directional and quantitative water and fertilizer operation for fruit trees at different positions. The flow and pressure sensors integrated on the spray pipes feed back the execution data to the control unit in real time, forming a closed-loop control system, thereby ensuring the accuracy of each irrigation and fertilization operation and providing data support for further model optimization, and ultimately achieving the goal of precise agricultural management of water and fertilizer saving, quality improvement and efficiency increase.

[0025] The overall structure is shown in Figure 1 The data acquisition module includes a network of meteorological sensors, a network of soil multi-parameter sensors, and a UAV equipped with a hyperspectral camera in the field. The meteorological data includes air temperature, air humidity, light intensity, wind speed, wind direction, and CO2 concentration. The soil data includes soil temperature, soil humidity, pH, conductivity, nitrogen, phosphorus, and potassium content. The fruit tree canopy image data includes visible light and spectral orthographic images covering the entire growth period of the fruit trees.

[0026] The data preprocessing module is also used for radiation correction, geometric correction, and atmospheric correction of the fruit tree canopy image data, and image stitching to generate a complete inter-orchard orthographic image. A semantic segmentation network based on deep learning is used to process the stitched image, separating the fruit tree canopy from the soil background. Various vegetation indices are extracted from the processed fruit tree canopy image, including normalized vegetation index, green band normalized vegetation index, excess green index, visible light atmospheric impedance index, and modified soil-adjusted vegetation index. Feature selection is performed on the extracted vegetation indices, and the Pearson correlation coefficient method is used to select a high-correlation vegetation index combination related to the fruit tree deficiency state to construct a feature subset.

[0027] The specific steps of data preprocessing are as follows: First, use the orchard orthographic image that has been corrected for radiation, geometry, and atmosphere as input data. Through pixel-level labeling of historical images, a training data set containing two labels "fruit tree canopy" and "soil background" is generated. A semantic segmentation network is used for training. The network extracts multi-level features through an encoder, and then recovers details through upsampling and skip connection in a decoder, finally outputs high-precision binary segmentation mask, thereby accurately separating the canopy pixels from the soil, weeds and other backgrounds.

[0028] After obtaining the pure fruit tree canopy area, the system calculates multiple vegetation indices in the area, including normalized difference vegetation index, green band normalized difference vegetation index, excess green index, visible atmosphere resistance index, and modified soil adjusted vegetation index, to generate a multi-dimensional index atlas reflecting the physiological state of the canopy. To further optimize the effectiveness of the features, the system matches the average vegetation index of the canopy area in each image with the measured fruit tree nutrition status data. Based on the Pearson correlation coefficient method, the correlation between each index and the nutrition status is analyzed one by one, and those with an absolute correlation coefficient exceeding a certain threshold are selected to form a high-value feature subset significantly related to the fruit tree fertility status. This feature subset is finally used to build a fertilizer deficiency diagnosis model, providing core data for generating accurate and reliable intelligent fertilization schemes, effectively improving the accuracy of water and fertilizer decision-making and the efficiency of system operation.

[0029] The specific steps for the host computer state monitoring and control software to receive and process the index and spectral features output by the data preprocessing module and generate the intelligent irrigation and fertilization schemes are as follows: The host computer state monitoring and control software receives and fuses multi-source feature data from the data preprocessing module in real time through the communication interface, including meteorological indices, soil condition indices, and a vegetation index feature subset strongly related to the nutrition status of fruit trees. After completing the spatiotemporal alignment of the data, the software first predicts future weather conditions based on the LSTM time series prediction model, calculates the crop water requirement (ETc) using the Penman-Monteith formula, and then formulates an irrigation strategy based on real-time soil moisture data to generate an intelligent irrigation scheme containing the irrigation timing, single water volume, and total water volume.

