Artificial Intelligence-Based Risk Classification and Early Warning Method and System for Landscape Plants

Abstract

This application provides a method and system for risk grading and early warning of garden plants based on artificial intelligence. Among them, vibration and temperature data of the operating components of the target device are collected in real time by environment-adaptive sensors as initial detection data. A compensation calculation model is established based on the temperature distribution characteristics of the sensor protective layer and the detection area, temperature compensation parameters are dynamically calculated, and standardized detection data that eliminates the influence of environmental temperature is generated. The standardized data is matched with the historical fault characteristics in the abnormal vibration feature database to form an analysis model. The system inputs the comprehensive feature set into the analysis model for matching analysis, and at the same time combines the data on changes in the material properties of the equipment components and the real-time lubrication state data, and finally generates early warning information with a grading response function. The solution provided by the embodiments of this application can realize the full-process automatic processing from data collection to intelligent early warning.

Description

Technical Field

[0001] This application relates to the technical field of intelligent device status monitoring and fault warning, and particularly relates to an artificial intelligence-based risk classification and warning method and system for garden plants. Background Art

[0002] In modern garden maintenance, early and accurate identification and risk warning of plant diseases, pests and abnormal growth problems are required. Traditional manual inspections are inefficient, and existing technologies are difficult to achieve automated anomaly detection of multiple plant species in complex natural environments (such as light changes, canopy occlusion, etc.). There is an urgent need for an intelligent monitoring solution that can integrate spectral feature analysis, has strong environmental adaptability and supports classification warning.

[0003] Currently, there is a plant health monitoring system based on unmanned aerial vehicle (UAV) multispectral imaging, which judges abnormal areas by collecting vegetation indices (such as NDVI) and combining preset thresholds. This system uses reflectance analysis of fixed bands, can identify typical problems such as abnormal chlorophyll content, and generates a preliminary abnormal distribution map.

[0004] However, this solution only relies on static threshold judgment, cannot adapt to spectral feature differences caused by different plant species, growth stages and dynamic environmental changes; and lacks the ability to predict the trend of anomaly evolution, making it difficult to distinguish between temporary stress responses and substantial risks, resulting in a high false alarm rate and insufficient warning timeliness. Summary of the Invention

[0005] This application provides an artificial intelligence-based risk classification and warning method and system for garden plants to solve the problems of high false alarm rate and insufficient warning timeliness in the prior art.

[0006] In a first aspect, this application provides an artificial intelligence-based risk classification and warning method for garden plants, including:

[0007] Collecting images of the target plant coverage area through a UAV cluster deployed in the garden area to obtain the reflectance spectral data of the plant canopy;

[0008] Analyzing the chlorophyll content data in the reflectance spectral data based on the spectral change characteristic curves and reflectance intensity threshold data corresponding to different plant types in abnormal growth states through a pre-constructed plant abnormal spectrum database to generate abnormal growth characteristic data;

[0009] Identifying abnormal areas in the plant canopy based on the continuous change characteristics of the abnormal growth characteristic data according to an artificial intelligence algorithm, and verifying the surrounding areas corresponding to the abnormal areas to screen out potential risk areas;

[0010] Match the spatial distribution density of the potential risk area with the evolution process data corresponding to the historical abnormal cases stored in the plant abnormal spectrum database, and correct the diffusion probability of the potential risk area under the local climate environment based on the matching result to generate a risk distribution map including multiple warning levels;

[0011] According to the risk distribution map, output a processing suggestion scheme matching each warning level.

[0012] Optionally, for each plant type, pre-construct a plant abnormal spectrum database including spectral change characteristic curves and reflection intensity threshold data within multiple consecutive band intervals in the abnormal growth state of the plant type;

[0013] Divide the reflection spectrum data into multiple sub-band data segments according to the band interval division rule of the spectral change characteristic curve in the plant abnormal spectrum database;

[0014] Judge whether the fluctuation direction of the reflection intensity sequence of each sub-band data segment is consistent with the preset direction of the spectral change characteristic curve, and compare whether the mean value of the reflection intensity data of each sub-band data segment exceeds the upper or lower limit of the reflection intensity threshold data to mark it as a first-level deviation segment or a second-level deviation segment;

[0015] Extract at least two specific sub-band data segments associated with the chlorophyll content data in the reflection spectrum data from the plant abnormal spectrum database, and judge whether the reflection intensity difference between the at least two specific sub-band data segments exceeds the reflection intensity threshold data to mark it as a chlorophyll abnormal segment;

[0016] Generate growth abnormal characteristic data including band deviation type and intensity information according to the proportion of the number of the first-level deviation segments, second-level deviation segments and chlorophyll abnormal segments.

[0017] Optionally, preset a first sub-band interval and a second sub-band interval associated with the chlorophyll content data in the reflection spectrum data in the plant abnormal spectrum database;

[0018] According to the first sub-band interval and the second sub-band interval, extract two corresponding specific sub-band data segments from the multiple sub-band data segments;

[0019] Calculate the reflection intensity difference between the two specific sub-band data segments, and judge whether the positive and negative signs of the reflection intensity difference are consistent with the sign direction in the reflection intensity threshold data range;

[0020] If the positive or negative sign of the reflection intensity difference is consistent with the preset sign direction, and the reflection intensity difference exceeds the upper or lower limit of the reflection intensity threshold data, then mark the two specific sub-band data segments as chlorophyll abnormal segments.

[0021] Optionally, calculate the continuous change characteristics of three consecutive time points and the deviation degree of the current characteristic value from the normal reference value for the growth abnormal characteristic data according to the time series using an artificial intelligence algorithm;

[0022] Mark the points where the continuous change characteristics exceed the first threshold and the deviation degree exceeds the second threshold as abnormal points, and perform spatial clustering on the abnormal points to form an abnormal area in the plant canopy;

[0023] Form a circular verification band by expanding the abnormal area by a preset distance, and count the proportion of the number of abnormal points within the circular verification band. When the proportion exceeds the verification threshold, it is determined as an effective abnormal area;

[0024] Calculate the risk index based on the abnormal intensity of the core area, the abnormal diffusion degree of the verification band, and the deterioration trend in the time series of the effective abnormal area, and screen out the areas where the risk index exceeds the warning value as potential risk areas.

[0025] Optionally, extract the evolution process data corresponding to each historical abnormal case from the plant abnormal spectrum database. The evolution process data includes the initial abnormal density distribution, the density change trend at each evolution stage, and the corresponding environmental parameter change curve;

[0026] Perform multi-dimensional comparison between the density distribution map of the potential risk area and the evolution process data of each historical case to determine the initial matching coefficient, and mark the historical cases with the initial matching coefficient exceeding the preset reference value as candidate cases;

[0027] Perform stage matching between the risk area characteristics of the candidate cases and the evolution process data corresponding to the historical cases to calculate the evolution matching degree of the candidate cases;

[0028] Take the evolution matching degree of the candidate cases as the matching result.

[0029] Optionally, obtain the reflection intensity sequence of the multiple sub-band data segments within the corresponding band interval, extract the change amount between adjacent data points in the reflection intensity sequence, and calculate the positive and negative sign distribution ratio of the change amount;

[0030] Compare the positive and negative sign distribution ratio with the preset allowable direction range in the spectral change characteristic curve. When the positive and negative sign distribution ratio exceeds the allowable direction range, it is determined that the direction is inconsistent, otherwise it is determined that the direction is consistent;

[0031] Calculate the average reflection intensity of the multiple sub-band data segments, and obtain the upper and lower limit values of the corresponding band intervals in the reflection intensity threshold data;

[0032] When the fluctuation direction is consistent with the preset direction and the average value exceeds the upper limit value or the lower limit value, it is marked as a first-level deviation segment. When the fluctuation direction is inconsistent with the preset direction and the average value exceeds the upper limit value or the lower limit value, it is marked as a second-level deviation segment.

[0033] Optionally, based on the matching result, correct the diffusion probability of the potential risk area under the local climate environment. According to the comparison relationship between the corrected diffusion probability and the preset risk threshold, divide the early warning levels into high, medium, and low levels, and generate a risk distribution map corresponding to the early warning levels.

[0034] In a second aspect, the present application provides an artificial intelligence-based risk grading and early warning system for garden plants, including:

[0035] An acquisition module, configured to collect images of the target plant coverage area through a drone cluster deployed in the garden area to obtain the reflection spectrum data of the plant canopy;

[0036] A generation module, configured to analyze the chlorophyll content data in the reflection spectrum data based on the spectral change characteristic curves and reflection intensity threshold data corresponding to different plant types in the abnormal growth state through a pre-constructed plant abnormal spectrum database, so as to generate abnormal growth characteristic data;

[0037] A screening module, configured to identify the abnormal areas in the plant canopy based on the continuous change characteristics of the abnormal growth characteristic data according to the artificial intelligence algorithm, and verify the surrounding areas corresponding to the abnormal areas to screen out potential risk areas;

[0038] The generation module is further configured to match the spatial distribution density of the potential risk area with the evolution process data corresponding to the historical abnormal cases stored in the plant abnormal spectrum database, and correct the diffusion probability of the potential risk area under the local climate environment based on the matching result, so as to generate a risk distribution map including multiple early warning levels;

[0039] An output module, configured to output a processing suggestion scheme matching each early warning level according to the risk distribution map.

[0040] In a third aspect, an embodiment of the present application provides a computing device, including a processing component and a storage component; the storage component stores one or more computer instructions; the one or more computer instructions are used to be called and executed by the processing component to implement an artificial intelligence-based risk grading and early warning method for garden plants as described in the first aspect above.

[0041] In a fourth aspect, an embodiment of the present application provides a computer storage medium storing a computer program, which when executed by a computer, implements a method for grading and warning of garden plant risks based on artificial intelligence as described in the first aspect.

[0042] In an embodiment of the present application, an unmanned aerial vehicle (UAV) cluster deployed in a garden area is used to collect images of the target plant coverage area to obtain the reflection spectrum data of the plant canopy; based on a pre-constructed plant abnormal spectrum database, and according to the spectral change characteristic curves and reflection intensity threshold data corresponding to different plant types in the abnormal growth state, the chlorophyll content data in the reflection spectrum data is analyzed to generate growth abnormal characteristic data; based on the continuous change characteristics of the growth abnormal characteristic data by using an artificial intelligence algorithm, the abnormal areas in the plant canopy are identified, and the surrounding areas corresponding to the abnormal areas are verified to screen out potential risk areas; the spatial distribution density of the potential risk areas is matched with the evolution process data corresponding to the historical abnormal cases stored in the plant abnormal spectrum database, and based on the matching result, the diffusion probability of the potential risk areas in the local climate environment is corrected to generate a risk distribution map including multiple warning levels; according to the risk distribution map, a processing suggestion scheme matching each warning level is output.

