Garden greening engineering project management method based on big data and AI

Abstract

The application discloses a garden greening engineering project management method based on big data and AI, and belongs to the technical field of garden greening engineering project management. The method comprises the following steps: constructing a multi-dimensional evaluation vector of a plant growth state according to historical maintenance records and real-time environmental monitoring data; generating a growth deviation index according to the multi-dimensional evaluation vector and a plant phenology; dividing a key intervention area through spatial clustering and generating differentiated maintenance instructions; matching the differentiated maintenance instructions with material inventory data to output a material allocation scheme; generating an operation sequence of each maintenance unit; adjusting an execution priority according to the operation sequence and real-time weather forecast data to obtain a dynamic operation schedule; and distributing maintenance tasks to operation terminals according to the dynamic operation schedule. Through multi-dimensional growth deviation evaluation and dynamic scheduling, the method realizes accurate allocation of maintenance resources and adaptive optimization of operation plans.

Description

Technical Field

[0001] This invention relates to the field of landscape greening project management technology, specifically a landscape greening project management method based on big data and AI. Background Technology

[0002] Landscape greening project maintenance and management involves numerous vegetation units, requiring timely pruning, fertilization, and pest and disease control based on the actual growth status of the plants. Current management methods largely rely on maintenance personnel conducting periodic on-site inspections, judging plant growth based on experience, and arranging subsequent work according to inventory levels. This approach struggles to integrate and analyze historical maintenance measures, real-time environmental factors, and the plants' own phenological patterns. The assessment of growth status remains at isolated points in time, lacking quantitative deviation measurement methods, resulting in delayed identification and blurred boundaries of abnormal areas.

[0003] After determining maintenance needs, existing solutions typically generate work plans using a fixed sequence or manual scheduling. This disconnects material allocation from work arrangement, making it impossible to dynamically adjust the execution sequence based on the actual arrival time of materials in each area. Furthermore, the work plans do not consider the impact of future weather changes on the feasibility of outdoor operations, easily leading to conflicts between tasks and adverse conditions such as rain and strong winds. This results in forced work interruptions or rework, preventing some maintenance units from completing their tasks within the appropriate window of opportunity.

[0004] To address the above issues, it is necessary to solve how to deeply integrate historical maintenance records, real-time environmental data, and plant phenological patterns to accurately identify the growth and development deviations of each maintenance unit and automatically delineate differentiated intervention ranges based on spatial clustering characteristics. It is also necessary to solve how to automatically generate an executable work schedule that takes into account both the timing of material delivery and meteorological suitability under the dual conditions of material supply constraints and dynamic weather changes, so that maintenance instructions can be distributed in an orderly manner and dynamically respond to on-site execution feedback. Summary of the Invention

[0005] This invention provides a landscape greening project management method based on big data and AI, aiming to solve the problems of inaccurate identification of abnormal plant growth status, subjective lag in the delineation of maintenance areas, and static rigidity of work plans under the influence of material allocation and weather changes in the existing technology, so as to realize the differentiated automatic generation of maintenance intervention and dynamic adaptive scheduling of work sequences.

[0006] To achieve the above objectives, the present invention provides the following technical solution: The present invention provides a method for managing landscaping projects based on big data and AI, comprising:

[0007] Based on historical maintenance records and real-time environmental monitoring data from landscaping projects, a multi-dimensional assessment vector for plant growth status is constructed. Preferably, historical pruning time, historical fertilization amount, and historical pest and disease occurrence frequency for each maintenance unit are obtained as the first type of static attributes, while real-time soil moisture, real-time light intensity, and real-time temperature for each maintenance unit are obtained as the first type of dynamic attributes. The first type of static attributes and the first type of dynamic attributes are aligned by timestamps to form an attribute sequence, which is then normalized to generate the multi-dimensional assessment vector. This integrates long-term maintenance experience with the current environmental situation, laying a data foundation for accurately judging plant growth status.

[0008] Based on the multidimensional evaluation vector and the preset plant phenological calendar, a growth deviation index for each maintenance unit is generated. As a technical solution of this invention, the standard leaf area index range and standard plant height range for each plant on the current date are extracted from the plant phenological calendar. The difference between the actual leaf area index and the standard leaf area index range in the multidimensional evaluation vector is calculated to obtain the leaf area deviation component. The difference between the actual plant height and the standard plant height range is calculated to obtain the plant height deviation component. The leaf area deviation component and the plant height deviation component are then weighted and summed to generate the growth deviation index. This introduces phenological rhythms into the comparison, making the judgment of growth abnormalities more biologically based and avoiding misjudgments caused by a single threshold.

[0009] Based on the growth deviation index, key intervention areas are delineated through spatial clustering, and differentiated maintenance instructions are generated according to these key intervention areas. Preferably, the geographical coordinates of each maintenance unit are combined with the corresponding growth deviation index to form a spatial deviation point set. Within this set, the neighborhood radius is expanded outward from the point with the maximum deviation index as the initial seed. Points within the neighborhood whose deviation index exceeds a preset threshold are grouped into the same candidate region. The number of maintenance units within the candidate region is counted, and when the number exceeds a threshold, the candidate region is marked as a key intervention area. For each key intervention area, the deviation type of each maintenance unit within the region is counted, and differentiated maintenance instructions corresponding to the deviation type are generated. This process can automatically identify large areas of abnormality, avoiding the multiple scheduling overhead caused by scattered handling, while ensuring precise matching of maintenance measures with abnormality types.

[0010] The differentiated maintenance instructions are matched with material inventory data to output a material allocation plan. Specifically, the required material types and quantities are extracted from the differentiated maintenance instructions; the current inventory of each warehouse for the required material types is retrieved from the material inventory data; the material satisfaction level of each warehouse is calculated based on the current inventory and demand; and supply warehouses are selected in descending order of material satisfaction level to generate a material allocation plan. Using this allocation logic, global inventory visualization and allocation can reduce material waiting time and significantly reduce the risk of some warehouses having stockpiles while others experience shortages.

