Crown feature monitoring system and method in litsea coreana tree crown cultivation process

Abstract

The invention discloses a crown feature monitoring system and method in the eagle tea tree crown cultivation process, and relates to the technical field of agricultural planting, and the method comprises the steps: carrying out the cultivation dimension hierarchy analysis of multi-dimensional monitoring data in the eagle tea tree crown cultivation process, and generating a crown feature dimension hierarchy map; performing fine splitting on the multi-dimensional monitoring data to form a plurality of crown feature information units; when the multi-dimensional monitoring data is not attached with a cultivation attribute identifier, generating a cultivation attribute identifier corresponding to the multi-dimensional monitoring data based on a preset large language model and a core feature summary of the multi-dimensional monitoring data; and building a target cultivation monitoring database. According to the method, through multi-database attribute merging and scenarized pruning optimization and in combination with a multi-level backtracking matching mechanism, the data matching success rate is decided, and the cultivation scheme is improved in a targeted manner; structured monitoring data is injected through an agricultural scene large language model, and a generated cultivation scheme is scientific and can be implemented.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of agricultural planting, and in particular to a tree crown feature monitoring system and method in a tree crown cultivation process of a Camellia sinensis var. assamica. BACKGROUND

[0002] In recent years, with the large-scale development of Camellia sinensis var. assamica planting industry, the scientificity of tree crown cultivation has an increasingly significant impact on yield and quality, and tree crown feature monitoring is a core prerequisite for precise cultivation. However, the following key problems exist in the prior art: Lack of unified structured organization of multi-source monitoring data: Camellia sinensis var. assamica tree crown monitoring data sources are complex (including sensor real-time collection, unmanned aerial vehicle remote sensing, manual measurement, etc.), data formats are different, dimensions are chaotic, and it is difficult to efficiently integrate and call according to cultivation needs; Cultivation attribute identification is missing or not unified: part of the monitoring data does not carry key attribute information such as cultivation stage and monitoring site, and the attribute terms of different data sources are inconsistent (such as mixed use of “tree crown outside” and “tree crown periphery”), resulting in poor data correlation; Low efficiency of data retrieval and decision matching: traditional monitoring data is mostly stored in scattered form, lacks hierarchical organization, precise data retrieval for specific cultivation decisions (such as high-yield fertilization in peak production period and stress resistance pruning) is difficult, and redundant data traversal volume is large; Cultivation scheme lacks data precision support: existing cultivation schemes mostly rely on experience and are difficult to dynamically adjust in combination with real-time, multi-dimensional tree crown feature data, and are insufficient in pertinence and implementability.

[0003] Therefore, there are still problems of low structured degree of Camellia sinensis var. assamica tree crown cultivation monitoring data, low retrieval efficiency, and inaccurate matching of cultivation scheme and data in the prior art, which restricts the scientificity and effectiveness of tree crown cultivation. SUMMARY

[0004] This section is intended to summarize some aspects of the embodiments of the present application and briefly introduce some preferred embodiments. Some simplifications or omissions may be made in this section and the abstract and title of the specification of the present application in order to avoid obscuring the purpose of this section, the abstract and the title, and such simplifications or omissions cannot be used to limit the scope of the present application.

[0005] In view of the above problems existing in the prior art, the present application is proposed.

[0006] To solve the above technical problems, the present application provides the following technical scheme: a tree crown feature monitoring method in a tree crown cultivation process of Camellia sinensis var. assamica, comprising the following steps: Performing cultivation dimension hierarchical analysis on multi-dimensional monitoring data in the tree crown cultivation process of Camellia sinensis var. assamica to generate a tree crown feature dimension hierarchical atlas; based on the tree crown feature dimension hierarchical atlas, the multi-dimensional monitoring data is finely split to form a plurality of tree crown feature information units; When the multi-dimensional monitoring data is not attached with the cultivation attribute identifier, based on the preset large language model and the core feature summary of the multi-dimensional monitoring data, the cultivation attribute identifier corresponding to the multi-dimensional monitoring data is generated; Based on the cultivation attribute identifier and the plurality of tree crown feature information units of at least one group of multi-dimensional monitoring data, a target cultivation monitoring database is built; the target cultivation monitoring database includes a tree crown cultivation attribute hierarchical atlas composed of the cultivation attribute identifier of the multi-dimensional monitoring data.

[0007] As a preferred scheme of the tree crown feature monitoring system and method in the process of cultivating the eagle tea tree crown, wherein: the multi-dimensional monitoring data in the process of cultivating the eagle tea tree crown is analyzed in the cultivation dimension hierarchical analysis, and a tree crown feature dimension hierarchical atlas is generated, which includes: Based on the structured identification information, data format features or self-defined analysis rules of the multi-dimensional monitoring data, the cultivation stage dimension, monitoring site dimension and feature type dimension corresponding to the multi-dimensional monitoring data are extracted; if the extraction fails, the collection basic information of the multi-dimensional monitoring data is used as the top layer analysis dimension; based on the cultivation stage dimension, monitoring site dimension and feature type dimension, or the top layer analysis dimension, the tree crown feature dimension hierarchical atlas is generated.

[0008] As a preferred scheme of the tree crown feature monitoring system and method in the process of cultivating the eagle tea tree crown, wherein: the multi-dimensional monitoring data in the process of cultivating the eagle tea tree crown is analyzed in the cultivation dimension hierarchical analysis, and a tree crown feature dimension hierarchical atlas is generated, which includes: converting the multi-dimensional monitoring data into a unified standardized data format; based on the standardized data format, the initial dimension information and the core feature summary of the multi-dimensional monitoring data are extracted; based on the preset large language model, the initial dimension information and the core feature summary, the tree crown feature dimension hierarchical atlas corresponding to the eagle tea tree crown cultivation is generated.

[0009] As a preferred scheme of the tree crown feature monitoring system and method in the process of cultivating the eagle tea tree crown, wherein: based on the tree crown feature dimension hierarchical atlas, the multi-dimensional monitoring data is finely split to form a plurality of tree crown feature information units, which includes: based on the preset time window and the preset sampling interval, the multi-dimensional monitoring data corresponding to each dimension branch in the tree crown feature dimension hierarchical atlas is sequentially cut to form a plurality of tree crown feature information units.

[0010] As a preferred embodiment of the canopy feature monitoring system and method for the cultivation process of the Eagle Tea Tree described in this invention, after constructing the target cultivation monitoring database based on at least one set of cultivation attribute identifiers of the multi-dimensional monitoring data and multiple canopy feature information units, the system further includes: In response to a canopy cultivation decision request, multiple target cultivation monitoring databases corresponding to the canopy cultivation decision request are determined; The cultivation attribute identifiers of multiple target cultivation monitoring databases are merged to obtain a dynamic cultivation monitoring database and a dynamic attribute identifier corresponding to the dynamic cultivation monitoring database; The tree canopy cultivation attribute hierarchy map composed of the dynamic attribute identifiers is optimized and pruned to obtain simplified attribute identifiers; Obtain the canopy feature information unit from the simplified attribute identifier that matches the canopy cultivation decision request; Based on the preset large language model, the canopy cultivation decision request, and the canopy feature information unit, a cultivation scheme suggestion corresponding to the canopy cultivation decision is obtained.

[0011] As a preferred embodiment of the canopy feature monitoring system and method for the cultivation of eagle tea tree canopies according to the present invention, wherein: the optimization and pruning of the canopy cultivation attribute hierarchy map composed of the dynamic attribute identifiers to obtain simplified attribute identifiers includes: Based on a pre-defined large language model, the semantic correlation degree of cultivation between any two nodes in the dynamic attribute identifier is calculated; the target cultivation scenario corresponding to the canopy cultivation decision request is identified, and the core priority dimension under the target cultivation scenario is determined; If the semantic correlation between two nodes is higher than the preset correlation threshold and neither belongs to the core priority dimension, then the two nodes will be merged. If the merged node still contains non-core priority child nodes, then remove the non-core priority child nodes and retain the core priority child nodes to obtain the simplified attribute identifier.

[0012] As a preferred embodiment of the canopy feature monitoring system and method for the cultivation of eagle tea tree canopies according to the present invention, wherein: the canopy feature information unit that matches the canopy cultivation decision request in the simplified attribute identifier includes: Obtain the dimension node in the simplified attribute identifier that matches the canopy cultivation decision request; obtain the canopy feature information unit under the dimension node that matches the canopy cultivation decision request; If the canopy feature information unit under the dimension node that matches the canopy cultivation decision request does not meet the preset conditions, then the parent node of the dimension node is selected, and the canopy feature information unit under the parent node that matches the canopy cultivation decision request is obtained. If the tree crown feature information unit under the parent node still does not meet the preset conditions, then continue to backtrack upwards to the root node until a tree crown feature information unit that meets the preset conditions is obtained.