[0043] The technical solution of the present application has the following beneficial effects:

[0044] In the present application, the UAV cluster is used to collect the reflection spectrum data of the plant canopy, combined with the pre-constructed plant abnormal spectrum database for chlorophyll content analysis to generate growth abnormal characteristic data, the AI algorithm is used to identify abnormal areas and verify and screen potential risk areas, the distribution density of the risk areas is matched with the evolution data of historical cases, and the diffusion probability is corrected based on the local climate, finally generating a hierarchical warning risk distribution map and corresponding processing schemes, realizing the full-process automated management from plant health monitoring, abnormal accurate identification, risk dynamic prediction to maintenance intelligent decision-making, and significantly improving the accuracy of garden plant disease warning and maintenance efficiency.

[0045] Furthermore, a database containing multi-band spectral change characteristic curves and reflection intensity threshold data under abnormal conditions is pre-constructed for different plant types. By splitting the collected reflection spectral data into sub-band data segments according to the database rules, the consistency of the fluctuation direction of each sub-band reflection intensity sequence and whether the mean value exceeds the threshold are judged in turn to mark the deviation level, and the chlorophyll-related sub-bands are extracted for difference comparison to mark the abnormal segments. Finally, based on the proportion of the number of various deviation segments, growth abnormal characteristic data containing band deviation characteristic information is generated. This method realizes the accurate identification and quantitative evaluation of different plant growth abnormalities through multi-band spectral feature analysis combined with dynamic threshold judgment, effectively distinguishes normal physiological changes from pathological abnormalities, provides an objective and quantifiable characteristic basis for plant health status diagnosis, and significantly improves the accuracy and reliability of abnormal detection.

[0046] These aspects or other aspects of the present application will be more clearly understood in the following description of the embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following will briefly introduce the drawings required for the description of the embodiments or the prior art. Obviously, the following drawings are some embodiments of the present application. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0048] Figure 1 Shows a flowchart of a method for risk classification and early warning of garden plants based on artificial intelligence provided by the present application;

[0049] Figure 2 Shows a schematic structural diagram of a system for risk classification and early warning of garden plants based on artificial intelligence provided by the present application;

[0050] Figure 3 Shows a schematic structural diagram of a computing device provided by the present application. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0051] In order to enable those skilled in the art to better understand the solution of the present application, the following will clearly and completely describe the technical solutions in the embodiments of the present application with reference to the drawings in the embodiments of the present application.

[0052] In some processes described in the specification, claims, and the above-mentioned drawings of the present application, a plurality of operations appear in a specific order. However, it should be clearly understood that these operations may not be executed in the order in which they appear herein or may be executed in parallel. The serial numbers of the operations, such as 101, 102, etc., are only used to distinguish different operations, and the serial numbers themselves do not represent any execution order. In addition, these processes may include more or fewer operations, and these operations may be executed in sequence or in parallel. It should be noted that the descriptions such as "first" and "second" in this article are used to distinguish different messages, devices, modules, etc., do not represent a sequence, and do not limit that "first" and "second" are different types.

[0053] Researchers have found that the current health monitoring of garden plants mainly relies on manual inspections or single-spectral index analysis, which has problems such as low detection efficiency, inability to distinguish plant species-specific abnormalities, and lack of the ability to predict the spread trend of diseases. Based on this, a risk classification and early warning method for garden plants based on artificial intelligence is provided. This method can achieve accurate identification of plant abnormalities, prediction of risk spread, and classification and early warning through multi-spectral feature analysis and historical case matching. The technical solution of the present application is applicable to the intelligent monitoring and maintenance management of plant health in scenarios such as urban gardens, botanical gardens, and ecological protection areas. It can achieve accurate identification and intelligent diagnosis of plant growth abnormalities, greatly improve the accuracy and reliability of disease detection; scientifically predict the spread trend of diseases and evaluate the risk level, providing a forward-looking warning for garden maintenance; through intelligent classification and early warning and disposal suggestions, significantly improve the accuracy and management efficiency of garden maintenance, and effectively solve the problems of low efficiency, high false alarm rate, and lack of predictability existing in traditional monitoring methods.

[0054] The entire R & D process reflects a series of steps from the construction of a plant-specific spectral database, multi-dimensional anomaly feature analysis to intelligent risk prediction and classification early warning, aiming to overcome the problems of low detection efficiency, high false alarm rate, and lack of predictability in existing monitoring schemes, so as to achieve accurate diagnosis of the health status of garden plants, scientific assessment of disease risks, and intelligent management of maintenance decisions.

[0055] Next, the technical solutions in the embodiments of the present application will be clearly and completely described in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative efforts belong to the scope of protection of the present application.

[0056] Figure 1 For the embodiments of the present application, a flowchart of a risk classification and early warning method for garden plants based on artificial intelligence is provided, as Figure 1 shown, the method includes:

[0057] 101. Use a drone swarm deployed in a garden area to collect images of the target plant coverage area to obtain the reflectance spectral data of the plant canopy.

[0058] In this step, the drone swarm refers to a cooperative operation system composed of multiple drones equipped with multispectral or hyperspectral imaging devices.

[0059] The target plant coverage area refers to the distribution range of specific garden plants that need to be monitored; the plant canopy refers to the leaf community structure at the uppermost layer of the plant.

[0060] The plant canopy refers to the leaf community structure at the uppermost layer of the plant that receives light, and its spectral characteristics can intuitively reflect the physiological state of the plant.

[0061] The reflectance spectral data refers to the data of the continuous spectral characteristic curve formed by the reflectance of the plant canopy to light of different wavelengths, and contains spectral information in multiple bands such as visible light and near-infrared.

[0062] In the embodiment of the present application, first, a drone swarm equipped with hyperspectral imagers is deployed inside the garden area, and systematic image collection is performed on the target plant coverage area through a preset flight path and cooperative control strategy. Then, the sensors carried by the drones obtain the multispectral reflectance data of the plant canopy in the visible light - near-infrared band range, and at the same time record environmental parameters such as the lighting conditions and shooting angles during collection to ensure data consistency. Finally, radiation correction and geometric correction processing are performed on the original spectral data to generate standardized reflectance spectral data, providing a data basis for subsequent analysis of the plant growth state.

[0063] In the ecological restoration area of a wetland park, multiple drones equipped with hyperspectral cameras are deployed, and periodic inspections are carried out on the reed community according to a preset grid. During the period of stable lighting, the drone swarm synchronously collects data in the visible light and near-infrared bands at a fixed altitude, and records the real-time lighting conditions at the same time. The collected original data is processed by professional software to eliminate the influence of atmospheric scattering, and finally a canopy reflectance dataset containing multiple spectral channels is generated to ensure that the spatial resolution meets the requirements of fine monitoring.

[0064] 102. Based on the spectral change characteristic curves and reflectance intensity threshold data corresponding to different plant types in abnormal growth states in a pre-constructed plant abnormal spectrum database, analyze the chlorophyll content data in the reflectance spectral data to generate abnormal growth characteristic data.

[0065] In this step, the plant abnormal spectrum database refers to an expert knowledge base established through long-term observation and containing the typical spectral characteristics of various garden plants in different abnormal growth states.

[0066] The plant type refers to different types of garden ornamental plants classified by family and genus; the abnormal growth state refers to the abnormal growth conditions such as plant suffering from pests, diseases, nutrient deficiency, water stress, etc.

[0067] The spectral change characteristic curve refers to the dynamic change law of the reflection spectral characteristics of plants during the abnormal development process.

[0068] The reflection intensity threshold data refers to the reference standard value for determining whether the spectral data is abnormal determined through big data statistical analysis.

[0069] The chlorophyll content data refers to the core physiological index reflecting the photosynthesis ability and health status of plants.

[0070] The growth abnormal characteristic data refers to the structured diagnostic data obtained through multi-dimensional analysis and quantitatively characterizing the abnormal type and degree of plants.

[0071] In the embodiment of the present application, first, the pre-constructed plant abnormal spectrum database is called, which contains the standard spectral change characteristic curves and reflection intensity threshold data of different plant types under various growth abnormal states. The reflection spectral data obtained in step 101 is input into the database comparison system, and the reflectance characteristics of the chlorophyll-sensitive band are mainly analyzed. The deviation degree between the current chlorophyll content data and the healthy reference value is calculated through a professional algorithm, and the abnormality is determined in combination with the reflection intensity threshold data. Finally, the growth abnormal characteristic data including the abnormal type, abnormal degree, and spatial position information is output, and the data format is completely compatible with the database storage structure.

[0072] For the reed community in the wetland park, the pre-established plant abnormal spectrum library is called by the system, which contains the characteristic spectral curves of reeds under various typical stress states. Through comparative analysis, it is found that the change in the reflectance of the currently collected data in a specific band is significantly different from that of the healthy samples, conforming to certain pollution or disease characteristics. The system automatically calculates the vegetation index, marks the area where the reflectance exceeds the normal range as abnormal, and generates a file containing the location information and abnormal score.

[0073] 103. Based on the continuous change characteristics of the growth abnormal characteristic data according to the artificial intelligence algorithm, identify the abnormal areas in the plant canopy, and verify the surrounding areas corresponding to the abnormal areas to screen out potential risk areas;

[0074] In this step, the artificial intelligence algorithm refers to a plant abnormal recognition model constructed by using deep learning technology, which can automatically learn the change law of plant abnormal characteristics.

[0075] The growth abnormal characteristic data refers to the structured data set after standardization processing, which contains plant physiological indicators in multiple dimensions.

[0076] The continuous change feature refers to the dynamic evolution of plant abnormalities in a time series, and the abnormal area in the plant canopy refers to the spatial range of possible health problems identified by the algorithm.

[0077] The peripheral area refers to the buffer zone that expands outward from the abnormal area and is used to verify the diffusion trend of the anomaly.

[0078] Potential risk areas refer to abnormal ranges that are confirmed to require special attention after multiple verifications.

[0079] In the embodiment of the present application, a convolutional neural network (CNN) is first used to perform spatiotemporal analysis on the continuous change characteristics of the growth anomaly feature data, and the abnormal patch area in the canopy is located by continuously monitoring the spectral change trend (such as the reflectivity decreasing day by day) within the period. Then, the spectral similarity of the pixels in the surrounding area corresponding to the abnormal area is verified, and the isolated noise points are eliminated in combination with the spatial gradient changes of the vegetation index (such as NDVI). Finally, the boundaries of the contiguous abnormal area are delineated through morphological processing, and the potential risk areas that need to be focused on are screened out based on the abnormal area ratio and aggregation, so as to avoid the interference of misjudgment of a single detection.

