Urban garden management system and method based on artificial intelligence

Abstract

The invention relates to the field of garden management, in particular to an artificial intelligence-based urban garden management system and method.The system comprises a vegetation surveying and mapping module, a garden evaluation module, a transpiration water demand module, an irrigation decision module and a planning control module, the vegetation surveying and mapping module is used for establishing a vegetation model, and the garden evaluation module is used for estimating the three-dimensional green quantity of a garden; the transpiration water demand module is used for predicting the vegetation water demand, the irrigation decision-making module is used for establishing a park water demand model, and the planning control module is used for establishing a flow control system for irrigation. According to the invention, the precision, practicability and automation degree of vegetation measurement and calculation can be improved, the soil moisture content can be maintained, precise irrigation can be realized, and the water consumption and operation cost can be reduced; the water resource utilization efficiency is optimized, garden fine management is promoted, the utilization efficiency of a pipe network system is improved, the urban landscaping management level is improved, and higher stability and application potential are achieved.

Description

Technical Field

[0001] This invention relates to the field of garden management, specifically to an artificial intelligence-based urban garden management system and method. Background Technology

[0002] Urban parks are natural spaces within a city, utilizing plants, topography, and architecture. As an important component of urban ecology and public services, they provide recreational, aesthetic, and ecological environment improvement functions. With the development of urban management, the level of automation in urban park construction is constantly improving. To address the increasing water demand during green space irrigation, intelligent irrigation systems using water-saving irrigation equipment such as sprinkler and drip irrigation for precise water supply have become the mainstream irrigation method for urban parks.

[0003] The management of urban parks relies on tedious on-site measurements. The process of assessing vegetation coverage in park areas is labor-intensive, and the evaluation of vegetation irrigation processes is not precise enough. Furthermore, the complex environment and heterogeneous vegetation distribution within parks increase the difficulty of measurement. Existing automated irrigation devices only have a single water volume regulation function, which is insufficient to cope with the irrigation needs of different vegetation in different times and spaces, reducing water use efficiency, wasting irrigation water, and potentially affecting the growth process of park plants.

[0004] In addition, vegetation irrigation is closely related to vegetation growth, soil and climate. However, under actual measurement conditions, soil structure varies, and irrigation results are easily affected by matrix effects, which reduces irrigation accuracy. Crop sensing systems are expensive and difficult to deploy, resulting in problems such as high cost, poor timeliness and large errors. Summary of the Invention

[0005] The purpose of this invention is to provide an artificial intelligence-based urban garden management system and method to solve the problems mentioned in the background art.

[0006] To solve the above-mentioned technical problems, the present invention provides the following technical solution: an artificial intelligence-based urban garden management system, comprising: a vegetation mapping module, a garden assessment module, a transpiration water demand module, an irrigation decision module, and a planning control module; The vegetation mapping module is used to take pictures of urban garden areas using drones. By pre-detecting the image frames, the module obtains the vegetation shooting frame. Based on the target vegetation image stream within the shooting frame, it generates time-series point cloud data. The point cloud data is centrally processed to calculate the vegetation branch radius. The point cloud is clustered and segmented according to the skeleton. The convergence effect is simulated according to the allometric growth algorithm. Quadrilateral leaf models are generated at the edge nodes of the skeleton to obtain the vegetation model. The garden assessment module is used to extract the leaf area index of vegetation from UAV images, remove the soil background by using the color difference of the vegetation canopy and stitch the images together, extract canopy data based on the image grayscale, determine the canopy height and vegetation coverage of the garden area, and estimate the three-dimensional green volume of the garden by voxel segmentation of the vegetation model. The transpiration water requirement module is used to collect soil index data, screen the minimum dataset through principal component analysis, calculate the soil quality index based on membership function and weight, calculate the water stress index based on soil quality index, leaf area index and vegetation type through plant growth model, input the water stress index and three-dimensional green volume into the Penman formula, calculate vegetation transpiration, and use long short-term memory neural network to predict vegetation water requirement. The irrigation decision module is used to train a BP neural network through meteorological data, establish a fuzzy decision system through the trained model, take precipitation, humidity and temperature difference as input, output meteorological evaporation, establish a park water demand model based on vegetation water demand, meteorological evaporation and soil moisture content, evaluate the water demand for irrigation in the park, and keep the soil moisture content in the park within a predetermined range. The planning and control module is used to establish an RBF prediction model for soil moisture, determine the irrigation amount based on water demand and soil moisture, divide the garden into fan-shaped sub-regions with different radii and equal central angles based on the central axis coordinates and irrigation radius of the irrigation device, and establish a flow control system for irrigation with the goal of minimizing the irrigation time difference of the equipment in each sub-region, and with flow rate, irrigation time and irrigation interval as constraints.

