Garden nursery stock intelligent management method and system based on intelligent maintenance

Abstract

The invention belongs to the technical field of garden management, and discloses a garden nursery stock intelligent management method and system based on intelligent maintenance, and the method comprises the steps: a data collection module collects environment data in real time; the fixed-point monitoring module collects high-definition images of nursery stocks in a garden at fixed points. The cruise monitoring module comprehensively collects high-definition images; the image fusion module carries out splicing fusion; the image preprocessing module preprocesses the image, and the feature extraction module extracts features of the image; the image analysis module identifies the growth stage, the growth condition and the abnormal condition of the nursery stock; the comprehensive evaluation module comprehensively evaluates the growth environment and growth condition of the garden seedlings and predicts potential risks; and the maintenance suggestion module gives maintenance suggestions of the garden seedlings. According to the invention, multi-dimensional real-time monitoring of garden meteorology and nursery stock growth conditions is realized; the seedling growth stage, growth vigor and disease and pest abnormal conditions are accurately identified, the seedling growth environment and condition are comprehensively evaluated, and potential risks are predicted based on historical and real-time data.

Description

Technical Field

[0001] The present application relates to the technical field of garden management, and more specifically, to a method and system for intelligent management of garden seedlings based on intelligent maintenance. Background Technique

[0002] Garden seedlings can purify the air, absorb carbon dioxide, release oxygen, and also adsorb dust and harmful gases in the air, improving air quality. For example, the common plane tree has large leaves with fluff on the surface, which can effectively adsorb dust. At the same time, the roots of seedlings can fix the soil and conserve water, prevent soil erosion, and reduce the risks of disasters such as heavy rain and floods. Reasonable seedling management can create beautiful landscapes. By matching seedlings with different seasons, colors, and shapes, a garden landscape with scenery in all seasons can be created. For example, cherry blossoms in spring, lotuses in summer, ginkgoes in autumn, and wintersweets in winter make the garden show unique charm with the change of seasons. Carefully pruned shaped seedlings, such as boxwoods in spherical or conical shapes, can also add an artistic sense to the garden. Healthy-growing garden seedlings can enhance the ornamental value and commercial value of the garden, attract more tourists, and bring economic benefits to tourist attractions, parks, etc. For urban real estate, high-quality garden landscapes can increase the property value and promote sales. Good seedling management can also reduce the cost of pest control, extend the lifespan of seedlings, and save resources. Rich garden seedling species provide habitats and food sources for various organisms, promoting the development of biodiversity. For example, birds build nests and inhabit on trees, and insects feed on plants, forming a complete ecological chain and maintaining ecological balance.

[0003] The prior art document with the publication number CN119228069A provides a method for intelligent management of garden seedlings based on three-dimensional perception. The invention obtains the three-dimensional point cloud data of the garden based on binocular stereo vision technology, and combines image segmentation and clustering algorithms to analyze the distribution of cherry trees in the garden and obtain each cherry management area; by obtaining the environmental data and image data of each area, calculates the water demand coefficient of the cherry blossoms in the area according to the first cherry data; secondly, constructs a cherry blossom anomaly recognition model, uses a deep convolutional neural network to process the regional image data, identifies the abnormal characteristics of pests in the area and calculates the cherry blossom anomaly coefficient of the area; finally, according to the water demand coefficient and the cherry blossom anomaly coefficient of the area, comprehensively calculates the urgency coefficient of cherry blossom scheduling in the area, and then realizes intelligent resource allocation and scheduling according to the urgency coefficient of cherry blossom scheduling in the area. It can achieve intelligent scheduling of garden resources, thereby improving the management efficiency of the garden.

[0004] Although the existing technical solutions described above can achieve relevant beneficial effects through the structures of the existing technologies, they still have the following defects: 1. It is impossible to achieve all-round and multi-dimensional real-time monitoring, resulting in blind spots in understanding the garden meteorology and the growth status of nursery stock, missing some key information and being unable to timely grasp the overall situation within the garden. 2. It is difficult to accurately identify the growth stage, growth trend, and abnormal conditions such as pests and diseases of the nursery stock. There may be misjudgments or missed judgments, and it is impossible to provide a reliable scientific basis for garden maintenance, making the maintenance work lack pertinence and reducing the garden maintenance effect. 3. It is impossible to fully integrate multi-source data, and the evaluation of the growth environment and status of the nursery stock is not comprehensive enough. It is also difficult to conduct effective potential risk prediction based on historical and real-time data, resulting in the inability to make preparations in advance when facing emergencies and being prone to losses. 4. It is impossible to give maintenance suggestions based on scientific analysis and evaluation results, combined with expert experience and knowledge bases. The maintenance suggestions may only be simple summaries of experience, lacking detailed planning and guidance in multiple aspects such as irrigation, fertilization, and pruning. It is difficult for maintenance personnel to carry out scientific operations, affecting the healthy growth of garden nursery stock.

[0005] In view of this, we propose a smart management method and system for garden nursery stock based on intelligent maintenance. Summary of the Invention

[0006] 1. Technical problems to be solved

[0007] The purpose of this application is to provide a smart management method and system for garden nursery stock based on intelligent maintenance, which solves the technical problems raised in the above background technology, and realizes multi-dimensional real-time monitoring of garden meteorology and the growth status of nursery stock by comprehensively collecting high-definition images of garden nursery stock through a fixed-point monitoring module and a cruise monitoring module; accurately identifying abnormal conditions such as the growth stage, growth trend, and pests and diseases of the nursery stock through an image analysis module to provide a scientific basis for garden maintenance; establishing an evaluation model by combining multi-source data through a comprehensive evaluation module to comprehensively evaluate the growth environment and status of the nursery stock, and predicting potential risks based on historical and real-time data.

[0008] 2. Technical solutions.

[0009] The technical solution of this application provides a smart management system for garden nursery stock based on intelligent maintenance, including:

[0010] Data collection module: Collect geological data, building data, and nursery stock data of the garden; label the data as reference samples; standardize the collected data to provide high-quality reference samples for subsequent data analysis and model training.

[0011] Data acquisition module: Reasonably deploy a variety of sensors according to the geological data of the garden and the distribution data of nursery stock species to collect environmental data in real time.

[0012] Meteorological data acquisition module: It includes a weather instrument to collect meteorological data in real time, including meteorological data such as temperature, humidity, wind speed, wind direction, rainfall, and air pressure.

[0013] Fixed-point monitoring module: It includes multiple high-definition cameras. According to the geological data of the garden and the distribution data of nursery stock species, using Geographic Information System (GIS) technology for analysis, multiple high-definition cameras are reasonably arranged to conduct fixed-point monitoring of the nursery stock in the garden.

[0014] Route planning module: According to the geological data of the garden and the distribution data of nursery stock species, plan the optimal flight route of the drone.

[0015] Cruise monitoring module: It includes a drone and high-definition cameras. According to the planned route, conduct cruise monitoring of the nursery stock in the garden to comprehensively collect high-definition images.

[0016] Image preprocessing module: Preprocess the images collected by the fixed-point monitoring module and the spliced and fused images, including filtering and denoising, image enhancement, grayscale conversion, and normalization, etc.

[0017] Feature extraction module: Extract features from the preprocessed images. The extracted features include color, texture, and shape.

[0018] Image fusion module: Splice and fuse the images collected by the cruise monitoring module and the images collected in real time by the fixed-point monitoring module.

[0019] Image analysis module: Construct a ResNet model based on deep learning to identify and analyze the images after feature extraction, timely identify the growth stage and growth situation of the nursery stock; timely identify abnormal situations such as pests and diseases; identify the situations where the nursery stock needs pruning.

[0020] Comprehensive evaluation module: Combine the environmental data (temperature, humidity, soil nutrient content, light) and meteorological data collected in real time by the data acquisition module with the analysis results of the image analysis module, use data analysis methods such as multiple linear regression and principal component analysis (PCA) to establish a comprehensive evaluation model to comprehensively evaluate the growth environment and growth status of the nursery stock in the garden. Based on historical data and real-time monitoring data, use time series analysis and support vector machine (SVM) methods to predict potential risks.

[0021] Maintenance advice module: According to the analysis results of the image analysis module and the evaluation results of the comprehensive evaluation module, combined with expert experience and maintenance knowledge base, use rule reasoning and intelligent decision-making algorithms to give maintenance advice for the nursery stock in the garden. The maintenance advice covers aspects such as irrigation, fertilization, pruning, and cleaning up fallen leaves, providing detailed and operable guidance for maintenance personnel.

[0022] Alarm module: When an abnormal situation is detected, it issues an alarm in a timely manner; it includes an audible and visual alarm. While issuing an alarm in the control center, it notifies relevant management personnel of the alarm information in a timely manner through methods such as text messages and APP push, ensuring that abnormal situations are dealt with in a timely manner.

[0023] PLC control center: It is network-connected to the data collection module, data acquisition module, meteorological data acquisition module, fixed-point monitoring module, route planning module, cruise monitoring module, image preprocessing module, feature extraction module, image analysis module, comprehensive evaluation module, maintenance suggestion module, and alarm module.