[0080] Taking a certain reed canopy as an example, the artificial intelligence model detected that the vegetation index in a specific area during continuous monitoring continued to decline, exceeding the health threshold. Through image segmentation technology, it was found that the abnormal area showed diffusion characteristics, and there was a similar spectral trend in the surrounding buffer zone. After morphological processing, it was confirmed that the area of ​​the core abnormal area accounted for a significant proportion and was highly overlapped with the diseased area in the visible light image. The system eventually listed the area as a high-risk area for priority treatment, while eliminating misjudgments caused by other short-term interference.

[0081] 104. Match the spatial distribution density of the potential risk area with the evolution process data corresponding to the historical abnormal cases stored in the plant abnormal spectrum database, and modify the diffusion probability of the potential risk area under the local climate environment based on the matching result to generate a risk distribution map containing multiple warning levels;

[0082] In this step, the spatial distribution density of potential risk areas refers to the distribution density of abnormal areas within a unit area.

[0083] The plant anomaly spectrum database refers to an expert knowledge base containing plant anomaly cases over the years.

[0084] Historical abnormal cases refer to typical plant abnormal events and their development processes recorded in the past.

[0085] The evolution process data refers to the complete historical record of the anomaly from its occurrence to its development, and the local climate environment refers to the real-time microclimate conditions such as temperature and humidity in the monitoring area.

[0086] The diffusion probability refers to the predicted value of the likelihood of abnormal spread based on multi-factor analysis; the early warning level refers to different levels of warnings divided according to the risk degree.

[0087] The risk distribution map is a spatial visualization chart that intuitively shows the risk levels of each region.

[0088] In the embodiment of the present application, first, the spatial distribution density index of the potential risk area (such as the number of abnormal patches per unit area) is calculated and similarity matched with historical diffusion cases of the same type of disease in the plant abnormal spectrum database (such as rust, mildew). Then, based on the matching result, the diffusion probability of the potential risk area under the local climate environment is corrected. By introducing local climate parameters (such as temperature, humidity, wind speed), a Bayesian network model is constructed to correct the limitations of simple spatial density assessment and predict the diffusion probability of the abnormal area under different environmental conditions. Finally, according to the probability values, high, medium, and low risk areas are divided, and a risk distribution map containing multiple early warning levels is generated. The reference data of the intervention effects of historical similar cases are synchronously marked in the map.

[0089] For the above case, the system retrieves historical similar case records and finds that the diffusion pattern of the current abnormal area is similar to a past pollution event. Combining meteorological forecasts and soil condition data, professional model simulations show that if no timely intervention is carried out, the abnormal area will expand rapidly. The model divides the surrounding areas into different risk levels according to environmental factors. In the finally generated thermal early warning map, the core area shows the highest risk, and a gradually changing risk buffer zone is formed around it.

[0090] 105. According to the risk distribution map, output a treatment suggestion plan that matches each early warning level.

[0091] In this step, the risk distribution map is a visual chart of the regional distribution marked with different risk levels.

[0092] The early warning level refers to the severity level divided according to the risk prediction result.

[0093] The treatment suggestion plan refers to differential maintenance measures formulated for different risk levels, including prevention and control methods, operation time sequences, and resource allocation suggestions.

[0094] In the embodiment of the present application, first, the early warning level division of each region is determined according to the risk distribution map. Then, the matching historical treatment plan is retrieved from the expert knowledge base as a reference benchmark. Next, the plan is optimized and adjusted in combination with factors such as the current plant phenological period, weather conditions, and maintenance resources. Finally, a treatment suggestion plan that matches each early warning level and includes specific operation specifications, implementation time windows, and expected effects is generated.

[0095] For the highest-risk areas, the system has pushed a multi-level response plan: immediately start the precise spraying of specific solutions by drones for treatment; transplant specific plants to the core pollution area in the short term; implement enhanced monitoring of surrounding risk areas. At the same time, according to the characteristics of the current growth stage of the plants, the plan specifically states to adopt the most suitable pesticide application method to reduce damage. All disposal instructions are transmitted to on-site intelligent devices in real time, and the system will continuously monitor the treatment effect to ensure the effectiveness of the disposal measures.

[0096] In summary, steps 101 to 105 achieve all-weather and all-round monitoring of the health status of landscape plants, establish a scientific and accurate health diagnosis system, greatly improve the accuracy and timeliness of anomaly recognition, and provide forward-looking decision-making support for disease prevention and control; the finally formed hierarchical early warning and precise maintenance plan realizes the transformation of landscape management from passive response to active prevention. The popularization and application of this system will significantly improve the intelligent level of urban landscape management, effectively protect precious landscape plant resources, and provide solid technical support for creating a sustainable ecological urban environment.

[0097] To improve the risk grading and early warning of landscape plants based on artificial intelligence, an intelligent analysis system based on a plant anomaly spectrum database has been developed. By constructing a database containing the spectral change characteristic curves and reflection intensity threshold data of various plant anomalies, this system realizes the multi-band fine analysis of reflection spectral data, provides an objective and accurate scientific basis for the health diagnosis of landscape plants, and significantly improves the accuracy and reliability of plant anomaly detection.

[0098] In some embodiments, in step 102, through the pre-constructed plant anomaly spectrum database, based on the spectral change characteristic curves and reflection intensity threshold data corresponding to different plant types in the growth anomaly state, the chlorophyll content data in the reflection spectral data is analyzed to generate growth anomaly characteristic data, including:

[0099] 201. For each plant type, pre-construct a plant anomaly spectrum database containing the spectral change characteristic curves and reflection intensity threshold data within multiple consecutive band intervals of the plant type in the growth anomaly state;

[0100] In step 201, the plant types refer to various garden plant varieties strictly classified according to the plant taxonomy system, including but not limited to Ginkgo biloba and Platanus hispanica in the tree category, Rosa chinensis and Rhododendron simsii in the shrub category, and Tulipa gesneriana and Lilium brownii in the herb category; the abnormal growth states refer to various pathological and physiological abnormalities that occur during the growth of plants, specifically including various abnormal growth conditions such as fungal diseases, bacterial infections, pest infestations, nutrient element deficiencies, water stress, and inappropriate light; the spectral change characteristic curve refers to the characteristic change trajectory with diagnostic significance presented by the reflection spectrum of a plant with respect to wavelength changes under a specific abnormal state, and this change trajectory can intuitively reflect the change in the physiological state of the plant; the reflection intensity threshold data refers to the normal value range of the reflectance of each band calculated by using statistical methods through collecting a large number of healthy samples, including the upper limit value and the lower limit value of the reflectance, and these threshold data have plant species specificity and seasonal variation characteristics.

[0101] In the embodiment of the present application, first, for each plant type, the spectral change characteristic curves in multiple consecutive band intervals under the abnormal growth state are obtained through experimental observation, including but not limited to typical abnormal modes such as chlorophyll degradation and water deficiency. Then, statistical analysis is performed on each band interval to determine the upper and lower limit ranges of the reflection intensity threshold data, and a standardized plant abnormal spectrum database is established. Finally, the spectral characteristic curves and the reflection intensity threshold data of different plant types under different abnormal growth states are associated and stored to form a reference data set for real-time comparison.

[0102] 202. Divide the reflection spectrum data into multiple sub-band data segments according to the band interval division rule of the spectral change characteristic curve in the plant abnormal spectrum database;

[0103] In step 202, the reflection spectrum data refers to the continuous reflectance data of the plant canopy collected by a professional hyperspectral imaging device, and these data have high spectral resolution and high spatial resolution; the band interval division rule refers to the optimal spectral segmentation scheme determined through expert demonstration according to the spectral response characteristics and abnormal diagnosis requirements of different plant varieties; the sub-band data segment refers to a set of reflectance data within a specific wavelength range obtained by precisely segmenting the original hyperspectral data by using digital signal processing methods according to the established division rule, and each sub-band corresponds to a specific physiological index diagnosis function.

[0104] In the embodiments of the present application, first, the band interval division rules of the spectral change characteristic curves pre-stored in the plant abnormal spectrum database are accurately read. This rule not only clearly stipulates the starting and ending wavelengths of each characteristic band, but also includes the overlapping settings and boundary processing methods of each band interval. Then, according to this division rule, the high-resolution reflection spectral data collected in real time is accurately segmented into multiple sub-band data segments that are exactly the same as the database structure, ensuring that the wavelength range of each sub-band data segment strictly corresponds to the definition in the database. Finally, professional normalization processing is performed on each sub-band data segment, and standardized radiation correction and atmospheric correction algorithms are used to eliminate the influence of factors such as environmental light conditions and observation angles, ensuring that its reflection intensity data can be directly and accurately compared with the threshold data in the database.

[0105] 203. Determine whether the fluctuation direction of the reflection intensity sequence of each sub-band data segment is consistent with the preset direction of the spectral change characteristic curve, and compare whether the mean value of the reflection intensity data of each sub-band data segment exceeds the upper or lower limit of the reflection intensity threshold data, so as to mark it as a first-level deviation segment or a second-level deviation segment;

[0106] In step 203, the reflection intensity sequence refers to a continuous sequence of reflectance values arranged in wavelength order within a specific sub-band data segment, and this sequence reflects the subtle change characteristics of the reflectance within this band range; the preset direction refers to the typical trend direction of the reflectance change of this band in a specific abnormal state recorded in the database, including various modes such as monotonically increasing, monotonically decreasing, increasing first and then decreasing, etc.; the first-level deviation segment refers to an abnormal data segment that satisfies two conditions at the same time: one is that the fluctuation direction of the reflection intensity sequence is completely consistent with the preset direction, and the other is that the mean reflectance significantly exceeds the threshold range; the second-level deviation segment refers to an abnormal data segment whose reflection intensity sequence fluctuation direction is inconsistent with the preset direction but the mean reflectance exceeds the threshold range.

[0107] In the embodiments of the present application, first, a time series analysis method is used to conduct a detailed trend analysis on the reflection intensity sequence of each sub-band data segment, and the sliding window algorithm and regression analysis are used to determine whether its fluctuation direction is completely consistent with the preset direction of the spectral change characteristic curve of the corresponding band in the plant abnormal spectrum database. Then, statistical calculation methods are used to accurately calculate the mean value of the reflection intensity data of each sub-band data segment, and it is strictly compared with the upper and lower limits of the reflection intensity threshold data stored in the database. Finally, accurate marking is performed according to the comparison results: when the fluctuation direction of the reflection intensity sequence is completely consistent with the preset direction and the mean value exceeds the threshold range, it is marked as a first-level deviation segment; when only the mean value exceeds the threshold range but the fluctuation direction is inconsistent, it is marked as a second-level deviation segment. This hierarchical marking method can more accurately reflect the degree and characteristics of plant growth abnormalities.