[0007] Furthermore, the vegetation mapping module includes: an image acquisition unit, a point cloud processing unit, and a branch modeling unit; The image acquisition unit is used to plan the gridded flight path of the drone in the garden area, check the color index of the captured images, generate the shooting frame, match the position of the shooting frame according to the SIFT feature of the drone's altitude change rate, so that the shooting frame contains complete vegetation data, and compensate for the image matching according to the speed change rate. The point cloud processing unit is used to register images from different time periods to the same coordinate system based on the depth images and timestamps captured by the UAV, sample and process the depth image frames in the coordinate system to generate a single vegetation point cloud, and extract the initial skeleton lines of branches and trunks through the Voronoi diagram. The branch modeling unit assigns weights proportionally to the length, point cloud density, and average curvature of each branch skeleton, eliminates low-weight skeletons, integrates adjacent skeletons, performs cylindrical fitting along the skeleton lines, and generates quadrilateral leaves of the same size at the end nodes and middle nodes of the skeleton. The leaf area is determined by the vegetation type.

[0008] Furthermore, the garden assessment module includes: an image stitching unit and a greening monitoring unit; The image stitching unit is used to calculate the NDVI index of vegetation through the captured spectral images, input the radiative transfer model to obtain the leaf area index, separate the ground point and vegetation point pixels, stitch all vegetation point pixels together, obtain a binary image of vegetation through semantic segmentation, calculate the proportion of vegetation pixels to total pixels, and obtain the vegetation coverage. The greening monitoring unit is used to voxelize the three-dimensional vegetation model, calculate the leaf surface area, branch surface area and volume of the vegetation in each voxel, and sum up the total surface area of ​​all vegetation to obtain the three-dimensional green volume of the urban garden area.

[0009] Furthermore, the evapotranspiration water demand module includes: a soil quality unit and a demand prediction unit; The soil quality unit is used to synchronize soil index data from meteorological stations or soil monitoring stations. The soil index data includes electrical conductivity, carbon-nitrogen ratio, sand content and moisture content. The spatiotemporally aligned data is screened using principal component analysis to form a minimal dataset. An S-shaped membership function is constructed for each index to calculate the soil quality index. The demand forecasting unit is used to fit the water stress index of each vegetation unit based on the vegetation stomatal conductance model, calculate the potential evapotranspiration using the Penman formula, train the forecasting model using a long short-term memory network, input historical meteorological data, soil moisture content, vegetation index and potential evapotranspiration, and output the vegetation water demand in the next cycle.

[0010] Furthermore, the irrigation decision module includes: a model training unit and a water use assessment unit; The model training unit is used to take each meteorological variable as the input layer node of the neural network, randomly combine the meteorological variables as the training dataset, determine the number of hidden layer nodes of the neural network, optimize the connection weights, and obtain the trained BP neural model. The water use assessment unit is used to weight and combine precipitation, irrigation, vegetation water demand and runoff replenishment into a soil moisture content prediction model, adjust the irrigation parameters to minimize irrigation water consumption, maintain soil moisture content within a predetermined range in the next cycle, and generate an irrigation scheduling plan.

[0011] Furthermore, the planning and control module includes: a soil moisture detection unit, a regional irrigation unit, and a flow control unit; The soil moisture detection unit is used to predict soil moisture based on urban phenological characteristics, including season, temperature, precipitation and humidity, using a radial basis function neural network. It uses the gradient descent method to adjust the center vector and width parameters to determine the availability of water demand. The regional irrigation unit is used to divide sub-regions with the irrigation center axis coordinates as the center and the effective range as the radius. In each sub-region, a regional irrigation coefficient is set according to the terrain and soil moisture. The irrigation time of the irrigation sprinkler equipment in each sub-region is adjusted according to the irrigation coefficient and the irrigation scheduling plan. The flow control unit is used to control the solenoid valve switch via the controller, adjust the pump station frequency and valve opening, irrigate according to the start and stop time sequence of the sprinklers in each sub-area, and display the irrigation time and water consumption on the interactive interface.

[0012] An artificial intelligence-based urban garden management method includes the following steps: Step S1. Use a drone to photograph the garden area, generate a time-series point cloud based on the vegetation image stream within the vegetation shooting frame, calculate the vegetation branch radius based on the point cloud data, cluster and segment the point cloud based on the skeleton, generate quadrilateral leaf structures at the edge nodes of the skeleton, and generate a vegetation model. Step S2. Extract the leaf area index of vegetation from the image, stitch the image according to the vegetation color, determine the vegetation coverage of the garden area, segment the voxels of the vegetation model, and estimate the three-dimensional green volume of the garden based on the vegetation coverage. Step S3. Collect soil index data, filter the minimum dataset, construct a membership function to calculate the soil quality index, calculate the water stress index based on the soil quality index, leaf area index and vegetation type, and then predict the vegetation water requirement based on the water stress index and three-dimensional green volume. Step S4. Train a neural network using meteorological data to establish a fuzzy decision system. Input precipitation, humidity, and temperature difference, and output meteorological evaporation. Establish a water demand model based on vegetation water demand, meteorological evaporation, and soil moisture content, and output the water demand for irrigation in the park. Step S5. Establish a soil moisture prediction model, determine the irrigation amount based on water demand and soil moisture, divide the garden into sub-regions based on the central axis and irrigation radius of the irrigation device, establish a flow control system, and adjust the flow, irrigation time, and irrigation interval in each sub-region.