[0024] Through the above technical solutions, the data acquisition module collects environmental data in real time; the meteorological data acquisition module collects meteorological data in real time, and the fixed-point monitoring module conducts fixed-point monitoring on the nursery stock in the garden; the route planning module plans the optimal flight route of the drone according to the geological data of the garden and the distribution data of the nursery stock types; the cruise monitoring module conducts cruise monitoring on the garden nursery stock according to the planned route, and comprehensively collects high-definition images; the image preprocessing module preprocesses the images collected by the fixed-point monitoring module and the spliced and fused images; the feature extraction module extracts features from the preprocessed images; the image fusion module splices and fuses the images collected by the cruise monitoring module and the images collected in real time by the fixed-point monitoring module; the image analysis module constructs a ResNet model based on deep learning, identifies and analyzes the images after feature extraction, and timely identifies the growth stage and growth condition of the nursery stock; timely identifies abnormal situations such as pests and diseases; identifies the situations where the nursery stock needs to be pruned; the comprehensive evaluation module combines the environmental data and meteorological data collected in real time by the data acquisition module with the analysis results of the image analysis module to comprehensively evaluate the growth environment and growth condition of the garden nursery stock. Predict potential risks. The maintenance suggestion module gives maintenance suggestions for the garden nursery stock according to the analysis results of the image analysis module and the evaluation results of the comprehensive evaluation module, combined with expert experience and the maintenance knowledge base, and using rule reasoning and intelligent decision-making algorithms.

[0025] Furthermore, the comprehensive evaluation module combines the environmental data and meteorological data collected in real time by the data acquisition module with the analysis results of the image analysis module to establish a comprehensive evaluation model, and comprehensively evaluates the growth environment and growth condition of the garden nursery stock. Based on historical data and real-time monitoring data, predicting potential risks includes the following steps:

[0026] 1. Data Sorting: Obtain real-time environmental data, including data such as temperature, humidity, soil nutrient content, and light intensity, which reflect the real-time environmental conditions for the growth of garden seedlings. Synchronously collect meteorological data, such as rainfall, wind speed, and air pressure, to understand the meteorological conditions in the area where the garden is located. Obtain analysis results such as the growth stage, growth trend, abnormal conditions such as pests and diseases, and whether pruning is required from the image analysis module. Classify and sort all the collected data according to time sequence and monitoring area to ensure the accuracy and integrity of the data for convenient subsequent analysis and use.

[0027] 2. Data Preprocessing: Check the integrity of the data. For missing data, adopt appropriate methods to fill it according to the data characteristics and actual situation. Perform normalization processing on the data to unify data of different magnitudes and units into a reasonable range, eliminate the dimensional differences between data, and make subsequent data analysis and model training more stable and accurate. Remove outliers from the data by setting a reasonable threshold range to identify and eliminate those data points that significantly deviate from the normal range, avoiding the large impact of outliers on the analysis results.

[0028] 3. Build a Comprehensive Evaluation Model:

[0029] 3.1. Use the multiple linear regression method. Take environmental data, meteorological data, and image analysis results as independent variables, and indicators of the growth status of seedlings (such as growth rate, health index, etc.) as dependent variables to establish a multiple linear regression model. By analyzing the linear relationship between independent variables and dependent variables, evaluate the influence degree of different factors on the growth status of seedlings.

[0030] 3.2. Adopt the principal component analysis (PCA) method to perform dimensionality reduction processing on a large amount of environmental data and meteorological data. Convert multiple related original variables into a few uncorrelated comprehensive variables (principal components).

[0031] 3.3. Combine the results of multiple linear regression and principal component analysis to build a comprehensive evaluation model to comprehensively evaluate the growth environment and growth status of garden seedlings. According to the output results of the model, score the growth environment of garden seedlings to judge whether the growth environment is suitable; classify the growth status to clarify whether the growth of seedlings is in a good, average, or poor state.

[0032] 4. Build a Risk Prediction Model:

[0033] 4.1. Collect historical data of garden seedlings, including past environmental data, meteorological data, occurrence of pests and diseases, and growth status of seedlings, etc.

[0034] 4.2. Adopt time series analysis method to analyze historical data and explore the laws and trends of data changes over time. For example, by analyzing the occurrence time and degree of pests and diseases in the past few years, predict the possibility and time nodes of future pests and diseases outbreaks.

[0035] 4.3. Use the support vector machine (SVM) method. Take environmental data, meteorological data, and historical pests and diseases occurrence data, etc. as input features, and the situation of pests and diseases outbreaks, the impact of extreme weather on seedlings, etc. as output labels to train the SVM model. Through the classification and prediction capabilities of the SVM model, evaluate and predict potential risks, such as predicting the probability of pests and diseases outbreaks under different meteorological conditions, and the degree of damage that extreme weather may cause to seedlings.

[0036] 5. Risk assessment: According to the results of the risk prediction model, evaluate the potential risks faced by landscape seedlings, and determine the level and scope of influence of the risks. For example, divide the possibility of pests and diseases outbreaks into three levels: high, medium, and low, and divide the degree of impact of extreme weather on seedlings into three levels: severe, moderate, and mild.

[0037] 6. Formulation of countermeasures: Develop corresponding countermeasures in advance for different levels of risks. For the high-risk possibility of pests and diseases outbreaks, formulate a detailed pest and disease control plan, including preparing control agents in advance, arranging professional personnel for regular inspections and control work; for areas that may be severely affected by extreme weather, take protective measures in advance, such as building windbreak sheds, setting up drainage systems, etc., to reduce the damage of risks to landscape seedlings.

[0038] 7. Model update and optimization: Regularly collect new environmental data, meteorological data, and image analysis results to update and optimize the comprehensive evaluation model and risk prediction model. As time goes by and data accumulates, the model can continuously learn new information and improve the accuracy of evaluation and prediction. According to actual landscape maintenance experience and feedback information, adjust the parameters and algorithms of the model to make the model more in line with the actual situation. For example, if it is found that there is a large deviation between the actual occurrence of pests and diseases and the model prediction, analyze the reasons and improve the risk prediction model, adjust the input features and model parameters to enhance the prediction ability of the model.

[0039] The present invention provides a smart management method for landscape seedlings based on intelligent maintenance, including the following steps:

[0040] S1. The data collection module collects the geological data, building data, and seedling data of the garden; annotate the data as reference samples; standardize the collected data. The data acquisition module reasonably arranges a variety of sensors according to the geological data of the garden and the seedling species distribution data to collect environmental data in real time.

[0041] S2. The meteorological data acquisition module collects meteorological data in real time, including meteorological data such as temperature, humidity, wind speed, wind direction, rainfall, and air pressure.

[0042] S3. The fixed-point monitoring module collects high-definition images of the seedlings in the garden at fixed points.

[0043] S4. The route planning module plans the optimal flight route of the drone according to the geological data of the garden and the distribution data of the seedling species.

[0044] S5. The cruise monitoring module conducts cruise monitoring on the garden seedlings according to the planned route and comprehensively collects high-definition images.

[0045] S6. The image fusion module stitches and fuses the images collected by the cruise monitoring module and the images collected by the fixed-point monitoring module in real time.

[0046] S7. The image preprocessing module preprocesses the images collected by the fixed-point monitoring module and the stitched and fused images, including filtering and denoising, image enhancement, grayscale conversion, and normalization, etc.; the feature extraction module extracts features from the preprocessed images, and the extracted features include color, texture, and shape.

[0047] S8. The image analysis module conducts recognition and analysis on the images after feature extraction, and promptly identifies the growth stage and growth condition of the seedlings; promptly identifies abnormal conditions such as pests and diseases; identifies the situation where the seedlings need to be pruned.

[0048] S9. The comprehensive evaluation module combines the environmental data and meteorological data collected in real time by the data acquisition module with the analysis results of the image analysis module to establish a comprehensive evaluation model, and comprehensively evaluates the growth environment and growth condition of the garden seedlings. Based on historical data and real-time monitoring data, time series analysis and support vector machine SVM methods are used to predict potential risks.

[0049] S10. The maintenance suggestion module gives maintenance suggestions for the garden seedlings according to the analysis results of the image analysis module and the evaluation results of the comprehensive evaluation module, combines expert experience and the maintenance knowledge base, and uses rule reasoning and intelligent decision-making algorithms. The maintenance suggestions cover aspects such as irrigation, fertilization, pruning, and cleaning up fallen leaves, providing detailed and operable guidance for the maintenance personnel.

[0050] S11. The alarm module: when an abnormal situation is detected, an alarm is issued in a timely manner.

[0051] 3. Beneficial effects

[0052] One or more technical solutions provided in the technical solution of this application have at least the following technical effects or advantages:

[0053] 1. The present invention comprehensively collects high-definition images of garden seedlings through a fixed-point monitoring module and a cruise monitoring module, realizing multi-dimensional real-time monitoring of garden meteorology and the growth status of seedlings.

[0054] 2. The present invention can accurately identify abnormal situations such as the growth stage, growth trend, and pests and diseases of seedlings through an image analysis module, providing a scientific basis for garden maintenance.

[0055] 3. The present invention establishes an evaluation model by combining multi-source data through a comprehensive evaluation module, comprehensively evaluates the growth environment and status of seedlings, predicts potential risks based on historical and real-time data, and makes preparations in advance.

[0056] 4. The present invention gives detailed maintenance suggestions covering multiple aspects such as irrigation, fertilization, and pruning based on the analysis and evaluation results through a maintenance suggestion module, combining expert experience and knowledge base, and guiding maintenance personnel to operate scientifically. BRIEF DESCRIPTION OF THE DRAWINGS

[0057] Figure 1 It is a schematic flow chart of a smart management method for garden seedlings based on intelligent maintenance disclosed in a preferred embodiment of the present application. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0058] The following further describes the present application in detail with reference to the accompanying drawings of the specification.