[0108] 204. Extract at least two specific sub-band data segments associated with the chlorophyll content data in the plant abnormal spectral database, and determine whether the difference in reflection intensity between the at least two specific sub-band data segments exceeds the reflection intensity threshold data, and mark it as a chlorophyll abnormal segment;

[0109] In step 204, the chlorophyll content data refers to the core physiological index data reflecting the photosynthesis ability and health status of plants obtained by inversion through professional spectral analysis methods; the specific sub-band data segment refers to a combination of characteristic wavelength intervals that have been scientifically verified to have a strong correlation with the chlorophyll content, usually including the chlorophyll absorption peak band and the reflection peak band; the difference in reflection intensity refers to the difference value between the average reflectance of the two characteristic sub-band data segments, and this difference value can effectively reflect the relative content change of chlorophyll.

[0110] In the embodiment of the present application, first, accurately extract the information of specific sub-band data segments closely associated with the chlorophyll content data from the plant abnormal spectral database. These bands are usually located in the red light region (such as 670 - 680nm) and the near-infrared region (such as 700 - 750nm) that are most sensitive to chlorophyll changes. Then, accurately locate the corresponding specific sub-band data segments in the real-time collected hyperspectral reflection spectral data, and use the differential algorithm to calculate the difference in reflection intensity between them. Finally, strictly compare this difference with the reflection intensity threshold data. When the difference exceeds the preset threshold range, mark this band combination as a chlorophyll abnormal segment. This marking method can effectively identify potential abnormalities in the photosynthesis system, including early abnormal symptoms such as reduced chlorophyll content and decreased photosynthetic efficiency.

[0111] 205. Generate growth abnormal characteristic data including band deviation type and intensity information according to the proportion of the number of the first-level deviation segments, the second-level deviation segments, and the chlorophyll abnormal segments.

[0112] In step 205, the proportion of the number refers to the percentage value of the number of abnormal marked sub-bands of each type in the total number of analyzed sub-bands. This index reflects the extent of the abnormality; the band deviation types include three basic types: first-level deviation, second-level deviation, and chlorophyll abnormality. Each type can be further divided into three grades: mild, moderate, and severe according to the deviation amplitude; the intensity information refers to the specific degree quantization value of each abnormal sub-band exceeding the normal threshold range, usually expressed by the standard deviation multiple or percentage.

[0113] In the embodiments of the present application, first, a statistical algorithm is used to accurately count the number of segments marked as first-level deviation segments, second-level deviation segments, and chlorophyll anomaly segments in all sub-band data segments. Then, a weighted calculation method is used to determine the proportion of each type of anomaly segment in the total sub-band data segments, taking into account the importance differences of different band intervals for plant health diagnosis. Finally, based on the proportion results and the band distribution characteristics of the anomaly segments, comprehensive growth anomaly feature data including detailed band deviation types (such as water anomaly, chlorophyll anomaly, nutrient deficiency, etc.) and accurate intensity information (such as mild anomaly, moderate anomaly, severe anomaly) is generated. This data not only includes quantitative indicators of the anomaly degree but also provides diagnostic information on the anomaly types, providing a scientific basis and decision support for subsequent plant health assessment and precision agriculture management.

[0114] The following is a specific example:

[0115] In the monitoring project of the key protected plant area in a botanical garden, the system first calls the pre-constructed and improved rare plant abnormal spectrum database, which contains the spectral characteristic data of this plant variety in 12 typical abnormal states. The reflected spectral data collected by the hyperspectral imager is accurately segmented according to the fine band division scheme of this plant variety (a total of 18 characteristic bands). In-depth analysis reveals that the reflection intensity sequence in the 650 - 680 nm band shows a typical change trend of leaf diseases (correlation coefficient 0.92) and the mean value is 2.3 standard deviations lower than the normal threshold, and the system marks it as a severe first-level deviation segment. Further analysis shows that the difference in reflectance between the 680 nm and 720 nm characteristic bands reaches 0.15, exceeding the normal range by 35%, which is confirmed as a severe chlorophyll anomaly segment. The final statistical results show that 35% of the analyzed bands have first-level deviations (20% of which are severe), and 25% of the bands show chlorophyll anomaly characteristics (15% of which are severe). The system generates a detailed feature report containing the diagnostic conclusion of "severe photosynthesis disorder, it is recommended to immediately carry out foliar nutrient supplementation and fungal control".

[0116] In summary, through steps 201 to 205, the accurate identification of specific anomalies of different plant varieties is achieved; by using the multi-band collaborative analysis and hierarchical marking method, the reliability of chlorophyll anomaly detection is significantly improved; by quantitatively evaluating the anomaly degree and type distribution characteristics, a scientific decision-making basis is provided for the health diagnosis and precision maintenance of garden plants, and overall, the technical problems existing in traditional monitoring methods are solved, and the intelligent level of garden plant health monitoring is greatly improved.

[0117] To solve the problems that existing plant health monitoring methods are difficult to accurately identify abnormal changes in chlorophyll and cannot distinguish between temporary fluctuations and substantial lesions, a chlorophyll anomaly detection system based on multi-band collaborative analysis has been developed. By pre-determining the chlorophyll characteristic band combination, the system realizes precise comparison and analysis of reflectance spectral data, provides a reliable basis for evaluating the chlorophyll status for plant health diagnosis, and significantly improves the accuracy of photosynthesis anomaly detection. In some embodiments, the step of extracting at least two specific sub-band data segments associated with the chlorophyll content data in the reflectance spectral data from the plant anomaly spectral database in step 204 and determining whether the difference in reflectance intensity between the at least two specific sub-band data segments exceeds the reflectance intensity threshold data to mark it as a chlorophyll anomaly segment includes:

[0118] 301. Predetermine a first sub-band interval and a second sub-band interval associated with the chlorophyll content data in the reflectance spectral data in the plant anomaly spectral database;

[0119] In step 301, the plant anomaly spectral database refers to a standardized spectral feature database containing various landscape plants at different growth stages and different health conditions established through long-term observations; the chlorophyll content data refers to the core physiological parameter reflecting the photosynthesis efficiency of plants obtained through professional spectral inversion algorithms; the first sub-band interval refers to the wavelength range where the chlorophyll absorption feature is the most significant, usually located in the red light band; the second sub-band interval refers to the near-infrared band range where the chlorophyll reflection feature is the most obvious; the reflectance spectral data refers to the spectral data set containing the continuous reflectance information of the plant canopy in the visible to near-infrared bands collected by the hyperspectral imaging system.

[0120] In the embodiments of the present application, first, a first sub-band interval and a second sub-band interval associated with the chlorophyll content data in the reflectance spectral data are predetermined in the plant anomaly spectral database. These two sub-band intervals are specifically selected as characteristic bands with strong correlation with the chlorophyll content data, and the selection basis includes: the first sub-band interval is located in the band range where the chlorophyll absorption feature is obvious, and the second sub-band interval is located in the band range sensitive to changes in the internal structure of the leaf. In the database construction stage, the correlation coefficients between these two sub-band intervals and the chlorophyll content changes are verified through a large number of experimental data to ensure the reliability of their use as chlorophyll content diagnosis indicators. At the same time, for different plant types, the parameter settings of the corresponding first sub-band interval and second sub-band interval are established respectively and stored in the plant anomaly spectral database.

[0121] 302. Extract the corresponding two specific sub-band data segments from the multiple sub-band data segments according to the first sub-band interval and the second sub-band interval;

[0122] In step 302, the multiple sub-band data segments refer to the dataset of characteristic wavelength intervals formed by the refined division of the original hyperspectral data according to the physiological and biochemical characteristics of plants. Each sub-band corresponds to a specific diagnostic value for plant physiological functions; the specific sub-band data segment refers to the set of key characteristic band data directly related to chlorophyll diagnosis selected from all sub-bands through professional spectral analysis techniques. These band combinations have clear plant physiological significance and stable diagnostic performance.

[0123] In the embodiments of the present application, first, according to the wavelength range parameters of the first sub-band interval and the second sub-band interval preset in step 301, precise extraction is performed from the multiple sub-band data segments after band division processing. The extraction process uses a strict wavelength matching algorithm to ensure that the wavelength ranges of the two specific sub-band data segments extracted are exactly the same as the definitions in the plant abnormal spectrum database. For the reflectance spectral data obtained in each monitoring period, this extraction process will be executed by the system to ensure the temporal continuity of the data. During the extraction process, triple verification of data quality will be implemented: wavelength range verification, signal-to-noise ratio verification, and continuity verification, to ensure that the two specific sub-band data segments extracted meet the quality requirements for subsequent analysis.

[0124] 303. Calculate the difference in reflectance intensity between the two specific sub-band data segments, and determine whether the positive or negative sign of the difference in reflectance intensity is consistent with the sign direction in the reflectance intensity threshold data range;

[0125] In step 303, the difference in reflectance intensity refers to the quantitative difference value between the mean reflectance rates of two characteristic sub-bands after standardization processing; the reflectance intensity threshold data range refers to the normal fluctuation interval of the difference in chlorophyll-related bands established by long-term monitoring of healthy plants; the sign direction refers to the specific physiological significance represented by the positive or negative of the difference value. For example, a negative value usually indicates an enhanced chlorophyll absorption characteristic.

[0126] In the embodiments of the present application, first, the difference in reflectance intensity between the two specific sub-band data segments extracted in step 302 is calculated. The calculation process uses a standardization algorithm: first, calculate the mean reflectance intensity of all sampling points in the first sub-band interval and the second sub-band interval respectively, and then calculate the difference between them. Then, compare the positive or negative sign of the difference in reflectance intensity with the reflectance intensity threshold data range stored in the plant abnormal spectrum database. The standard sign direction of this difference under different growth states is preset in the database as the judgment basis. The system will perform a strict judgment of sign direction consistency, which is a key step in identifying chlorophyll abnormalities. The influence of measurement errors will be considered during the judgment process, and a reasonable error tolerance range will be set.

[0127] 304. If the positive and negative signs of the reflection intensity difference are consistent with the preset sign direction, and the reflection intensity difference exceeds the upper limit or lower limit of the reflection intensity threshold data, the two specific sub-band data segments are marked as chlorophyll anomaly segments.

[0128] In step 304, the preset symbol direction refers to the typical variation pattern of chlorophyll abnormality summarized by plant pathologists based on a large number of clinical observations; the upper and lower limits refer to the dynamic threshold boundaries set in consideration of seasonal variation factors; and the chlorophyll abnormality segment refers to the spectral data segment that has been confirmed through multiple verifications to have significant chlorophyll metabolism abnormalities.