[0013] A monitoring system for monitoring canopy characteristics in the above-mentioned method of eagle tea tree canopy cultivation process includes: a cultivation dimension analysis module, used to perform cultivation dimension hierarchical analysis on multi-dimensional monitoring data in the eagle tea tree canopy cultivation process, and generate a canopy characteristic dimension hierarchical map; The feature data splitting module is used to perform fine splitting of the multi-dimensional monitoring data based on the tree canopy feature dimension hierarchy map, forming multiple tree canopy feature information units; The attribute identifier generation module is used to generate the cultivation attribute identifier corresponding to the multi-dimensional monitoring data based on a preset large language model and a summary of the core features of the multi-dimensional monitoring data when the multi-dimensional monitoring data does not have a cultivation attribute identifier. The monitoring database construction module is used to construct a target cultivation monitoring database based on at least one set of cultivation attribute identifiers of the multi-dimensional monitoring data and multiple canopy feature information units; the target cultivation monitoring database includes a canopy cultivation attribute hierarchy map composed of cultivation attribute identifiers of the multi-dimensional monitoring data.

[0014] The present invention also discloses a computer device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps of the above-described method for monitoring crown characteristics during the cultivation of eagle tea tree crowns.

[0015] The present invention also discloses a computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the steps of the above-described method for monitoring crown characteristics during the cultivation of eagle tea tree crowns.

[0016] The beneficial effects of this invention are: 1. Through standardized preprocessing of multi-source data and construction of dimensional maps adapted to multiple methods, structured analysis of data with different collection specifications was achieved, resulting in a significant improvement in dimensional analysis coverage and robustness of data structure processing; 2. By dynamically configuring time windows and sampling intervals according to the cultivation stage and combining attribute index optimization, the monitoring data is refined and efficiently retrieved, the semantic coherence of information units is improved, and the retrieval response time is shortened; by generating cultivation attribute labels in both manual and automatic modes and combining the semantic understanding capabilities of the domain's large language model, the matching degree between labels and data features is improved, reducing the cost of manual annotation. 3. By merging multiple database attributes and optimizing pruning in a scenario-based manner, combined with a multi-level backtracking matching mechanism, the success rate of decision data matching and the pertinence of cultivation plans are improved; by injecting structured monitoring data into the agricultural scenario big language model, the generated cultivation plan is scientific and feasible, significantly improving the scientificity and effectiveness of eagle tea tree canopy cultivation. Attached Figure Description

[0017] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort. Wherein: Figure 1 This is a schematic diagram of the overall process of a method for monitoring crown characteristics during the cultivation of eagle tea tree crowns, as proposed in this invention. Figure 2 This is a logic block diagram of a preferred embodiment of a method for monitoring crown characteristics during the cultivation of Eagle Tea Trees proposed in this invention. Detailed Implementation

[0018] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0019] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.

[0020] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in different places in this specification does not necessarily refer to the same embodiment, nor is it a single or selective embodiment that is mutually exclusive with other embodiments.

[0021] Reference Figure 1 As an embodiment of the present invention, a canopy characteristic monitoring system and method for the cultivation process of Eagle Tea Trees are provided. The method includes the following steps: Step 1: Analyze the multi-dimensional monitoring data during the cultivation of the Eagle Tea tree canopy at different cultivation dimensions to generate a hierarchical map of canopy features. Specifically, the canopy feature dimensional hierarchy map can be a tree-like data structure describing the hierarchical relationship of monitoring data on the canopy cultivation of Eagle Tea trees. The canopy feature dimensional hierarchy map includes dimension nodes, which can correspond to specific stages in the cultivation process, canopy monitoring locations, or feature indicator types. For example, the canopy feature dimensional hierarchy map can be obtained by parsing the metadata identifiers and collection parameter labels of multi-dimensional monitoring data, and can also be constructed by combining hierarchical logical traversal of the cultivation cycle.

[0022] For example, a canopy feature dimension hierarchy map can include one or more generated hierarchical structures such as acquisition stage identifiers for real-time sensor monitoring data, canopy region partitioning for UAV remote sensing data, and feature index classification for manually measured data. In a specific embodiment, parent-child dimension node relationships can be established layer by layer by identifying cultivation stage labels, monitoring site annotations, and their data classification attributes in multi-dimensional monitoring data, forming a canopy feature dimension hierarchy map with a nested cultivation logic structure. By extracting the inherent cultivation association information from the multi-dimensional monitoring data, a structured expression of the overall framework of the monitoring data can be achieved, thereby supporting subsequent division and organization of the monitoring data according to cultivation dimension logic.

[0023] Step 2: Based on the hierarchical map of tree canopy features, the multi-dimensional monitoring data is finely divided into multiple tree canopy feature information units; Specifically, a canopy feature information unit can be the smallest information unit carrying single feature data under a specific cultivation scenario. Its content may include, but is not limited to, feature value sequences within a time period, feature change trend curves, and feature distribution heatmaps. It can be generated by identifying the natural segmentation of data based on the dimensional node relationships of the canopy feature dimensional hierarchy map, combined with time window division or feature threshold filtering techniques. For example, the division of canopy feature information units can be based on one or more of the following: cultivation stage nodes, canopy location nodes, feature indicator nodes, or data collection time periods. In a specific embodiment, when the canopy feature dimensional hierarchy map contains dimensional nodes such as "peak production period → canopy periphery → bud and leaf density," the monitoring data within the corresponding range can be extracted into independent canopy feature information units, retaining their cultivation dimensional attribution information. By deconstructing multi-dimensional monitoring data according to both cultivation dimensions and data features, information units with clear boundaries and well-defined dimensions can be formed, providing structurally consistent data input for subsequent cultivation attribute identification generation and monitoring database organization.

[0024] Step 3: When the multi-dimensional monitoring data does not include a breeding attribute identifier, generate the breeding attribute identifier corresponding to the multi-dimensional monitoring data based on the preset large language model and the core feature summary of the multi-dimensional monitoring data; Specifically, cultivation attribute identifiers can be classification identifiers or scene labels used to describe the cultivation-related characteristics of the Eagle Tea tree canopy monitoring data. For example, the type of cultivation attribute identifier can be pre-set for the Eagle Tea tree cultivation scene, or it can be dynamically generated based on a summary of the core characteristics of the monitoring data. For example, if information such as the cultivation stage and monitoring location has been labeled via sensors or manually filled in during monitoring data collection, then the multi-dimensional monitoring data can be considered to have accompanying cultivation attribute identifiers.

[0025] When multi-dimensional monitoring data does not include corresponding cultivation attribute identifiers, the core features of the monitoring data can be extracted as input, and a pre-defined large language model can be used for semantic understanding and feature association of the cultivation scenario. The large language model then extracts and generates cultivation attribute identifiers. In this embodiment, the pre-defined large language model may include, but is not limited to, pre-trained models in the agricultural field based on the Transformer architecture, such as the agricultural version of GPT, the BERT model for crop condition analysis, etc.

[0026] In one specific embodiment, a core feature overview (such as "the average monthly growth of peripheral branches of mountain eagle tea trees during the peak production period in May-June 2025 was 5cm") can be extracted from multi-dimensional monitoring data and input into a large language model. The model identifies core cultivation elements through understanding the context of the cultivation scenario and performs semantic normalization processing in conjunction with the eagle tea tree cultivation knowledge base, outputting a set of attribute labels that conform to the preset cultivation classification system. For example, for the core feature overview "the average monthly growth of peripheral branches of mountain eagle tea trees during the peak production period in May-June 2025 was 5cm", the system can generate one or more cultivation attribute labels such as "high-yield cultivation target", "peak production period", "canopy periphery", and "branch growth rate". By leveraging the cultivation scenario semantic modeling capability of the large language model, automated attribute labeling of monitoring data can be achieved, reducing manual labeling costs while improving the matching degree between cultivation attribute labels and cultivation features of monitoring data.

[0027] Furthermore, during the process of calling the large language model, the dimensions of the cultivation attribute identifiers can be defined. For example, the large language model can be instructed to label the cultivation target attributes, growth stage attributes, and canopy part attributes from the monitoring data, thereby obtaining multi-dimensional cultivation attribute identifiers.

[0028] Step 4: Based on the cultivation attribute identifiers of at least one set of multi-dimensional monitoring data and multiple canopy feature information units, construct a target cultivation monitoring database; the target cultivation monitoring database includes a canopy cultivation attribute hierarchy map composed of cultivation attribute identifiers of multi-dimensional monitoring data.