[0059] Refer to Figure 1 , an embodiment of the present application provides a smart management system for garden seedlings based on intelligent maintenance, including:

[0060] Data collection module: Collect geological data of the garden (including soil type, pH value, compactness, shape boundary, etc.), building data (such as location and structure information of pavilions, terraces, and roads), and seedling data (seedling variety, quantity, planting time, initial growth status, etc.); label the data as reference samples; standardize the collected data to provide high-quality reference samples for subsequent data analysis and model training.

[0061] Data acquisition module: Reasonably deploy a variety of sensors (temperature sensors, humidity sensors, soil nutrient content sensors, light sensors) according to the geological data of the garden and the seedling species distribution data, and collect environmental data in real time; reasonably deploy sensors in different regions according to the growth needs of seedlings and the key points of environmental monitoring.

[0062] Meteorological data acquisition module: Including a weather station, which collects meteorological data in real time, including meteorological data such as temperature, humidity, wind speed, wind direction, rainfall, and air pressure.

[0063] Fixed-point monitoring module: It includes multiple high-definition cameras. According to the geological data of the garden and the distribution data of seedling species, using Geographic Information System (GIS) technology for analysis, multiple high-definition cameras are reasonably arranged to conduct fixed-point monitoring of the seedlings in the garden.

[0064] Route planning module: According to the geological data of the garden and the distribution data of seedling species, plan the optimal flight route of the drone; ensure that the drone can comprehensively cover the garden area during flight and avoid obstacles, and efficiently complete the cruise monitoring task.

[0065] Cruise monitoring module: It includes a drone and high-definition cameras. According to the planned route, conduct cruise monitoring of the garden seedlings and comprehensively collect high-definition images.

[0066] Image preprocessing module: Preprocess the images collected by the fixed-point monitoring module and the spliced and fused images, including filtering and denoising, image enhancement, grayscale conversion, and normalization, etc.

[0067] Feature extraction module: Extract features from the preprocessed images. The extracted features include color, texture, and shape.

[0068] Color feature extraction: Use methods such as color histogram and color moment to extract the color features of the seedlings in the image, such as the green degree of the leaves and the color distribution of the flowers, etc., for judging the growth status of the seedlings and variety identification.

[0069] Texture feature extraction: Adopt algorithms such as Gray-Level Co-Occurrence Matrix (GLCM) and Local Binary Pattern (LBP) to extract the texture features of the seedlings, such as the thickness of the leaf texture and the texture direction of the branches, etc., to assist in judging whether there are abnormal situations such as pests and diseases and nutrient deficiencies in the seedlings.

[0070] Shape feature extraction: Through technologies such as edge detection and contour extraction, extract the shape features of the seedlings, such as the shape of the tree crown and the curvature of the branches, etc., for analyzing the growth form and growth trend of the seedlings.

[0071] Image fusion module: Splice and fuse the images collected by the cruise monitoring module and the images collected in real time by the fixed-point monitoring module.

[0072] Image analysis module: Build a ResNet model based on deep learning to identify and analyze the images after feature extraction, and timely identify the growth stage (seedling stage, growth stage, flowering stage, fruiting stage, etc.) and growth trend (vigorous, general, weak) of the seedlings; timely identify abnormal situations such as pests and diseases (nutrient deficiency, water anomaly, environmental pollution, and human damage, etc.); identify the situation where the seedlings need to be pruned.

[0073] Comprehensive Evaluation Module: Combine the environmental data (temperature, humidity, soil nutrient content, light) and meteorological data collected in real time by the data collection module with the analysis results of the image analysis module, and use data analysis methods such as multiple linear regression and principal component analysis (PCA) to establish a comprehensive evaluation model to comprehensively evaluate the growth environment and growth status of garden seedlings. Based on historical data and real-time monitoring data, use time series analysis and support vector machine (SVM) methods to predict potential risks, such as the possibility of pest and disease outbreaks and the impact of extreme weather on seedlings, and formulate countermeasures in advance.

[0074] Maintenance Suggestion Module: According to the analysis results of the image analysis module and the evaluation results of the comprehensive evaluation module, combined with expert experience and the maintenance knowledge base, use rule reasoning and intelligent decision-making algorithms to give maintenance suggestions for garden seedlings. The maintenance suggestions cover aspects such as irrigation (irrigation time, water volume), fertilization (fertilizer type, fertilization amount, fertilization time), pruning (pruning part, degree, time), and leaf cleaning (cleaning frequency, method), providing detailed and operable guidance for maintenance personnel.

[0075] Alarm Module: When abnormal situations are detected, send out alarms in a timely manner; including audible and visual alarms. While sending alarms in the control center, notify relevant management personnel of the alarm information in a timely manner through methods such as text messages and APP push to ensure that abnormal situations are handled in a timely manner.

[0076] PLC Control Center: Connect to the data collection module, data acquisition module, meteorological data acquisition module, fixed-point monitoring module, route planning module, cruise monitoring module, image preprocessing module, feature extraction module, image analysis module, comprehensive evaluation module, maintenance suggestion module, and alarm module through a network.

[0077] In this technical solution, the data acquisition module collects environmental data in real time; the meteorological data acquisition module collects meteorological data in real time, and the fixed-point monitoring module conducts fixed-point monitoring on the seedlings in the garden; the route planning module plans the optimal flight route of the drone according to the geological data of the garden and the distribution data of the seedling species; the cruise monitoring module conducts cruise monitoring on the garden seedlings according to the planned route, and comprehensively collects high-definition images; the image preprocessing module preprocesses the images collected by the fixed-point monitoring module and the stitched and fused images; the feature extraction module extracts features from the preprocessed images; the image fusion module stitches and fuses the images collected by the cruise monitoring module and the images collected in real time by the fixed-point monitoring module; the image analysis module constructs a ResNet model based on deep learning, identifies and analyzes the images after feature extraction, and promptly identifies the growth stage and growth condition of the seedlings; promptly identifies abnormal conditions such as pests and diseases; identifies the situation where the seedlings need to be pruned; the comprehensive evaluation module combines the environmental data and meteorological data collected in real time by the data acquisition module with the analysis results of the image analysis module, uses data analysis methods such as multiple linear regression and principal component analysis (PCA) to establish a comprehensive evaluation model, and comprehensively evaluates the growth environment and growth condition of the garden seedlings. Based on historical data and real-time monitoring data, time series analysis and support vector machine (SVM) methods are used to predict potential risks. The maintenance advice module gives maintenance advice for the garden seedlings according to the analysis results of the image analysis module and the evaluation results of the comprehensive evaluation module, combines expert experience and the maintenance knowledge base, and uses rule reasoning and intelligent decision-making algorithms. When an abnormal situation is detected, the alarm module issues an alarm in a timely manner.

[0078] Furthermore, the data acquisition module reasonably arranges a variety of sensors according to the geological data of the garden and the distribution data of the seedling species, and collects environmental data in real time; according to the growth requirements of the seedlings and the key points of environmental monitoring, sensors are reasonably arranged in different regions; including the following steps:

[0079] 1. Data collection and analysis: Comprehensively collect the geological data of the garden, including information such as soil type, acidity and alkalinity, topography and landform, etc.; at the same time, collect the distribution data of the seedling species, and clarify the planting areas and quantities of different seedling varieties. Use geographic information system (GIS) technology to analyze these data, draw maps of the garden geology and seedling distribution, and provide an intuitive basis for subsequent sensor layout. According to the growth requirements of the seedlings and the key points of environmental monitoring, select sensors of appropriate types and accuracies.

[0080] 2. Determine key monitoring areas: Divide key monitoring areas according to the functional zoning of the garden and the growth characteristics of the seedlings. In areas with dense seedling planting, due to the large number of seedlings and intense growth competition, they are more sensitive to environmental changes and need to be monitored closely; near water sources, soil moisture and nutrients are easily affected by the water source, which is also a key monitoring area; for rare seedling planting areas, due to their high value and strict growth conditions, environmental monitoring needs to be strengthened. For example, in the flower planting area of the garden, since flowers have relatively strict requirements for light and temperature, this area can be set as a key monitoring area to densify the sensor layout.

[0081] 3. Sensor layout design: In key monitoring areas, make a reasonable layout according to the monitoring ranges and effective distances of different sensors. At the same time, consider the mutual interference between sensors to avoid the distance between different types of sensors being too close, which may affect the data accuracy.

[0082] 4. Sensor installation and debugging: Carry out the installation work of the sensors according to the layout design plan. After installation, debug all sensors to check their working status and whether the data transmission is normal, and calibrate the sensors through calibration equipment to ensure the accuracy of the collected data.

[0083] 5. Data collection and transmission settings: Set a reasonable data collection frequency. Adopt wireless transmission technologies such as LoRa, NB-IoT, etc. to transmit the data collected by the sensors to the data aggregation node in real time, and then transmit the data to the PLC control center through the data aggregation node. Set up a data caching mechanism on the sensors and the data aggregation node. When the network fails, temporarily store the data and automatically resend it after the network resumes to ensure the integrity and continuity of the data.