[0129] In an embodiment of the present application, first, based on the judgment result of step 303, when the positive and negative signs of the reflection intensity difference are consistent with the preset sign direction in the reflection intensity threshold data range, the threshold comparison stage is entered. The system will strictly compare the absolute value of the reflection intensity difference with the upper and lower limits of the reflection intensity threshold data stored in the plant abnormal spectrum database. A dynamic threshold mechanism is used for comparison, taking into account influencing factors such as plant type and growth stage. When the reflection intensity difference exceeds the upper threshold, it is marked as an abnormal increase in chlorophyll content; when it exceeds the lower threshold, it is marked as an abnormal decrease in chlorophyll content. The system will mark two specific sub-band data segments that meet the conditions as chlorophyll abnormality segments and record the abnormality level. The marking process implements a double verification mechanism to ensure the accuracy of the marking results.

[0130] Here is a specific example:

[0131] In an ecological monitoring project in a forest park, the system conducted a health assessment of the protected southern yew. First, based on the unique spectral characteristics of the tree species, the optimal band division scheme (a total of 22 sub-bands) was called from the database. Then the hyperspectral data collected by the drone was processed to accurately extract the chlorophyll a characteristic absorption band and the near-infrared reflectance platform band. Analysis found that the reflectivity differences of some plants in the monitored area in these two bands were significantly abnormal, which was highly consistent with the magnesium deficiency symptoms found in recent soil tests. The system automatically marked these abnormal areas and generated a detailed diagnostic report, suggesting precise maintenance through a combination of foliar spraying and soil improvement. After the management staff implemented maintenance according to the suggestion, the review data one month later showed that the chlorophyll index had improved significantly.

[0132] In summary, steps 301 to 304 achieve accurate assessment of the photosynthesis status of plants by scientifically selecting the chlorophyll characteristic band combination; adopt the differential analysis and double verification mechanism to significantly improve the reliability of chlorophyll anomaly detection; provide an objective basis for chlorophyll status assessment for plant health management by establishing a complete labeling and diagnosis process, effectively solve the deficiencies of traditional methods in chlorophyll anomaly identification, and provide important technical support for the precise maintenance of landscape plants.

[0133] To solve the problems that traditional plant health monitoring methods are difficult to accurately identify early anomalies and cannot effectively distinguish local anomalies from systemic risks, a plant health risk assessment system based on spatio-temporal feature analysis is developed. The early warning ability and management efficiency of plant health monitoring are significantly improved. In some embodiments, in step 103, the artificial intelligence algorithm based on the continuous change characteristics of the growth anomaly feature data identifies the abnormal areas in the plant canopy and verifies the surrounding areas corresponding to the abnormal areas to screen out potential risk areas, including:

[0134] 401. Calculate the continuous change characteristics of three consecutive time points and the deviation degree between the current feature value and the normal reference value of the growth anomaly feature data in time series using an artificial intelligence algorithm;

[0135] In step 401, the growth anomaly feature data refers to a structured data set containing multi-dimensional physiological indicators such as plant chlorophyll content, water status, and nutrient level obtained through hyperspectral analysis; the time series refers to a set of plant growth monitoring data collected continuously at fixed time intervals; the continuous change characteristics refer to the change trend and fluctuation characteristics of plant physiological indicators in three consecutive monitoring cycles; the normal reference value refers to a dynamic reference value range established through long-term observation of healthy plants considering seasonal changes and growth stages; the deviation degree refers to the significance level of the difference between the current observed value and the normal reference range quantified by statistical analysis methods.

[0136] In the embodiment of the present application, first, the growth anomaly feature data is arranged in time series, and an artificial intelligence algorithm is used to analyze the continuous change characteristics of three consecutive time points. Specifically, it includes: calculating the ratio change rate of the deviation segments of each band at each time point, the intensity fluctuation value of the chlorophyll anomaly segment, and the evolution trend of the reflection intensity difference. Then, the feature value of the current time point is compared with the normal reference value stored in the plant anomaly spectral database, and the deviation degree is quantified and calculated. This process strictly maintains the structural consistency with the growth anomaly feature data generated in step 205 to ensure the standardization of the input data for time series analysis. Finally, the analysis result including the continuous change characteristic parameters and the deviation degree score is output.

[0137] 402. Mark the points where the continuously varying feature exceeds the first threshold and the deviation degree exceeds the second threshold as abnormal points, and perform spatial clustering on the abnormal points to form abnormal regions in the plant canopy;

[0138] In step 402, that the continuously varying feature exceeds the first threshold means that the change trend of the plant physiological index reaches a preset significance level, indicating the existence of a risk of continuous deterioration; that the deviation degree exceeds the second threshold means that the degree to which the current observed value exceeds the normal reference range reaches the warning standard; the abnormal point refers to the spatial position coordinates where a health problem is confirmed through double - condition verification; spatial clustering refers to aggregating adjacent abnormal points in spatial distribution into a continuous abnormal region with a clear boundary based on density and distance algorithms.

[0139] In the embodiment of the present application, first, a first threshold is set to judge the significance of the continuously varying feature, and a second threshold is set to judge the severity of the deviation degree. The system scans all monitoring points. When the continuously varying feature of a certain point exceeds the first threshold and the deviation degree exceeds the second threshold, the point is marked as an abnormal point. Then, a spatial clustering algorithm is used to perform aggregation analysis on all abnormal points, and the marking rules of the first - level deviation segment and the second - level deviation segment in step 203 are kept consistent during the clustering process. For each formed abnormal region, record its spatial range, the number of abnormal points, and the average deviation degree, and these parameters directly come from the calculation results of step 401.

[0140] 403. Form a circular verification zone by expanding the abnormal region by a preset distance, and count the proportion of the number of abnormal points in the circular verification zone. When the proportion exceeds the verification threshold, it is determined as an effective abnormal region;

[0141] In step 403, the circular verification zone refers to a circular buffer zone formed by expanding outward based on the boundary of the abnormal region, considering the law of plant disease spread; the proportion of the number of abnormal points refers to the proportion of the number of points determined to be abnormal within the range of the verification zone to the total number of points in this region; the verification threshold refers to the critical value for abnormal diffusion determination determined based on historical data analysis; the effective abnormal region refers to the range of the abnormal region that is confirmed to have a diffusion risk through surrounding environment verification.

[0142] In the embodiment of the present application, first, based on the boundary of the abnormal region determined in step 402, expand outward by a preset distance to form a circular verification zone. This preset distance is set according to the characteristics of the plant canopy and is usually 1.2 times the average radius of the abnormal region. Then, count the abnormal characteristics of all points within the verification zone and calculate the proportion of the number of points marked as abnormal points among them. When this proportion exceeds the pre - set verification threshold, determine the original abnormal region as an effective abnormal region. During the determination process, strictly adopt the abnormal point marking standard in step 402 to ensure the consistency of the verification logic.

[0143] 404. Calculate a risk index according to the anomaly intensity of the core area of ​​the effective anomaly area, the anomaly diffusion degree of the verification zone, and the deterioration trend in the time series, and screen out areas where the risk index exceeds the warning value as potential risk areas.

[0144] In step 404, the abnormal intensity of the core area refers to the quantitative value of the health risk level inside the abnormal area obtained by fusion calculation of multiple indicators; the abnormal diffusion degree of the verification zone refers to the spatial gradient characteristics of the spread of abnormal characteristics to the surrounding areas; the deterioration trend refers to the accelerated development trend of abnormal characteristics in the time dimension; the risk index is a comprehensive risk assessment value obtained by weighted fusion calculation of multi-dimensional indicators; the warning value is the risk warning trigger threshold dynamically adjusted according to the characteristics of the plant variety and the growth stage.

[0145] In the embodiment of the present application, firstly, based on the effective abnormal area confirmed in step 403, three core indicators are calculated: the abnormal intensity of the core area (taken from the deviation degree data of step 401); the abnormal diffusion degree of the verification zone (calculated according to the verification result of step 403); the deterioration trend in the time series (from the continuous change characteristic analysis of step 401). Then, the comprehensive risk index is calculated according to the preset formula, and the weight coefficient in the formula matches the historical cases in the plant abnormal spectrum database. Finally, the area with a risk index exceeding the warning value is screened out as a potential risk area. The warning value is dynamically adjusted according to the plant variety and growth stage to ensure the accuracy of the early warning.

[0146] Here is a specific example:

[0147] In the monitoring of the subtropical plant area of ​​a large botanical garden, the system found yellowing leaves in the banyan community through three consecutive weeks of monitoring. Analysis shows that the reflectivity of the 680nm band continues to rise, and the moisture index has a clear downward trend (step 401). Spatial analysis identified two abnormal clusters, among which the abnormality in the east of the main scenic area was the most significant (step 402). Peripheral verification of the area found that 25% of the plants within 50 meters to the northwest had early symptoms (step 403). Comprehensively evaluating the high intensity of the core abnormality in the area, the fast diffusion speed, and the high temperature and high humidity season, the system determined it to be a high-risk area (step 404). The garden immediately took isolation and prevention measures based on the early warning information, and a re-examination two weeks later showed that the disease was effectively controlled.

[0148] In summary, steps 401 to 404 achieve the early accurate identification and risk assessment of plant health anomalies by constructing a spatio-temporal multi-dimensional feature analysis system; adopt a technical route that combines spatial clustering with surrounding environment verification, significantly improving the accuracy and reliability of risk area detection; and provide a scientific basis for garden maintenance decision-making by establishing a dynamic risk assessment model. This method effectively solves the problems existing in traditional monitoring means in aspects such as early anomaly identification, risk assessment accuracy, and warning timeliness, provides key technical support for the digital transformation of plant health management, and has important application value for improving garden maintenance levels and protecting plant resources.

[0149] To solve the problems of difficult accurate prediction of anomaly diffusion trends and lack of historical experience reference in existing plant health monitoring, a risk prediction system based on historical case matching is developed. It provides an objective and reliable reference basis for risk prediction, significantly improves the accuracy and scientific nature of anomaly diffusion trend prediction, and provides important support for precise decision-making in garden maintenance. In some embodiments, the matching of the spatial distribution density of the potential risk area with the evolution process data corresponding to the historical anomaly cases stored in the plant anomaly spectrum database in step 104 to obtain a matching result includes:

[0150] 501. Extract the evolution process data corresponding to each historical anomaly case from the plant anomaly spectrum database, where the evolution process data includes the initial anomaly density distribution, the density change trend at each evolution stage, and the corresponding environmental parameter change curve;

[0151] In step 501, the plant anomaly spectrum database refers to a professional database system established through long-term monitoring and containing complete records of various plant anomaly events; historical anomaly cases refer to typical plant health anomaly events that have been verified by experts and completely recorded, including various types such as diseases, pests, and physiological disorders; evolution process data refers to a complete spatio-temporal record data set of the anomaly event from its initial occurrence to its final resolution; the initial anomaly density distribution refers to the spatial distribution density and morphological characteristics when the anomaly is first detected; the density change trend refers to the area expansion speed and spatial distribution change law shown by the anomaly area in the subsequent development process; and the environmental parameter change curve refers to the micro-environment monitoring data recorded synchronously with the anomaly development process, including continuous change indicators such as temperature, humidity, light intensity, and soil parameters.