[0030] Meanwhile, in terms of fertilization decision-making, the software uses the pre-established relationship model between fruit tree growth and vegetation index to diagnose the current tree nutrition status, queries the standard fertilizer requirement threshold values for different growth stages in the knowledge base, calculates the types and quantities of fertilizers to be supplemented based on the fertilizer deficiency model, and allocates the total fertilizer amount to multiple irrigation events following the "fertilizer follows water" principle to form an executable intelligent fertilization scheme. Finally, the software integrates the irrigation and fertilization schemes into control instructions executable by the water and fertilizer device to drive field equipment to complete precision work, thereby realizing closed-loop intelligent management of water and fertilizer integration.

[0031] The structure and specific working steps of the water and fertilizer device are as follows: Figure 3As shown: mainly by the storage and supply unit, the mixing unit and the execution unit. The core of the storage and supply unit is a plurality of independent fertilizer storage tank 01, for storing different types of solid or liquid fertilizer; each fertilizer storage tank bottom is provided with a fertilizer outlet 02, all the fertilizer outlet is connected to a set of common fertilizer outlet pipe 08, so as to realize the collection and delivery of multiple nutrients. The mixing unit includes a water and fertilizer tank 07, the fertilizer from the fertilizer outlet pipe 08 and the irrigation water from the water inlet pipe 06 are mixed to form a solution.

[0032] In addition to supplying water to the water and fertilizer tank, the water inlet pipe 06 also needs to be connected to the ground water pipe to meet other irrigation needs, and the maintenance opening 09 provided at the top of the water and fertilizer tank facilitates daily maintenance and cleaning. The execution unit is composed of the water and fertilizer outlet pipe 05, the rotating assembly 03 and the lifting assembly 04; the rotating assembly drives the water and fertilizer outlet pipe to rotate horizontally to adjust the spraying direction, and the lifting assembly controls the vertical height to adapt to different crops or irrigation modes, and the two constitute a precision adjustable sprinkler system, which finally realizes the directional and quantitative water and fertilizer operation for different positions of crops.

[0033] Those skilled in the art should understand that the embodiments of the present application can be provided as a method, a system or a computer program product. Therefore, the present application can adopt a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware aspects. Moreover, the present application can adopt the form of a computer program product implemented on one or more computer usable storage media including but not limited to disk storage, CD-ROM, optical storage, etc., which contains computer usable program code.

[0034] The present application is described with reference to flowcharts and / or block diagrams according to the method, equipment (system) and computer program product of the embodiments of the present application. It should be understood that each flow and / or block in the flowchart and / or block diagram can be realized by computer program instructions, and the combination of the flows and / or blocks in the flowchart and / or block diagram. These computer program instructions can be provided to the processor of a general-purpose computer, a special-purpose computer, an embedded processor or other programmable data processing equipment to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing equipment produce a means for implementing the functions specified in the flow Figure 1 The functions specified in one flow or multiple flows and / or blocks Figure 1 The functions specified in one flow or multiple flows and / or blocks

[0035] These computer program instructions can also be stored in a computer readable storage medium which can guide the computer or other programmable data processing equipment to work in a specific way, so that the instructions stored in the computer readable storage medium produce a product including instruction means, which implements the functions specified in the flow Figure 1 The functions specified in one flow or multiple flows and / or blocks Figure 1 The functions specified in one flow or multiple flows and / or blocks

[0036] These computer program instructions can also be loaded into a computer or other programmable data processing devices, so that a series of operational steps are performed on the computer or other programmable data processing devices to generate computer-implemented processes, thus the instructions executed on the computer or other programmable data processing devices provide operational steps for implementing the functions of the flow Figure 1 One flow or multiple flows and / or the functions specified in the block Figure 1 One block or multiple blocks.

[0037] Finally, it should be noted that: the above examples are only used to illustrate the technical solutions of the present application, but not to limit it, although the above embodiments of the present application have been described in detail, those skilled in the art should understand: the specific embodiments of the present application can be modified or replaced by the same, without departing from the spirit and scope of the present application, any modification or equivalent replacement, which should be covered within the scope of protection of the claims of the present application.