[0084] 6. Real-time monitoring and maintenance: After the system runs, monitor the working status of the sensors and the quality of the collected data in real time. Regularly maintain the sensors, check whether the sensors are damaged, aged, etc., and replace the faulty sensors in time; clean the dust and debris on the surface of the sensors to ensure the normal operation of the sensors. At the same time, optimize and adjust the sensor layout and collection parameters according to the seasonal changes of the garden, the adjustment of the seedling growth stage, and new monitoring requirements to ensure that the data collection module can always provide accurate and effective environmental data for the maintenance of garden seedlings.

[0085] Furthermore, the route planning module plans the optimal flight route of the drone according to the geological data of the garden and the data on the distribution of seedling species; it includes the following steps:

[0086] 1. Data collection and integration: Comprehensively collect the geological data of the garden, including topographical and geomorphic information such as the existence of mountains, hills, lakes, etc.; building distribution data covering the locations and structures of pavilions, terraces, towers, management buildings, roads, bridges, etc.; and the layouts of various infrastructure such as power lines and communication towers. Summarize the distribution data of nursery stock species, and clarify the planting areas, areas, and densities of different varieties of nursery stock.

[0087] 2. Obstacle and key area marking: Based on the collected data, accurately mark all obstacles on the Geographic Information System (GIS) map, including physical buildings, water bodies, tall trees, overhead lines, etc. Mark the areas in the garden that need to be monitored key points, such as rare nursery stock planting areas, newly planted nursery stock areas, and areas with high incidence of pests and diseases.

[0088] 3. Preliminary route planning: Select the Dijkstra path planning algorithm, with the garden boundary and key monitoring areas as constraints, and set the starting and ending points of the drone. The algorithm preliminarily plans multiple possible flight routes based on the terrain of the garden, obstacle distribution, and nursery stock areas, ensuring that the routes can cover all key monitoring areas and most of the garden area.

[0089] 4. Route optimization and adjustment: Optimize the preliminarily planned routes, considering the drone's battery life, flight speed, and image acquisition requirements. Ensure that the routes are within the allowable battery range of the drone and the flight speed is suitable for high-definition image acquisition. Further avoid obstacles, adjust the routes close to obstacles, and increase the safety distance to prevent the drone from colliding during flight. Combine the actual situation of the garden, such as wind direction and lighting conditions, to optimize the flight direction, so that the drone can obtain better shooting angles and lighting conditions during flight, improving the quality of image acquisition.

[0090] 5. Real-time dynamic adjustment: During the drone flight, use real-time monitored meteorological data, personnel activities in the garden, and emergencies (such as temporarily built facilities, unexpectedly appearing obstacles, etc.) to dynamically adjust the flight route. When encountering bad weather such as strong winds and heavy rains, automatically adjust the route to find a relatively safe flight area or return to the take-off point. If it is detected that there are personnel activities in the garden that may affect the flight safety of the drone, change the route in time to avoid crowded areas.

[0091] 6. Route Verification and Update: After each UAV completes the cruise monitoring task, analyze and evaluate the actual flight route and the collected data. Check for any missing monitoring areas and verify the safety and efficiency of the route. Based on the evaluation results, update and optimize the route planning model to provide a more reasonable route for the next flight. As the garden develops and changes (such as new buildings, adjustment of nursery stock planting areas, etc.), update the geological and nursery stock distribution data in a timely manner and re-plan the flight route to ensure that the UAV can always efficiently complete the cruise monitoring task.

[0092] Furthermore, the image fusion module stitches and fuses the images collected by the cruise monitoring module and the images collected in real time by the fixed-point monitoring module; it includes the following steps:

[0093] 1. Image Classification: From the storage device of the cruise monitoring module, extract high-definition panoramic cruise images in chronological order and monitoring area numbers; at the same time, obtain close-up images of different fixed-point positions within the same time period from the corresponding storage path of the fixed-point monitoring module.

[0094] Classify and store the images in different folders according to the shooting time of the images accurate to seconds, the longitude and latitude information of the shooting location, and the garden sub-region to which they belong, for convenient subsequent quick retrieval and call.

[0095] 2. Image Preprocessing: It includes denoising processing, image enhancement, and grayscale processing.

[0096] Denoising Processing: Use professional image processing software (such as the OpenCV library) to perform Gaussian filtering operations on each image. Set the Gaussian kernel size to 3×3 or 5×5 in the software, and adjust the standard deviation between 0.5 - 1.5 according to the image noise situation to remove random noise points in the image and make the image smoother.

[0097] Image Enhancement: Through the histogram adjustment function of image editing tools (such as Adobe Photoshop), manually stretch the grayscale histogram of the image to enhance the overall contrast of the image.

[0098] Grayscale Processing: If it is a color image, in a programming environment (such as the PIL library in Python), convert the RGB color space to the grayscale space through an algorithm. Select the weighted average method, and perform weighted summation on the three RGB channels according to the weights of 0.299, 0.587, and 0.114 to obtain a single grayscale value.

[0099] 3. Feature Extraction: With the help of a dedicated SIFT algorithm implementation library (such as the SIFT module in OpenCV), feature extraction is performed on the preprocessed image. Set the number of groups of the Gaussian pyramid to 4 - 6 groups, and the number of image layers within each group to 3 - 5 layers to detect the extreme points of the image at different scales. For each extreme point, calculate its main direction and a 128 - dimensional feature descriptor, which contains the gradient direction and amplitude information within the neighborhood of this point. Use the SURF module in OpenCV, set the threshold of the Hessian matrix between 300 - 500, and adjust it according to the complexity of the image. Use the integral image to quickly calculate the determinant of the Hessian matrix and detect the feature points in the image. Based on the Haar wavelet response, calculate the direction and a 64 - dimensional feature descriptor of the feature points to obtain the stable features of the image.

[0100] 4. Feature Matching: Adopt a data structure based on kd - tree, call the KDTree class in the scikit - learn library of Python to construct an index of feature points. Calculate the Euclidean distance between the feature descriptors of two images, set a distance threshold, and regard the feature point pairs with a distance less than the threshold as matching points to initially screen out possible matching relationships. For the initially matched point pairs, further eliminate the incorrect matching points through cross - validation.

[0101] 5. Image Stitching and Fusion:

[0102] 5.1. Homography Matrix Calculation: Use the OpenCV library in Python to call the RANSAC algorithm function ransacReprojThreshold. According to the matched feature point pairs, set the maximum number of iterations to 500 - 1000 times, and the reprojection error threshold to be between 3 - 5 pixels. Calculate the homography matrix through random sampling, remove the incorrect matching points, and determine the projective transformation relationship between the two images.

[0103] 5.2. Image Stitching: Based on the calculated homography matrix, use the warpPerspective function in the OpenCV library to project the fixed - point monitoring image into the coordinate system of the cruise monitoring image. Set the size of the target image, adjust it according to the size of the cruise monitoring image and the range of projective transformation to ensure that the stitched image is complete and distortion - free.

[0104] 5.3. Fusion Processing: For the overlapping area after stitching, adopt a weighted average fusion algorithm. According to the distance of the pixel points to the edges of the two images, write code in Python to implement weight assignment. The closer a pixel is to the edge of an image, the higher the weight of the corresponding pixel in that image, making the transition in the overlapping area more natural. Perform image fusion processing according to the following formula: I(x,y)=w1(x,y)I1(x,y)+w2(x,y)I2(x,y).

[0105] w1(x,y) = a * d1 / (d1 + d2) + (1 - a) * [I1 - / (I1 - + I2 - )].

[0106] W2(x,y) = (1 - a) * d2 / (d1 + d2) + a * [I2 - / (I1 - + I2 - )]; w1(x,y) + w2(x,y) = 1; Wherein, I(x,y) is the pixel value of the fused image at the coordinate (x,y), which is the final result calculated by the weighted average fusion algorithm, comprehensively reflecting the information of the cruise monitoring image and the fixed-point monitoring image at this position. I1(x,y) refers to the pixel value of the cruise monitoring image at the coordinate (x,y), reflecting the image information at this position from the perspective of cruise monitoring. I2(x,y) refers to the pixel value of the fixed-point monitoring image at the coordinate (x,y), representing the image information at this position from the perspective of fixed-point monitoring. w1(x,y) and w2(x,y) are the weights of the cruise monitoring image I1(x,y) and the fixed-point monitoring image I2(x,y) respectively when calculating the fused pixel value. They are functions of the pixel coordinates (x,y), indicating the contribution degree of the two images to the fused pixel value at this pixel position. a is the visual importance parameter, and its value range is from 0 to 1. d1 is the distance from the pixel (x,y) to the edge of the cruise monitoring image. d2 is the distance from the pixel (x,y) to the edge of the fixed-point monitoring image. I1 is the gray mean value of the neighborhood centered on the pixel (x,y) (such as set as a 3×3 neighborhood) in the cruise monitoring image, reflecting the average gray level of the cruise monitoring image in this neighborhood. I2 is the gray mean value of the neighborhood centered on the pixel (x,y) (such as set as a 3×3 neighborhood) in the fixed-point monitoring image, reflecting the average gray level of the fixed-point monitoring image in this neighborhood.