[0152] In the embodiments of the present application, first, the evolution process data corresponding to each historical abnormal case is extracted from the plant abnormal spectrum database. The specific extraction content includes the abnormal density distribution map at the initial stage of each case, which contains the spatial density values of abnormal points and their distribution characteristics. At the same time, the density change trend data at each time node during the case evolution process is extracted, and these data record the expansion rate of the abnormal area and the fluctuation of the core density. The environmental parameter change curve corresponding to the case is also extracted, including the time series records of parameters such as temperature, humidity, and light. During the extraction process, the data fields are strictly kept consistent with the original storage structure of the database to ensure the integrity and accuracy of the evolution process data. All the extracted data is converted into a unified space-time coordinate system to establish a standardized data basis for subsequent comparison and analysis.

[0153] 502. Compare the density distribution map of the potential risk area with the evolution process data of each historical case in multiple dimensions to determine the initial matching coefficient, and mark the historical cases with the initial matching coefficient exceeding the preset benchmark value as candidate cases; perform staged matching between the risk area characteristics of the candidate cases and the evolution process data corresponding to the historical cases, and calculate the evolution matching degree of the candidate cases;

[0154] In step 502, the potential risk area refers to the set of abnormal areas with diffusion risk confirmed through preliminary analysis; the density distribution map refers to the rasterized thematic map reflecting the distribution density of abnormal points generated by the spatial interpolation algorithm; the multi-dimensional comparison refers to the comprehensive similarity evaluation from multiple independent dimensions such as spatial morphological characteristics, time evolution laws, and environmental response characteristics; the evolution matching degree refers to the overall similarity between the current risk situation and the historical cases quantified by a mathematical model, which is a standardized scoring value between 0 and 1.

[0155] In the embodiments of the present application, first, the evolution process data corresponding to each historical abnormal case is extracted from the plant abnormal spectrum database. These data include the initial abnormal density distribution, the density change trend at each evolution stage, and the corresponding environmental parameter change curve. Compare the density distribution map of the currently identified potential risk area with the evolution process data in the historical cases. This step involves data comparison in multiple dimensions, such as the density distribution pattern, the density change rate and direction, and the changes in environmental parameters (such as temperature, humidity, etc.). Based on the results of the above multi-dimensional comparison, an initial matching coefficient is calculated for each historical case. This coefficient reflects the similarity or matching degree between the current potential risk area and the historical case in each dimension. Set a preset benchmark value, and mark the historical cases with the initial matching coefficient exceeding this benchmark value as candidate cases. This means that only when the similarity between a historical case and the current potential risk area is high enough will it be further considered.

[0156] Secondly, for each candidate case, deeply analyze the characteristics of its risk areas and make a detailed match with the data of this historical case at different evolution stages. This includes evaluating the risk diffusion patterns at different times, the changes in the affected ranges, and other relevant factors. Compare with the current potential risk areas stage by stage according to the different development stages of the historical case. The purpose of doing this is to better understand the possible development trajectories of the potential risk areas and their potential impact scales. Based on the results of the above stage-by-stage matching, calculate an evolution matching degree for each candidate case. This metric comprehensively considers the similarities and differences in the development of risk areas over time and provides a basis for predicting future development trends.

[0157] Through these two steps, the historical data can be effectively utilized to predict and evaluate the risk levels of current garden plants, thereby providing a scientific basis for taking timely and effective early warning and management measures. This method not only improves the accuracy of risk management but also provides support for the rational allocation of resources.

[0158] 503. Use the evolution matching degree of the candidate case as the matching result.

[0159] In step 503, the candidate case refers to a set of historical abnormal cases with reference value obtained through preliminary screening and similarity calculation; the evolution matching degree refers to a comprehensive similarity score obtained through normalization processing and weighted calculation; the matching result refers to a professional report finally output by the system, including the best matching case and its detailed similarity analysis.

[0160] In the embodiment of the present application, first, sort the evolution matching degrees of each historical case calculated according to step 502, and select the cases with matching degrees exceeding the preset threshold as candidate cases. The preset threshold is set by default, and then conduct a secondary verification on the candidate cases to check the matching situation between the key turning points in their evolution process data and the current potential risk areas. Verify whether the phase difference of the environmental parameter change curve is within the allowable range. Finally, output the candidate cases that pass the verification and their evolution matching degrees as the formal matching results. This result maintains the corresponding relationship with the fields of the plant abnormal spectrum database, including the complete case numbers, matching degree scores, and descriptions of key matching features, providing a reliable basis for subsequent early warning analysis.

[0161] The following is a specific example:

[0162] In the pine forest health monitoring project of a scenic spot, the system detected abnormal phenomena such as yellowing of pine needles in many places. Through historical case matching analysis, the system first screened out 15 eligible pine tree abnormal cases from the database (step 501). After detailed comparison, it was found that the current abnormal distribution highly coincided with the initial characteristics of a pine wilt disease five years ago. The similarity of the spatial distribution pattern reached the "very high" level, and the similarity of the environmental response characteristics was "relatively high". The comprehensive matching degree ranked first (step 502). The system automatically generated a warning report, indicating that the disease might spread to adjacent areas within the next month under similar meteorological conditions. The scenic spot management department deployed prevention and control measures in advance according to this prediction, including setting up isolation belts and targeted chemical treatments, and successfully controlled the disease in the initial occurrence area (step 503).

[0163] In summary, steps 501 to 503 achieved scientific prediction and risk assessment of the abnormal development of plant health by establishing a perfect historical case matching mechanism; adopted a comprehensive similarity evaluation system with multiple dimensions and multiple indicators, significantly improving the accuracy and reliability of the prediction results; through intelligent matching result analysis and visual presentation, provided an intuitive and reliable reference basis for garden maintenance decision-making. This technology effectively solved the problems of insufficient experience and strong subjectivity in the prediction of abnormal development trends by traditional monitoring methods, realized the transformation from passive response to active prevention, greatly improved the scientific level of plant health management and risk prevention and control capabilities, and had important application value for protecting rare plant resources and maintaining ecological security.

[0164] To solve the problems in existing plant health monitoring that it is difficult to accurately identify abnormal types and unable to quantitatively evaluate the severity of abnormalities, a plant abnormal grading and marking system based on multi-dimensional spectral analysis was developed. It provided differential identification criteria for different types of abnormalities, significantly improving the accuracy and operability of plant health status assessment, and providing a scientific basis for precise maintenance decision-making. In some embodiments, step 203 of determining whether the fluctuation direction of the reflection intensity sequence of each sub-band data segment is consistent with the preset direction of the spectral change characteristic curve, and comparing whether the mean value of the reflection intensity data of each sub-band data segment exceeds the upper or lower limit of the reflection intensity threshold data to mark it as a first-level deviation segment or a second-level deviation segment includes:

[0165] 601. Obtain the reflection intensity sequence of the multiple sub-band data segments within the corresponding band interval, extract the change amount between adjacent data points in the reflection intensity sequence, and calculate the positive and negative sign distribution ratio of the change amount;

[0166] In step 601, the sub-band data segment refers to a dataset of characteristic wavelength intervals formed by scientifically dividing the original hyperspectral data according to plant physiological characteristics. Each sub-band corresponds to a specific physiological diagnosis function; the band interval refers to a continuous wavelength range segment divided within the visible to near-infrared spectral range, and its width is determined according to the spectral response characteristics of plant varieties; the reflection intensity sequence refers to a sequence of numbers composed of continuous reflectance measurement values arranged in wavelength order within a specific sub-band interval; the change amount refers to the difference in reflectance values between adjacent wavelength points in the reflection intensity sequence, reflecting the local change characteristics of the spectral curve; the positive and negative sign distribution ratio refers to the ratio of the number of positive and negative reflectance change amounts to the total change amount within a specific sub-band interval, used to quantify the overall change trend of the spectral curve.

[0167] In the embodiment of the present application, first, the reflection intensity sequence of the multiple sub-band data segments within the corresponding band intervals is obtained, and this sequence contains continuous reflection intensity data points arranged in wavelength order. The adjacent data point difference calculation is performed on each sub-band data segment to extract the change amount of the reflection intensity between every two adjacent data points. Then, the positive and negative sign distribution of all change amounts is counted, and the ratio of the number of positive change amounts to the number of negative change amounts is calculated. The ratio calculation process uses a sliding window method, and the window size is dynamically adjusted according to the length of the band interval to ensure the representativeness of the statistical results. Finally, the positive and negative sign distribution ratio value of each sub-band data segment is output as the basic data for direction consistency judgment.

[0168] 602. Compare the positive and negative sign distribution ratio with the preset allowable direction range in the spectral change characteristic curve. When the positive and negative sign distribution ratio exceeds the allowable direction range, it is determined that the direction is inconsistent; otherwise, it is determined that the direction is consistent.

[0169] In step 602, the allowable direction range refers to the reasonable fluctuation interval of the reflectance change trend of a specific plant variety in a specific abnormal state recorded in the plant abnormal spectral database, and this range is obtained through statistical analysis of a large amount of experimental data; direction inconsistency means that the change trend of the measured reflection intensity sequence has a statistically significant difference from the typical abnormal characteristic pattern recorded in the database; direction consistency means that the change trend of the measured reflection intensity sequence conforms to one or more typical abnormal characteristic patterns in the database within the allowable error range.

[0170] In the embodiments of the present application, first, the preset allowable direction range parameter in the spectral change characteristic curve is read from the plant abnormal spectrum database, and this parameter defines the normal direction distribution interval of the change in reflection intensity in the healthy state. Then, the positive and negative sign distribution ratios calculated in step 601 are strictly compared with this allowable direction range. The interval judgment method is used in the comparison process. When the measured ratio value completely falls outside the allowable direction range, it is determined that the directions are inconsistent; when the ratio value wholly or partly falls within the allowable direction range, it is determined that the directions are consistent. The system will record the direction consistency judgment results of each sub-band data segment and mark the specific wavelength positions exceeding the allowable range.

[0171] 603. Calculate the mean value of the reflection intensity of the multiple sub-band data segments, and obtain the upper limit value and the lower limit value of the corresponding band interval in the reflection intensity threshold data;

[0172] In step 603, the mean value of the reflection intensity refers to the arithmetic mean of the reflectivities of all valid wavelength points within the sub-band data segment, representing the overall reflection level of this band; the upper limit value refers to the higher boundary value of the normal distribution range of the reflectivity of a specific band established through long-term monitoring of healthy plants; the lower limit value refers to the lower boundary value of the normal distribution range of the reflectivity of a specific band.

