Method for classifying grass cultivars for approval based on use characteristics

Abstract

The application discloses a grass variety approval classification method based on use characteristics, relates to the technical field of variety approval data classification, and comprises the following steps: acquiring a to-be-classified grass plant data set, establishing a plant number for each plant, and uniformly connecting text, pictures, tests, detections, genetic and application data to form a five-module plant characteristic package of basic information, botanical characteristics, biological characteristics, genetic information, resource characteristics and application; an use classification index library is established, the five-module plant characteristic package is mapped into a use evidence matrix corresponding to a plant-use category-evidence field pair; the use adaptation degrees of each use category are calculated based on the use evidence matrix, and the main use category, the auxiliary use category, the review state and the missing field list are determined; and the classification report and the material package are generated according to the classification result by calling the approval template. The grass variety data forms a reviewable use classification basis, and the risks of scattered data, misjudgment of composite use and missing materials are reduced.

Description

Technical Field

[0001] This invention relates to the field of variety approval data classification technology, specifically a method for classifying grass varieties based on their intended use. Background Technology

[0002] The approval of grass varieties covers various uses, including forage grass, turfgrass, ornamental grass, ecological restoration grass, and energy grass. Applicants need to submit relevant information such as variety origin, botanical characteristics, biological characteristics, regional trials, production trials, quality testing, photographic data, and supporting documentation. The current standard is GB / T30395-2013, "Technical Regulations for Grass Variety Approval," published on December 31, 2013, and implemented on July 16, 2014. Application materials also require test reports detailing the test site overview, control varieties, test design, measurement indicators, and test results. Currently, the prevailing method is to manually compile the application text, photographs, test reports, and inspection reports, which are then checked and categorized by the approval personnel according to the intended use.

[0003] Chinese patent document CN111931700A discloses a method and system for identifying genuine and counterfeit maize varieties based on a multi-classifier. The basic idea of ​​this technology is to first acquire a two-dimensional color image of the target maize kernel, then preprocess the acquired two-dimensional color image by grayscale conversion, denoising, and binarization; finally, color feature extraction, shape feature extraction, and texture feature input are used to RGB and HSV classifiers. The color features include area, diameter, perimeter, roundness, ellipticity, and extension; the texture features are derived from parameters such as energy, entropy, moment of inertia, and correlation produced by the grayscale co-occurrence matrix; finally, the phenotypic feature results are input to three classifiers—KNN, neural network, and support vector machine—and the target kernel's variety is determined by voting. This system is configured according to the signal flow, including a color image acquisition module, grayscale processing module, denoising module, binarization module, color feature extraction module, texture feature extraction module, shape feature extraction module, and classifiers. It belongs to the variety identification technology based on seed image phenotypic features.

[0004] The aforementioned existing technologies are suitable for identifying grain images where the target object has a relatively stable morphology and the input data type is relatively simple. Their core judgment focuses on image preprocessing, phenotypic feature extraction, and multi-classifier voting results. However, in the context of grass variety approval, the object to be processed is not limited to a single organ image. Application materials generally include textual descriptions, photographs of the whole plant and its organs, variety comparison trials, regional trials, production trials, quality testing, proof of origin, and introduction materials. The same herbaceous plant may also have clues regarding its use in pasture, ornamental purposes, turf use, ecological restoration, food, or energy utilization. If a variety identification method based primarily on single image features is still used, it can only answer the question "Does the object resemble a certain preset variety?" and cannot solve approval business issues such as use classification, coexistence of multiple uses, missing data, and material supplementation.

[0005] The underlying mechanism is that the features obtained by image classifiers are mainly composed of the pixels, morphology, and texture of the image. There is no direct correspondence between the feature space and the resistance, adaptability, yield, quality, safety certification, and test origin required for the approved use. Furthermore, the voting output of multi-classifiers is generally a category or a true / false judgment, which cannot naturally express the primary use, secondary use, or insufficient data. If this problem is not solved, the compilation of approval materials will continue to rely on manual cross-checking, easily leading to unclear attribution of multiple data sets for the same plant, scattered use judgments, and delayed discovery of missing materials. Therefore, the technical problem to be solved in this field is how to form a stable and verifiable basis for use classification in grass variety approval materials that are from multiple sources, heterogeneous, and have coexisting use clues. Summary of the Invention

[0006] (a) Technical problems to be solved To address the shortcomings of existing technologies, this invention provides a method for classifying and approving grass varieties based on their intended uses. This method comprises five modules: basic information, botanical characteristics, biological characteristics, genetic information, resource characteristics, and applications. A use classification index library is established, mapping these five modules to a use evidence matrix corresponding to plant, use category, and evidence field. Based on the use evidence matrix, the use fit of each use category is calculated to determine the primary use category, secondary use category, verification status, and a list of missing fields. Based on the classification results, an approval template is invoked to generate a classification report and a materials package. This method enables grass variety data to form a verifiable basis for use classification, reducing the risks of data dispersion, misjudgment of combined uses, and omissions in materials. It also solves the technical problems described in the background art.