[0107] 6. Fusion image quality assessment:

[0108] 6.1 Peak signal-to-noise ratio (PSNR): Write a function in Python. According to the calculation formula of PSNR, calculate the mean square error (MSE) between the original images (cruise monitoring image and fixed-point monitoring image) and the fused image, and then substitute the MSE into the PSNR formula to obtain the PSNR value. Set the threshold of PSNR. If it is lower than the threshold, it is considered that the image quality does not meet the standard.

[0109] 6.2 Structural Similarity Index (SSIM): Use the compare_ssim function in the scikit-image library of Python to calculate the similarity between the fused image and the original image in terms of brightness, contrast, and structure, and obtain the SSIM value. Set the SSIM threshold between 0.8 - 0.9. If it is lower than this threshold, adjust the fusion parameters (such as the weight distribution of weighted average fusion) or reselect the feature extraction and matching algorithm (such as changing to the ORB algorithm), and perform image fusion again until the quality requirements are met.

[0110] Furthermore, the image analysis module constructs a ResNet model based on deep learning to identify and analyze the images after feature extraction, and promptly identify the growth stage and growth trend of the seedlings; promptly identify abnormal situations such as pests and diseases; identify the situations where the seedlings need to be pruned; including the following steps:

[0111] 1. Data collection and annotation: Obtain a large number of images of landscape seedlings with extracted features from the image acquisition module, covering seedlings in different seasons, different growth stages, and different health conditions. Organize professional personnel to carefully annotate the images.

[0112] 2. Data preprocessing: Perform normalization processing and data augmentation operations on the collected images; expand the dataset by rotating, flipping, scaling, cropping, etc., to increase the diversity of the data and improve the generalization ability of the model.

[0113] 3. Construct the ResNet model: Import a deep learning framework (such as TensorFlow or PyTorch), and use the relevant functions and classes in the framework to construct the ResNet model. Define the input layer of the model, and set the input shape according to the number of channels of the image (such as the number of RGB image channels is 3) and the size (such as 224×224). Add each module of ResNet in turn, including convolutional layers, batch normalization layers, activation function layers, residual blocks, etc. In the convolutional layer, reasonably set parameters such as the convolutional kernel size, stride, and padding to extract different features of the image; the batch normalization layer is used to accelerate model training and improve stability; the activation function (such as the ReLU function) increases the non-linear expression ability of the model; the residual block is the core structure of ResNet, which solves the gradient vanishing problem of deep neural networks through skip connections, enabling the model to learn more complex features. Define the output layer of the model, and set the number of output nodes according to the number of categories to be recognized. Select the cross-entropy loss function. Select the Adam optimizer; set evaluation metrics such as accuracy, recall, F1 value, etc., to evaluate the performance of the model during training and validation.

[0114] The loss function is: L = -Σ N i＝1 λ iΣ M j＝1 [w j *y ij *log(y ij ^)]; In the formula, L represents the value of the loss function, which is used to measure the difference between the model prediction result and the true label. During the model training process, by continuously adjusting the model parameters, the value of is minimized, thereby improving the prediction accuracy of the model. N is the number of samples, that is, the total number of image samples participating in the model training or evaluation. Each sample contains image information about garden nursery stock and the corresponding true label. λ i is the importance weight of the sample, corresponding to the i-th sample. It reflects the relative importance of this sample in the model training, and the value range is usually greater than 0. For example, for key samples containing rare nursery stock or serious pests and diseases, λ i can be set to a larger value, so that the model pays more attention to these samples during training, and avoids insufficient learning of key samples by the model due to sample imbalance. M is the number of categories. In the garden nursery stock image analysis task, it covers all classification categories such as the growth stage of the nursery stock (such as seedling stage, growth stage, etc.), growth vigor (vigorous, general, weak), abnormal situations such as pests and diseases (nutrient deficiency, water anomaly, etc.), and whether pruning is required. w j is the feature complexity weight, corresponding to the j-th category. It can be determined by calculating indicators such as the variance and entropy of the image features of this category. y ij represents the value of the true label of the i-th sample on the j-th category, using the one-hot encoding form. If the i-th sample belongs to the j-th category, then y ij = 1; otherwise y ij = 0. y ij ^ represents the prediction probability of the model for the i-th sample on the j-th category, and the value range is between 0 and 1. It reflects the likelihood that the model thinks this sample belongs to the -th category.

[0115] 4. Model Training: Divide the preprocessed data into a training set, a validation set, and a test set. Use the training set to train the model. In each round of training, the model will traverse all batches of data in the training set, calculate the error between the predicted value and the true value according to the loss function, and update the model parameters through backpropagation of the optimizer. Use the validation set to validate the model and monitor the performance metrics of the model.

[0116] 5. Model Optimization: Adjust the hyperparameters of the model, such as learning rate, regularization coefficient, Dropout rate, etc. Use methods like cross-validation to find the optimal combination of hyperparameters and improve the performance of the model. Adopt the method of transfer learning, use the pre-trained ResNet model on large-scale image datasets (such as ImageNet), and then fine-tune it on the garden nursery stock image dataset to accelerate the convergence speed of the model and improve the accuracy of the model. Perform pruning operations on the model to remove unimportant connections and parameters in the model, reduce the size of the model, improve the inference speed of the model, and at the same time keep the performance of the model basically unchanged.

[0117] 6. Model Evaluation: Use the test set to comprehensively evaluate the optimized model, calculate indicators such as accuracy, recall rate, F1 value, etc. of the model in identifying abnormal situations such as the growth stage, growth condition, pests and diseases of nursery stocks, and the pruning situation of nursery stocks, and evaluate the performance of the model. Draw a confusion matrix to visually display the prediction situation of the model in each category, analyze the error types and reasons of the model, and further understand the model performance.

[0118] 7. Image Recognition and Analysis: Input the garden nursery stock images after real-time acquisition and feature extraction into the trained model. The model performs recognition and analysis on the images according to the learned features and classification rules, and outputs the growth stage, growth condition, whether there are abnormal situations such as pests and diseases, and whether pruning is required of the nursery stocks. Feed the recognition results back to the comprehensive evaluation module and the maintenance advice module to provide a decision-making basis for the maintenance management of garden nursery stocks.

[0119] Furthermore, the comprehensive evaluation module combines the environmental data and meteorological data collected in real time by the data collection module with the analysis results of the image analysis module to establish a comprehensive evaluation model to comprehensively evaluate the growth environment and growth status of garden nursery stocks. Based on historical data and real-time monitoring data, predict potential risks, including the following steps:

[0120] 1. Data Sorting: Obtain real-time environmental data, including data such as temperature, humidity, soil nutrient content, light intensity, etc., which reflect the real-time environmental conditions for the growth of garden nursery stocks. Synchronously collect meteorological data, such as rainfall, wind speed, air pressure, etc., to understand the meteorological conditions in the area where the garden is located. Obtain the analysis results of the growth stage, growth condition, pests and diseases and other abnormal situations of nursery stocks and whether pruning is required from the image analysis module. Sort all the collected data according to time sequence and monitoring area to ensure the accuracy and integrity of the data for convenient subsequent analysis and use.

[0121] 2. Data preprocessing: Check the integrity of the data. For missing data, appropriate methods are used for filling according to the data characteristics and actual situation. Normalize the data to unify data of different magnitudes and units into a reasonable range, eliminate the dimensional differences between data, and make subsequent data analysis and model training more stable and accurate. Remove outliers from the data. By setting a reasonable threshold range, identify and eliminate those data points that significantly deviate from the normal range to avoid the large impact of outliers on the analysis results.

[0122] 3. Build a comprehensive evaluation model;

[0123] 3.1. Use the multiple linear regression method. Take environmental data, meteorological data, and image analysis results as independent variables, and the growth status indicators of nursery stock (such as growth rate, health index, etc.) as the dependent variable to establish a multiple linear regression model. By analyzing the linear relationship between the independent variables and the dependent variable, evaluate the influence degree of different factors on the growth status of nursery stock. The multiple linear regression model is:

[0124] y t =β0+Σ n j＝1 [a ij (t,s)β j x jt +γ it e t .

[0125] a ij (t,s)=I ij (t,s) / {Σ n j＝1 [I ij (t,s)]}; γ it =1+ζ it ; In the formula, y t represents the growth status indicator of landscape nursery stock at time t, which is the dependent variable we hope to predict and evaluate through the model. For example, it can be a quantitative indicator that can reflect the growth status such as the growth height and health index of the nursery stock. β0 is the intercept term, which represents the value of y jt when all independent variables x t are 0, representing the basic state of nursery stock growth or the part that is not affected by the considered independent variables. n is the number of independent variables, covering various index quantities of environmental data (such as temperature, humidity, soil nutrient content, etc.), meteorological data (such as rainfall, wind speed, air pressure, etc.), and image analysis results (such as nursery stock growth stage, pest and disease situation, etc.). a ij (t,s) is the adaptive weight, related to the independent variable x jis related to time t and the seedling growth stage s. It reflects the relative importance of the j-th independent variable on the growth status of seedlings at a specific growth stage s and time t. This weight changes dynamically with time and growth stage and is updated regularly through the Analytic Hierarchy Process (AHP) combined with expert experience. β j is the regression coefficient, representing the j-th independent variable x j 's impact on the dependent variable y t . x jt represents the j-th independent variable at time t, that is, various environmental data, meteorological data, or image analysis results. γ it is the dynamic error correction factor, related to sample i and time t. It is determined based on the trend of the prediction error of the model over a past period and is used to correct the error of the model at different times. e t is the error term, representing the random part in the model that cannot be explained by the independent variables, including measurement errors, unconsidered influencing factors, etc. I ij (t, s) is the importance score of the independent variable x j at the growth stage s and time t. It is obtained through the Analytic Hierarchy Process (AHP) combined with expert experience analysis and is used to quantify the importance of this independent variable on the growth status of seedlings at this time. ζ it is the error trend coefficient obtained through linear regression analysis, reflecting the change trend of the prediction error of the model over a past period and is used to adjust the dynamic error correction factor γ it .