[0011] Based on the material allocation plan, a work sequence for each maintenance unit is generated. Preferably, the material arrival time for each maintenance unit is extracted from the material allocation plan, and the corresponding growth deviation index for each maintenance unit is obtained. The maintenance units are then sorted from largest to smallest according to the growth deviation index to obtain a basic work sequence. Maintenance units in the basic work sequence whose materials have not yet arrived are then postponed based on the material arrival time, thus generating the work sequence. This ensures that the work sequence responds to both urgency and material availability constraints, preventing personnel and equipment from running idle unnecessarily.

[0012] Based on the work sequence and real-time weather forecast data, the execution priority of each task in the work sequence is adjusted to obtain a dynamic work schedule. As a technical solution of this invention, the probability of rainfall and wind speed level for a preset future period are extracted from the real-time weather forecast data. For each task in the work sequence, the conflict relationship between the task type and the probability of rainfall and wind speed level is determined. When a conflict is found, the execution priority of the task is lowered to obtain an adjusted priority. The work order is then rearranged according to the adjusted priority to obtain a dynamic work schedule. This method enables maintenance operations to automatically avoid unsuitable weather conditions such as strong winds and rainfall, significantly reducing personnel injuries and ineffective work caused by weather.

[0013] According to the dynamic work schedule, maintenance tasks are distributed to corresponding work terminals. Specifically, the first task to be executed in the current time period and its corresponding first maintenance unit are extracted from the dynamic work schedule. The first geographical coordinates of the first maintenance unit and the operation procedure document of the first task are obtained. The first geographical coordinates, the operation procedure document, and the estimated duration of the first task are packaged into a first task package, which is then sent to the work terminal associated with the first maintenance unit via a wireless communication link. Workers can obtain accurate positioning and standardized operation guidance without the need for experience-based judgment, improving the standardization and efficiency of on-site execution.

[0014] After task distribution, the system receives start and end signals from the work terminal. Based on these signals, it calculates the actual execution time of the first task, compares it to the estimated time to obtain the time deviation. If the deviation exceeds a preset threshold, it extracts subsequent tasks from the dynamic task schedule and updates their estimated start times based on the deviation. This closed loop enables the schedule to automatically adjust in real-time to keep pace with on-site progress, maintaining the reliability and consistency of the overall work plan.

[0015] After receiving the work completion signal, a review instruction for the first maintenance unit is triggered based on the signal. The review image corresponding to the instruction is acquired, and a pixel-by-pixel comparison is performed between the review image and the standard reference image of the first maintenance unit to generate a pixel difference heatmap. The actual maintenance coverage rate is calculated based on the heatmap. If the actual maintenance coverage rate is lower than the coverage threshold, the first work is re-added to the execution queue. Through image-based automated review, maintenance quality can be objectively quantified, incomplete maintenance can be automatically detected, and rework can be triggered, ensuring that the maintenance effect in key intervention areas meets the predetermined standards.

[0016] The technical effects and advantages provided by the present invention in the above technical solution are as follows:

[0017] This method extracts the standard leaf area index (LAI) and standard plant height ranges for each plant on the current date from the plant phenological calendar. The actual LAI and actual plant height in the multidimensional evaluation vector are then subtracted from these standard ranges to generate LAI and plant height deviation components, which are then weighted and summed to obtain the growth deviation index. Using the plant's phenological stage as a reference, this approach shifts the assessment of growth status from a single absolute value to reflecting the expected growth level deviation at the current developmental stage, avoiding misjudging phenological differences as growth abnormalities. Based on this, the geographical coordinates of each maintenance unit are combined with the growth deviation index to form a spatial deviation point set. Using the point with the maximum deviation index as the initial seed, a neighboring region is expanded outwards. Points within this region with a deviation index exceeding a preset threshold are grouped into the same candidate region. The number of maintenance units is counted to mark key intervention areas, and differentiated maintenance instructions are generated based on the deviation type within each region. This approach utilizes spatial autocorrelation to aggregate discrete outlier points, forming continuous key intervention areas. This eliminates the arbitrariness of manually defining boundaries based on experience. At the same time, it generates targeted instructions based on the actual deviation type within the area, ensuring that pruning, fertilization, or pest and disease control measures precisely correspond to specific problems, reducing resource consumption caused by excessive or misaligned maintenance.

[0018] After receiving differentiated maintenance instructions, the required material types and quantities are extracted. Material inventory data is queried to calculate the material availability of each warehouse. Supply warehouses are selected from highest to lowest availability to generate a material allocation plan. The material arrival time for each maintenance unit is extracted from the allocation plan. Maintenance units are then sorted from highest to lowest growth deviation index to form a basic work sequence. Maintenance units whose materials have not yet arrived are then postponed to generate a new work sequence. Subsequently, the probability of rainfall and wind speed levels for a preset future period are extracted from real-time weather forecast data. The conflict between the work type and the probability of rainfall and wind speed levels for each operation in the work sequence is determined. When a conflict occurs, the execution priority is lowered, and the work is rearranged according to the adjusted priority to obtain a dynamic work schedule. This approach uses material availability as a hard constraint for initiating operations, prioritizing tasks with significant growth deviations and readily available materials. It also uses weather suitability as a priority adjustment factor, proactively downgrading operations such as open-air spraying and high-altitude pruning that conflict with high rainfall probabilities or high wind speeds. This ensures that the final work schedule is feasible in terms of both resource timing and meteorological safety, avoiding work stoppages and rework due to material shortages or weather conditions after tasks are issued, and reducing ineffective scheduling and terminal waiting time. Attached Figure Description

[0019] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in this invention. For those skilled in the art, other drawings can be obtained based on these drawings.

[0020] Figure 1 This is a flowchart of a landscaping project management methodology based on big data and AI.