[0007] (II) Technical Solution To achieve the above objectives, the present invention provides the following technical solution: The herbaceous plant variety classification method based on use characteristics is executed by a data processing server. This includes receiving a dataset of herbaceous plants to be classified, generating a plant ID for each plant, associating text, images, experimental, detection, genetic, and application data with the plant ID, and extracting fields to form a five-module plant feature package. A pre-set use classification index library is then invoked to map the five-module plant feature package into a use evidence matrix, with plant IDs as rows, use categories as columns, and evidence fields as cell contents, according to use categories and evidence levels. The use fit of each use category is calculated and ranked based on the use evidence matrix; When the highest-level suitability reaches the classification threshold and the difference between the highest-level suitability and the second-highest-level suitability reaches the difference threshold, the primary-level suitability category is output; when the highest-level suitability reaches the classification threshold but the difference does not reach the difference threshold, the primary-level suitability category, secondary-level suitability category, and review status are output; when the highest-level suitability does not reach the classification threshold, the data insufficiency status and missing field list are output. Use the corresponding approved template to generate a classification report.

[0008] Furthermore, the data processing server extracts the classification field group, source field group, source proof field group, and batch field from the herbaceous plant dataset to be classified; when the classification field group, source field group, and source proof field group are all consistent, the corresponding data are merged into the same plant number; When at least one field group is inconsistent, generate different plant IDs for the corresponding data; and write the plant IDs into each field record of the five-module plant feature package.

[0009] Furthermore, when the data processing server forms the five-module plant feature package, it records the source module, data type, original page segment, manual confirmation mark, and species-level generalization mark for each field. When a field record points to the plant itself corresponding to the plant number, the field record is written to the corresponding module. When a field record only points to the general characteristics of the plant species to which it belongs, the field record is retained as a species-level generalization field and excluded from the mapping input of the use evidence matrix; and the module packaging degree of each module is generated according to the field weight, the effective field label and the source credibility coefficient.

[0010] Furthermore, each cell of the use evidence matrix is ​​associated with a plant number and a use category; Each cell includes evidence of direct use, morphological support, biological support, genetic stability, specific quality, experimental evidence, and approval materials. Each piece of evidence records the field name, field value, source module, data type, original page segment, and manual confirmation mark; the purpose suitability only calls the evidence fields that have been written in the corresponding cell.

[0011] Furthermore, the use evidence matrix uses a ternary association index to store a fixed correspondence between plant ID, use category, and evidence field; the ternary association index includes a plant ID index, a use category index, and an evidence field index; A ternary join index maintains the same field names, field sources, and evidence layer markers in at least one of the following data structures: a two-dimensional table, a ternary table, a graph database edge relation, and a key-value index; the data processing server reads the evidence fields used to calculate the suitability of use based on the ternary join index.

[0012] Furthermore, when the data processing server calls the usage classification index library, it sequentially reads the usage trigger field, supporting field, and material field for each usage category. When all three fields exist, a candidate usage category is generated. When the usage trigger field is missing but the supporting and material fields exist, a candidate usage category awaiting review is generated. When only the usage trigger field exists and the supporting and material fields are missing, a usage clue is generated and writing to the candidate usage category sequence is prohibited. When other field combinations appear, a corresponding list of missing fields is generated and the usage category remains in an undetermined state.

[0013] Furthermore, when the candidate use category belongs to at least one of medicinal herbs, spice herbs, edible herbs, and energy herbs, the data processing server reads the security compliance field and the special quality field. When both the safety compliance field and the special quality field exist, the candidate use category is allowed to enter the use suitability ranking; If at least one field is missing, retain the evidence field for that candidate use category, output the review status and a list of missing fields, and prohibit that candidate use category from being directly written into the main use category.

[0014] Furthermore, the data processing server generates a classification results page when generating the classification report; the classification results page displays the plant number, candidate use category sequence, primary use category, secondary use category, review status, data insufficiency status, missing field list, and approval template number; When an auxiliary use category exists, the classification results page displays the auxiliary use column; when no auxiliary use category exists, the classification results page hides the auxiliary use column and retains the primary use category column.