[0173] In the embodiments of the present application, first, the mean value of the reflection intensity sequence of the multiple sub-band data segments is calculated using the weighted average algorithm, where a higher weight is assigned to the reflection intensity at the center position of the band interval. Then, the reflection intensity threshold data is accurately retrieved from the plant abnormal spectrum database to obtain the upper limit value and the lower limit value of the corresponding band interval. This threshold data is dynamically adjusted according to the plant variety, growth stage, and environmental conditions to ensure the applicability of the threshold setting. The system will compare the mean value of the reflection intensity of each sub-band data segment with the corresponding threshold interval, and record the specific band positions and the exceeding amplitudes that exceed the threshold.

[0174] 604. When the fluctuation direction is consistent with the preset direction and the mean value exceeds the upper limit value or the lower limit value, it is marked as a first-level deviation segment; when the fluctuation direction is inconsistent with the preset direction and the mean value exceeds the upper limit value or the lower limit value, it is marked as a second-level deviation segment.

[0175] In step 604, the first-level deviation segment refers to a significant anomaly that simultaneously satisfies the following two conditions: the fluctuation direction of the reflection intensity sequence is consistent with a certain typical anomaly characteristic pattern in the database; the mean value of the reflection intensity significantly exceeds the normal threshold range; the second-level deviation segment refers to a potential anomaly that satisfies the following condition: the mean value of the reflection intensity exceeds the normal threshold range, but the fluctuation direction is inconsistent with the typical anomaly characteristic pattern.

[0176] In the embodiments of the present application, first, based on the direction consistency determination result in step 602 and the threshold comparison result in step 603, a hierarchical marking operation is performed. When a certain sub-band data segment simultaneously meets two conditions: the fluctuation direction is consistent with the preset direction and the average reflection intensity exceeds the upper limit value or the lower limit value, the sub-band data segment is marked as a first-level deviation segment. When the fluctuation direction is inconsistent with the preset direction but the average reflection intensity still exceeds the threshold, it is marked as a second-level deviation segment. The marking process adopts a double-verification mechanism, and for boundary cases, the analysis results of adjacent bands are referred to for review. Finally, a complete report containing deviation marks for all sub-band data segments is generated, where the first-level deviation segments and the second-level deviation segments are distinguished by different identifiers, providing a standardized input for subsequent comprehensive analysis.

[0177] The following is a specific example:

[0178] In an ecological monitoring project in a certain forest park, the system conducts a health assessment of the Korean pine forest in the key protection area. When analyzing the spectral data of Korean pine in a certain area, the system first detects an obvious upward trend in the reflection intensity sequence in the 690 - 710 nm band (step 601). After comparing with the database, this change direction highly coincides with the initial characteristics of pine needle rust (step 602). The average reflection intensity of this band is significantly higher than the normal upper limit value of healthy Korean pine in this season (step 603). The system marks it as a first-level deviation segment and diagnoses it as "initial infection of pine needle rust" (step 604). According to this diagnosis result, the management staff implemented a preventive spraying of low-concentration pesticides, effectively controlling the development of the disease. The re-inspection data two weeks later showed that the spectral characteristics of Korean pine in this area had returned to the normal range.

[0179] In summary, steps 601 to 603 achieve the accurate identification and classification assessment of plant health anomalies by establishing a scientific and perfect spectral anomaly hierarchical marking system; adopt a technical route combining direction feature analysis and threshold deviation detection, significantly improving the accuracy and reliability of anomaly diagnosis; and provide a scientific basis for differential and precise plant health management by establishing a standardized anomaly grading standard. This method effectively solves the technical bottlenecks existing in traditional monitoring means in aspects such as early anomaly identification, type differentiation, and degree assessment, greatly improving the intelligent level and decision-making support ability of plant health monitoring, and providing important technical support for the digital transformation of garden maintenance and the scientific protection of plant resources.

[0180] To solve the problems that existing plant health risk warning methods are difficult to accurately predict the abnormal diffusion trend and cannot dynamically adjust the warning level, a risk warning classification system based on multi-factor coupling analysis is developed. It intuitively presents the risk level distribution in different regions, provides decision-making support for precise prevention and control, and significantly improves the accuracy and timeliness of garden plant health risk warning. In some embodiments, in step 105, the diffusion probability of the potential risk region under the local climate environment is corrected based on the matching result to generate a risk distribution map including multiple warning levels, including:

[0181] 701. Correct the diffusion probability of the potential risk region under the local climate environment based on the matching result. According to the comparison relationship between the corrected diffusion probability and the preset risk threshold, divide it into three warning levels: high, medium, and low, and generate a risk distribution map corresponding to the warning level.

[0182] In step 701, the matching result refers to the comparison conclusion between the current potential risk region and historical abnormal cases obtained through multi-dimensional similarity evaluation, including core indicators such as spatial distribution form similarity and evolution speed matching degree; the local climate environment specifically refers to the microclimate data set collected in real time in the monitoring area, including key environmental parameters directly affecting the spread of plant diseases such as temperature, relative humidity, rainfall, and wind speed; the diffusion probability is a quantitative value of the expansion possibility of the abnormal region calculated based on the historical statistical model and real-time environmental data; the risk threshold is the warning level division standard determined through expert demonstration and historical data analysis, including three key critical values: high-risk threshold, medium-risk threshold, and low-risk threshold; the risk distribution map is a spatial visualization chart generated using geographic information system technology, which respectively identifies high, medium, and low warning regions through the red, yellow, and blue color gradients, and supports multi-scale display and interactive query functions.

[0183] In the embodiment of the present application, first, the confidence level of the matching result is evaluated, and high-confidence historical cases are selected as the main reference. Then, a diffusion probability correction model including environmental factors such as temperature and humidity is established, and the influence weights of each factor are quantified through multiple regression analysis. Next, according to the current real-time environmental monitoring data, the environmental correction coefficient is calculated to adjust the basic diffusion probability. Subsequently, the corrected diffusion probability is compared with the preset risk threshold, and the fuzzy comprehensive evaluation method is used to determine the final warning level. Finally, spatial interpolation technology is used to generate a continuous risk distribution map, and different risk regions are highlighted through visual rendering. The entire analysis process uses the sliding time window technology to realize the dynamic update of risk warning.

[0184] The following is a specific example:

[0185] In the monitoring of lotus diseases in a wetland park, the system found that the matching degree between the current abnormal area distribution and the "brown spot disease of lotus leaves" in the historical case database reached 85%. Considering the current high-temperature and high-humidity environment (temperature 32°C, humidity 85%), the system corrected the basic diffusion probability. According to the preset thresholds (high risk > 70%, medium risk 40 - 70%, low risk < 40%), this area was marked as a high-risk area. The generated risk distribution map clearly showed that the core area was a red warning, the 50-meter buffer zone around it was a yellow warning, and the outermost area was a blue warning. The management staff implemented hierarchical prevention and control based on this warning map, focused on dealing with high-risk areas, and the re-inspection after one week showed that the spread of the disease was effectively curbed.

[0186] In summary, step 701 realizes the scientific quantitative assessment of plant health risks by establishing a dynamic probability correction model; adopts a three-level warning mechanism, significantly improving the rationality and operability of risk classification; through spatial visualization, it provides an intuitive decision-making basis for precise prevention and control. This technology effectively solves the deficiencies of traditional methods in aspects such as risk prediction accuracy, warning timeliness, and prevention and control pertinence, realizes the transformation from passive response to active prevention, and provides key technical support for the digital transformation of garden plant health management.

[0187] Figure 2 The following is a schematic structural diagram of a garden plant risk classification and warning system based on artificial intelligence provided by an embodiment of the present application. As Figure 2 shown, the system includes:

[0188] An acquisition module 21, configured to collect images of the target plant coverage area through a drone cluster deployed in the garden area to obtain the reflection spectral data of the plant canopy;

[0189] A generation module 22, configured to analyze the chlorophyll content data in the reflection spectral data through a pre-constructed plant abnormal spectrum database based on the spectral change characteristic curves and reflection intensity threshold data corresponding to different plant types in the abnormal growth state to generate abnormal growth characteristic data;

[0190] A screening module 23, configured to identify the abnormal areas in the plant canopy based on the continuous change characteristics of the abnormal growth characteristic data according to an artificial intelligence algorithm, and verify the surrounding areas corresponding to the abnormal areas to screen out potential risk areas;

[0191] A correction module 24, further configured to match the spatial distribution density of the potential risk areas with the evolution process data corresponding to the historical abnormal cases stored in the plant abnormal spectrum database, and correct the diffusion probability of the potential risk areas in the local climate environment based on the matching result to generate a risk distribution map including multiple warning levels;

[0192] An output module 25, configured to output a processing suggestion scheme matching each warning level according to the risk distribution map.

[0193] Figure 2 The above-mentioned artificial intelligence-based risk grading and early warning system for garden plants can execute Figure 1 The artificial intelligence-based risk grading and early warning method for garden plants described in the embodiments shown, and its implementation principle and technical effects will not be elaborated. For the above-mentioned artificial intelligence-based risk grading and early warning system for garden plants in the embodiments, the specific manners in which each module and unit perform operations have been described in detail in the embodiments related to the method, and will not be elaborated here.

[0194] In a possible design, Figure 2 The artificial intelligence-based risk grading and early warning system of the embodiments shown can be implemented as a computing device, such as Figure 3 shown, the computing device may include a storage component 31 and a processing component 32;

[0195] The storage component 31 stores one or more computer instructions, wherein the one or more computer instructions are called and executed by the processing component 32.

[0196] The processing component 32 is used for the above Figure 1 The artificial intelligence-based risk grading and early warning method of the embodiments.

[0197] Among them, the processing component 32 may include one or more processors to execute computer instructions to complete all or part of the steps in the above method. Of course, the processing component may also be implemented by one or more application-specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DSPDs), programmable logic devices (PLDs), field programmable gate arrays (FPGAs), controllers, microcontrollers, microprocessors or other electronic components for executing the above method.

[0198] The storage component 31 is configured to store various types of data to support operations on the terminal. The storage component may be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic memory, flash memory, magnetic disk or optical disk.

[0199] Of course, the computing device may also necessarily include other components, such as input / output interfaces, display components, communication components, etc.

[0200] The input / output interface provides an interface between the processing component and the peripheral interface module, and the peripheral interface module may be an output device, an input device, etc.

[0201] The communication component is configured to facilitate communication between the computing device and other devices in a wired or wireless manner, etc.

[0202] Among them, the computing device may be a physical device or an elastic computing host provided by a cloud computing platform, etc. At this time, the computing device may refer to a cloud server, and the above-mentioned processing component, storage component, etc. may be basic server resources leased or purchased from a cloud computing platform.