[0029] The target cultivation monitoring database can be a structured collection of monitoring information used to store, organize, and retrieve data on the canopy characteristics of *Tea lanceolata* trees. This database includes a hierarchical map of canopy cultivation attributes composed of cultivation attribute identifiers from multi-dimensional monitoring data. For example, it can be constructed by integrating canopy feature information units, cultivation attribute indexes, and the hierarchical map structure. The cultivation attribute index can be established using time-series indexes or hash mapping to achieve the association mapping between cultivation attribute identifiers and corresponding canopy feature information units, supporting rapid location of target monitoring data according to cultivation scenarios. The hierarchical map of canopy cultivation attributes can be a tree-like cultivation dimension classification system composed of cultivation attribute identifiers, organized hierarchically through scenario associations between cultivation attributes. In an exemplary embodiment, "high-yield cultivation target" can be set as the parent node, with "peak production period → canopy periphery → bud and leaf density" as its subordinate hierarchical child nodes. Furthermore, in one specific embodiment, graph database technology (such as Neo4j) can be used to store the hierarchical nodes of cultivation attributes and their interrelationships to support multi-dimensional path queries (such as "peak production period + outer canopy → monitoring data of corresponding fertilization schemes") and dynamic expansion of monitoring data in the Eagle Tea tree cultivation scenario. Through the integration of multi-dimensional cultivation data, a canopy monitoring data management system that combines semantic relevance of the cultivation scenario with navigable data structure can be formed.

[0030] This embodiment provides a method for monitoring the canopy features of Eagle Tea trees. It obtains a canopy feature hierarchy map by performing cultivation dimension hierarchical analysis on multi-dimensional monitoring data. Based on this map, the monitoring data is finely segmented to obtain multiple canopy feature information units. In the absence of cultivation attribute identifiers in the monitoring data, a pre-set large language model is used in conjunction with a core feature overview to generate corresponding cultivation attribute identifiers. The cultivation attribute identifiers and canopy feature information units are integrated to complete the construction of a target cultivation monitoring database containing a canopy cultivation attribute hierarchy map composed of cultivation attribute identifiers. Through the synergistic effect of structured analysis of cultivation dimensions, scenario-driven attribute identifier generation, and cultivation dimension organization mechanisms, the redundant data traversal during canopy feature data retrieval can be effectively reduced, improving the matching accuracy between cultivation attribute identifiers and monitoring data features. Furthermore, by optimizing the data retrieval path for cultivation decisions through the joint architecture of the canopy cultivation attribute hierarchy map and the cultivation attribute index, the method achieves the technical effect of improving the utilization efficiency and decision-making accuracy of Eagle Tea tree canopy cultivation data.

[0031] In one embodiment, multi-dimensional monitoring data during the cultivation process of the Eagle Tea Tree canopy is analyzed at the cultivation dimension level to generate a canopy feature dimension level map, including: Based on the structured identification information, data format characteristics, or custom parsing rules of multi-dimensional monitoring data, extract the cultivation stage dimension, monitoring part dimension, and feature type dimension corresponding to the multi-dimensional monitoring data; If extraction fails, the basic information collected from the multi-dimensional monitoring data will be used as the top-level analytical dimension. Based on the dimensions of cultivation stage, monitoring location, and feature type, or the top-level analysis dimension, a hierarchical map of tree canopy features is generated.

[0032] Specifically, structured identification information can be preset metadata tags used to identify cultivation-related attributes in multi-dimensional monitoring data, which can be obtained by parsing the attribute configuration fields of the monitoring data acquisition terminal. For example, structured identification information can include, but is not limited to, one or more preset attribute tags such as "seedling stage", "upper canopy", and "canopy width", which can reflect the explicit annotation of the cultivation characteristics of Eagle Tea Tree during data collection.

[0033] Data format features can be format attributes that reflect differences in cultivation dimensions in multi-dimensional monitoring data. They can be determined by analyzing one or more of the following: data timestamp format, numerical classification range, and data file naming rules. For example, data format features can also include one or more of the following: "2025Y05M" in the timestamp corresponds to "peak production period", the numerical range "5-8cm" corresponds to "twig growth rate", and the file name containing "outer leaf area" corresponds to "canopy periphery". These features are used to infer potential cultivation dimensions in monitoring data that do not carry explicit identifiers.

[0034] Custom parsing rules can be data analysis strategies used to identify semantic or format patterns of cultivation dimensions in monitoring data. These rules can be implemented through regular expression matching, feature keyword association, data classification models, and other methods. For example, custom parsing rules can include matching the keyword "seedling stage" in timestamps, associating the "bud and leaf density" feature type corresponding to the numerical sequence, identifying one or more of the "canopy interior" locations corresponding to the data collection coordinates, thereby extracting implicit cultivation dimension information from monitoring data without explicit attribute identifiers.

[0035] The cultivation stage dimension can be a stage marker in the growth cycle of the Eagle Tea Tree, the monitoring part dimension can be a regional division marker of the canopy, and the feature type dimension can be a specific growth indicator marker corresponding to the monitoring data. The three together constitute a hierarchical structural unit with cultivation scenario association (such as "peak production period → outer canopy → bud and leaf density").

[0036] Extracting dimensions such as cultivation stage, monitoring location, and feature type can be done by applying one or more of structured identification information, data format features, and custom parsing rules in parallel. For example, applying these three types of information and rules in parallel can perform semantic fusion and redundancy removal on multi-source extraction results, thereby identifying cultivation-related dimensions in monitoring data from different sources and improving the completeness and matching accuracy of dimension extraction.

[0037] The basic information collected can be essential metadata identifiers for multi-dimensional monitoring data collection, which can be obtained by reading log fields from the data collection terminal (such as collection batch, tea tree number, and collection date). When it is determined that the aforementioned dimension extraction process has not output a valid cultivation dimension node and the determination fails, the default dimension generation mechanism is triggered, using the basic information collected as the top-level parsing dimension. This ensures that even in the extreme case of a complete lack of cultivation association identifiers, a cultivation dimension framework containing at least the top-level node can still be generated, avoiding the interruption of subsequent data splitting processes due to missing dimensions.

[0038] Based on the dimensions of cultivation stage, monitoring location, and feature type, or the top-level analysis dimension, a hierarchical map of tree canopy features is generated. For example, a tree-like data structure can be constructed, with the cultivation stage dimension as the root node, the monitoring location dimension as its child nodes, and the feature type dimension as the child nodes of the monitoring location dimension, forming a multi-level map structure of "cultivation stage - monitoring location - feature type". If only some dimensions are extracted (e.g., only the cultivation stage dimension) or only the top-level dimension is generated from collected basic information, a simplified map structure of the corresponding level is constructed. In a specific embodiment, if the monitoring data successfully extracts "peak production period" as the cultivation stage dimension, "canopy periphery" as the monitoring location dimension, and "bud and leaf density" as the feature type dimension, the map can contain three levels of nodes; if a batch of data can only generate the top-level dimension from the collected basic information "batch 202506 tea trees", the map only contains that top-level node as the sole dimension. Simultaneously, the hierarchical depth and branch structure of the map are dynamically adjusted according to the actual extraction results, preserving the cultivation-related logic of the monitoring data while adapting to data source forms with different collection specifications.

[0039] This embodiment provides a method for monitoring the canopy features of *Tea lanceolata* trees. By performing hierarchical analysis of multi-dimensional monitoring data on cultivation dimensions, a hierarchical map of canopy features is generated. This includes extracting cultivation stages, monitoring locations, and feature types based on structured identifiers, data format features, or custom analysis rules. If extraction fails, the collected basic information is used as the top-level analysis dimension, and a corresponding hierarchical map of canopy features is constructed based on the extraction results. Multi-dimensional cultivation information is identified by fusing preset identifiers, format attributes, and semantic rules with the collected basic information as a fallback strategy to ensure the continuity of the analysis process. This allows for the generation of stable and reasonable cultivation dimension maps when processing monitoring data from both standard and non-standard sources. It exhibits strong adaptability, especially when dealing with monitoring data lacking cultivation identifiers or with chaotic formats, significantly improving the coverage and reliability of cultivation dimension analysis. This, in turn, enhances the accuracy of subsequent data segmentation and avoids data organization deviations caused by missing or misjudged dimensions. This method effectively improves the robustness of structured processing of *Tea lanceolata* monitoring data.

[0040] In one embodiment, multi-dimensional monitoring data during the cultivation process of the Eagle Tea Tree canopy is analyzed at the cultivation dimension level to generate a canopy feature dimension level map, including: Convert multi-dimensional monitoring data into a unified standardized data format; based on the standardized data format, extract the initial dimensional information and core feature overview of the multi-dimensional monitoring data; Based on the pre-set large language model, initial dimensional information, and core feature overview, a hierarchical map of canopy features corresponding to the cultivation of Eagle Tea Tree canopy is generated.