[0015] (III) Beneficial Effects This invention provides a method for classifying and approving grass varieties based on their intended use, which has the following beneficial effects: By using plant identification numbers, textual data, image data, experimental data, testing data, genetic data, and application data are linked to a single plant object, forming a five-module plant characteristic package. This transforms the scattered documents for grass variety approval into a structured collection of fields, reducing errors caused by identical or different species, or unclear data attribution. Through the continuous merging of basic information, botanical characteristics, biological characteristics, genetic information, and resource characteristics and application modules, the characteristics of the applicant plant, its origin type, usage clues, and supporting materials are stored in the same data chain, preventing the substitution of general characteristics of the plant species for the specific characteristics of the applicant variety.

[0016] By mapping the use classification index library to the five-module plant feature package, a use evidence matrix corresponding to plant-use category-evidence field is obtained. The criteria for judging ecological restoration grass, forage grass, ornamental grass, turfgrass, medicinal grass, spice grass, edible grass, and energy grass can be expanded according to the evidence layer, improving the verifiability of use classification results. Through use suitability calculation and candidate use category sequence sorting, direct use evidence, morphological supporting evidence, biological supporting evidence, genetic stability evidence, specific quality evidence, experimental evidence, and safety compliance evidence are evaluated collaboratively, enabling the separate output of single strong use, multi-use, and insufficient data states.

[0017] By generating a list of primary uses, secondary uses, verification status, and missing fields in a linked manner, the system prevents a single use term from locking a plant when it has multiple uses such as ornamental, landscaping, edible, medicinal, or energy applications. It also prompts for additional experimental, testing, and supporting documentation.

[0018] The classification report and material package are generated by calling the approval template based on the primary use category and the auxiliary use category. This ensures that the supporting evidence fields, evidence sources, recommended supplementary tests, recommended control varieties and the list of application materials are all included in the approval application framework, reducing the possibility of omissions in manual processing and making it easier for subsequent expert review and material correction. Attached Figure Description

[0019] Figure 1 This is a schematic diagram of the overall process of the present invention; Figure 2 This is a schematic diagram of the multi-source data aggregation and five-module plant feature package construction of the present invention; Figure 3 This is a schematic diagram showing the mapping between the application classification index library and the application evidence matrix of the present invention; Figure 4 This is a schematic diagram illustrating the application suitability calculation and primary / secondary application determination of the present invention; Figure 5 This is a schematic diagram illustrating the security pre-judgment and missing field handling of the present invention; Figure 6 This is a schematic diagram illustrating the application of the approved template and the generation of the material package for this invention. Detailed Implementation

[0020] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0021] Please see Figures 1-6 This invention provides a method for classifying and approving grass varieties based on their intended use, comprising: Step 1: Construct a five-module plant feature package based on a single plant as the basic object, so that each use judgment can be returned to the specific plant number, specific data source and specific field location.

[0022] Data for the approval of herbaceous plants typically comes from variety descriptions, local standards, photographs, regional trials, production trials, test reports, and introduction certificates. The name, origin, morphology, resistance, and uses of the same plant are often scattered across different documents. Directly searching by use terminology would mix general species characteristics with the specific characteristics of the applying variety, and would also fragment the evidence for the same plant at different stages. Step one involves establishing plant numbers and then linking the data according to these numbers. The purpose is not to increase the amount of data, but to solidify the smallest unit of evidence required for subsequent classification.

[0023] After receiving the dataset of herbaceous plants to be classified, the data processing server first reads the Chinese name, Latin name, synonyms, family, genus, species, subspecies classification, resource source, collection location, place of origin, and introduction area. Then, it generates plant IDs in the order of name layer, classification layer, source layer, and data layer. Once the plant ID is generated, textual data, image data, experimental data, detection data, genetic data, and application data are all indexed by this plant ID, no longer by the file name. After this processing, even if a plant appears in different reports with abbreviations, local names, or variant Latin names, the system still merges them by family, genus, species, place of origin, and data time, avoiding the same plant being split into multiple objects.

[0024] For example, staff upload a text document containing a field collection record, a set of photos of roots, stems, leaves, and flowers, and a drought resistance observation record. The data processing server first extracts the collection location, collection time, ecological environment, and altitude range from the text, then creates part labels for the photos by individual plant, population, root, stem, leaf, flower, fruit, or seed, and finally groups the drought resistance observation record under the same plant ID. On-site operators see a single plant object on the interface, with textual evidence, photographic evidence, and experimental evidence listed sequentially below it, instead of three separate files. If the data comes from a bred variety, the system records the parentage, breeding method, and generation; if the data comes from a wild-cultivated and domesticated variety, the system records the wild resource collection location, collection time, and domestication process; if the data comes from a local variety, the system records the origin, cultivation history, and current area; if the data comes from an introduced variety, the system records the country of origin, introduction time, registration certificate, and variety rights holder information. The above source types do not change the plant ID; they only participate in subsequent integrity judgments as source-level fields. When the same plant has both local variety records and subsequent breeding records, the server first retains the local origin field, then adds the breeding process field, and places the two types of fields in different data layers under the same plant number. After the on-site data entry personnel select the data type on the upload interface, the system sends the scanned copy, photo, table, and text description to the corresponding parsing channel to form initial fields that can be further verified.