[0013] Furthermore, step S1 includes: Step S11. Plan a gridded flight path for the drone in the garden area, check the color index of the captured images, generate a shooting frame, match the shooting frame position according to the SIFT feature of the drone's altitude change rate, so that the shooting frame contains complete vegetation data, and compensate for the image matching according to the speed change rate. Step S12. Based on the depth images and timestamps captured by the UAV, register the images from different time periods to the same coordinate system, sample and process the depth image frames within the coordinate system to generate a single vegetation point cloud, and extract the initial skeleton lines of branches and trunks through the Voronoi diagram. Step S13. Based on the length, point cloud density and average curvature of each branch skeleton, weights are proportionally weighted and assigned. Low-weight skeletons are eliminated, adjacent skeletons are integrated, and cylindrical fitting is performed along the skeleton line. Quadrilateral leaves of the same size are generated at the end nodes and middle nodes of the skeleton. The leaf area is determined by the vegetation type.

[0014] Furthermore, step S2 includes: Step S21. Calculate the NDVI index of vegetation through the captured spectral image, input the radiative transfer model to obtain the leaf area index, separate the ground point and vegetation point pixels, stitch all vegetation point pixels together, obtain the vegetation binary image through semantic segmentation, calculate the proportion of vegetation pixels to total pixels, and obtain the vegetation coverage. Step S22. Convert the three-dimensional vegetation model into voxels, calculate the leaf surface area, branch surface area and volume of vegetation in each voxel, and sum up the total surface area of ​​all vegetation to obtain the three-dimensional green volume of the urban garden area.

[0015] Furthermore, step S3 includes: Step S31. Synchronize soil index data from meteorological stations or soil monitoring stations. The soil index data includes electrical conductivity, carbon-nitrogen ratio, sand content and moisture content. Spatiotemporally aligned data are used. Principal component analysis is used to screen the indicators to form a minimal dataset. An S-shaped membership function is constructed for each indicator to calculate the soil quality index. Step S32. Fit the water stress index of each vegetation unit based on the vegetation stomatal conductance model, calculate the potential evapotranspiration using the Penman formula, train the prediction model using a long short-term memory network, input historical meteorological data, soil moisture content, vegetation index and potential evapotranspiration, and output the vegetation water demand in the next cycle.

[0016] Furthermore, step S4 includes: Step S41. Use each meteorological variable as the input layer node of the neural network, randomly combine the meteorological variables as the training dataset, determine the number of hidden layer nodes of the neural network, optimize the connection weights, and obtain the trained BP neural model. Step S42. Weightedly combine precipitation, irrigation, vegetation water demand and runoff replenishment into a soil moisture content prediction model, adjust the irrigation parameter to minimize irrigation water consumption, maintain soil moisture content within a predetermined range in the next cycle, and generate an irrigation scheduling plan.

[0017] Furthermore, step S5 includes: Step S51. Based on the phenological characteristics of the city, including season, temperature, precipitation and humidity, soil moisture is predicted by radial basis function neural network. The center vector and width parameters are adjusted by gradient descent method to determine the availability of water demand. Step S52. Divide the area into sub-regions with the irrigation center axis coordinates as the center and the effective range as the radius. Set the regional irrigation coefficient for each sub-region based on the terrain and soil moisture. Adjust the irrigation time of the irrigation sprinkler equipment in each sub-region according to the irrigation coefficient and the irrigation scheduling plan. Step S53. Control the solenoid valve switch through the controller, adjust the pump station frequency and valve opening, irrigate according to the start and stop time sequence of the sprinklers in each sub-area, and display the irrigation time and water consumption on the interactive interface.

[0018] Compared with the prior art, the beneficial effects achieved by the present invention are: 1. This invention uses drones to photograph garden areas, matches the position of the shooting frame, generates time-series point clouds, generates branches and leaves according to the allometric growth algorithm, and obtains a vegetation model. This enables automated monitoring and intelligent digital archiving of garden vegetation, assists in landscape reconstruction, ensures the accuracy of vegetation canopy distribution data extraction, improves the accuracy, practicality and automation of vegetation measurement, and enhances the management capabilities of urban green areas.

[0019] 2. This invention collects soil index data synchronously from meteorological stations or soil monitoring stations, uses long short-term memory neural networks to predict vegetation water demand, establishes a fuzzy decision system, assesses the water demand for irrigation in the park, can more accurately predict soil moisture, adjust irrigation strategies in real time, maintain soil moisture content, achieve precision irrigation, reduce water consumption and operating costs, optimize water resource utilization efficiency, and promote refined management of gardens.