[0203] The embodiment of the present application also provides a computer storage medium storing a computer program, and when the computer program is executed by a computer, it can implement the above-mentioned Figure 1 risk classification and early warning method for garden plants based on artificial intelligence shown in the embodiment.

[0204] Those skilled in the art can clearly understand that for the convenience and simplicity of description, the specific working processes of the above-described systems, devices, and units can refer to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0205] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed to multiple network units. Some or all of the modules can be selected according to actual needs to achieve the purpose of the solution of this embodiment. Those of ordinary skill in the art can understand and implement it without creative labor.

[0206] Through the description of the above embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus a necessary general hardware platform, and of course, it can also be implemented by hardware. Based on such an understanding, the essence of the above technical solution or the part that contributes to the prior art can be embodied in the form of a software product. The computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to enable a computer device (which may be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or some parts of the embodiments.

[0207] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present application, rather than limiting them; although the present application 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 on 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 application.

Claims

1. An artificial intelligence-based risk classification and early warning method for garden plants, characterized in that, Including: Using a drone cluster deployed in a garden area to collect images of the target plant coverage area to obtain the reflection spectral data of the plant canopy; Based on a pre-constructed plant abnormal spectrum database, analyzing the chlorophyll content data in the reflection spectral data based on the spectral change characteristic curves and reflection intensity threshold data corresponding to different plant types in abnormal growth states to generate abnormal growth characteristic data; Based on an artificial intelligence algorithm, identifying the abnormal areas in the plant canopy according to the continuous change characteristics of the abnormal growth characteristic data, and verifying the surrounding areas corresponding to the abnormal areas to screen out potential risk areas; Matching the spatial distribution density of the potential risk areas with the evolution process data corresponding to the historical abnormal cases stored in the plant abnormal spectrum database, and correcting the diffusion probability of the potential risk areas in the local climate environment based on the matching result to generate a risk distribution map including multiple warning levels; According to the risk distribution map, outputting a processing suggestion scheme matching each warning level; Based on a pre-constructed plant abnormal spectrum database, analyzing the chlorophyll content data in the reflection spectral data based on the spectral change characteristic curves and reflection intensity threshold data corresponding to different plant types in abnormal growth states to generate abnormal growth characteristic data, including: For each plant type, pre-constructing a plant abnormal spectrum database containing the spectral change characteristic curves and reflection intensity threshold data in multiple continuous band intervals of the plant type in abnormal growth states; Dividing the reflection spectral data into multiple sub-band data segments according to the band interval division rule of the spectral change characteristic curve in the plant abnormal spectrum database; Judging whether the fluctuation direction of the reflection intensity sequence of each sub-band data segment is consistent with the preset direction of the spectral change characteristic curve, and comparing whether the mean value of the reflection intensity data of each sub-band data segment exceeds the upper or lower limit of the reflection intensity threshold data to mark it as a first-level deviation segment or a second-level deviation segment; Extracting at least two specific sub-band data segments associated with the chlorophyll content data in the reflection spectral data from the plant abnormal spectrum database, and judging whether the reflection intensity difference between the at least two specific sub-band data segments exceeds the reflection intensity threshold data to mark it as a chlorophyll abnormal segment; Generating abnormal growth characteristic data including band deviation type and intensity information according to the proportion of the number of the first-level deviation segments, second-level deviation segments and chlorophyll abnormal segments.

2. The method according to claim 1, wherein Extracting at least two specific sub-band data segments associated with the chlorophyll content data in the reflection spectral data from the plant abnormal spectrum database, and judging whether the reflection intensity difference between the at least two specific sub-band data segments exceeds the reflection intensity threshold data to mark it as a chlorophyll abnormal segment, including: Pre-setting a first sub-band interval and a second sub-band interval associated with the chlorophyll content data in the reflection spectral data in the plant abnormal spectrum database; Extract two corresponding specific sub-band data segments from the multiple sub-band data segments according to the first sub-band interval and the second sub-band interval; Calculate the difference in reflection intensity between the two specific sub-band data segments, and determine whether the positive / negative sign of the difference in reflection intensity is consistent with the sign direction in the reflection intensity threshold data range; If the positive / negative sign of the difference in reflection intensity is consistent with the preset sign direction, and the difference in reflection intensity exceeds the upper or lower limit of the reflection intensity threshold data, mark the two specific sub-band data segments as chlorophyll anomaly segments.

3. The method according to claim 1, wherein Based on the artificial intelligence algorithm, identify the abnormal areas in the plant canopy according to the continuous change characteristics of the growth anomaly feature data, and verify the surrounding areas corresponding to the abnormal areas to screen out potential risk areas, including: Use the artificial intelligence algorithm to calculate the continuous change characteristics of three consecutive time points and the deviation degree of the current feature value from the normal reference value for the growth anomaly feature data in time series; Mark the points where the continuous change characteristics exceed the first threshold and the deviation degree exceeds the second threshold as abnormal points, and perform spatial clustering on the abnormal points to form abnormal areas in the plant canopy; Form an annular verification zone by expanding the abnormal area by a preset distance, and count the proportion of the number of abnormal points in the annular verification zone. When the proportion exceeds the verification threshold, it is determined as an effective abnormal area; Calculate the risk index according to the abnormal intensity of the core area, the abnormal diffusion degree of the verification zone, and the deterioration trend in the time series of the effective abnormal area, and screen out the areas where the risk index exceeds the warning value as potential risk areas.

4. The method according to claim 1, wherein Match the spatial distribution density of the potential risk areas with the evolution process data corresponding to the historical abnormal cases stored in the plant abnormal spectrum database to obtain a matching result, including: Extract the evolution process data corresponding to each historical abnormal case from the plant abnormal spectrum database. The evolution process data includes the initial abnormal density distribution, the density change trend at each evolution stage, and the corresponding environmental parameter change curve; Perform multi-dimensional comparison between the density distribution map of the potential risk areas and the evolution process data of each historical case to determine the initial matching coefficient, and mark the historical cases with the initial matching coefficient exceeding the preset reference value as candidate cases; perform stage matching between the risk area characteristics of the candidate cases and the evolution process data corresponding to the historical cases to calculate the evolution matching degree of the candidate cases; Take the evolution matching degree of the candidate cases as the matching result.

5. The method according to claim 1, wherein Judge whether the fluctuation direction of the reflection intensity sequence of each sub-band data segment is consistent with the preset direction of the spectral change characteristic curve, and compare whether the mean value of the reflection intensity data of each sub-band data segment exceeds the upper or lower limit of the reflection intensity threshold data to mark it as a first-level deviation segment or a second-level deviation segment, including: Obtain the reflection intensity sequence of the multiple sub-band data segments in the corresponding band interval, extract the change amount between adjacent data points in the reflection intensity sequence, and calculate the positive / negative sign distribution ratio of the change amount; Compare the positive and negative sign distribution ratio with the preset allowable direction range in the spectral change characteristic curve. When the positive and negative sign distribution ratio exceeds the allowable direction range, it is determined that the directions are inconsistent; otherwise, it is determined that the directions are consistent. Calculate the mean reflection intensity of the multiple sub-band data segments, and obtain the upper limit and lower limit values of the corresponding band intervals in the reflection intensity threshold data. When the fluctuation direction is consistent with the preset direction and the mean value exceeds the upper limit or the lower limit, it is marked as a first-level deviation segment. When the fluctuation direction is inconsistent with the preset direction and the mean value exceeds the upper limit or the lower limit, it is marked as a second-level deviation segment.

6. The method according to claim 1, wherein The method for correcting the diffusion probability of the potential risk area in the local climate environment based on the matching result to generate a risk distribution map including multiple warning levels includes: Correct the diffusion probability of the potential risk area in the local climate environment based on the matching result. According to the comparison relationship between the corrected diffusion probability and the preset risk threshold, divide it into three warning levels: high, medium, and low, and generate a risk distribution map corresponding to the warning levels.

7. An artificial intelligence-based risk classification and early warning system for garden plants, characterized in that, Including: An acquisition module, configured to collect images of the target plant coverage area through a drone cluster deployed in the garden area to obtain the reflection spectrum data of the plant canopy. A generation module, configured to analyze the chlorophyll content data in the reflection spectrum data through a pre-constructed plant abnormal spectrum database based on the spectral change characteristic curves and reflection intensity threshold data corresponding to different plant types in the abnormal growth state, so as to generate abnormal growth characteristic data. A screening module, configured to identify the abnormal areas in the plant canopy based on the continuous change characteristics of the abnormal growth characteristic data according to the artificial intelligence algorithm, and verify the surrounding areas corresponding to the abnormal areas to screen out potential risk areas. A correction module, which matches the spatial distribution density of the potential risk area with the evolution process data corresponding to the historical abnormal cases stored in the plant abnormal spectrum database, and corrects the diffusion probability of the potential risk area in the local climate environment based on the matching result to generate a risk distribution map including multiple warning levels. An output module, configured to output a processing suggestion scheme matching each warning level according to the risk distribution map. Analyze the chlorophyll content data in the reflection spectrum data through a pre-constructed plant abnormal spectrum database based on the spectral change characteristic curves and reflection intensity threshold data corresponding to different plant types in the abnormal growth state, so as to generate abnormal growth characteristic data, including: For each plant type, pre-construct a plant abnormal spectrum database including the spectral change characteristic curves and reflection intensity threshold data in multiple consecutive band intervals of the plant type in the abnormal growth state. Divide the reflection spectrum data into multiple sub-band data segments according to the band interval division rule of the spectral change characteristic curve in the plant abnormal spectrum database. Determine whether the fluctuation direction of the reflection intensity sequence of each sub-band data segment is consistent with the preset direction of the spectral change characteristic curve, and compare whether the mean value of the reflection intensity data of each sub-band data segment exceeds the upper or lower limit of the reflection intensity threshold data, so as to mark it as a first-level deviation segment or a second-level deviation segment; Extract at least two specific sub-band data segments associated with the chlorophyll content data in the reflection spectral data from the plant abnormal spectral database, and determine whether the difference in reflection intensity between the at least two specific sub-band data segments exceeds the reflection intensity threshold data, so as to mark it as a chlorophyll abnormal segment; Generate growth abnormal characteristic data including band deviation type and intensity information according to the proportion of the number of the first-level deviation segments, the second-level deviation segments and the chlorophyll abnormal segments.

8. A computing device, characterized in that, It includes a processing component and a storage component; the storage component stores one or more computer instructions; the one or more computer instructions are used to be called and executed by the processing component to implement an artificial intelligence-based risk classification and early warning method for garden plants according to any one of claims 1 to 6.

9. A computer storage medium, characterized in that, A computer program is stored, and when the computer program is executed by a computer, it implements an artificial intelligence-based risk classification and early warning method for garden plants according to any one of claims 1 to 6.

Images