[0025] Preferably, the plant identification uses fixed-length segmented characters, consisting of a classification identifier segment, a source identifier segment, and a sequence identifier segment; a database encoding method with equivalent functionality can also use an incremental primary key plus hash verification. During data access, the server retains the original file name, upload time, data type, data page segment, and manual verification marks. This allows subsequent module merging to pinpoint field evidence, reducing erroneous and missed merges caused by name variations, and providing stable row identifiers for establishing the plant-use category-evidence field matrix in step two.

[0026] To supplement the enforceable rules for plant numbering, the data processing server executes a plant numbering generation operator when generating plant numbers. This operator does not simply use sequential numbering; instead, it integrates category identity, source identity, supporting documentation, and data batch into a single object identifier. Its specific form is:

[0027] In the formula: plant number Indicates the first The object number of each plant to be classified is used to uniformly link textual data, image data, experimental data, test data, genetic data, and application data. Classification field group This field represents the Chinese name, Latin name, alternative names, family, genus, species, and infraspecific classification, used to determine the plant's taxonomic identity; Source field group This indicates the resource source, collection location, place of origin, introduction area, and resource type, used to distinguish plants with the same name from different sources. Source verification field group. This field indicates the parentage, breeding method, generation, introduction certificate, or description of local variety origin, used to help identify the same plant species. (Batch field) This indicates the batch of data uploaded or the batch of data submitted, and is used to form a sequence segment.

[0028] In the category field group Source field group and source proof field group When all data are consistent, the data processing server will merge the corresponding data into the same plant ID. When Chinese names are the same but the family, genus, species, provenance field, or resource type are inconsistent, the data processing server generates different plant numbers for the corresponding data. The rules record the differences in the conflict alert field. This rule enables the uniform processing of homonyms, homonyms, local names, Latin variants, and citation numbers, and provides a definite object for the subsequent five module fields.

[0029] Furthermore, the plant identification number is formed by sequentially concatenating a classification identification segment, a source identification segment, a batch identification segment, and a sequence identification segment. The classification identification segment is generated after standardization of family, genus, species, and subspecies classification; the source identification segment is generated from resource type, collection location, place of origin, or introduction area; the batch identification segment is generated from the batch of data uploaded or the batch of approval application; and the sequence identification segment is generated by the data processing server according to the receiving order within the same batch. For data with different Chinese names, aliases, or place names but consistent classification field group, source field group, and source certification field group, the data processing server writes the original name into the alias field and merges them into the same plant identification number; for data with the same Chinese name but inconsistent in any field group, the data processing server generates different plant identification numbers and retains the differences.

[0030] After plant identification is completed, the data processing server performs field segmentation on the data under the same plant identification number. First, family, genus, species, distribution, resource source, and cultivation area are categorized into the basic information module. Then, life cycle, life form, plant height, crown width, root morphology, stem morphology, leaf morphology, inflorescence type, flower color, flowering period, fruit morphology, seed morphology, pollination method, and reproduction method are categorized into the botanical characteristics module. Subsequently, suitable climate, suitable soil, drought resistance, salt and alkali tolerance, waterlogging tolerance, heat resistance, cold resistance, disease resistance, insect resistance, regeneration, growth period, green period, turf establishment rate, ornamental period, overwintering rate, summering rate, coverage, biomass, and reproductive capacity are categorized into the biological characteristics module. Finally, parental origin, germplasm number, breeding method, molecular marker information, genotype information, ploidy, genetic stability, population uniformity, and variety specificity are categorized into the genetic information module. Finally, the main uses, auxiliary uses, application scenarios, parts used, methods of use, feed quality, medicinal components, aromatic volatile oils, food safety, nutritional value, calorific value, cellulose content, biomass conversion efficiency, ecological restoration goals, maintenance costs, and industrialization value are categorized into the resource characteristics and application module.