[0041] Specifically, a unified standardized data format can be the text of monitoring data for the eagle tea tree, expressed using structured fields. Fixed fields are used to annotate the data collection attributes, numerical information, and associated notes. For example, this can be achieved by converting the original monitoring data using a data conversion tool. The original monitoring data includes, but is not limited to, one or more of the following: real-time sensor data, UAV remote sensing data, and digitized data from handwritten records. In one specific embodiment, the monitoring data format conversion can utilize existing agricultural data standardization libraries or conversion scripts to uniformly convert multi-dimensional monitoring data from different sources into a unified standardized data format (such as JSON format, containing fixed fields such as "collection time," "tea tree number," "characteristic value," and "collection location"). This standardizes the representation of the monitoring data, eliminates processing compatibility issues caused by differences in data collection terminals, and retains the cultivation-related attribute information in the original monitoring data.

[0042] Initial dimension information can be the basic dimension content of cultivation association clearly identified according to fixed field specifications in a standardized data format. For example, the content corresponding to fields such as "collection stage" and "monitoring site" can be identified by field matching or based on parsing tools to extract the corresponding initial dimensions such as cultivation stage and monitoring site and their correlation.

[0043] The core feature summary can be an information summary reflecting the core growth characteristics of the monitoring data. For example, it can be generated by statistically analyzing the range and mean trend of the "feature value" field in the standardized data, or by identifying the growth status description contained in the "remarks" field, such as "In May 2025, the average monthly growth of the outer branches of tea tree A01 during its peak production period was 5cm". In a specific embodiment, based on the converted unified standardized data format, the initial dimensional information and the core feature summary can be separated and extracted to obtain the basic cultivation dimensional framework and core growth characteristic information of the monitoring data.

[0044] The pre-set large language model can be a pre-trained model for agricultural scenarios based on the Transformer architecture. For example, it can be one or more models such as the agricultural version of GPT or the BERT model for crop condition analysis. The large language model receives initial dimensional information and a core feature overview as input. Through the correlation analysis of the cultivation scenario in the initial dimensions and the understanding of the growth feature distribution in the core feature overview, it outputs a hierarchical map of tree crown feature dimensions with cultivation logic coherence. In a specific embodiment, when there are missing correlations in the initial dimensions (such as only "monitoring part = outer canopy" is extracted without extracting the cultivation stage), the model can automatically supplement the cultivation stage dimension based on the keyword "peak production period" in the core feature overview; if the core feature overview mentions two related features, "twig growth + bud and leaf density", the model can infer and suggest adding "canopy growth feature" as a parent dimension node to associate the two child feature dimensions. Furthermore, by using prompt word engineering in agricultural cultivation scenarios (such as "based on the high-yield cultivation target of Laoying tea tree, generate a dimensional map of the 'cultivation stage-monitoring part-feature type' hierarchy"), the large language model can output text with a specified hierarchical structure. Then, through type conversion tools, the formatted text can be converted into node relationship objects that the graph database can recognize, thereby improving the efficiency of graph construction. By reconstructing the dimensional structure of monitoring data through a semantic-driven approach in cultivation scenarios, the limitations of traditional parsing methods that rely solely on explicit field identifiers can be overcome, and the ability to restore the dimensions of non-standard monitoring data can be improved.

[0045] This embodiment provides a method for monitoring the canopy features of Eagle Tea trees. By converting multi-dimensional monitoring data into a unified standardized data format, initial dimensional information and core feature summaries of the monitoring data are extracted based on the unified standardized data format. A hierarchical map of canopy features is generated based on a preset large language model, initial dimensional information, and core feature summaries. Data format standardization processing is used to eliminate the parsing differences of multi-source monitoring data. The dual information input of structured initial dimensions and unstructured core feature summaries enhances the model's understanding of the cultivation scenario. Furthermore, semantic analysis mechanisms are used to supplement missing dimensions and optimize the correlation between dimensions. This method can achieve high-precision restoration of the cultivation dimensional structure of the monitoring data, thereby generating a hierarchical map of canopy features that is more in line with the cultivation needs of Eagle Tea trees and is logically coherent.

[0046] In one embodiment, based on the hierarchical map of tree canopy features, the multi-dimensional monitoring data is finely segmented to form multiple tree canopy feature information units, including: Based on a preset time window and a preset sampling interval, the multi-dimensional monitoring data corresponding to each dimension branch in the tree canopy feature dimension hierarchy map are segmented sequentially to form multiple tree canopy feature information units.

[0047] Specifically, the preset time window can be a data processing mechanism used to control the time granularity of monitoring data for Eagle Tea trees. The window duration can be the length of the cultivation time interval covered by a single canopy feature information unit, and the time interval length can be measured using cultivation cycle-related units such as days, weeks, and months. The window duration can be set through cultivation scenario configuration files, user-inputted cultivation stage requirements, or system-preset default cycle values. In an exemplary embodiment, a 15-day time window can be set when processing monitoring data during the seedling stage of Eagle Tea; a 30-day time window can be set when processing monitoring data during the peak production period.

[0048] The preset sampling interval can be the length of time that the time window spans as it moves across the time series of monitored data. Its value is less than or equal to the time window length, and it is used to control the degree of temporal overlap between adjacent canopy feature information units. The sampling interval can be configured through system cultivation parameters or adaptively adjusted according to the rate of feature change at different growth stages of the Eagle Tea tree. When the sampling interval is less than the time window length, there will be a temporal overlap area between adjacent canopy feature information units, thus preserving the temporal continuity of the Eagle Tea tree growth data.

[0049] Multi-dimensional monitoring data is segmented based on preset time windows and preset sampling intervals. For example, the time region of monitoring data covered by each dimension branch can be determined according to the hierarchical map of canopy features. Preset time window duration and sampling interval parameters are read; these parameters can be adjusted according to the ideal data density of canopy feature information units or the needs of the cultivation decision-making scenario. By iterating through the time series of monitoring data, the time window is gradually slid along with the sampling interval, encapsulating the monitoring data within the window into independent canopy feature information units. By adjusting the time window duration and sampling interval, the temporal granularity of canopy feature information units can be flexibly controlled, avoiding data redundancy caused by excessively long cultivation stages. Overlapping time regions allow growth characteristic changes across time windows to be covered in multiple information units, improving the fault tolerance of data analysis during cultivation decision-making. Furthermore, appropriate segmentation strategies can be selected based on different cultivation scenarios of the Eagle Tea Tree. For example, for the seedling stage (fast growth rate), a smaller sampling interval (such as 7 days) can be set to capture subtle growth changes; for the peak production stage (stable growth rate), a sampling interval of the same length as the time window can be set to avoid data duplication and ensure data utilization efficiency.

[0050] This embodiment provides a method for monitoring the canopy features of *Tea lanceolata* var. *tigerica*. Based on a hierarchical map of canopy features, it finely segments multi-dimensional monitoring data to form multiple canopy feature information units. This includes: based on a preset time window and a preset sampling interval, sequentially segmenting the multi-dimensional monitoring data corresponding to each dimension branch in the hierarchical map of canopy features to form multiple canopy feature information units. By combining the hierarchical structure of canopy features with a parameterized time window mechanism to achieve fine-grained data segmentation, controlling the cultivation time coverage of information units using the time window duration, and adjusting the time overlap of adjacent information units through the sampling interval to preserve the temporal continuity of growth data, the method improves the overall semantic coherence of growth and the adaptability of cultivation decisions for canopy feature information units. It reduces the fragmentation of growth information caused by unreasonable time segmentation, achieving the technical effect of enhancing the accuracy of data support for cultivation decisions and improving the effectiveness of canopy cultivation measures for *Tea lanceolata* var. *tigerica*. In one embodiment, after constructing the target cultivation monitoring database based on cultivation attribute identifiers and multiple canopy feature information units from at least one set of multi-dimensional monitoring data, the method further includes: In response to canopy cultivation decision requests, multiple target cultivation monitoring databases corresponding to the canopy cultivation decision requests are identified; The cultivation attribute identifiers of multiple target cultivation monitoring databases are merged to obtain a dynamic cultivation monitoring database and the dynamic attribute identifiers corresponding to the dynamic cultivation monitoring database; The tree canopy cultivation attribute hierarchy map composed of dynamic attribute identifiers is optimized and pruned to obtain simplified attribute identifiers; Obtain the canopy feature information units from the simplified attribute identifiers that match the canopy cultivation decision request; Based on the pre-set large language model, tree canopy cultivation decision request, and tree canopy feature information unit, cultivation scheme suggestions corresponding to the tree canopy cultivation decision are obtained.

[0051] Specifically, the canopy cultivation decision request can be a user-input natural language command or demand description related to the cultivation of Eagle Tea trees, used to express the decision-making needs for specific cultivation measures. This can be obtained through methods such as receiving input from the cultivation management terminal interface, API calls from the agricultural IoT platform, or speech recognition conversion. Multiple target cultivation monitoring databases can be collections of monitoring data storing structured feature information units of Eagle Tea tree canopies and their cultivation attribute identifiers, used to support canopy feature retrieval and cultivation decision support for different cultivation scenarios, varieties, or planting areas. Each target cultivation monitoring database can be constructed using the Eagle Tea tree canopy feature monitoring method described in the above embodiment, which will not be elaborated upon in this embodiment.