[0021] Figure 2 This is a flowchart of the growth deviation index calculation process;

[0022] Figure 3 This is a flowchart for identifying key intervention areas for garden maintenance and generating maintenance instructions based on spatial deviation point sets. Detailed Implementation

[0023] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0024] See Figure 1This invention provides a method for managing landscaping projects based on big data and AI. The method includes: constructing a multi-dimensional evaluation vector of plant growth status based on historical maintenance records and real-time environmental monitoring data of the landscaping project; generating a growth deviation index for each maintenance unit based on the multi-dimensional evaluation vector and a preset plant phenological calendar; delineating key intervention areas through spatial clustering based on the growth deviation index, and generating differentiated maintenance instructions based on the key intervention areas; matching the differentiated maintenance instructions with material inventory data to output a material allocation plan; generating a work sequence for each maintenance unit based on the material allocation plan; adjusting the execution priority of each operation in the work sequence based on the work sequence and real-time weather forecast data to obtain a dynamic work schedule; and distributing maintenance tasks to corresponding work terminals according to the dynamic work schedule.

[0025] In practice, the steps of obtaining the historical pruning time, historical fertilization amount, and historical pest and disease occurrence frequency for each maintenance unit are achieved by accessing the landscape engineering project management database. The landscape engineering project management database stores the historical maintenance records for each maintenance unit. From these records, the number of days since the most recent pruning operation for that maintenance unit is extracted as the historical pruning time; the total amount of fertilizer applied within a preset historical period is extracted as the historical fertilization amount; and the number of pest and disease events occurring within that preset historical period is extracted as the historical pest and disease occurrence frequency. These historical pruning time, historical fertilization amount, and historical pest and disease occurrence frequency constitute the first type of static attributes.

[0026] In practice, the acquisition of real-time soil moisture, real-time light intensity, and real-time air temperature for each maintenance unit is achieved through sensor nodes deployed within each unit. Soil moisture sensors measure soil volumetric water content at a fixed acquisition period, and this measured volumetric water content is used as the real-time soil moisture. Light intensity sensors measure photosynthetically active radiation flux at a fixed acquisition period, and this measured photosynthetically active radiation flux is used as the real-time light intensity. Air temperature sensors measure air temperature at a fixed acquisition period, and this measured air temperature is used as the real-time air temperature. These real-time soil moisture, real-time light intensity, and real-time air temperature constitute the first type of dynamic attribute. The sensor nodes transmit the acquired first type of dynamic attribute to the data center via a wireless transmission protocol, and the data center timestamps the acquisition time for each of the first type of dynamic attributes.

[0027] In practice, the process of aligning the first type of static attributes with the first type of dynamic attributes by timestamp is as follows: For each maintenance unit, the data center reads the first type of static attributes of that maintenance unit from the landscaping project management database, and simultaneously extracts the first type of dynamic attributes of that maintenance unit at a specified timestamp from the received first type of dynamic attributes. The specified timestamp is the collection time of the first type of dynamic attributes. For attribute items in the first type of static attributes that do not change frequently over time, the attribute values ​​of the read first type of static attributes are directly concatenated with the attribute values ​​of the first type of dynamic attributes to form an attribute vector for that maintenance unit, and this attribute vector is associated with the specified timestamp. The above concatenation operation is performed on all maintenance units at the same specified timestamp to obtain an attribute sequence composed of the attribute vectors of all maintenance units.

[0028] In some embodiments, the attribute sequence is normalized using a maximum-minimum normalization method. The maximum and minimum attribute values ​​for each attribute item in the attribute sequence are pre-determined. For each attribute vector in the attribute sequence, the th... The attribute value of each attribute item Through normalization transformation Calculate the normalized attribute values ,in Indicates the first in the attribute sequence The minimum attribute value of each attribute item. Indicates the first in the attribute sequence The maximum attribute value of each attribute item. The normalized attribute values ​​of all attribute items in the attribute vector are then recombine to form a normalized attribute vector. This normalized attribute vector is the multidimensional evaluation vector.

[0029] In specific implementation, please refer to Figure 2 The step of extracting the standard leaf area index (SLA) range and standard plant height range for each plant species on the current date from the plant phenological calendar is achieved by querying a pre-defined plant phenological calendar database. This database stores the lower limit, upper limit, lower limit, and upper limit of the SLA for each plant species on every day of the year, using plant species identification and date as a joint index. Based on the plant species identification and the current date, a query is initiated into the plant phenological calendar database to retrieve the returned lower and upper limits of the SLA, as well as the lower and upper limits of the standard plant height. The closed interval formed by the lower and upper limits of the SLA represents the standard leaf area index range, and the closed interval formed by the lower and upper limits of the standard plant height represents the standard plant height range.

[0030] In practical implementation, the process of calculating the difference between the actual leaf area index and the standard leaf area index range in the multidimensional evaluation vector is as follows: The actual leaf area index is read from the multidimensional evaluation vector; when the actual leaf area index is less than the lower limit of the standard leaf area index, the absolute value of the difference between the lower limit and the actual leaf area index is calculated as the leaf area deviation component; when the actual leaf area index is greater than the upper limit of the standard leaf area index, the absolute value of the difference between the actual leaf area index and the upper limit is calculated as the leaf area deviation component; when the actual leaf area index falls within the range of the standard leaf area index, the leaf area deviation component is zero. The method for reading the actual leaf area index from the multidimensional evaluation vector is as follows: During the construction of the multidimensional evaluation vector, a fixed index position is assigned to each attribute item. The leaf area index attribute item corresponds to a predefined leaf area index number. The corresponding normalized attribute value is extracted from the multidimensional evaluation vector through the leaf area index number, and the normalized attribute value is restored to the original leaf area index through an inverse normalization operation. The inverse normalization operation uses the inverse transformation of maximum and minimum value normalization. The maximum and minimum attribute values ​​used in the inverse transformation formula are the same as the maximum and minimum attribute values ​​of leaf area index recorded when constructing the multidimensional evaluation vector.