[0031] Before field merging, the data processing server performs text data extraction and semantic normalization operators. The input to this operator is the plant number. The following text data is output as a collection of text fields. The data processing server pre-builds a usage dictionary, a field synonym table, a negation context table, and a species-level generalization prompt table. The usage dictionary stores terms such as mowing, grazing, landscaping, mine restoration, rocky desertification control, volatile oil, medicinal parts, and calorific value. The field synonym table unifies fresh grass weight and forage yield into the fresh grass yield field. The negation context table identifies exclusion statements such as "unsuitable for consumption" and "no medicinal value found." The species-level generalization prompt table identifies expressions such as "usually applicable to this species" and "generally applicable to this genus." Only sentences that simultaneously point to the plant number... When the target field and data source are specified, the field is included in the five-module plant feature package.

[0032] Furthermore, during text data extraction, the data processing server first forms sentence segments by paragraph, period, semicolon, and table cell. Then, within each sentence segment, it simultaneously verifies the plant ID's corresponding name, target field, and data source marker. Data source markers include the test report name, detection report name, image number, collection record page number, and manual confirmation marker. If a sentence segment contains exclusion statements from the negative context table, the data processing server writes that sentence segment into the exclusion record; if the sentence segment only describes the general characteristics of the plant species, the data processing server writes that field into the species-level generalization field, without writing a field validity marker.

[0033] For image data, the data processing server executes image data part annotation and visual feature extraction operators. The input to this operator is the plant number. The image data below is output as a collection of image evidence. Each image generates an image number, which is then compared to the plant's ID number. The process involves binding and setting the image location, subject, shooting stage, field name, and confirmation flag. Image locations can be selected from single plants, populations, roots, stems, leaves, flowers, fruits, seeds, and vegetative propagations; subjects can be selected from the applicant plant and control plants; images not manually verified serve only as candidate evidence, while manually verified images are included in the evidence binding strength criteria. The calculation.

[0034] Furthermore, the image number consists of the plant number, the image part code, the subject code, and the upload sequence number. The image part code corresponds one-to-one with a single plant, a group, root, stem, leaf, flower, fruit, seed, and vegetative propagation body. The subject code is used to distinguish between the declared plant and the control plant. The data processing server simultaneously stores the original image name, image number, image part, subject, shooting stage, corresponding field name, and confirmation mark in the image evidence set; when the image part does not correspond to the target field, the image is only saved with the plant number and is not included in the effective calculation of the evidence binding strength.

[0035] Five-module field merging operator Plant numbering Text field collection Collection of photographic evidence and source field set As input, output a five-module plant feature package Its specific form is: In the formula: five modules of plant characteristic package Indicates plant number The corresponding set of structured fields. The set of text fields. This represents the text data extraction and semantic normalization operator. The set of fields to be output. A set of image evidence. This indicates the operator for part labeling and visual feature extraction from image data. The output set of image evidence. Source field set. It indicates the source of the resource, the place of collection, the place of origin, the area of ​​introduction, and the type of resource.

[0036] The data processing server stores only one value for the same original field, but allows that field to take on multiple uses when entering step two. For example, the original value of a covered grass might be stored in the biological characteristics module, used as biological supporting evidence when entering the ecological restoration grass category, and used as a clue to the turfgrass status when entering the lawn grass category. Furthermore, to ensure that field merging can be directly invoked in subsequent steps, the server calculates the module packaging degree for each module. : In the formula: module packaging degree Indicates module number The field formation level of the corresponding module, with a value from 0 to 1, is used to determine whether the module can enter the purpose mapping in step two; module number. This indicates a basic information module, a botanical characteristics module, a biological characteristics module, a genetic information module, or a resource characteristics and application module; fields Represents a single field within a module; a collection of fields. Indicates module number The set of fields that need to be merged for the corresponding module; field weights. Representation field The contribution to this module is preferably between 0.1 and 1; field validity flag. Representation field Whether it was explicitly extracted, with a value of 0 or 1; source credibility coefficient. Representation field The source strength is preferably between 0.2 and 1. The strength of the test report and experimental report is higher than that of the general description text, and the image annotation after manual proofreading is higher than that of the unproofread image annotation.

[0037] For example, a plant description might state that it is salt-tolerant, rapidly spreading, and used for the remediation of abandoned mining areas. Images might show the population's cover status, and experimental data might record the cover observation process. The server categorizes salt tolerance and rapid cover into the biological characteristics module, abandoned mining area remediation into the resource characteristics and applications module, and the cover status of the population photos into the image evidence annotations in the botanical characteristics module. If the text only states that the plant is drought-tolerant without specifying the observation results for the declared variety, the server will mark this field as a species-level generalization field and will not include it as a variety-specific field in the module's inclusion degree. Valid items.

[0038] Furthermore, field merging not only identifies words but also distinguishes data levels and evidence objects, avoiding the misidentification of common species descriptions as usable material for classification. If the same field conflicts in two sources—for example, one report states strong cold resistance while another experimental record only indicates a low overwintering rate—the server does not delete any other records but generates a conflict marker next to the field value, placing the experimental record first for review. After manual review, the field value is updated to confirm the result. The current conflict is still saved as a field source note so that subsequent mappings will know that the field has undergone review.