[0052] Identifying multiple target cultivation monitoring databases corresponding to canopy cultivation decision requests can be achieved by semantically parsing the canopy cultivation decision requests, extracting cultivation keywords (such as "high-yielding type" and "peak-production pruning"), identifying the decision intent, and comparing them with the topic attribute identifiers of each cultivation monitoring database based on semantic matching or vector similarity calculation. This allows for the selection of a set of databases with high relevance, enabling dynamic association of multiple monitoring data sources according to cultivation needs. The dynamic cultivation monitoring database can be a monitoring data organization structure temporarily constructed to respond to specific canopy cultivation decision requests. It can contain canopy feature information units from multiple target cultivation monitoring databases and an integrated cultivation attribute system, forming a composite monitoring data network covering multiple cultivation scenarios through merging operations. The dynamic attribute identifier can be a unified cultivation attribute system used to organize and index canopy feature information units in the dynamic cultivation monitoring database. It can be generated through attribute terminology normalization, cultivation dimension hierarchy reconstruction, and scenario semantic fusion.

[0053] It is understandable that the cultivation attribute identifiers of multiple target cultivation monitoring databases may differ in structure and terminology. Merging the cultivation attribute identifiers of multiple target cultivation monitoring databases can include measures such as terminology standardization (e.g., unifying "twig growth amount" and "new shoot growth rate" into "twig growth characteristics"), semantic alignment (e.g., mapping "outer canopy" and "canopy periphery" to the same location dimension), and redundancy elimination (e.g., removing duplicate "collection time" attributes). For example, this can be achieved by constructing a unified attribute space for Eagle Tea Tree cultivation, mapping the original cultivation attribute identifiers of each database to this space, and merging and reorganizing them according to the semantic equivalence or dimensional hierarchy of cultivation scenarios. This can form a dynamic cultivation monitoring data organization structure that covers multiple cultivation scenarios and has a consistent structure.

[0054] Simplified attribute identifiers can be a subset of attributes that are highly relevant to the current canopy cultivation decision request after screening. This can be obtained by removing low-relevance branches (e.g., removing the "ornamental canopy morphology" attribute when cultivating a "high-yield" tree) or irrelevant dimensions (e.g., removing the non-canopy attribute "soil pH value" when making pruning decisions). The canopy cultivation attribute hierarchy map composed of dynamic attribute identifiers can be optimized and pruned. For example, the semantic similarity of each attribute node to the canopy cultivation decision request can be calculated, and a threshold can be set to filter low-scoring nodes. Alternatively, an algorithm based on cultivation decision information gain can be used to evaluate the contribution of attribute nodes to the decision and remove redundant branches. This can reduce the scope of monitoring data retrieval and improve the efficiency of subsequent cultivation decision data matching.

[0055] Canopy feature information units that match the canopy cultivation decision request can be obtained through a dual matching mechanism of cultivation attribute identification guidance and monitoring data semantics. For example, a hybrid retrieval strategy can be used to screen candidate information units with the best decision support by combining the cultivation attribute path weight (e.g., the path weight of "high-yielding type → peak production period → canopy periphery" is higher than that of "canopy periphery" alone) with the growth semantic similarity score between the canopy feature information unit data and the cultivation decision request.

[0056] Based on a pre-set large language model, tree canopy cultivation decision requests, and tree canopy feature information units, cultivation scheme suggestions corresponding to tree canopy cultivation decisions are obtained. This can be achieved through the Eagle Tea Tree Cultivation Scene Prompt Engineering method, which concatenates the cultivation decision request and the matched tree canopy feature information units into a structured input prompt (such as "Based on monitoring data of bud and leaf density <5 / cm² on the periphery of the canopy during the peak production period, provide a fertilization scheme for high-yielding Eagle Tea Trees"). This prompt is then fed into a pre-trained large language model in the agricultural field for reasoning and generation, allowing the model to reference real tree canopy monitoring data during the generation process.

[0057] This embodiment provides a method for monitoring the canopy characteristics of eagle tea trees. In response to canopy cultivation decision requests, it identifies multiple target cultivation monitoring databases. The cultivation attribute identifiers of these databases are grouped to form a dynamic cultivation monitoring database and corresponding dynamic attribute identifiers. The hierarchical graph formed by these dynamic attribute identifiers is optimized and pruned to obtain simplified attribute identifiers. Canopy feature information units matching the canopy cultivation decision requests are then extracted. Based on a preset large language model, the cultivation decision request, and the information units, cultivation plan suggestions are generated. This method enables cross-scenario integration of the cultivation attribute system, optimizes the scope of monitoring data retrieval by pruning the attribute hierarchical graph, and enhances the factual reliability and scenario adaptability of cultivation plan suggestions by injecting structured canopy monitoring data into the agricultural large language model. This achieves the technical effect of reducing the computational cost of monitoring data retrieval and improving the canopy cultivation targeting and implementation effectiveness of cultivation plan suggestions when handling decision-making tasks involving multiple cultivation scenarios and complex canopy characteristics.

[0058] In one embodiment, optimizing and pruning the tree canopy cultivation attribute hierarchy map composed of dynamic attribute identifiers yields a simplified attribute identifier, including: Based on a pre-defined large language model, the semantic correlation degree of cultivation between any two nodes in the dynamic attribute identifier is calculated; the target cultivation scenario corresponding to the tree crown cultivation decision request is identified, and the core priority dimension under the target cultivation scenario is determined. If the semantic correlation between two nodes is higher than the preset correlation threshold and neither belongs to the core priority dimension, then the two nodes will be merged. If the merged node still contains non-core priority child nodes, then remove the non-core priority child nodes and retain the core priority child nodes to obtain a simplified attribute identifier.

[0059] Specifically, nodes can be specific cultivation dimension identifiers in the tree canopy cultivation attribute hierarchy map, used to organize and represent tree canopy feature information units in the dynamic cultivation monitoring database. For example, "peak production period," "canopy periphery," and "bud and leaf density" are all independent nodes. Cultivation semantic relevance can be an indicator reflecting the thematic relevance of any two cultivation dimension nodes in the Eagle Tea tree cultivation scenario. It measures the semantic similarity of nodes and can be obtained by calculating the semantic similarity of node attribute identifiers using a pre-set large language model. For example, cultivation semantic relevance can be calculated based on the cosine similarity of word vectors in a pre-trained model in the agricultural field, including but not limited to semantic matching of nodes such as "twig growth rate" and "bud and leaf density," obtaining a relevance score in the 0-1 range.

[0060] The target cultivation scenario can be the core cultivation requirement type corresponding to the canopy cultivation decision request, such as "high-yield canopy cultivation," "resistance-resistant canopy pruning," and "ornamental canopy shaping," used to clarify the priority orientation of cultivation decisions. Core priority dimensions are cultivation dimension nodes that play a key supporting role in the decision-making process under the target cultivation scenario, and can be determined through a pre-defined cultivation scenario-core dimension mapping table. For example, in the "high-yield canopy cultivation" scenario, core priority dimensions may include "bud and leaf density," "twig growth rate," and "leaf area index," etc.

[0061] The preset association threshold can be a critical score used to determine whether two cultivation dimension nodes are semantically similar, and is used to filter out the merging operation of irrelevant nodes. It can be configured through the experience of agricultural cultivation experts or adaptively adjusted by the system based on historical decision data. For example, in the Eagle Tea Tree cultivation scenario, the preset association threshold can be set to 0.7, that is, nodes with a cultivation semantic association degree ≥ 0.7 are judged as semantically similar.

[0062] Node merging can adjust the hierarchical relationship between two semantically similar nodes in the tree canopy cultivation attribute hierarchy graph, integrating them into a single parent node that encompasses both of their semantics. This simplifies the graph structure and reduces redundant branches in non-core dimensions. For example, if the cultivation semantic correlation between "twig growth rate" and "bud and leaf density" is 0.8 (above the threshold) and neither belongs to the core priority dimension of the "high-yield" scenario, then the two can be merged into a single parent node called "canopy growth characteristics".

[0063] Non-core priority child nodes can be child nodes under the merged parent node that do not belong to the core priority dimensions of the target cultivation scenario. The removal operation removes these child nodes and their associated canopy feature information units, retaining only the core priority child nodes to further simplify the map structure. For example, if the merged "canopy growth feature" parent node contains "branching angle" (a non-core dimension), then that child node is removed, and only core child nodes such as "bud and leaf density" and "twig growth rate" are retained.