[0020] 3. Based on the phenological characteristics of the city, this invention predicts soil moisture, divides the garden area according to the irrigation device, determines the irrigation time and water consumption, and establishes a flow control system for irrigation. This realizes the optimization of irrigation time and flow control, ensuring an efficient and precise park irrigation management process, improving the utilization efficiency of the pipeline system, alleviating the city's water shortage problem, and improving the level of urban greening management. It has higher stability and application potential. Attached Figure Description

[0021] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used in conjunction with embodiments of the invention to explain the invention and do not constitute a limitation thereof. In the drawings: Figure 1 This is a schematic diagram of the structure of an artificial intelligence-based urban garden management system according to the present invention; Figure 2 This is a schematic diagram illustrating the steps of an artificial intelligence-based urban garden management method according to the present invention. Detailed Implementation

[0022] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0023] Please see Figures 1 to 2 The present invention provides a technical solution: an artificial intelligence-based urban garden management system, comprising: a vegetation mapping module, a garden assessment module, a transpiration water requirement module, an irrigation decision module, and a planning control module; The vegetation mapping module is used to take pictures of urban garden areas using drones. By pre-detecting the image frames, the module obtains the vegetation shooting frame. Based on the target vegetation image stream within the shooting frame, it generates time-series point cloud data. The point cloud data is centrally processed to calculate the vegetation branch radius. The point cloud is clustered and segmented according to the skeleton. The convergence effect is simulated according to the allometric growth algorithm. Quadrilateral leaf models are generated at the edge nodes of the skeleton to obtain the vegetation model. The vegetation mapping module includes: an image acquisition unit, a point cloud processing unit, and a branch modeling unit; The image acquisition unit is used to plan the gridded flight path of the drone in the garden area, check the color index of the captured images, generate the shooting frame, match the position of the shooting frame according to the SIFT feature of the drone's altitude change rate, so that the shooting frame contains complete vegetation data, and compensate for the image matching according to the speed change rate. The point cloud processing unit is used to register images from different time periods to the same coordinate system based on the depth images and timestamps captured by the UAV, sample and process the depth image frames in the coordinate system to generate a single vegetation point cloud, and extract the initial skeleton lines of branches and trunks through the Voronoi diagram. The branch modeling unit assigns weights proportionally to the length, point cloud density, and average curvature of each branch skeleton, eliminates low-weight skeletons, integrates adjacent skeletons, performs cylindrical fitting along the skeleton lines, and generates quadrilateral leaves of the same size at the end nodes and middle nodes of the skeleton. The leaf area is determined by the vegetation type.

[0024] The garden assessment module is used to extract the leaf area index of vegetation from UAV images, remove the soil background by using the color difference of the vegetation canopy and stitch the images together, extract canopy data based on the image grayscale, determine the canopy height and vegetation coverage of the garden area, and estimate the three-dimensional green volume of the garden by voxel segmentation of the vegetation model. The garden assessment module includes: an image stitching unit and a greening monitoring unit; The image stitching unit is used to calculate the NDVI index of vegetation through the captured spectral images, input the radiative transfer model to obtain the leaf area index, separate the ground point and vegetation point pixels, stitch all vegetation point pixels together, obtain a binary image of vegetation through semantic segmentation, calculate the proportion of vegetation pixels to total pixels, and obtain the vegetation coverage. The greening monitoring unit is used to voxelize the three-dimensional vegetation model, calculate the leaf surface area, branch surface area and volume of the vegetation in each voxel, and sum up the total surface area of ​​all vegetation to obtain the three-dimensional green volume of the urban garden area.

[0025] The transpiration water requirement module is used to collect soil index data, screen the minimum dataset through principal component analysis, calculate the soil quality index based on membership function and weight, calculate the water stress index based on soil quality index, leaf area index and vegetation type through plant growth model, input the water stress index and three-dimensional green volume into the Penman formula, calculate vegetation transpiration, and use long short-term memory neural network to predict vegetation water requirement. The evapotranspiration water demand module includes: a soil quality unit and a demand prediction unit; The soil quality unit is used to synchronize soil index data from meteorological stations or soil monitoring stations. The soil index data includes electrical conductivity, carbon-nitrogen ratio, sand content and moisture content. The spatiotemporally aligned data is screened using principal component analysis to form a minimal dataset. An S-shaped membership function is constructed for each index to calculate the soil quality index. The demand forecasting unit is used to fit the water stress index of each vegetation unit based on the vegetation stomatal conductance model, calculate the potential evapotranspiration using the Penman formula, train the forecasting model using a long short-term memory network, input historical meteorological data, soil moisture content, vegetation index and potential evapotranspiration, and output the vegetation water demand in the next cycle.