Claims

1. A deep learning-based fruit planting water and fertilizer management system, characterized in that, The system comprises a data acquisition module, a data preprocessing module, a water and fertilizer state monitoring module, a control module and a water and fertilizer device, wherein, The data acquisition module comprises a weather sensor network, a soil multi-parameter sensor network and a high-spectrum camera-carrying unmanned aerial vehicle arranged in a field, for acquiring original data in real time or periodically, the original data comprising field weather data, soil environment data and fruit tree canopy image data, and sending the original data to the data preprocessing module; The data preprocessing module receives the original data and sequentially performs standardization, correction and fusion processing on the acquired original data based on a multi-sensor fusion algorithm, and extracts indexes for representing the growth of fruit trees, soil conditions and weather characteristics; The water and fertilizer state monitoring module fuses the indexes for representing the growth of fruit trees, soil conditions and weather characteristics based on a machine learning algorithm, constructs a soil-plant-weather matching relationship model, and outputs an intelligent irrigation scheme and an intelligent fertilization scheme; The control module is used for controlling the water and fertilizer device to perform precise irrigation and fertilization operation according to the decision prescription output by the intelligent decision model; The water and fertilizer device comprises a water and fertilizer barrel, a plurality of spray pipes are circumferentially arranged on the side of the water and fertilizer barrel, and an adjusting assembly is further arranged between the water and fertilizer barrel and the spray pipes, the adjusting assembly adjusts the displacement and opening and closing of the spray pipes according to the control module, so as to realize precise irrigation and fertilization.

2. The deep learning-based fruit planting water and fertilizer management system according to claim 1, characterized in that, The data preprocessing module sequentially performs standardization, correction and fusion processing on the original data by using an extended Kalman filter algorithm, and the specific steps are as follows: The weather data and soil data are dimensionally unified, the parameters such as air temperature, air humidity and light intensity in the weather data are converted to the interval [0, 1], the parameters such as soil temperature and soil humidity in the soil data are converted to the interval [0, 1], and the fruit tree canopy image data is normalized by pixel value to the range [0, 255]; The fusion results of the standardized multi-source data are taken as system state quantities, a state equation is established to describe the dynamic change of the data with time, and a state transition matrix is trained based on the time sequence characteristics of historical data; The measured values of each sensor are taken as observation values, an observation equation is established, and an observation matrix is configured according to the accuracy of the sensor; Filter parameters are initialized, including the initial value of the state, the covariance matrix and the process noise and observation noise variance; the current state quantity and the covariance matrix are predicted based on the state equation, the Kalman gain is calculated combined with the observation equation, and the predicted value is corrected to obtain the corrected fusion data; The indexes for representing the weather data, soil conditions and growth of fruit trees are extracted from the corrected fusion data, including the humidity index, light accumulation index calculated based on the weather data, the soil moisture index and nutrient comprehensive index calculated based on the soil data, and the normalized vegetation index and green band normalized vegetation index extracted based on the fruit tree canopy image data.

3. The deep learning-based fruit planting water and fertilizer management system according to claim 1, characterized in that: The processing of the fruit tree canopy image data by the data preprocessing module comprises radiation correction, geometric correction and atmospheric correction, and image stitching is performed to generate a complete field orthophoto image; The stitched image is processed by using a semantic segmentation network based on deep learning to separate the fruit tree canopy from the soil background. A plurality of vegetation indices are extracted from the processed fruit tree canopy image, including normalized difference vegetation index, green normalized difference vegetation index, excess green index, visible atmosphere resistance index, and modified soil adjusted vegetation index; Feature selection is performed on the extracted vegetation indices, and a high correlation between the fruit tree and the lack of fertilizer status is screened out based on the Pearson correlation coefficient method to construct a feature subset.