[0064] This embodiment provides a method for monitoring the canopy features of eagle tea trees. It calculates the semantic correlation between any two nodes in a dynamic attribute identifier based on a preset large language model, identifies the target cultivation scenario corresponding to the canopy cultivation decision request, and determines the core priority dimension. If the semantic correlation between two nodes is higher than a preset correlation threshold and neither belongs to the core priority dimension, the nodes are merged. Then, non-core priority child nodes under the merged nodes are removed to obtain a simplified attribute identifier. The scenario relevance of the nodes is quantified using the semantic correlation of cultivation. Pruning decisions are made in conjunction with the core dimension priority of the target cultivation scenario. This simplifies the hierarchical graph structure of canopy cultivation attributes while retaining key decision data, avoiding low retrieval efficiency due to redundancy in non-core dimensions. This achieves the technical effect of synergistic optimization of improving the accuracy of cultivation decision data retrieval and the compactness of the graph structure.

[0065] In one embodiment, obtaining the canopy feature information unit in the simplified attribute identifier that matches the canopy cultivation decision request includes: Obtain the dimension node that matches the canopy cultivation decision request from the simplified attribute identifier; obtain the canopy feature information unit that matches the canopy cultivation decision request under the dimension node; If the canopy feature information unit under the dimension node that matches the canopy cultivation decision request does not meet the preset conditions, then select the parent node of the dimension node and obtain the canopy feature information unit under the parent node that matches the canopy cultivation decision request. If the tree crown feature information unit under the parent node still does not meet the preset conditions, then continue to backtrack upwards to the root node until a tree crown feature information unit that meets the preset conditions is obtained.

[0066] Specifically, the method involves obtaining the dimension nodes that match the canopy cultivation decision request from the simplified attribute identifiers. This can be achieved by loading a semantic lexicon specifically for the cultivation of eagle tea trees, inputting the text content of the canopy cultivation decision request and each cultivation attribute node into a semantic similarity calculation model, obtaining the scene semantic similarity value between the two, and filtering out nodes with similarity values ​​higher than a preset matching threshold (e.g., ≥0.8). This operation relies on the tree-like hierarchical structure of the canopy cultivation attribute hierarchy graph, enabling rapid traversal of dimension nodes and accurate semantic comparison.

[0067] The preset conditions are evaluation criteria set for the "cultivation decision support capability" of the tree canopy feature information unit set. Their core function is to screen out effective information units sufficient to support the generation of scientific cultivation plans, serving as the core screening basis to ensure the rationality of subsequent cultivation plan recommendations. For example, the preset conditions need to adapt to the differences in growth characteristics of different cultivation stages of the Lao Ying tea tree, specifically including: 1. Information unit quantity threshold: During the seedling stage (when growth status fluctuates greatly), the number of information units corresponding to each 15-day monitoring window must be ≥8; during the peak production stage (when growth status is relatively stable), the number of information units corresponding to each 30-day monitoring window must be ≥5; 2. Data timeliness requirements: During the seedling stage, the proportion of information units collected within the last 15 days must be ≥80%; during the peak production stage, the proportion of information units collected within the last 30 days must be ≥70%; 3. Cultivation feature relevance score: The semantic similarity between the growth data carried by the information unit and the growth of the tree canopy cultivation decision request must be ≥0.85; 4. Data collection accuracy level: Branch growth data must reach centimeter-level accuracy, and bud and leaf density data must reach "units / square centimeter" level accuracy.

[0068] When selecting the parent node of a dimension node, it is necessary to traverse the dimensional hierarchy of the canopy cultivation attribute hierarchy graph, back up one level, and verify the legality of the path of the parent node in the graph. To obtain the canopy feature information units that match the canopy cultivation decision request under the parent node, the refined retrieval logic of the canopy feature information units is reused in the broader cultivation dimension corresponding to the parent node (e.g., back up from the "bud and leaf density" dimension to the "canopy periphery" dimension), and the set of information units that meet the decision request requirements is re-selected, thereby expanding the retrieval range of effective data.

[0069] If the tree crown feature information unit under the parent node still does not meet the aforementioned preset conditions, it can continue to back up to a higher-level dimension node (for example, back up from the "tree crown periphery" dimension to the "peak production period" dimension), until it back up to the root node of the tree crown cultivation attribute hierarchy map (i.e. the "Eagle Tea Tree Cultivation Core Feature" node), thereby exhaustively monitoring the resources in the database to ensure that effective information units that can support cultivation decisions are obtained.

[0070] The above operations in this embodiment, through a coherent process of "precise matching of dimensional nodes → filtering of information unit conditions → backtracking of parent node hierarchy → backup of root node resources," can automatically expand to a broader range of cultivation dimensions to supplement effective data when the number of information units under the initially matched dimensional nodes is insufficient, lacks timeliness, or does not meet accuracy standards. It achieves efficient retrieval by relying on the structured features of the tree canopy cultivation attribute hierarchy map, and ensures the reliability of data acquisition through a multi-level backtracking mechanism. Ultimately, it can improve the success rate of matching tree canopy cultivation decision data, avoid deviations in cultivation measures such as fertilization and pruning due to insufficient data support, and significantly enhance the practical feasibility of cultivation plan recommendations.

[0071] Reference Figure 2 To more clearly illustrate the technical solution of this invention, a detailed embodiment of a method for monitoring canopy characteristics during the cultivation of Eagle Tea Trees is provided, including: I. Establishment of a target cultivation and monitoring database.

[0072] (I) Construction of tree canopy feature dimension hierarchy map and splitting of monitoring data.

[0073] 1. Multi-source monitoring data acquisition and preprocessing: Based on the needs of the Laoying tea tree cultivation scenario, three types of core acquisition equipment were selected, including centimeter-level laser rangefinder (for collecting branch growth data), high-resolution drone (for canopy area zonal imaging), and manual digital recorder (for recording manually measurable indicators such as bud and leaf density).

[0074] The data collection frequency was dynamically adjusted according to the cultivation stage: data was collected every 3 days during the seedling stage, every 7 days during the peak production stage, and every 15 days during the dormancy stage, ensuring that the data covered key growth nodes. Preprocessing steps included: removing sensor outliers (using the 3σ principle), standardizing the data timestamp format (YYYY-MM-DD HH:MM:SS), and supplementing missing data (using linear interpolation), providing a standardized data foundation for subsequent analysis.

[0075] 2. Construction of a hierarchical map of tree canopy features (multiple methods adapted): (1) Parsing method based on structured identifiers and data format features (applicable to standard collected data): Step 1: Read the structured identification information of the monitoring data and extract the preset labels "cultivation stage", "monitoring location" and "feature type", such as "cultivation stage = seedling period", "monitoring location = inner canopy" and "feature type = number of new shoots" in the sensor metadata.

[0076] Step 2: Analyze the data format features to assist in verification. For example, the timestamp "2025-03" corresponds to "seedling period", the numerical range "0-3cm" corresponds to "new shoot growth", and the file name "outer periphery-leaf area-202505" corresponds to "canopy periphery → leaf area index".

[0077] Step 3: If the analysis results of the two types of information are consistent, directly construct a three-level map of "cultivation stage - monitoring location - characteristic type"; if there are differences, the structured identifier shall prevail, and the data format characteristics shall be used as a supplementary reference.

[0078] (2) Parsing method based on custom parsing rules (applicable to non-standard collected data): Custom rules include three core logics: 1. Keyword matching (e.g., "frost resistance" in data notes corresponds to "stress-resistant cultivation target"); 2. Coordinate association (e.g., the collection latitude and longitude correspond to "mountainous planting area → sun-facing side of tree canopy"); 3. Numerical mapping (e.g., "leaf thickness ≥ 0.3 mm" corresponds to "stress resistance characteristics").

[0079] In one specific embodiment, the dimensionality extraction is completed by using regular expression matching of the keyword "seedling period", associating the collection coordinate "28° North latitude" with the mountainous planting area, and mapping "2cm of branch growth" with "new shoot growth characteristics" to the digital data recorded by hand.

[0080] (3) Parsing method based on large language model + standardized format (applicable to unlabeled data): Step 1: Convert the multi-source monitoring data into a standardized JSON format, with fixed fields including "collection time", "tea tree number", "collection device", "feature value", and "remarks" to eliminate format compatibility issues caused by device differences.

[0081] Step 2: Extract initial dimension information and core feature overview. For example, from "feature value = 4cm" and "remarks = May new shoots", extract the initial dimension "new shoot growth" and generate the core feature overview "In May 2025, the average monthly growth of new shoots of mountain eagle tea trees was 4cm".

[0082] Step 3: Construct a large language model for agricultural scenarios. Prompt words: "You are an expert in the dimensional analysis of Eagle Tea Tree cultivation. Based on the data: core feature overview, generate a dimensional map at the level of 'cultivation stage - monitoring part - feature type', supplement missing dimensions, and return the results in JSON format."

[0083] Step 4: Analyze the model output results, such as supplementing the model with "cultivation stage = peak production period" and "monitoring location = outer canopy", and finally construct a complete three-level map.