[0126] 3.2. Use the principal component analysis (PCA) method to perform dimensionality reduction on a large amount of environmental data and meteorological data. Convert multiple related original variables into a few uncorrelated comprehensive variables (principal components), which can retain the main information of the original data while reducing the complexity of the data, facilitating subsequent analysis and modeling.

[0127] 3.3. Combine the results of multiple linear regression and principal component analysis to construct a comprehensive evaluation model to comprehensively evaluate the growth environment and growth status of landscape seedlings. According to the output results of the model, score the growth environment of landscape seedlings to determine whether the growth environment is suitable; classify the growth status to clarify whether the growth of seedlings is in a good, average, or poor state.

[0128] 4. Construct a risk prediction model:

[0129] 4.1. Collect historical data of landscape seedlings, including past environmental data, meteorological data, occurrence of pests and diseases, and the growth status of seedlings, etc.

[0130] 4.2. Use the time series analysis method to analyze historical data and explore the laws and trends of data changes over time. For example, by analyzing the occurrence time and degree of pests and diseases in the past few years, predict the likelihood and time nodes of future pest and disease outbreaks. The risk prediction model is:

[0131] λ s (t)φ(B)▽ d y t =θ(B)τ t +I t ;I t =Σ K k＝1 (w k I tk );In the formula, λ s (t) is the growth stage adjustment factor, which is related to the growth stage of the seedlings at a certain time. It is a coefficient with different values according to different growth stages of the seedlings, used to adjust the parameter performance of the model in different growth stages. For example, in the seedling stage, growth stage, flowering stage, and fruiting stage, different values will be determined through the analysis of historical data and expert experience. For example, it is set to in the seedling stage and in the growth stage, etc. The purpose is to make the model more conform to the time series characteristics of different growth stages. φ(B) = 1 - φ1B - φ2B 2 -...-φ p B p ;B is the backward shift operator, φ i is the autoregressive coefficient, and p is the autoregressive order. The autoregressive part reflects the linear relationship between the time series y t and its past values, and the model is constructed through these coefficients and shift operators. ▽ d is the difference operator, and d is the difference order; θ(B) is the moving average part, θ(B) = 1 + θ1B + θ2B 2 +...+θ q B q ;θ i is the moving average coefficient, and q is the moving average order. The moving average part is used to describe the random interference term in the time series, and the random fluctuations are modeled through these coefficients and shift operators. τ t is the white noise sequence, representing the random error part that cannot be explained in the model, reflecting the unpredictable factors in the time series. I t is the external intervention factor, which comprehensively considers the impacts of various artificial measures in garden maintenance on the growth of seedlings. k is used to distinguish the serial numbers of different intervention factors, and K is the total number of intervention factors; the intervention factors include the amount of fertilizer to be applied, the amount of irrigation, pest and disease control measures, etc. w k is the weight of each intervention factor, and its value is determined through expert evaluation or experiments. This weight reflects the relative importance of each intervention factor in affecting the growth of seedlings. Itk It is the quantified value of specific intervention factors.

[0132] 4.3. Use the support vector machine (SVM) method. Take environmental data, meteorological data, and historical pest and disease occurrence data, etc. as input features, and the pest and disease outbreak situation, the impact of extreme weather on seedlings, etc. as output labels to train the SVM model. Through the classification and prediction capabilities of the SVM model, evaluate and predict potential risks, such as predicting the probability of pest and disease outbreaks under different meteorological conditions, and the degree of damage that extreme weather may cause to seedlings.

[0133] 5. Risk assessment: According to the results of the risk prediction model, evaluate the potential risks faced by landscape seedlings, and determine the level and scope of influence of the risks. For example, divide the possibility of pest and disease outbreaks into three levels: high, medium, and low, and divide the degree of impact of extreme weather on seedlings into three levels: severe, moderate, and mild.

[0134] 6. Formulate countermeasures: For different levels of risks, formulate corresponding countermeasures in advance. For the high-risk possibility of pest and disease outbreaks, formulate a detailed pest and disease control plan, including preparing control agents in advance, arranging professional personnel for regular inspections and control work; for areas that may be severely affected by extreme weather, take protective measures in advance, such as building windbreak sheds, setting up drainage systems, etc., to reduce the damage of risks to landscape seedlings.

[0135] 7. Model update and optimization: Regularly collect new environmental data, meteorological data, and image analysis results to update and optimize the comprehensive evaluation model and the risk prediction model. As time goes by and data accumulates, the model can continuously learn new information and improve the accuracy of evaluation and prediction. According to actual landscape maintenance experience and feedback information, adjust the parameters and algorithms of the model to make the model more in line with the actual situation. For example, if it is found that there is a large deviation between the actual pest and disease occurrence situation and the model prediction, analyze the reasons and improve the risk prediction model, adjust the input features and model parameters to enhance the prediction ability of the model.

[0136] Furthermore, the maintenance suggestion module, based on the analysis results of the image analysis module and the evaluation results of the comprehensive evaluation module, combines expert experience and the maintenance knowledge base, and uses rule reasoning and intelligent decision-making algorithms to give maintenance suggestions for landscape seedlings, including the following steps:

[0137] 1. Data collection and integration: Obtain the analysis results of the growth stage, growth condition, abnormal conditions such as pests and diseases, and whether pruning is required of the nursery stock from the image analysis module. Obtain the evaluation results of the growth environment of landscape nursery stock from the comprehensive evaluation module, including environmental data such as temperature, humidity, soil nutrient content, light, and the analysis of the impact of meteorological data on the growth of nursery stock. Organize the expert experience database, and collect the maintenance suggestions and treatment methods accumulated by experts in long-term landscape maintenance work for different growth conditions and environmental conditions. Extract relevant knowledge from the maintenance knowledge base, covering the biological characteristics of various landscape nursery stock, suitable growth environment conditions, and common maintenance measures and standards.

[0138] 2. Rule base formulation: Formulate irrigation rules based on the collected data and knowledge. Formulate fertilization rules, and determine the fertilizer type, fertilization amount, and fertilization time according to the growth stage of the nursery stock and the soil nutrient content. Formulate pruning rules, and determine the pruning part, degree, and time according to the growth stage of the nursery stock, tree shape requirements, and the presence of pests and diseases. Formulate rules for cleaning fallen leaves, and determine the cleaning frequency and method according to the season and the growth of the nursery stock.

[0139] 3. Application of intelligent decision-making algorithms: Apply rule inference algorithms to match the results of image analysis and comprehensive evaluation with the formulated rule base. For example, when it is detected that a certain nursery stock is in the flowering period and the phosphorus element in the soil nutrient content is low, according to the fertilization rule, it is inferred that high-phosphorus fertilizer should be applied and the time.

[0140] Combine intelligent decision-making algorithms, such as decision tree algorithms, neural network algorithms, etc., to conduct comprehensive analysis and decision-making on complex situations. For nursery stock affected by multiple growth problems and environmental factors at the same time, through the decision tree algorithm, according to the priorities and weights of different factors, gradually analyze and obtain the optimal maintenance suggestions. In the process of constructing the decision tree, use the information gain IG to select the optimal feature for splitting. The information gain calculation formula is:

[0141] IG(D, A) = u i H(D) - Σ V v＝1 {|D v | / |d| * [u i H i (D v )] + λ j H j (D v )}, where IG(D, A) is the information gain, which represents the information gain value calculated based on the data set and features considering the weights of nursery stock growth characteristics and the influence coefficients of environmental factors, and is used to select the optimal splitting feature when constructing the decision tree. The larger the information gain value, the stronger the classification ability of the feature for the data set. u iis the weight of the seedling growth characteristics, related to the \(i\)-th seedling growth characteristic. It reflects the relative importance of this seedling growth characteristic in the maintenance decision-making, and is determined by the Analytic Hierarchy Process (AHP) combined with expert experience. \(H(D)\) is the entropy of the original dataset. Entropy is a concept in information theory, used to measure the uncertainty or chaos degree of the dataset. The larger the value of \(H(D)\), the more dispersed the class distribution of the samples in the dataset, and the higher the uncertainty; conversely, the smaller the value of \(H(D)\), the more "pure" the dataset, and the more concentrated the class distribution. \(V\) is the number of feature values, that is, the number of different values that the feature can take. \(|D v | / |d| is the proportion of the subset \(D v in the original dataset, where \(|D v | represents the number of samples in the subset \(D v , and \(|d|\) represents the total number of samples in the original dataset. This proportion is used to weight the entropy of each subset when calculating the information gain. \(H i (D v ) is the entropy when only considering the seedling growth characteristics in the subset. It measures the uncertainty presented only based on the seedling growth characteristics in this subset. \(\lambda j is the environmental factor influence coefficient, related to the \(j\)-th environmental factor characteristic. It reflects the influence degree of this environmental factor on the seedling growth and maintenance decision-making, and its value range is usually between 0 and 1, determined according to the correlation analysis between the environmental factors and the seedling growth. \(H j (D v ) is the entropy when only considering the environmental factor characteristics in the subset \(D v . It measures the uncertainty presented only based on the environmental factor characteristics in this subset. By calculating \(H j (D v ), the influence degree of the environmental factors on the sample classification under this subset can be understood.