4. The deep learning-based fruit planting water and fertilizer management system according to claim 1 or 3, characterized in that, The water and fertilizer state monitoring module includes host computer state monitoring and control software and a control unit, wherein: The host computer state monitoring and control software receives and processes the index and spectral features output by the data preprocessing module, and generates the intelligent irrigation scheme and the intelligent fertilization scheme; The control unit coordinates the operation of each module according to the host computer state monitoring and control software to ensure the continuity of data transmission and scheme execution.

5. The deep learning-based fruit planting water and fertilizer management system according to claim 4, characterized in that, The water and fertilizer state monitoring module also includes a knowledge base that stores the nitrogen, phosphorus, potassium, and trace element demand thresholds of different fruit tree varieties at different growth stages, and can be dynamically updated according to field measurement data and historical planting records.

6. The deep learning-based fruit planting water and fertilizer management system according to claim 1, characterized in that, The water and fertilizer state monitoring module predicts future periods of meteorological data based on the LSTM time series prediction model, and combines real-time soil moisture data to output an intelligent irrigation scheme including predicted irrigation frequency, single irrigation water depth, and total irrigation water volume; The intelligent fertilization scheme is based on the vegetation index combination after feature selection, and a relationship model between tree growth and vegetation index is established, and a lack of fertilizer amount model is further constructed to represent the growth difference, and finally the relationship model between tree growth and vegetation index and the lack of fertilizer amount model are fused, and the standard fertilizer demand in the knowledge base is combined to generate an intelligent fertilization scheme.

7. The deep learning-based fruit planting water and fertilizer management system according to claim 6, characterized in that, The specific process of the LSTM time series prediction model for predicting future periods of meteorological data includes: Data input: divide the preprocessed meteorological data by time series, and use a continuous preset time length of meteorological data segment as input sample, which contains time series features of temperature, humidity, light intensity, and wind speed; Model training: use sliding window method to construct training set, window step is set to 1 hour, through correlation analysis of historical meteorological data and corresponding period of measured soil moisture, optimize the number of hidden layer neurons, dropout coefficient and iteration times of LSTM model, so that the prediction error of future 72 hours of meteorological parameters is controlled within the preset threshold; Prediction output: output the predicted value of hourly temperature, air humidity, light intensity and wind speed in the next 72 hours, and calculate the cumulative reference crop evapotranspiration in this period. 8.The deep learning-based fruit planting water and fertilizer management system according to claim 1 or 7, characterized in that, The water and fertilizer device also includes a driving assembly and an execution assembly, the driving assembly receives instructions from the control unit and drives the execution assembly, and the execution assembly adjusts the spatial angle, telescopic length and on-off state of the spray pipe to realize directional and quantitative irrigation and fertilization of fruit trees at different positions. 9.The deep learning-based fruit planting water and fertilizer management system according to claim 8, characterized in that, The spray pipe of the water and fertilizer device is provided with a flow sensor and a pressure sensor for real-time feedback of irrigation and fertilization amount and pressure parameters to the control unit to form a closed loop control to ensure execution accuracy.

10. A deep learning-based fruit planting water and fertilizer management method using the system according to any one of claims 1-9. The method comprises the following steps: Real-time or periodic acquisition of raw data, including meteorological data, soil data and fruit tree canopy image data; The original data are sequentially subjected to standardization, correction and fusion processing based on a preset algorithm, and indices for representing meteorological data, soil conditions and growth of fruit trees are extracted therefrom; According to a preset algorithm, a soil-fruit tree-climate matching relationship model is constructed based on the indices of meteorological data, soil conditions and growth of fruit trees, and an intelligent irrigation and fertilization scheme for predicting irrigation requirements is outputted; The water and fertilizer device is controlled to perform precise irrigation and fertilization operation, wherein the water and fertilizer device comprises a water and fertilizer barrel, a plurality of spray pipes are circumferentially installed on a side portion of the water and fertilizer barrel, and an adjusting assembly is further arranged between the water and fertilizer barrel and the spray pipes, the adjusting assembly adjusts displacement and opening and closing of the spray pipes according to a control instruction to realize precise irrigation and fertilization.

Images