[0026] The irrigation decision module is used to train a BP neural network through meteorological data, establish a fuzzy decision system through the trained model, take precipitation, humidity and temperature difference as input, output meteorological evaporation, establish a park water demand model based on vegetation water demand, meteorological evaporation and soil moisture content, evaluate the water demand for irrigation in the park, and keep the soil moisture content in the park within a predetermined range. The irrigation decision module includes: a model training unit and a water use assessment unit; The model training unit is used to take each meteorological variable as the input layer node of the neural network, randomly combine the meteorological variables as the training dataset, determine the number of hidden layer nodes of the neural network, optimize the connection weights, and obtain the trained BP neural model. The water use assessment unit is used to weight and combine precipitation, irrigation, vegetation water demand and runoff replenishment into a soil moisture content prediction model, adjust the irrigation parameters to minimize irrigation water consumption, maintain soil moisture content within a predetermined range in the next cycle, and generate an irrigation scheduling plan.

[0027] The planning and control module is used to establish an RBF prediction model for soil moisture, determine the irrigation amount based on water demand and soil moisture, divide the garden into fan-shaped sub-regions with different radii and equal central angles based on the central axis coordinates and irrigation radius of the irrigation device, and establish a flow control system for irrigation with the goal of minimizing the irrigation time difference of the equipment in each sub-region, and with flow rate, irrigation time and irrigation interval as constraints.

[0028] The planning and control module includes: a soil moisture detection unit, a regional irrigation unit, and a flow control unit; The soil moisture detection unit is used to predict soil moisture based on urban phenological characteristics, including season, temperature, precipitation and humidity, using a radial basis function neural network. It uses the gradient descent method to adjust the center vector and width parameters to determine the availability of water demand. The regional irrigation unit is used to divide sub-regions with the irrigation center axis coordinates as the center and the effective range as the radius. In each sub-region, a regional irrigation coefficient is set according to the terrain and soil moisture. The irrigation time of the irrigation sprinkler equipment in each sub-region is adjusted according to the irrigation coefficient and the irrigation scheduling plan. The flow control unit is used to control the solenoid valve switch via the controller, adjust the pump station frequency and valve opening, irrigate according to the start and stop time sequence of the sprinklers in each sub-area, and display the irrigation time and water consumption on the interactive interface.

[0029] An artificial intelligence-based urban garden management method includes the following steps: Step S1. Use a drone to photograph the garden area, generate a time-series point cloud based on the vegetation image stream within the vegetation shooting frame, calculate the vegetation branch radius based on the point cloud data, cluster and segment the point cloud based on the skeleton, generate quadrilateral leaf structures at the edge nodes of the skeleton, and generate a vegetation model. Step S1 includes: Step S11. Plan a gridded flight path for the drone in the garden area, check the color index of the captured images, generate a shooting frame, match the shooting frame position according to the SIFT feature of the drone's altitude change rate, so that the shooting frame contains complete vegetation data, and compensate for the image matching according to the speed change rate. Step S12. Based on the depth images and timestamps captured by the UAV, register the images from different time periods to the same coordinate system, sample and process the depth image frames within the coordinate system to generate a single vegetation point cloud, and extract the initial skeleton lines of branches and trunks through the Voronoi diagram. Step S13. Based on the length, point cloud density and average curvature of each branch skeleton, weights are proportionally weighted and assigned. Low-weight skeletons are eliminated, adjacent skeletons are integrated, and cylindrical fitting is performed along the skeleton line. Quadrilateral leaves of the same size are generated at the end nodes and middle nodes of the skeleton. The leaf area is determined by the vegetation type.

[0030] Step S2. Extract the leaf area index of vegetation from the image, stitch the image according to the vegetation color, determine the vegetation coverage of the garden area, segment the voxels of the vegetation model, and estimate the three-dimensional green volume of the garden based on the vegetation coverage. Step S2 includes: Step S21. Calculate the NDVI index of vegetation through the captured spectral image, input the radiative transfer model to obtain the leaf area index, separate the ground point and vegetation point pixels, stitch all vegetation point pixels together, obtain the vegetation binary image through semantic segmentation, calculate the proportion of vegetation pixels to total pixels, and obtain the vegetation coverage. Step S22. Convert the three-dimensional vegetation model into voxels, calculate the leaf surface area, branch surface area and volume of vegetation in each voxel, and sum up the total surface area of ​​all vegetation to obtain the three-dimensional green volume of the urban garden area.

[0031] Step S3. Collect soil index data, filter the minimum dataset, construct a membership function to calculate the soil quality index, calculate the water stress index based on the soil quality index, leaf area index and vegetation type, and then predict the vegetation water requirement based on the water stress index and three-dimensional green volume. Step S3 includes: Step S31. Synchronize soil index data from meteorological stations or soil monitoring stations. The soil index data includes electrical conductivity, carbon-nitrogen ratio, sand content and moisture content. Spatiotemporally aligned data are used. Principal component analysis is used to screen the indicators to form a minimal dataset. An S-shaped membership function is constructed for each indicator to calculate the soil quality index. Step S32. Fit the water stress index of each vegetation unit based on the vegetation stomatal conductance model, calculate the potential evapotranspiration using the Penman formula, train the prediction model using a long short-term memory network, input historical meteorological data, soil moisture content, vegetation index and potential evapotranspiration, and output the vegetation water demand in the next cycle.