[0039] Furthermore, field conflict handling does not overwrite the original data. The data processing server establishes a conflict field set for conflicting fields, storing the field values, source modules, data types, original page segments, and manual confirmation marks for each conflicting field. After the reviewer selects a confirmation value on the review interface, the data processing server only appends the confirmation value and confirmation time to the five-module plant feature package, without deleting the original fields from the conflict field set. This ensures that when reading the purpose evidence matrix in step two, both confirmed and conflicting fields can be obtained simultaneously, avoiding breaks in the preceding data chain.

[0040] After image data enters the system, it is first manually or using image annotation tools to classify it into individual plants, groups, roots, stems, leaves, flowers, fruits, seeds, or vegetative propagations, and then cross-linked with text fields under the same plant ID. The server does not directly convert images into usage conclusions, but rather attaches images as morphological or state evidence to the corresponding modules.

[0041] For example, leaf color photos are linked to leaf morphology in the botanical characteristics module; inflorescence photos are linked to inflorescence type and flower color; canopy cover photos are linked to cover evidence annotations; lawn establishment photos are linked to lawn establishment status evidence annotations, etc. If photos of the declared variety and control variety exist simultaneously, the server sets separate tags for the declared photo and control photo, enabling the differentiation of its own traits from the comparison object in subsequent steps.

[0042] To handle inconsistencies between text, images, and reports, the server calculates the strength of evidence binding. : In the formula: the strength of evidence binding Indicates the evidence number The degree of connection between the corresponding evidence and the target field is preferably set to 0 to 1; Evidence number This refers to textual evidence, photographic evidence, experimental evidence, or test evidence under the same plant number; textual consistency is indicated. This indicates the semantic consistency between the evidence text and the target field, with a value ranging from 0 to 1; image matching score... This indicates the degree of match between the image part annotation and the target field, with a value ranging from 0 to 1; (Experimental source division) Indicates whether the evidence comes from comparative testing, regional testing, production testing, or third-party testing, with a value ranging from 0 to 1; manual confirmation score. Indicates the degree of confirmation after manual review, with a value ranging from 0 to 1; text weight Image weights Experimental weights and artificial weight These represent the contributions of the four types of evidence to the binding process, with the sum of all four being 1. The optimal weighting for the trial is determined by these four factors. and artificial weight Higher than individual text weights .

[0043] Field weight Source credibility coefficient Text weight Image weights Experimental weights and artificial weight Stored in a weight configuration table; the weight configuration table must include at least the weight object, corresponding module, corresponding field, weight value, data entry personnel, and activation status. Field validity flag. The value is 1 when the field explicitly refers to the plant being declared, and 0 when the field is missing or is only a species-level generalization. Text consistency score. Image matching is obtained by matching the usage dictionary and the field thesaurus. The source of the experiment was determined by the correspondence between image parts and target fields. The determination is based on whether the evidence comes from variety comparison trials, regional trials, production trials, or test reports, and is subject to manual confirmation. Obtained through manual verification of the markings. This configuration method enables module packaging. Strength of Evidence All calculations can be performed using defined fields, rather than relying on subjective human judgment.

[0044] For example, if a plant description mentions the beauty of its inflorescence, but the image only shows a distant view of the plant and not a close-up of the inflorescence, the server will leave the viewing-related fields as pending completion. However, if a close-up of the inflorescence, a record of the flowering period, and a description of the garden's layout are uploaded simultaneously, the server will categorize the leaf color, inflorescence, flowering period, and viewing period into the botanical characteristics module and the resource characteristics and application module, thus increasing the strength of the evidence binding. For example, if a plant is described as edible but no information on its safety for consumption or processing is provided, the server will only record its edibility as a clue to its use, without marking the safety for consumption field as complete. Similarly, for candidate turfgrass data, if a photo shows low, dense growth, but the text does not describe its tolerance to mowing and trampling, the server will output the missing fields of tolerance to mowing and trampling, without providing a definitive conclusion about the turfgrass type beforehand.

[0045] Finally, the server outputs a five-module plant feature package. This feature package includes at least the plant ID, module name, field name, field value, field source, and evidence binding strength. Module package Missing fields and review prompts.

[0046] Preferably, field values ​​are stored as structured fields, while the original sentences and image numbers are stored as supplementary evidence. Equivalent implementations also include relational databases, document databases, or graph databases, as long as a fixed mapping between plant numbers, fields, and evidence can be maintained. This output allows step two to directly read available evidence for the same plant under different use categories and incorporate any missing fields into the use evidence matrix.