[0031] In practice, the process of calculating the difference between the actual plant height and the standard plant height range in the multidimensional evaluation vector is as follows: The actual plant height is read from the multidimensional evaluation vector; when the actual plant height is less than the lower limit of the standard plant height, the absolute value of the difference between the lower limit and the actual plant height is calculated as the plant height deviation component; when the actual plant height is greater than the upper limit of the standard plant height, the absolute value of the difference between the actual plant height and the upper limit of the standard plant height is calculated as the plant height deviation component; when the actual plant height falls within the standard plant height range, the plant height deviation component is zero. The method of reading the actual plant height from the multidimensional evaluation vector is the same as the method of reading the actual leaf area index. After extracting the normalized attribute value through the plant height index number corresponding to the plant height attribute item, the actual plant height is restored using an inverse normalization operation.

[0032] In some embodiments, the process of weighted summing of leaf area deviation and plant height deviation components is used to calculate the growth deviation index according to the following formula. :

[0033]

[0034] in, This indicates the growth deviation index of the maintenance unit. Indicates the leaf area deviation component. This indicates the deviation of plant height. The weighting coefficients representing the leaf area deviation components. This represents the weighting coefficient for the plant height deviation component. The weighting coefficient for the leaf area deviation component. The value range is a real number greater than or equal to 0 and less than or equal to 1, and the weighting coefficient of the plant height deviation component. The range of values ​​for is a real number greater than or equal to 0 and less than or equal to 1, and satisfies . Weighting coefficients for leaf area deviation components Weighting coefficients for plant height deviation component The method of obtaining the weights is as follows: They are retrieved from the maintenance weight configuration table based on the plant species identifier. The maintenance weight configuration table pre-stores the leaf area deviation weight value and plant height deviation weight value for each plant species. These weight values ​​are set by experts in the field of landscape engineering based on plant physiological characteristics. When the plant is a foliage plant, the weight coefficient of the leaf area deviation component is... The weighting coefficient for the plant height deviation component is set to 0.7. Set to 0.3; when the plant is a non-foliage plant, the weighting coefficient of the leaf area deviation component is... The weighting coefficient for the plant height deviation component is set to 0.4. The value is set to 0.6. The growth deviation index for each maintenance unit is obtained through the above weighted summation operation.

[0035] In specific implementation, please refer to Figure 3 The process of combining the geographic coordinates of each maintenance unit with its corresponding growth deviation index to form a spatial deviation point set is as follows: The maintenance unit identifier and geographic coordinates of each maintenance unit are extracted one by one from the maintenance unit information table. The geographic coordinates of the maintenance unit include longitude and latitude coordinates. Simultaneously, the growth deviation index corresponding to that maintenance unit is obtained from the calculated growth deviation index records based on the maintenance unit identifier. The longitude, latitude, and growth deviation index of the same maintenance unit are combined into a three-dimensional spatial point. The coordinate format of the three-dimensional spatial point is longitude, latitude, and growth deviation index. All three-dimensional spatial points corresponding to all maintenance units are summarized to form a spatial deviation point set. The spatial deviation point set is stored in a point set data structure in memory. Each element in the point set data structure contains a longitude coordinate field, a latitude coordinate field, and a deviation index field.

[0036] In practice, the process of expanding the neighborhood radius outward from the point with the maximum deviation index in the spatial deviation point set is as follows: The deviation index field of each three-dimensional spatial point in the spatial deviation point set is traversed. The three-dimensional spatial point with the largest deviation index value is identified through comparison and calculation, and this point is designated as the point with the maximum deviation index. Using the point with the maximum deviation index as the initial seed point, and with the longitude and latitude coordinates of the initial seed point as the center, and a preset neighborhood radius as the search radius, all three-dimensional spatial points in the spatial deviation point set that satisfy the condition that the spatial distance is less than or equal to the preset neighborhood radius are searched. The spatial distance is calculated using the great circle distance between two points on the Earth's surface, based on the longitude and latitude coordinates of the two points using the spherical cosine law or the Havesing formula. The preset neighborhood radius is set according to the distribution density of maintenance units in the landscaping project. When the average spacing between maintenance units is 50 meters, the preset neighborhood radius is set to 80 meters to ensure that adjacent maintenance units within the same vegetation community can be included in the same analysis scope. Each searched 3D spatial point that meets the condition of having a spatial distance less than or equal to a preset neighborhood radius is subjected to a deviation index threshold. When the deviation index value of a 3D spatial point is greater than the preset threshold, the 3D spatial point is classified into the same candidate region. The preset threshold is set to 0.5. The basis for setting the preset threshold is that when the growth deviation index is greater than 0.5, it indicates that the actual growth indicators of the maintenance unit deviate significantly from the phenological calendar standard range and need to be included in the intervention assessment scope.

[0037] In practice, the process of counting the number of maintenance units within a candidate area and marking key intervention areas is as follows: The number of three-dimensional spatial points within the candidate area is counted to obtain the number of maintenance units. This number is then compared to a pre-configured threshold. The pre-configured threshold is set based on the project scale, with a threshold of 5 maintenance units. The basis for setting the threshold to 5 is that when 5 or more maintenance units exhibit significant and continuous spatial deviations, it indicates the presence of systemic growth stress factors in the area, requiring overall intervention at the regional level, rather than fragmented treatment targeting individual maintenance units. When the number of maintenance units within a candidate area exceeds the threshold, the candidate area is marked as a key intervention area, assigned a unique key intervention area identifier, and all maintenance unit identifiers within the key intervention area are recorded. When the number of maintenance units within a candidate area is less than or equal to the threshold, the candidate area is abandoned, and the maintenance units within it are not included in the key intervention area category.

[0038] In practice, the process of statistically analyzing the deviation types of each maintenance unit within each key intervention area is as follows: The growth deviation index of each maintenance unit within the key intervention area is obtained, along with the leaf area deviation component and plant height deviation component generated during the calculation of the growth deviation index. The values ​​of the leaf area deviation component and plant height deviation component are determined. When the leaf area deviation component is greater than zero and the plant height deviation component is equal to zero, the deviation type of the maintenance unit is recorded as leaf area deviation; when the leaf area deviation component is equal to zero and the plant height deviation component is greater than zero, the deviation type of the maintenance unit is recorded as plant height deviation; when both the leaf area deviation component and plant height deviation component are greater than zero, the deviation type of the maintenance unit is recorded as a composite deviation. The frequency of deviation types for all maintenance units within the key intervention area is statistically analyzed, and the deviation type with the highest frequency is taken as the dominant deviation type for that key intervention area.