[0084] (4) Last resort analysis method: If the above three methods fail to extract the data, the basic information “collection batch + tea tree number + collection date” will be used as the top-level dimension, such as “202506 batch - tea tree A08-20250510”, to ensure that the map construction is not interrupted.

[0085] 3. Time window-based monitoring data splitting: The preset time windows and sampling intervals are configured according to the cultivation stages: 1. Seedling stage: 15-day window, 7-day sampling interval (to capture rapid growth changes); 2. Peak production stage: 30-day window, 30-day sampling interval (to avoid data redundancy); 3. Dormant stage: 60-day window, 30-day sampling interval (to balance data density and efficiency).

[0086] The splitting process is as follows: 1. Divide the data range according to the dimensional branches of the graph (e.g., "peak production period → outer canopy → bud and leaf density"); 2. Read the window and interval parameters of the corresponding stage; 3. Slide the window along the time series, encapsulate the data within the window as independent information units, and retain the dimension attribution label.

[0087] For example, the data on "outer canopy → bud and leaf density" during the peak production period, after being divided into 30-day windows, includes a numerical sequence of bud and leaf density and a distribution heatmap for "May 1, 2025 to May 30, 2025".

[0088] (II) Generation of Cultivation Attribute Identifiers (Manual + Automatic Dual Mode): 1. Manual Tagging Mode: This mode presets three attribute dimensions, including cultivation target attributes (high-yielding, stress-resistant, ornamental), growth stage attributes (seedling stage, growing season, peak production stage, dormancy stage), and canopy location attributes (inner canopy, central canopy, outer canopy). When uploading monitoring data, users select the corresponding attributes through the cultivation management terminal interface, such as "high-yielding + peak production stage + outer canopy," and the system automatically associates them with the data tag field.

[0089] 2. Automatic Labeling Mode: Triggering Condition: When no manual labels are selected for the data, the large language model labeling process is automatically started. Model input includes: an overview of the core features of the monitoring data and a semantic lexicon for the cultivation of Laoying tea trees (containing 200+ cultivation-related keywords). Construction prompt: "Based on the overview of the core features of the monitoring data summary, generate labels from four dimensions: cultivation goals, growth stages, canopy parts, and feature types. The labels must fit the Laoying tea tree cultivation scenario, and the results should be returned in JSON format."

[0090] For example, if the input summary is “In June 2025, the growth of the outer branches of the mountain tea tree was 5cm, and the density of buds and leaves was 4 / cm²”, the model output labels are: “Cultivation goal = high yield type, growth stage = peak production period, canopy location = outer periphery, feature type = branch growth + bud and leaf density”.

[0091] (III) Establishment of a target cultivation and monitoring database: 1. Data integration logic: Integrate tree canopy feature information units, cultivation attribute indexes, and hierarchical graphs, and store them using the graph database Neo4j. Nodes are cultivation attribute identifiers (such as "peak production period"), edges are dimensional relationships (such as "peak production period → outer canopy"), and attributes are the storage paths of information units.

[0092] 2. Cultivation Attribute Index Construction: A combination of time-series indexing and hash mapping is adopted. The time-series index is associated with "collection time - attribute identifier", and the hash mapping is associated with "attribute identifier - information unit", supporting fast retrieval by "cultivation stage + time range". For example, when retrieving data on "peak production period in 2025 + outer canopy", the time range is first located through the time-series index, and then the information unit is directly accessed through the hash mapping, with a retrieval response time of ≤0.5 seconds.

[0093] II. Decision-making response and cultivation plan generation for tree canopy cultivation.

[0094] (a) Multi-database merging and pruning optimization: 1. Multi-database merging: When user decision requests involve multiple scenarios (such as "mountainous + hilly + high-yield cultivation"), the system filters three corresponding target cultivation monitoring databases. The merging logic adopts "attribute union + semantic alignment": 1. Merging hierarchical attributes, such as merging "canopy periphery" in database A and "canopy outer side" in database B into the same attribute; 2. Standardizing terminology, such as unifying "new shoot growth" and "branch growth rate" into "branch growth characteristics". A dynamic cultivation monitoring database and dynamic attribute identifiers are generated, for example, after merging, a joint attribute "region = mountainous / hilly" is added to adapt to cross-regional decision-making needs.

[0095] 2. Tree canopy cultivation attribute hierarchy map pruning: (1) Cultivation semantic association degree calculation: The agricultural version of BERT pre-trained model is used. The text of two attribute nodes is input, the cosine similarity of word vectors is calculated, and the association degree score in the 0-1 interval is output. The preset association threshold is adjusted according to the cultivation scenario: 0.75 for high-yield scenario, 0.7 for stress-resistant scenario, and 0.65 for ornamental scenario, to ensure that pruning is adapted to different decision-making needs.

[0096] (2) Determining the core priority dimension: By pre-setting the “scenario-core dimension” mapping table, for example, the core dimension of the high-yield scenario is “bud and leaf density, branch growth, and leaf area index”, and the core dimension of the stress-resistant scenario is “leaf thickness, branching angle, and resistance to pests and diseases”.

[0097] (3) Pruning operation process: 1. Calculate the correlation between any two nodes. If the correlation is higher than the threshold and neither is a core dimension, merge them (e.g., merge “twig growth” and “bud density” into “canopy growth characteristics”); 2. Remove non-core sub-nodes under the merged node (e.g., remove “branching angle” under “canopy growth characteristics”); 3. Retain core dimension nodes and related sub-nodes.

[0098] (II) Cultivating decision-making data matching and solution generation: 1. Data Matching Process: Step 1: Perform semantic parsing on the decision request "High-yield fertilization plan for mountain eagle tea trees in peak production period", extracting keywords "peak production period, mountain, high-yield, fertilization", and matching the "peak production period → outer canopy → branch growth + bud and leaf density" dimension nodes in the simplified attribute identifier. Step 2: Retrieve information units under this node and verify whether they meet the preset conditions: ① Quantity ≥ 5; ② Data percentage in the last 30 days ≥ 70%; ③ Semantic similarity ≥ 0.85; ④ Data accuracy reaches centimeter level / unit / cm². Step 3: If the conditions are met, use a mixed search of "attribute path weight + semantic similarity" to select the top 3 information units; if not, backtrack to the parent node "peak production period → outer canopy" and search again until the root node.

[0099] 2. Cultivation Plan Generation: Construct a prompt template: "Based on the following monitoring data: [Information Unit 1: Density of peripheral buds and leaves during peak production period: 4 buds / cm²; Information Unit 2: Monthly shoot growth: 5cm], and in combination with the high-yield cultivation needs of Laoying tea trees, generate a targeted fertilization plan, specifying the fertilizer type, application amount, and fertilization time. The plan must be scientific and feasible."

[0100] Input the prompts and information units into the agricultural version of the GPT model to generate a sample solution: "It is recommended to use nitrogen, phosphorus and potassium compound fertilizer (N:P:K=15:10:12), with an application rate of 0.5kg / tree. Apply it after rain in mid-June. Dig a shallow trench (10cm deep) along the drip line of the tree canopy, apply the fertilizer, and then cover it with soil. In conjunction with foliar spraying of 0.2% potassium dihydrogen phosphate solution, once every 10 days, for a total of 3 times."

[0101] (III) Handling Abnormal Scenarios: 1. Insufficient Data Scenarios: When the number of retrieved information units is less than 5 or the similarity is less than 0.85, the system automatically backtracks to the parent node and prompts the user that "the current subdivision data is insufficient, and the search has been expanded to a broader dimension." 2. Search Failure Scenarios: If no valid data is obtained after backtracking to the root node, the system generates an "Insufficient Information Explanation," recommends a similar cultivation scenario, and prompts the user to supplement the collection of corresponding dimension data.

[0102] The method for monitoring the canopy features of *Tea lanceolata* provided in this embodiment achieves structured analysis of data from different collection specifications through standardized preprocessing of multi-source data and construction of dimensional maps adapted to multiple methods. This significantly improves the coverage of cultivation dimension analysis and the robustness of data structured processing. By dynamically configuring time windows and sampling intervals according to cultivation stages and combining attribute index optimization, it achieves refined splitting and efficient retrieval of monitoring data, improves the semantic coherence of information units, and shortens retrieval response time. Through manual and automatic dual-mode cultivation attribute label generation, combined with the semantic understanding capabilities of a domain-specific large language model, the matching degree between labels and data features is improved, reducing manual annotation costs. Through multi-database attribute merging and scenario-based pruning optimization, combined with a multi-level backtracking matching mechanism, the success rate of decision data matching and the targeting of cultivation plans are improved. By injecting structured monitoring data into an agricultural scenario large language model, the generated cultivation plan is scientific and feasible, significantly improving the scientificity and effectiveness of *Tea lanceolata* canopy cultivation.