[0142] 4. Generate maintenance suggestions: According to the results of rule reasoning and intelligent decision-making, generate a detailed maintenance suggestion report. The report clearly lists specific suggestions in aspects such as irrigation (irrigation time, water volume), fertilization (fertilizer type, fertilization amount, fertilization time), pruning (pruning part, degree, time), and cleaning fallen leaves (cleaning frequency, method). Present the maintenance suggestions to the maintenance personnel in an intuitive and easy-to-understand way, such as in the form of a table or a document with pictures and texts. At the same time, provide a detailed description and operation guide for the maintenance suggestions to ensure that the maintenance personnel can accurately understand and execute.

[0143] 5. Feedback and Optimization: Collect feedback information from maintenance personnel during the implementation of maintenance suggestions to understand the problems and difficulties encountered in actual operations. Based on the feedback information, optimize and adjust the maintenance suggestions and the rule base. Regularly evaluate and improve the maintenance suggestion module, and continuously improve the rule base and intelligent decision-making algorithms by combining new expert experience, maintenance knowledge, and actual maintenance effects to improve the accuracy and operability of maintenance suggestions.

[0144] The present invention provides a method for intelligent management of garden nursery stock based on intelligent maintenance, including the following steps:

[0145] S1. The data collection module collects geological data, building data (such as location and structure information of pavilions, terraces, and roads, etc.) and nursery stock data of the garden; annotates the data as reference samples; and standardizes the collected data. The data acquisition module reasonably arranges a variety of sensors according to the geological data and nursery stock type distribution data of the garden to collect environmental data in real time.

[0146] S2. The meteorological data acquisition module collects meteorological data in real time, including meteorological data such as temperature, humidity, wind speed, wind direction, rainfall, and air pressure.

[0147] S3. The fixed-point monitoring module fixes points to collect high-definition images of nursery stock in the garden.

[0148] S4. The route planning module plans the optimal flight route of the drone according to the geological data and nursery stock type distribution data of the garden.

[0149] S5. The cruise monitoring module conducts cruise monitoring on garden nursery stock according to the planned route and comprehensively collects high-definition images.

[0150] S6. The image fusion module splices and fuses the images collected by the cruise monitoring module and the images collected in real time by the fixed-point monitoring module.

[0151] S7. The image preprocessing module preprocesses the images collected by the fixed-point monitoring module and the spliced and fused images, including filtering and denoising, image enhancement, grayscale conversion, and normalization, etc.; the feature extraction module extracts features from the preprocessed images, and the extracted features include color, texture, and shape.

[0152] S8. The image analysis module conducts recognition and analysis on the images after feature extraction, and timely identifies the growth stage (seedling stage, growth stage, flowering stage, fruiting stage, etc.) and growth trend (vigorous, general, weak) of the nursery stock; timely identifies abnormal situations such as pests and diseases (nutrient deficiency, water anomaly, environmental pollution, and human damage, etc.); and identifies the situations where the nursery stock needs to be pruned.

[0153] S9. The comprehensive evaluation module combines the environmental data (temperature, humidity, soil nutrient content, light) and meteorological data collected in real time by the data acquisition module with the analysis results of the image analysis module, and uses data analysis methods such as multiple linear regression and principal component analysis (PCA) to establish a comprehensive evaluation model to comprehensively evaluate the growth environment and growth status of garden seedlings. Based on historical data and real-time monitoring data, time series analysis and support vector machine (SVM) methods are used to predict potential risks.

[0154] S10. The maintenance suggestion module gives maintenance suggestions for garden seedlings according to the analysis results of the image analysis module and the evaluation results of the comprehensive evaluation module, combines expert experience and maintenance knowledge base, and uses rule reasoning and intelligent decision-making algorithms. The maintenance suggestions cover aspects such as irrigation (irrigation time, water volume), fertilization (fertilizer type, fertilization amount, fertilization time), pruning (pruning part, degree, time), and leaf litter cleaning (cleaning frequency, method), providing detailed and operable guidance for maintenance personnel.

[0155] S11. Alarm module: When abnormal situations are detected, alarms are issued in a timely manner; including audible and visual alarms. While the alarm is issued at the control center, the alarm information is notified to relevant management personnel in a timely manner through text messages, APP push, etc., to ensure that abnormal situations are handled in a timely manner.

[0156] The present invention comprehensively collects high-definition images of garden seedlings through the fixed-point monitoring module and the cruise monitoring module, realizing multi-dimensional real-time monitoring of garden meteorology and the growth status of seedlings. Through the image analysis module, the growth stage, growth trend, and abnormal situations such as pests and diseases of seedlings can be accurately identified, providing a scientific basis for garden maintenance. The comprehensive evaluation module combines multi-source data to establish an evaluation model, comprehensively evaluates the growth environment and status of seedlings, predicts potential risks based on historical and real-time data, and makes preparations in advance. The maintenance suggestion module gives detailed maintenance suggestions covering multiple aspects such as irrigation, fertilization, and pruning according to the analysis and evaluation results, combined with expert experience and knowledge base, guiding maintenance personnel to operate scientifically.

[0157] As described above, the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that: they can still modify the technical solutions recorded in the foregoing embodiments, or perform equivalent replacements for some of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for intelligent management of garden seedlings based on intelligent maintenance, characterized in that: The following steps are involved: S1, data collection module collects geological data, building data and seedling data of the garden; The data acquisition module collects environmental data in real time; S2, meteorological data collection module collects meteorological data in real time; S3, the fixed-point monitoring module collects high-definition images of the seedlings in the garden at fixed points; S4, the route planning module plans the optimal flight route of the UAV; S5, the cruise monitoring module collects high-definition images of garden seedlings according to the planned route; S6, the image fusion module splices and fuses the image collected by the cruise monitoring module and the image collected by the fixed-point monitoring module; S7, the image preprocessing module preprocesses the image collected by the fixed-point monitoring module and the spliced ​​and fused image, and the feature extraction module extracts features from the preprocessed image; S8, the image analysis module identifies the growth stage, growth condition and abnormality of the seedlings; S9, Comprehensive Assessment Module Establish a comprehensive assessment model to comprehensively assess the growth environment and growth status of garden seedlings and predict potential risks; S10, the maintenance suggestion module gives maintenance suggestions for the garden seedlings according to the analysis results of the image analysis module and the evaluation results of the comprehensive evaluation module; S11. Alarm module: When an abnormal situation is detected, an alarm will be issued in time.

2. The intelligent management method for garden seedlings based on intelligent maintenance according to claim 1, characterized in that: Step S1 includes the following steps: S11. Data collection and analysis: Comprehensively collect the geological data and wood species distribution data of the garden, use geographic information system (GIS) technology to analyze the data, and draw a map of the garden geology and seedling distribution; S12. Determine key monitoring areas: divide key monitoring areas according to the functional zoning of the garden and the growth characteristics of seedlings; S13. Sensor layout design: In key monitoring areas, reasonable layout should be carried out according to the monitoring range and effective distance of different sensors; S14, sensor installation and debugging: install and debug the sensor according to the layout design plan; S15, Data collection and transmission settings: set a reasonable data collection frequency; S16. Real-time monitoring and maintenance: Monitor the working status of the sensor and the quality of collected data in real time, and perform regular maintenance on the sensor.

3. The garden seedling intelligent management method based on intelligent maintenance according to claim 1, characterized in that: Step S4 includes the following steps: S41. Data collection and integration: Comprehensively collect geological data of the garden and summarize the distribution data of seedling species; S42. Marking of obstacles and key areas: Accurately mark all obstacles on the GIS map and mark the areas in the garden that need to be monitored; S43, preliminary route planning: select Dijkstra path planning algorithm to preliminarily plan the flight route; S44, route optimization and adjustment: optimizing the initially planned route; S45, Real-time dynamic adjustment: Real-time monitoring of meteorological data, personnel activities in the garden and emergencies to dynamically adjust the flight route; S46, Route verification and update: Analyze and evaluate the actual flight route and collected data, and update and optimize the route planning model based on the evaluation results.

4. The method for intelligent management of garden seedlings based on intelligent maintenance according to claim 1, characterized in that: Step S6 includes the following steps: S61, image classification: extract high-definition panoramic cruise images; obtain close-up images of different fixed-point positions within the same time period from the corresponding storage path of the fixed-point monitoring module; classify the images and store them in different folders; S62, image preprocessing: including denoising, image enhancement and grayscale processing; S63, feature extraction: extracting color, texture and shape features from the preprocessed image; S64, feature matching: construct an index of feature points, calculate the Euclidean distance between the feature descriptors of the two images, take the feature point pairs with a distance less than a threshold as matching points, and perform cross-validation on the preliminary matched point pairs; S65, image stitching and fusion: stitching and fusion of images after feature matching; S66. Fusion image quality assessment: The quality of the fusion image is assessed based on the peak signal-to-noise ratio (PSNR) and the structural similarity index (SSIM).