[0047] Preferably, the output file also retains the field lock status. Fields that have been manually confirmed are mainly referenced when entering step two and are not rewritten during the usage matching process; fields that are in the pending supplementation status are transmitted as prompts along with the plant number.

[0048] Step 2: Transform the five-module plant feature package into a plant-use category-evidence field use evidence matrix, so that the evidence position of each herbaceous plant in ecological restoration grass, pasture grass, ornamental grass, turf grass, medicinal grass, spice grass, edible grass and energy grass is presented in a structured way.

[0049] In the five-module plant feature package obtained in step one, the data processing server plays different roles for the same field. For example, fast coverage is a directly supporting field for ecological restoration grasses, but it can only serve as a clue to the turf establishment status; prominent inflorescences are a core field for ornamental grasses, but do not constitute evidence of feed quality for forage grasses.

[0050] Therefore, step two does not use a single keyword for classification. Instead, it first establishes a usage classification index library and then rearranges the fields according to usage category, field level, and evidence source. Furthermore, the usage judgment process changes from simply identifying a word for classification to a continuous series of actions: identifying the field, confirming the module, verifying evidence, and filling in the cell. This avoids mistaking a general description of a species for the approved usage of the plant being declared.

[0051] When the data processing server establishes the usage classification index library, it first sets up six field layers for each usage category: usage determination characteristics, morphological support characteristics, biological support characteristics, special quality characteristics, experimental evidence characteristics, and approved material characteristics. Each field layer stores the field name, source module, allowed evidence type, missing information, and dependencies on other fields. Use-based characteristics are used to receive direct applications such as mine wasteland remediation, mowing, landscaping, medicinal parts, volatile oils, food safety, and calorific value; morphological characteristics are used to receive plant type, leaf color, inflorescence, dwarfism, dense growth, and root, stem, leaf, flower, fruit, and seed morphology; biological characteristics are used to receive drought resistance, salt tolerance, green period, overwintering rate, summering rate, turf establishment rate, growth period, and regeneration; specific quality characteristics are used to receive fresh grass yield, dry grass yield, nutritional value, extractives, active ingredients, aromatic volatile oils, protein, cellulose, calorific value, and biomass conversion efficiency; experimental evidence characteristics are used to receive variety comparison trials, regional trials, production trials, and test reports; and approved material characteristics are used to receive source reports, photographic data, authenticity commitment letters, and descriptions of propagation materials.

[0052] To prevent different use categories from crowding out fields, the data processing server sets category anchors for the same field. Category anchors do not copy the field to multiple categories; instead, they record the field's identity when entering different use categories, while keeping the original value unchanged. For example, low-growing, dense grasses are marked as morphological support features when entering the turfgrass category, but only as a group landscape morphology annotation when entering the ornamental grass category; volatile oils are marked as specific quality features when entering the aromatic grass category, but only as specific quality features when both the medicinal part and active ingredient are recorded in the medicinal grass category.

[0053] For example, when staff select the ecological restoration grass category on the server interface, the interface displays six input areas: damaged habitat, stress resistance, cover status, restoration efficiency, biomass, and experimental source. When switching to the forage grass category, the same plant number displays mowing, grazing, fresh grass yield, hay yield, nutritional value, palatability, and feeding scenario. The fields are not deleted; they are simply placed into different field layers for different usage categories.

[0054] Preferably, the use classification index library is stored in two relational tables: one stores the use category and field layer, and the other stores the correspondence between the field layer and the field names in step one. Each time the server receives a new five-module plant characteristic package, it first reads the plant ID, then searches for the fields item by item according to the correspondence, without manually copying the fields. If the review rules add new specific quality fields, staff only need to add the field name, source module, and missing information to the relational table. When the existing plant ID re-enters step two, new missing information and candidate evidence will be generated. The technical effect of this process is that field attribution has clear boundaries, and the subsequent step three, when reading the use suitability, can distinguish between direct use evidence and background trait evidence; at the same time, plants with multiple uses will not be prematurely excluded from other candidate categories simply because a certain use term appears first.

[0055] To ensure the usage classification index library can be directly accessed by the data processing server, the library further includes a usage category table, a field mapping table, a necessary evidence table, a candidate rule table, and a threshold table. The usage category table stores ecological restoration grasses, pasture grasses, ornamental grasses, turfgrass, medicinal grasses, spice grasses, edible grasses, and energy grasses; the field mapping table stores usage categories, indicator fields, source modules, and evidence layers; the necessary evidence table stores the minimum material type required for primary usage determination; the candidate rule table stores the combination relationships between usage trigger fields, supporting fields, and material fields; and the threshold table stores classification thresholds. Sum and difference thresholds .