[0103] This embodiment also provides a canopy feature monitoring system during the cultivation process of Eagle Tea Trees. The system includes: a cultivation dimension analysis module, used to perform cultivation dimension hierarchical analysis on multi-dimensional monitoring data during the cultivation process of Eagle Tea Trees canopies, and generate a canopy feature dimension hierarchical map; and a feature data splitting module, used to perform fine splitting of the multi-dimensional monitoring data based on the canopy feature dimension hierarchical map, forming multiple canopy feature information units. An attribute identifier generation module is used to generate a cultivation attribute identifier corresponding to the multi-dimensional monitoring data based on a preset large language model and a summary of the core features of the multi-dimensional monitoring data when the multi-dimensional monitoring data does not have a cultivation attribute identifier. A monitoring database construction module is used to construct a target cultivation monitoring database based on at least one set of cultivation attribute identifiers of the multi-dimensional monitoring data and multiple canopy feature information units. The target cultivation monitoring database includes a canopy cultivation attribute hierarchy map composed of cultivation attribute identifiers of the multi-dimensional monitoring data.

[0104] This embodiment also provides a computer device applicable to a method for monitoring crown characteristics during the cultivation of eagle tea tree crowns, comprising: a memory and a processor; the memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions to realize the method for monitoring crown characteristics during the cultivation of eagle tea tree crowns as proposed in the above embodiment.

[0105] The computer device can be a terminal, comprising a processor, memory, communication interface, display screen, and input devices connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The communication interface is used for wired or wireless communication with external terminals; wireless communication can be achieved through Wi-Fi, carrier networks, NFC (Near Field Communication), or other technologies. The display screen can be an LCD screen or an e-ink screen. The input devices can be a touch layer covering the display screen, buttons, a trackball, or a touchpad on the computer device's casing, or an external keyboard, touchpad, or mouse.

[0106] This embodiment also provides a storage medium storing a computer program, which, when executed by a processor, implements a method for monitoring canopy characteristics during the cultivation of eagle tea tree canopies as proposed in the above embodiment. The storage medium can 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 Red-Only Memory (PROM), Read-Only Memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk.

[0107] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A method for monitoring canopy characteristics during the cultivation of Eagle Tea Trees, characterized in that, Includes the following steps: Multi-dimensional monitoring data during the cultivation process of the Eagle Tea Tree canopy were analyzed at the cultivation dimension level to generate a canopy feature dimension level map. Based on the tree canopy feature dimension hierarchy map, the multi-dimensional monitoring data is finely divided to form multiple tree canopy feature information units; When the multi-dimensional monitoring data does not include a cultivation attribute identifier, a cultivation attribute identifier corresponding to the multi-dimensional monitoring data is generated based on a preset large language model and a summary of the core features of the multi-dimensional monitoring data. A target cultivation monitoring database is constructed based on at least one set of cultivation attribute identifiers from the multi-dimensional monitoring data and multiple canopy feature information units; the target cultivation monitoring database includes a canopy cultivation attribute hierarchy map composed of cultivation attribute identifiers from the multi-dimensional monitoring data.

2. The method for monitoring canopy characteristics during the cultivation of Eagle Tea Trees according to claim 1, characterized in that: The process of analyzing the multi-dimensional monitoring data during the cultivation of the Eagle Tea tree canopy to generate a canopy feature dimension hierarchy map includes: Based on the structured identification information, data format features, or custom parsing rules of the multi-dimensional monitoring data, extract the cultivation stage dimension, monitoring part dimension, and feature type dimension corresponding to the multi-dimensional monitoring data; If extraction fails, the basic information collected from the multi-dimensional monitoring data will be used as the top-level analytical dimension. Based on the cultivation stage dimension, monitoring location dimension, and feature type dimension, or the top-level analysis dimension, the canopy feature dimension hierarchy map is generated.

3. The method for monitoring canopy characteristics during the cultivation of Eagle Tea Trees according to claim 1, characterized in that: The process of analyzing the multi-dimensional monitoring data during the cultivation of the Eagle Tea tree canopy to generate a canopy feature dimension hierarchy map includes: The multi-dimensional monitoring data is converted into a unified standardized data format; based on the standardized data format, the initial dimensional information and core feature summary of the multi-dimensional monitoring data are extracted. Based on the preset large language model, the initial dimensional information, and the core feature overview, a hierarchical map of canopy features corresponding to the cultivation of the Eagle Tea Tree canopy is generated.

4. The method for monitoring canopy characteristics during the cultivation of Eagle Tea Trees according to claim 1, characterized in that: The step of refining the multi-dimensional monitoring data based on the hierarchical map of tree canopy features to form multiple tree canopy feature information units includes: Based on a preset time window and a preset sampling interval, the multi-dimensional monitoring data corresponding to each dimension branch in the tree canopy feature dimension hierarchy map are sequentially segmented to form multiple tree canopy feature information units.

5. The method for monitoring canopy characteristics during the cultivation of Eagle Tea Trees according to claim 1, characterized in that: After constructing the target cultivation monitoring database based on at least one set of cultivation attribute identifiers from the multi-dimensional monitoring data and multiple canopy feature information units, the method further includes: In response to a canopy cultivation decision request, determine multiple target cultivation monitoring databases corresponding to the canopy cultivation decision request; The cultivation attribute identifiers of multiple target cultivation monitoring databases are merged to obtain a dynamic cultivation monitoring database and a dynamic attribute identifier corresponding to the dynamic cultivation monitoring database; The tree canopy cultivation attribute hierarchy map composed of the dynamic attribute identifiers is optimized and pruned to obtain simplified attribute identifiers; Obtain the canopy feature information unit from the simplified attribute identifier that matches the canopy cultivation decision request; Based on the preset large language model, the canopy cultivation decision request, and the canopy feature information unit, a cultivation scheme suggestion corresponding to the canopy cultivation decision is obtained.

6. The method for monitoring canopy characteristics during the cultivation of Eagle Tea Trees according to claim 5, characterized in that: The optimization and pruning of the tree canopy cultivation attribute hierarchy map composed of the dynamic attribute identifiers to obtain simplified attribute identifiers includes: Based on a pre-defined large language model, the semantic correlation degree of cultivation between any two nodes in the dynamic attribute identifier is calculated; the target cultivation scenario corresponding to the canopy cultivation decision request is identified, and the core priority dimension under the target cultivation scenario is determined; If the semantic correlation between two nodes is higher than the preset correlation threshold and neither belongs to the core priority dimension, then the two nodes will be merged. If the merged node still contains non-core priority child nodes, then remove the non-core priority child nodes and retain the core priority child nodes to obtain the simplified attribute identifier.

7. The method for monitoring canopy characteristics during the cultivation of Eagle Tea Trees according to claim 5, characterized in that: The unit for obtaining the canopy feature information that matches the canopy cultivation decision request from the simplified attribute identifier includes: Obtain the dimension node in the simplified attribute identifier that matches the canopy cultivation decision request; obtain the canopy feature information unit under the dimension node that matches the canopy cultivation decision request; If the canopy feature information unit under the dimension node that matches the canopy cultivation decision request does not meet the preset conditions, then the parent node of the dimension node is selected, and the canopy feature information unit under the parent node that matches the canopy cultivation decision request is obtained. If the tree crown feature information unit under the parent node still does not meet the preset conditions, then continue to backtrack upwards to the root node until a tree crown feature information unit that meets the preset conditions is obtained.

8. A monitoring system applied to the crown characteristic monitoring method in the crown cultivation process of the Eagle Tea Tree as described in any one of claims 1-7, characterized in that: The system includes: The cultivation dimension analysis module is used to perform cultivation dimension hierarchical analysis on multi-dimensional monitoring data during the cultivation process of the Eagle Tea Tree canopy, and generate a canopy feature dimension hierarchy map. The feature data splitting module is used to perform fine splitting of the multi-dimensional monitoring data based on the tree canopy feature dimension hierarchy map, forming multiple tree canopy feature information units; The attribute identifier generation module is used to generate the cultivation attribute identifier corresponding to the multi-dimensional monitoring data based on a preset large language model and a summary of the core features of the multi-dimensional monitoring data when the multi-dimensional monitoring data does not have a cultivation attribute identifier. The monitoring database construction module is used to construct a target cultivation monitoring database based on at least one set of cultivation attribute identifiers of the multi-dimensional monitoring data and multiple canopy feature information units; the target cultivation monitoring database includes a canopy cultivation attribute hierarchy map composed of cultivation attribute identifiers of the multi-dimensional monitoring data.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that: When the processor executes the computer program, it implements the steps of the crown feature monitoring method in the crown cultivation process of the Eagle Tea Tree as described in any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by the processor, it implements the steps of the crown feature monitoring method in the crown cultivation process of the Eagle Tea Tree as described in any one of claims 1 to 7.

Images