[0039] In some embodiments, the method for generating differentiated maintenance instructions corresponding to deviation types is as follows: a mapping table between deviation types and maintenance instruction templates is pre-established. In the mapping table, leaf area deviation types correspond to foliar fertilization instruction templates and pruning and shaping instruction templates; plant height deviation types correspond to growth regulator spraying instruction templates and support and correction instruction templates; and compound deviation types correspond to comprehensive maintenance instruction templates. The comprehensive maintenance instruction template includes a combination of foliar fertilization instruction templates, pruning and shaping instruction templates, growth regulator spraying instruction templates, and support and correction instruction templates. Based on the dominant deviation type of the key intervention area, the corresponding maintenance instruction template is queried from the mapping table, and the maintenance instruction template is instantiated into a differentiated maintenance instruction. The differentiated maintenance instruction includes a key intervention area identifier, a dominant deviation type, a maintenance measure name, maintenance measure execution parameters, and a maintenance measure execution range. The maintenance measure execution range is determined by the geographical coordinate boundaries of all maintenance units within the key intervention area.

[0040] In practice, the process of extracting the required material types and quantities from differentiated maintenance instructions is as follows: The structured data fields of the differentiated maintenance instructions are parsed. These instructions contain maintenance measure execution parameters, which are composed of a list of material items. Each material item in the list contains a material type identifier and a unit area usage. The material type identifier of each item is read sequentially from the list, and all read material type identifiers are summarized into a set of required material types. For each required material type in the set, the maintenance measure execution range of the key intervention area in the differentiated maintenance instruction is obtained. The area value of the execution range is obtained by calculating the area of ​​the circumscribed polygon formed by all maintenance units within the key intervention area. The area value is multiplied by the corresponding unit area usage in the material item to obtain the total demand for that material type in the key intervention area. The total demand for all key intervention areas is summed for each required material type to obtain the demand for each specific material type.

[0041] In practice, the process of querying material inventory data to obtain the current inventory quantity of each warehouse for the required material types is as follows: An inventory query request is sent to the warehouse management database, carrying the set of required material types. The warehouse management database maintains a warehouse inventory table, which includes a warehouse identifier field, a material type identifier field, and a current inventory quantity field. After receiving the inventory query request, the warehouse management database, based on each required material type in the required material type set, retrieves records in the warehouse inventory table whose material type identifier field matches the required material type, and returns the retrieved warehouse identifier and current inventory quantity. The returned results are organized into a warehouse material inventory mapping structure, where the corresponding current inventory quantity is stored for each combination of warehouse identifier and required material type.

[0042] In some embodiments, the process of calculating the material satisfaction level of each warehouse based on the current inventory level and demand level uses the following formula to calculate the material satisfaction level. :

[0043]

[0044] in, The warehouse identifier is The warehouse's satisfaction rate for a certain type of required materials. The warehouse identifier is The current inventory level of the required material type in the warehouse. This indicates the required quantity of this type of material. Material Satisfaction Rate The value of is a real number greater than or equal to 0, when the material satisfaction level When the value is greater than or equal to 1, it indicates that the warehouse is identified as... The current inventory level in the warehouse can fully meet the demand; when the material satisfaction rate When the value is less than 1, it indicates that the warehouse identifier is... The current inventory of the warehouses can only meet a portion of the demand. Warehouses are sorted in descending order of material satisfaction. Starting with the warehouse with the highest satisfaction, the current inventory is accumulated sequentially. Selection stops when the accumulated inventory reaches or exceeds the demand, and the selected warehouse is marked as a supply warehouse. For each required material type, a material allocation record is generated, containing the required material type, demand quantity, supply warehouse identifier, and the quantity allocated from that supply warehouse. The collection of material allocation records for all required material types constitutes the material allocation plan. When the accumulated current inventory of all supply warehouses for a certain required material type is still less than the demand, the shortage quantity for that required material type is marked in the material allocation plan.

[0045] In practice, the process of extracting the material arrival time for each maintenance unit from the material allocation plan is as follows: Each material allocation record in the material allocation plan also includes the geographical coordinates of the supply warehouse corresponding to the supply warehouse identifier. For each maintenance unit, the material allocation record corresponding to the required material type for that maintenance unit is determined, and the geographical coordinates of the supply warehouse corresponding to the supply warehouse identifier in that material allocation record are obtained. Based on the geographical coordinates of the supply warehouse and the geographical coordinates of the maintenance unit, the road network transportation distance between the supply warehouse and the maintenance unit is calculated. The road network transportation distance is divided by the preset average transportation speed to obtain the transportation time. The current system time is added to the transportation time to obtain the material arrival time corresponding to that maintenance unit. The average transportation speed is set based on the urban road traffic statistics of the project location, and is set at 30 kilometers per hour, based on the average speed of freight vehicles during off-peak hours in the city.

[0046] In practice, the process of obtaining the growth deviation index corresponding to each maintenance unit and sorting them from largest to smallest is as follows: The growth deviation index of all maintenance units is read from the growth deviation index storage table. A descending sorting algorithm is used to arrange the maintenance units according to the value of their growth deviation index from largest to smallest, forming a maintenance unit sequence. The maintenance unit with the largest growth deviation index is placed first in the sequence. This maintenance unit sequence is the basic operation sequence.