5. The garden seedling intelligent management method based on intelligent maintenance according to claim 4 is characterized in that: Step S65 includes the following steps: S65.

1. Homography matrix calculation: according to the matched feature point pairs, set the maximum number of iterations, calculate the homography matrix by random sampling, and determine the projection transformation relationship between the two images; S65.2, image stitching: project the fixed-point monitoring image into the coordinate system of the cruise monitoring image, and adjust it according to the size of the cruise monitoring image and the projection transformation range; S65.3, fusion processing: for the overlapping areas after stitching, a weighted average fusion algorithm is used for fusion processing; image fusion processing is performed according to the following formula: I(x,y)=w1(x,y)I1(x,y)+w2(x,y)I2(x,y); w1(x,y)=a*d1 / (d1+d2)+(1-a)[I1 - / (I1 - +I2 - )]; W2(x,y)=(1-a)*d2 / (d1+d2)+a[I2 - / (I1 - +I2 - )]; w1(x,y)+w2(x,y)=1; where I(x,y) is the pixel value of the fused image at the coordinate (x,y); I1(x,y) refers to the pixel value of the cruise monitoring image at the coordinate (x,y); I2(x,y) refers to the pixel value of the fixed-point monitoring image at the coordinate (x,y); w1(x,y) and w2(x,y) are the weights of the cruise monitoring image I1(x,y) and the fixed-point monitoring image I2(x,y) in calculating the fused pixel value, respectively; a is the visual importance parameter; d1 is the distance from the pixel (x,y) to the edge of the cruise monitoring image; d2 is the distance from the pixel (x,y) to the edge of the fixed-point monitoring image; I1 is the grayscale mean of the neighborhood centered on the pixel (x,y) in the cruise monitoring image; I2 is the grayscale mean of the neighborhood centered on the pixel (x,y) in the fixed-point monitoring image.

6. The garden seedling intelligent management method based on intelligent maintenance according to claim 1, characterized in that: Step S8 includes the following steps: S81, data collection and annotation: obtaining images of garden seedlings with extracted features and annotating the images; S82, data preprocessing: performing normalization and data enhancement operations on the collected images; S83. Build ResNet model: Build ResNet image analysis model, select cross entropy loss function, select Adam optimizer; set evaluation index; loss function is: L = -Σ N i＝1 λ i Σ M j＝1 [w j *y ij *log(y ij ^)], where L represents the loss function value; N is the number of samples; λ i is the importance weight of the sample; M is the number of categories; w j is the feature complexity weight; y ij Represents the value of the true label of the i-th sample in the j-th category; S84, model training: divide the preprocessed data into training set, validation set and test set; use the training set to train the model, use the validation set to validate the model, and monitor the performance indicators of the model; S85, Model optimization: Adjust the model's hyperparameters and find the optimal hyperparameter combination through cross-validation to improve the model's performance; S86. Model evaluation: Use the test set to comprehensively evaluate the optimized model and evaluate the performance of the model; S87, Image recognition and analysis: The garden seedling images after real-time collection and feature extraction are input into the trained model. The model recognizes and analyzes the images based on the learned features and classification rules.

7. The method for intelligent management of garden seedlings based on intelligent maintenance according to claim 1, characterized in that: Step S9 includes the following steps: S91, data collation: obtain real-time environmental data, collect meteorological data, obtain analysis results of the image analysis module; classify and collate the collected data according to time sequence and monitoring area; S92, data preprocessing: normalizing the data; S93. Build a comprehensive evaluation model: S93.

1. Use the multiple linear regression method to establish a multiple linear regression model with environmental data, meteorological data and image analysis results as independent variables and seedling growth indicators as dependent variables; analyze the linear relationship between independent variables and dependent variables, and evaluate the influence of different factors on seedling growth; S93.

2. Use the principal component analysis (PCA) method to reduce the dimensionality of a large amount of environmental and meteorological data; S93.

3. Combine the results of multiple linear regression and principal component analysis to construct a comprehensive evaluation model to comprehensively evaluate the growth environment and growth status of garden seedlings; S94. Constructing risk prediction model: S94.

1. Collect historical data on garden seedlings; S94.

2. Use time series analysis methods to analyze historical data and explore the patterns and trends of data changes over time; S94.

3. Use the support vector machine (SVM) method to train the SVM model by taking environmental data, meteorological data, and historical pest and disease occurrence data as input features, and pest and disease outbreaks and the impact of extreme weather on seedlings as output labels; S95, Risk assessment: Based on the results of the risk prediction model, evaluate the potential risks faced by garden seedlings and determine the level and scope of risk; S96. Response measures formulation: formulate corresponding response measures in advance for different levels of risks; S97. Model update and optimization: Regularly collect new environmental data, meteorological data and image analysis results, and update and optimize the comprehensive assessment model and risk prediction model.

8. The method for intelligent management of garden seedlings based on intelligent maintenance according to claim 7, characterized in that: In step S93.1, the multiple linear regression model is: y t =β0+Σ n j＝1 [a ij (t,s)β j x jt +γ it and t ];a ij (t,s)=I ij (t,s) / {Σ n j＝1 [I ij (t,s)]};γ it =1+ζ it ; In the formula, y t represents the growth status index of garden seedlings at time t; β0 is the intercept term; n is the number of independent variables; a ij (t,s) is the adaptive weight; β j is the regression coefficient, representing the jth independent variable x j For the dependent variable y t The degree of influence; x jt represents the j-th independent variable at time t; γ it is the dynamic error correction factor; e t is the error term; I ij (t,s) is the independent variable x at growth stage s and time t. j The importance score of it is the error trend coefficient obtained by linear regression analysis; In step S94.2, the risk prediction model is: I t =Σ K k＝1 (w k I tk ), where λ s (t) is the growth stage regulating factor; φ(B)=1-φ1B-φ2B 2 -...-φ p B p ; B is the backward shift operator, φ i is the autoregressive coefficient, p is the autoregressive order; is the difference operator, d is the difference order; θ(B) is the moving average part, θ(B)=1+θ1B+θ2B 2 +...+θ q B q ; θ i is the moving average coefficient, q is the moving average order; τ t is a white noise sequence; I t is the external intervention factor; k is used to distinguish the serial numbers of different intervention factors, K is the total number of intervention factors; w k is the weight of each intervention factor; I tk It is the quantitative value of the specific intervention factor.

9. The garden seedling intelligent management method based on intelligent maintenance according to claim 1, characterized in that: Step S10 includes the following steps: S10.

1. Data collection and integration: Obtain analysis results from the image analysis module and obtain growth environment assessment results of garden seedlings from the comprehensive assessment module; collect maintenance suggestions and treatment methods under different growth conditions and environmental conditions; S10.

2. Formulate rule base: Formulate irrigation rules based on the collected data and knowledge; S10.3, Application of intelligent decision-making algorithm: Use rule-based reasoning algorithms to match the results of image analysis and comprehensive evaluation with the established rule base, and combine intelligent decision-making algorithms to conduct comprehensive analysis and decision-making on complex situations; for seedlings with multiple growth problems and environmental factors at the same time, use decision tree algorithms to gradually analyze and derive the best maintenance recommendations based on the priorities and weights of different factors; S10.

4. Generate maintenance suggestions: Generate a detailed maintenance suggestion report based on the results of rule reasoning and intelligent decision-making; S10.5, Feedback and Optimization: Collect feedback information from maintenance personnel during the implementation of maintenance suggestions, and optimize and adjust the maintenance suggestions and rule base based on the feedback information.

10. A garden seedling intelligent management system based on intelligent maintenance, comprising: PLC control center, data collection module, data acquisition module, meteorological data acquisition module, fixed-point monitoring module, route planning module, cruise monitoring module, image preprocessing module, feature extraction module, image analysis module, comprehensive evaluation module, maintenance suggestion module and alarm module; characterized in that: Data collection module: collect geological data, construction data and seedling data of the garden; annotate and standardize the data as a reference sample; Data collection module: Based on the geological data of the garden and the distribution data of seedling species, various sensors are reasonably deployed to collect environmental data in real time; Meteorological data collection module: real-time collection of meteorological data; Fixed-point monitoring module: multiple high-definition cameras are rationally deployed to carry out fixed-point monitoring of seedlings in the garden; Route planning module: plans the optimal flight route for the drone based on the geological data of the garden and the distribution data of seedling species; Cruise monitoring module: collects high-definition images of garden seedlings according to the planned route; Image preprocessing module: preprocess the images collected by the fixed-point monitoring module and the spliced ​​and fused images; Feature extraction module: extract features from preprocessed images; Image fusion module: stitching and fusing the images collected by the cruise monitoring module and the images collected in real time by the fixed-point monitoring module; Image analysis module: recognize and analyze the image after feature extraction to identify the growth stage, growth status and abnormal conditions of the seedlings; Comprehensive evaluation module: Establish a comprehensive evaluation model to comprehensively evaluate the growth environment and growth status of garden seedlings, and predict potential risks based on historical data and real-time monitoring data; Maintenance suggestion module: provides maintenance suggestions for garden seedlings; Alarm module: when abnormal conditions are detected, an alarm will be issued in time; PLC control center: network connected with data collection module, data acquisition module, meteorological data acquisition module, fixed-point monitoring module, route planning module, cruise monitoring module, image preprocessing module, feature extraction module, image analysis module, comprehensive evaluation module, maintenance suggestion module and alarm module.

Images