[0032] Step S4. Train a neural network using meteorological data to establish a fuzzy decision system. Input precipitation, humidity, and temperature difference, and output meteorological evaporation. Establish a water demand model based on vegetation water demand, meteorological evaporation, and soil moisture content, and output the water demand for irrigation in the park. Step S4 includes: Step S41. Use each meteorological variable as the input layer node of the neural network, randomly combine the meteorological variables as the training dataset, determine the number of hidden layer nodes of the neural network, optimize the connection weights, and obtain the trained BP neural model. Step S42. Weightedly combine precipitation, irrigation, vegetation water demand and runoff replenishment into a soil moisture content prediction model, adjust the irrigation parameter to minimize irrigation water consumption, maintain soil moisture content within a predetermined range in the next cycle, and generate an irrigation scheduling plan.

[0033] Step S5. Establish a soil moisture prediction model, determine the irrigation amount based on water demand and soil moisture, divide the garden into sub-regions based on the central axis and irrigation radius of the irrigation device, establish a flow control system, and adjust the flow, irrigation time, and irrigation interval in each sub-region.

[0034] Step S5 includes: Step S51. Based on the phenological characteristics of the city, including season, temperature, precipitation and humidity, soil moisture is predicted by radial basis function neural network. The center vector and width parameters are adjusted by gradient descent method to determine the availability of water demand. Step S52. Divide the area into sub-regions with the irrigation center axis coordinates as the center and the effective range as the radius. Set the regional irrigation coefficient for each sub-region based on the terrain and soil moisture. Adjust the irrigation time of the irrigation sprinkler equipment in each sub-region according to the irrigation coefficient and the irrigation scheduling plan. Step S53. Control the solenoid valve switch through the controller, adjust the pump station frequency and valve opening, irrigate according to the start and stop time sequence of the sprinklers in each sub-area, and display the irrigation time and water consumption on the interactive interface.

[0035] Example: A drone is used to photograph urban garden areas to capture vegetation image data, identify the shooting frame area containing vegetation, align the drone's flight speed change rate with the image frame, obtain depth information from each image angle, generate a temporal point cloud, assign skeleton weights, remove low-weight skeletons, integrate adjacent skeletons, extract the main skeleton and branch structure, and use surface fitting to generate leaf contours to complete the 3D modeling of garden vegetation. The leaf area index of vegetation is extracted from drone images. Soil background is removed by canopy color difference and images are stitched together. Canopy height and vegetation coverage are measured, three-dimensional green volume is estimated, soil quality index is calculated, plant growth model is established, water demand is predicted, park water demand model is established, the park irrigation water demand is assessed, soil moisture is predicted, and equipment irrigation time is determined.

[0036] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0037] Finally, it should be noted that the above descriptions are merely preferred embodiments of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some of the technical features. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. An artificial intelligence-based urban garden management method, characterized in that, The method includes the following steps: Step S1. Use a drone to photograph the garden area, generate a time-series point cloud based on the vegetation image stream within the vegetation shooting frame, calculate the vegetation branch radius based on the point cloud data, cluster and segment the point cloud based on the skeleton, generate quadrilateral leaf structures at the edge nodes of the skeleton, and generate a vegetation model. Step S2. Extract the leaf area index of vegetation from the image, stitch the image according to the vegetation color, determine the vegetation coverage of the garden area, segment the voxels of the vegetation model, and estimate the three-dimensional green volume of the garden based on the vegetation coverage. Step S3. Collect soil index data, filter the minimum dataset, construct a membership function to calculate the soil quality index, calculate the water stress index based on the soil quality index, leaf area index and vegetation type, and then predict the vegetation water requirement based on the water stress index and three-dimensional green volume. Step S4. Train a neural network using meteorological data to establish a fuzzy decision system. Input precipitation, humidity, and temperature difference, and output meteorological evaporation. Establish a water demand model based on vegetation water demand, meteorological evaporation, and soil moisture content, and output the water demand for irrigation in the park. Step S5. Establish a soil moisture prediction model, determine the irrigation amount based on water demand and soil moisture, divide the garden into sub-regions based on the central axis and irrigation radius of the irrigation device, establish a flow control system, and adjust the flow, irrigation time, and irrigation interval in each sub-region.