[0047] In practice, the process of postponing maintenance units in the basic work sequence whose materials have not yet arrived, based on material arrival time, is as follows: Each maintenance unit in the basic work sequence is traversed. For the currently traversed maintenance unit, the material arrival time corresponding to that unit is obtained and compared with the current system time. If the material arrival time is later than the current system time, it indicates that the required materials for the maintenance unit have not yet arrived, and the maintenance unit is marked as a material-unarrived maintenance unit. If the material arrival time is earlier than or equal to the current system time, it indicates that the required materials for the maintenance unit have arrived, and the maintenance unit is kept in its current sorting position. All material-unarrived maintenance units are sorted from earliest to latest according to their corresponding material arrival times. The sorted material-unarrived maintenance units are then appended sequentially after the maintenance units that have maintained their sorting positions, forming the work sequence.

[0048] In practice, the steps of extracting the probability of rainfall and wind speed level for a preset time period from real-time weather forecast data are achieved by accessing a meteorological service data interface. A weather forecast query request is sent to the meteorological service data interface, carrying the geographic area to be queried and the preset time period parameter. The preset time period parameter is set as a 24-hour window from the current moment, based on the longest expected duration coverage of a single work cycle in the garden maintenance work sequence. The meteorological service data interface returns structured weather forecast response data. The time series fields in the weather forecast response data are parsed, and the rainfall probability field and wind speed level field are extracted hourly. The rainfall probability field records the probability of rainfall for each future hour as a percentage, and the wind speed level field records the wind speed intensity for each future hour as an integer value of the Beaufort scale.

[0049] In practice, the process of determining the conflict between the job type and the probability of rainfall and wind speed level for each job in the job sequence is as follows: Each job in the job sequence is traversed, and the job type of the currently traversed job is read. Job types are pre-classified into aerial trimming, ground spraying, soil operations, and general inspection, each corresponding to a set of weather constraint thresholds. The weather constraint thresholds for aerial trimming are a rainfall probability threshold of 60% and a wind speed level threshold of 4; for ground spraying, they are 50% and 4; for soil operations, they are 70% and 5; and no weather constraint thresholds are set for general inspection. The weather constraint thresholds corresponding to the current job type are compared with the probability of rainfall and wind speed level of the hour in which the corresponding job is expected to start in the weather forecast response data. When the probability of rainfall is greater than the rainfall probability threshold or the wind speed level is greater than the wind speed level threshold, the conflict between the job type and the probability of rainfall and wind speed level is determined to be conflict.

[0050] In practice, when a conflict is identified, the method for adjusting the execution priority of a task to obtain the adjusted priority is as follows: Each task in the task sequence is assigned an execution priority value, initially set to its sequential position number within the sequence; a smaller sequential position number indicates a higher execution priority. When a task is determined to conflict with weather conditions, a preset penalty value is added to its execution priority value. This penalty value is set to twice the total number of tasks in the sequence. The basis for setting the penalty value to twice the total number of tasks is that the penalized execution priority value will be greater than the execution priority values ​​of all unadjusted tasks in the sequence, thus automatically moving the conflicting task to the end of the sequence. The adjusted execution priority value is the adjusted priority.

[0051] In practice, the process of rearranging the job sequence according to the adjusted priority to obtain a dynamic job schedule is as follows: All jobs in the job sequence are sorted in ascending order according to their adjusted priority values, with the job having the lowest adjusted priority value placed first. An estimated start time and estimated end time are assigned to each job in the sorted job sequence. The assignment method is to sequentially add the estimated duration of the previous job starting from the current time to calculate the estimated start and end times for each job. The job sequence with assigned estimated start and end times is serialized into a dynamic job schedule. Each record in the dynamic job schedule includes a job identifier, job type, maintenance unit identifier, estimated start time, estimated end time, and adjusted priority.

[0052] In practice, the process of extracting the first job to be executed in the current time period and its corresponding first maintenance unit from the dynamic job schedule is as follows: Using the current system time as a reference, find the job in the dynamic job schedule whose estimated start time has the smallest time difference with the current system time, and extract this job as the first job. Then, using the maintenance unit identifier contained in the first job record in the dynamic job schedule, query the maintenance unit information table for the maintenance unit corresponding to that identifier, and designate that maintenance unit as the first maintenance unit.

[0053] In practice, the process of obtaining the first geographic coordinates of the first maintenance unit and the operation procedure file for the first operation is as follows: The longitude and latitude coordinates associated with the first maintenance unit identifier are read from the maintenance unit information table, and these coordinates are combined to form the first geographic coordinates. The operation procedure file corresponding to the operation type of the first operation is read from the maintenance operation procedure database. This operation procedure file is a standard operating procedure document containing textual operation steps and illustrations, and is in portable document format. The read operation procedure file is the operation procedure file for the first operation.

[0054] In practice, the process of packaging the first geographic coordinates, the operation procedure file, and the estimated duration of the first operation into the first task package is as follows: A structured data packet object is created, containing a coordinate field, an operation procedure file field, and a duration field. The first geographic coordinates are assigned to the coordinate field, the binary data stream of the operation procedure file is assigned to the operation procedure file field, and the estimated duration of the first operation recorded in the dynamic job schedule is assigned to the duration field. The structured data packet object is then serialized into a compact binary format or text format; the serialized data constitutes the first task package.

[0055] In practical implementation, the process of sending the first task packet to the work terminal associated with the first maintenance unit via a wireless communication link is as follows: The work terminal identifier bound to the first maintenance unit identifier is queried from the terminal registry, and the communication address corresponding to the work terminal identifier is obtained. The system's wireless communication module is invoked, and a transmission channel is established through a fourth-generation mobile communication network or wireless local area network link, using the communication address corresponding to the work terminal identifier as the target address. The first task packet is then sent to the target work terminal through the transmission channel. After receiving the first task packet, the work terminal parses the coordinate field, operation file field, and duration field in the first task packet, and displays the operation procedure file for the first operation, the first geographical coordinates of the first maintenance unit, and the estimated duration of the first operation on the work terminal interface.

[0056] In practice, the process of receiving the job start and job end signals from the job terminals is as follows: The system message receiving module listens to the message queues from each job terminal. When a job terminal executes the first job, upon initiating the job task, the job terminal collects the job start instruction and generates a job start signal, which includes a job identifier and a job start timestamp, and sends the job start signal to the system. When the job task is completed, the job terminal collects the job end instruction and generates a job end signal, which includes a job identifier and a job end timestamp, and sends the job end signal to the system. After receiving the job start and job end signals, the system message receiving module associates the signals with the first job based on the job identifier in the signals.