2. The urban garden management method based on artificial intelligence according to claim 1, characterized in that: Step S1 includes: Step S11. Plan a gridded flight path for the drone in the garden area, check the color index of the captured images, generate a shooting frame, match the shooting frame position according to the SIFT feature of the drone's altitude change rate, so that the shooting frame contains complete vegetation data, and compensate for the image matching according to the speed change rate. Step S12. Based on the depth images and timestamps captured by the UAV, register the images from different time periods to the same coordinate system, sample and process the depth image frames within the coordinate system to generate a single vegetation point cloud, and extract the initial skeleton lines of branches and trunks through the Voronoi diagram. Step S13. Based on the length, point cloud density and average curvature of each branch skeleton, weights are proportionally weighted and assigned. Low-weight skeletons are eliminated, adjacent skeletons are integrated, and cylindrical fitting is performed along the skeleton line. Quadrilateral leaves of the same size are generated at the end nodes and middle nodes of the skeleton. The leaf area is determined by the vegetation type.

3. The urban garden management method based on artificial intelligence according to claim 2, characterized in that: Step S2 includes: Step S21. Calculate the NDVI index of vegetation through the captured spectral image, input the radiative transfer model to obtain the leaf area index, separate the ground point and vegetation point pixels, stitch all vegetation point pixels together, obtain the vegetation binary image through semantic segmentation, calculate the proportion of vegetation pixels to total pixels, and obtain the vegetation coverage. Step S22. Convert the three-dimensional vegetation model into voxels, calculate the leaf surface area, branch surface area and volume of vegetation in each voxel, and sum up the total surface area of ​​all vegetation to obtain the three-dimensional green volume of the urban garden area. Step S3 includes: Step S31. Synchronize soil index data from meteorological stations or soil monitoring stations. The soil index data includes electrical conductivity, carbon-nitrogen ratio, sand content and moisture content. Spatiotemporally aligned data are used. Principal component analysis is used to screen the indicators to form a minimal dataset. An S-shaped membership function is constructed for each indicator to calculate the soil quality index. Step S32. Fit the water stress index of each vegetation unit based on the vegetation stomatal conductance model, calculate the potential evapotranspiration using the Penman formula, train the prediction model using a long short-term memory network, input historical meteorological data, soil moisture content, vegetation index and potential evapotranspiration, and output the vegetation water demand in the next cycle.

4. The urban garden management method based on artificial intelligence according to claim 3, characterized in that: Step S4 includes: Step S41. Use each meteorological variable as the input layer node of the neural network, randomly combine the meteorological variables as the training dataset, determine the number of hidden layer nodes of the neural network, optimize the connection weights, and obtain the trained BP neural model. Step S42. Weightedly combine precipitation, irrigation, vegetation water demand and runoff replenishment into a soil moisture content prediction model, adjust the irrigation parameter to minimize irrigation water consumption, maintain soil moisture content within a predetermined range in the next cycle, and generate an irrigation scheduling plan.

5. The urban garden management method based on artificial intelligence according to claim 4, characterized in that: Step S5 includes: Step S51. Based on the phenological characteristics of the city, including season, temperature, precipitation and humidity, soil moisture is predicted by radial basis function neural network. The center vector and width parameters are adjusted by gradient descent method to determine the availability of water demand. Step S52. Divide the area into sub-regions with the irrigation center axis coordinates as the center and the effective range as the radius. Set the regional irrigation coefficient for each sub-region based on the terrain and soil moisture. Adjust the irrigation time of the irrigation sprinkler equipment in each sub-region according to the irrigation coefficient and the irrigation scheduling plan. Step S53. Control the solenoid valve switch through the controller, adjust the pump station frequency and valve opening, irrigate according to the start and stop time sequence of the sprinklers in each sub-area, and display the irrigation time and water consumption on the interactive interface.

6. An artificial intelligence-based urban garden management system, characterized in that, The system includes the following modules: vegetation mapping module, landscape assessment module, transpiration water requirement module, irrigation decision module, and planning control module; The vegetation mapping module is used to take pictures of urban garden areas using drones. By pre-detecting the image frames, the module obtains the vegetation shooting frame. Based on the target vegetation image stream within the shooting frame, it generates time-series point cloud data. The point cloud data is centrally processed to calculate the vegetation branch radius. The point cloud is clustered and segmented according to the skeleton. The convergence effect is simulated according to the allometric growth algorithm. Quadrilateral leaf models are generated at the edge nodes of the skeleton to obtain the vegetation model. The garden assessment module is used to extract the leaf area index of vegetation from UAV images, remove the soil background by using the color difference of the vegetation canopy and stitch the images together, extract canopy data based on the image grayscale, determine the canopy height and vegetation coverage of the garden area, and estimate the three-dimensional green volume of the garden by voxel segmentation of the vegetation model. The transpiration water requirement module is used to collect soil index data, screen the minimum dataset through principal component analysis, calculate the soil quality index based on membership function and weight, calculate the water stress index based on soil quality index, leaf area index and vegetation type through plant growth model, input the water stress index and three-dimensional green volume into the Penman formula, calculate vegetation transpiration, and use long short-term memory neural network to predict vegetation water requirement. The irrigation decision module is used to train a BP neural network through meteorological data, establish a fuzzy decision system through the trained model, take precipitation, humidity and temperature difference as input, output meteorological evaporation, establish a park water demand model based on vegetation water demand, meteorological evaporation and soil moisture content, evaluate the water demand for irrigation in the park, and keep the soil moisture content in the park within a predetermined range. The planning and control module is used to establish an RBF prediction model for soil moisture, determine the irrigation amount based on water demand and soil moisture, divide the garden into fan-shaped sub-regions with different radii and equal central angles based on the central axis coordinates and irrigation radius of the irrigation device, and establish a flow control system for irrigation with the goal of minimizing the irrigation time difference of the equipment in each sub-region, and with flow rate, irrigation time and irrigation interval as constraints.

7. The urban garden management system based on artificial intelligence according to claim 6, characterized in that: The vegetation mapping module includes: an image acquisition unit, a point cloud processing unit, and a branch modeling unit; The image acquisition unit is used to plan the gridded flight path of the drone in the garden area, check the color index of the captured images, generate the shooting frame, match the position of the shooting frame according to the SIFT feature of the drone's altitude change rate, so that the shooting frame contains complete vegetation data, and compensate for the image matching according to the speed change rate. The point cloud processing unit is used to register images from different time periods to the same coordinate system based on the depth images and timestamps captured by the UAV, sample and process the depth image frames in the coordinate system to generate a single vegetation point cloud, and extract the initial skeleton lines of branches and trunks through the Voronoi diagram. The branch modeling unit assigns weights proportionally to the length, point cloud density, and average curvature of each branch skeleton, eliminates low-weight skeletons, integrates adjacent skeletons, performs cylindrical fitting along the skeleton lines, and generates quadrilateral leaves of the same size at the end nodes and middle nodes of the skeleton. The leaf area is determined by the vegetation type.

8. The urban garden management system based on artificial intelligence according to claim 7, characterized in that: The garden assessment module includes: an image stitching unit and a greening monitoring unit; The image stitching unit is used to calculate the NDVI index of vegetation through the captured spectral images, input the radiative transfer model to obtain the leaf area index, separate the ground point and vegetation point pixels, stitch all vegetation point pixels together, obtain a binary image of vegetation through semantic segmentation, calculate the proportion of vegetation pixels to total pixels, and obtain the vegetation coverage. The greening monitoring unit is used to voxelize the three-dimensional vegetation model, calculate the leaf surface area, branch surface area and volume of the vegetation in each voxel, and sum up the total surface area of ​​all vegetation to obtain the three-dimensional green volume of the urban garden area. The evapotranspiration water demand module includes: a soil quality unit and a demand prediction unit; The soil quality unit is used to synchronize soil index data from meteorological stations or soil monitoring stations. The soil index data includes electrical conductivity, carbon-nitrogen ratio, sand content and moisture content. The spatiotemporally aligned data is screened using principal component analysis to form a minimal dataset. An S-shaped membership function is constructed for each index to calculate the soil quality index. The demand forecasting unit is used to fit the water stress index of each vegetation unit based on the vegetation stomatal conductance model, calculate the potential evapotranspiration using the Penman formula, train the forecasting model using a long short-term memory network, input historical meteorological data, soil moisture content, vegetation index and potential evapotranspiration, and output the vegetation water demand in the next cycle.

9. The urban garden management system based on artificial intelligence according to claim 8, characterized in that: The irrigation decision module includes: a model training unit and a water use assessment unit; The model training unit is used to take each meteorological variable as the input layer node of the neural network, randomly combine the meteorological variables as the training dataset, determine the number of hidden layer nodes of the neural network, optimize the connection weights, and obtain the trained BP neural model. The water use assessment unit is used to weight and combine precipitation, irrigation, vegetation water demand and runoff replenishment into a soil moisture content prediction model, adjust the irrigation parameters to minimize irrigation water consumption, maintain soil moisture content within a predetermined range in the next cycle, and generate an irrigation scheduling plan.

10. The urban garden management system based on artificial intelligence according to claim 9, characterized in that: The planning and control module includes: a soil moisture detection unit, a regional irrigation unit, and a flow control unit; The soil moisture detection unit is used to predict soil moisture based on urban phenological characteristics, including season, temperature, precipitation and humidity, using a radial basis function neural network. It uses the gradient descent method to adjust the center vector and width parameters to determine the availability of water demand. The regional irrigation unit is used to divide sub-regions with the irrigation center axis coordinates as the center and the effective range as the radius. In each sub-region, a regional irrigation coefficient is set according to the terrain and soil moisture. The irrigation time of the irrigation sprinkler equipment in each sub-region is adjusted according to the irrigation coefficient and the irrigation scheduling plan. The flow control unit is used to control the solenoid valve switch via the controller, adjust the pump station frequency and valve opening, irrigate according to the start and stop time sequence of the sprinklers in each sub-area, and display the irrigation time and water consumption on the interactive interface.

Images