[0057] In practice, the process of calculating the actual execution time of the first task based on the task start signal and the task end signal is as follows: extract the task end timestamp from the task end signal, extract the task start timestamp from the task start signal, calculate the time difference between the task end timestamp and the task start timestamp, the time difference is measured in minutes, and the calculated time difference is the actual execution time of the first task.

[0058] In practice, the process of comparing the actual execution time with the estimated time to obtain the time deviation is as follows: The estimated time of the first job is read from the task record corresponding to the first task package; the estimated time is subtracted from the actual execution time to obtain the time deviation in minutes. A positive time deviation indicates that the actual execution time exceeds the estimated time; a negative time deviation indicates that the actual execution time is less than the estimated time; and a time deviation of zero indicates that the actual execution time is the same as the estimated time.

[0059] In practice, when the duration deviation exceeds a preset deviation threshold, the process of extracting subsequent tasks from the dynamic work schedule and updating their estimated start times based on the duration deviation is as follows: The preset deviation threshold is set to twenty minutes, based on the reasonable fluctuation tolerance range of the smallest granularity of a single garden maintenance operation. The absolute value of the duration deviation is compared with the preset deviation threshold. When the absolute value of the duration deviation is greater than the preset deviation threshold, all tasks following the first task in the work sequence are extracted from the dynamic work schedule as a set of subsequent tasks. For each subsequent task in the set, the estimated start time of the subsequent task is extended by the duration deviation value from the original estimated start time. When the actual execution time exceeds the estimated time, the extension direction is to postpone; when the actual execution time is less than the estimated time, the extension direction is to advance. The updated estimated start time covers the original estimated start time of the subsequent task in the dynamic work schedule.

[0060] In practice, the process of triggering a review instruction for the first maintenance unit based on the work completion signal is as follows: Upon receiving the work completion signal, the system automatically generates a review instruction. This instruction includes the first maintenance unit identifier, a trigger timestamp, and a review type field, which is set to vegetation coverage review. The review instruction is then published to the task instruction queue, where the image acquisition scheduling module retrieves the instruction and distributes it to the image acquisition device.

[0061] In practice, the process of obtaining the review image corresponding to the review instruction is as follows: The image acquisition scheduling module queries the address of the image acquisition device corresponding to the first maintenance unit based on the first maintenance unit identifier in the review instruction. The image acquisition device is a network camera installed on a fixed pole around the first maintenance unit or a drone that is linked and invoked. A shooting instruction is sent to the image acquisition device, requesting the acquisition of an orthographic image of the first maintenance unit. After the image acquisition device executes the shooting, it returns a digital image file, which is then marked as the review image.

[0062] In the specific implementation, the process of generating a pixel difference heatmap by comparing the verification image with the standard reference image of the first maintenance unit pixel by pixel is as follows: A standard reference image corresponding to the identifier of the first maintenance unit and taken from the same shooting angle is extracted from the maintenance reference image library. The standard reference image is a pre-taken standard image of the first maintenance unit under the condition of meeting maintenance standards. The verification image and the standard reference image are registered to align the same plant objects in the two images at pixel positions. The color space difference value is calculated pixel by pixel between the registered verification image and the standard reference image. The color space difference value is calculated using Euclidean distance in the CIELab color space. A blank canvas of the same size as the verification image is created. For each pixel position, when the color space difference value is greater than a preset difference threshold, a warm color value is assigned to the corresponding pixel on the blank canvas; when the color space difference value is less than or equal to the preset difference threshold, a cool color value is assigned, thereby generating a pixel difference heatmap. The preset difference threshold is set to the Euclidean distance value of 15 in the CIELab color space, based on the perception threshold of plant leaf color changes.

[0063] In some embodiments, the process of calculating the actual maintenance coverage rate based on the pixel difference heatmap uses the following formula:

[0064]

[0065] in, This indicates the actual maintenance coverage rate. This indicates the number of pixel columns in the pixel difference heatmap. This indicates the number of pixel rows in the pixel difference heatmap. Indicates the pixel difference heatmap at the th position Liede The pixel difference value at the row pixel position. This indicates a preset difference threshold. Indicates an indicator function, when The time indicator function takes a value of 1, otherwise it takes a value of 0. This represents the total number of pixels in the pixel difference heatmap whose pixel difference value is less than a preset difference threshold. This represents the total pixel area of ​​the target maintenance region of the first maintenance unit in the standard reference image. The method for obtaining the plant cover foreground region is to extract the plant cover foreground region from the standard reference image using an image segmentation model, and then count the total number of pixels contained in the plant cover foreground region. Actual maintenance coverage rate. The value range is a real number greater than or equal to 0 and less than or equal to 1. The closer the actual maintenance coverage rate is to 1, the more sufficient the maintenance operation coverage is.

[0066] In practice, the process of re-adding the first task to the execution queue when the actual maintenance coverage rate is lower than the coverage rate threshold is as follows: The coverage rate threshold is set to 0.85, based on the minimum acceptable maintenance coverage rate stipulated in the standards for acceptance of landscaping maintenance quality. The calculated actual maintenance coverage rate is compared with the coverage rate threshold. When the actual maintenance coverage rate is less than the threshold, the first task is marked as a substandard task, and its task identifier, first maintenance unit identifier, and current timestamp are reinserted to the end of the execution queue. Tasks in the execution queue will be rescheduled for execution during the subsequent dynamic task schedule generation process.

[0067] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application.

Claims

1. A method for managing landscaping projects based on big data and AI, characterized in that: The method includes: Based on historical maintenance records and real-time environmental monitoring data of landscaping projects, a multi-dimensional evaluation vector of plant growth status is constructed. Based on the multidimensional evaluation vector and the preset plant phenological calendar, the growth deviation index of each maintenance unit is generated; Based on the growth deviation index, key intervention areas are delineated through spatial clustering, and differentiated maintenance instructions are generated based on the key intervention areas; The differentiated maintenance instructions are matched with the material inventory data to output a material allocation plan; Based on the aforementioned material allocation plan, generate the work sequence for each maintenance unit; Based on the work sequence and real-time weather forecast data, the execution priority of each work in the work sequence is adjusted to obtain a dynamic work schedule. According to the dynamic work schedule, maintenance tasks are distributed to the corresponding work terminals.

2. The method for managing landscaping projects based on big data and AI according to claim 1, characterized in that, The steps for constructing a multidimensional assessment vector of plant growth status based on historical maintenance records and real-time environmental monitoring data of landscaping projects are as follows: The historical pruning time, historical fertilization amount, and historical number of pest and disease occurrences of each maintenance unit are obtained as the first type of static attributes. Real-time soil moisture, real-time light intensity, and real-time air temperature of each maintenance unit are obtained as the first type of dynamic attributes; Align the first type of static attributes with the first type of dynamic attributes according to timestamps to form an attribute sequence for each maintenance unit; The attribute sequence is normalized to generate the multidimensional evaluation vector.

3. The method for managing landscaping projects based on big data and AI according to claim 2, characterized in that, The step of generating the growth deviation index of each maintenance unit based on the multidimensional evaluation vector and the preset plant phenological calendar is as follows: Extract the standard leaf area index range and standard plant height range for each plant on the current date from the plant phenological calendar; The leaf area deviation component is obtained by performing a difference calculation between the actual leaf area index in the multidimensional evaluation vector and the standard leaf area index range. The difference between the actual plant height in the multidimensional evaluation vector and the standard plant height range is calculated to obtain the plant height deviation component. The growth deviation index is generated by weighted summation of the leaf area deviation component and the plant height deviation component.

4. The method for managing landscaping projects based on big data and AI according to claim 3, characterized in that, The step of delineating key intervention areas through spatial clustering based on the growth deviation index, and generating differentiated maintenance instructions based on the key intervention areas, specifically includes: The geographical coordinates of each maintenance unit are combined with the corresponding growth deviation index to form a spatial deviation point set; In the set of spatial deviation points, the point with the maximum deviation index is used as the initial seed, and the neighborhood radius is expanded outward. Points in the neighborhood with a deviation index greater than a preset threshold are classified into the same candidate region. The number of maintenance units within the candidate area is counted. When the number of maintenance units exceeds a threshold, the candidate area is marked as the key intervention area. For each key intervention area, the deviation type of each maintenance unit within the area is statistically analyzed, and the differentiated maintenance instructions corresponding to the deviation type are generated.

5. The landscaping project management method based on big data and AI according to claim 4, characterized in that, The step of matching the differentiated maintenance instructions with the material inventory data and outputting the material allocation plan is as follows: Extract the required material types and quantities from the differentiated maintenance instructions; Query the material inventory data to obtain the current inventory of the required material type in each warehouse; Based on the current inventory level and the demand level, calculate the material satisfaction level of each warehouse; Based on the material satisfaction level from high to low, supply warehouses are selected to generate the material allocation plan.

6. The method for managing landscaping projects based on big data and AI according to claim 5, characterized in that, The step of generating the work sequence for each maintenance unit based on the material allocation plan is as follows: Extract the material arrival time of each maintenance unit from the aforementioned material allocation plan; Obtain the growth deviation index corresponding to each maintenance unit, and sort the maintenance units from largest to smallest according to the growth deviation index to obtain the basic operation sequence; Based on the arrival time of the materials, the maintenance units in the basic operation sequence that have not yet received the materials are arranged later to generate the operation sequence.

7. The method for managing landscaping projects based on big data and AI according to claim 6, characterized in that, The step of adjusting the execution priority of each task in the task sequence based on the task sequence and real-time weather forecast data to obtain a dynamic task schedule is as follows: Extract the probability of rainfall and wind speed level for a future preset time period from the real-time weather forecast data; For each task in the task sequence, determine the conflict relationship between the task type and the rainfall probability and the wind speed level; When the conflict relationship is conflicting, the execution priority of the job is downgraded to obtain the adjusted priority; The job order in the job sequence is rearranged according to the adjusted priority to obtain the dynamic job schedule.

8. The method for managing landscaping projects based on big data and AI according to claim 7, characterized in that, The step of distributing maintenance tasks to corresponding work terminals according to the dynamic work schedule is as follows: Extract the first task to be executed in the current time period and its corresponding first maintenance unit from the dynamic work schedule; Obtain the first geographic coordinates of the first maintenance unit and the operation procedure file of the first operation; Package the first geographic coordinates, the operation procedure file, and the estimated duration of the first operation into a first task package; The first task packet is sent to the work terminal associated with the first maintenance unit via a wireless communication link.

9. The method for managing landscaping projects based on big data and AI according to claim 8, characterized in that, After sending the first task packet to the work terminal associated with the first maintenance unit via the wireless communication link, the method further includes: Receive the job start signal and job end signal fed back by the job terminal; Calculate the actual execution time of the first job based on the job start signal and the job end signal; The actual execution time is compared with the estimated time to obtain the time deviation; When the duration deviation exceeds a preset deviation threshold, subsequent tasks are extracted from the dynamic task schedule, and the estimated start time of the subsequent tasks is updated based on the duration deviation.

10. The method for managing landscaping projects based on big data and AI according to claim 9, characterized in that, After receiving the job start signal and job end signal from the job terminal, the method further includes: The operation completion signal triggers a verification instruction for the first maintenance unit; Obtain the verification image corresponding to the verification instruction, compare the verification image with the standard reference image of the first maintenance unit pixel by pixel, and generate a pixel difference heat map; The actual maintenance coverage rate is calculated based on the pixel difference heatmap. When the actual maintenance coverage rate is lower than the coverage rate threshold, the first job is added back to the queue to be executed.

Images