[0056] When the same field enters different use categories, its original value remains unchanged; only its identity changes. Furthermore, the candidate rule table generates candidate use categories using three types of fields: use trigger fields, supporting fields, and material fields. Use trigger fields come from the resource characteristics and application module, supporting fields from the botanical characteristics and biological characteristics modules, and material fields from experimental data, testing data, or image data. If a use category simultaneously possesses use trigger fields, supporting fields, and material fields, the data processing server generates a candidate use category; if it lacks a use trigger field but possesses at least two supporting fields and one piece of experimental evidence, the data processing server generates a candidate use category awaiting review; if it only has a use trigger field but lacks supporting and material fields, the data processing server only generates use clues and does not write them into the candidate use category sequence.

[0057] For example, when a staff member views the "Covering Grain" field under the "Ecological Restoration Grass" category, the data processing server will display this field as supporting biological evidence; while when the staff member views the "Covering Grain" field under the "Turfgrass" category, the data processing server will display this field as a clue to turf establishment status. The identities of these fields are fixed by the field mapping table; the manual staff only determines whether a field is acceptable and does not re-copy the field content.

[0058] After the use classification index library is completed, the data processing server reads the five-module plant feature package corresponding to each plant ID and performs mapping in the order of field source module - use category - field layer - evidence type. This mapping does not change the field values ​​in step one, but rather modifies the field values, field source, and module package degree. Strength of Evidence The usage category is also included. To ensure that text, images, experimental, and testing data are entered into the matrix at the same scale, the server calculates the matching strength of the usage field. : In the formula: the strength of fit of the application field. Indicates plant number Category of use The field fit, with values ​​ranging from 0 to 1, is used to fill the base strength of the use evidence matrix; plant number. This indicates the ID of the single plant object generated in step one, and its purpose is to fix the matrix row; usage category. This represents ecological restoration grasses, pasture grasses, ornamental grasses, turfgrass, medicinal grasses, spice grasses, edible grasses, or energy grasses; its function is to define fixed matrix columns; a set of indicator fields. Indicates the category of use All indicator fields involved in the mapping; Field weight Indicator field In terms of usage category The hierarchical contribution, ranging from 0.1 to 1, prioritizes the use-determination feature and specific quality feature over the morphological annotation field; the field reach coefficient... Indicates plant number Does it have an indicator field? Available records, with values ​​ranging from 0 to 1; higher values ​​are used for explicitly recorded records, and lower values ​​are used for missing records or records that are only generalized descriptions at the species level; field source module. Indicator field The first step of this module is to package the modules. Introducing a purpose mapping; evidence set Indicates plant number Supported indicator fields A collection of textual, pictorial, experimental, or testing evidence; Evidence binding strength Indicates the evidence number With indicator fields The degree of integration, following the definition in step one, serves to suppress the influence of uncorrected images or generalized text on the usage category; maximum selection function This means selecting the item with the strongest evidence binding from the evidence set, so that the same field participates in the mapping with the strongest valid evidence; When there are multiple pieces of evidence for the same indicator field, the data processing server first generates the field evidence fusion value. Then use the field evidence fusion value Replace the single item with the maximum strength of evidence and the strength of fit of the usage field. The calculation of the field evidence fusion value. The specific form is: In the formula: field evidence fusion value Indicates plant number Lower indicator fields The overall strength of evidence is determined by a value ranging from 0 to 1. (Evidence set) Indicates plant number Supported indicator fields A collection of textual, pictorial, experimental, or testing evidence; strength of evidence binding. Indicates the evidence number With indicator fields The degree of integration. This formula allows multiple pieces of moderately strong evidence to jointly support the same field, while avoiding a single weak piece of evidence directly raising the application category.

[0059] Accordingly, in implementations with multiple pieces of evidence, the strength of fit of the use field. Calculate using the following formula: In the formula: the strength of fit of the application field. Indicates plant number Category of use The degree of field fit, with a value ranging from 0 to 1; indicator field set Indicates the category of use All indicator fields involved in the mapping; field weights Indicator field In terms of usage category The hierarchical contribution, ranging from 0.1 to 1, is preset by the usage classification indicator library; the field reach coefficient... Indicates plant number Does it have an indicator field? Available records are set to 1 when they explicitly point to the plant being declared, and to 0 when they are missing or are only species-level generalization fields; Field Source Module Indicator field Belongs to the first step module; module package degree Indicates the degree of field formation in the field source module; field evidence fusion value. Indicates multiple pieces of evidence for the indicator field The overall support strength; For example, a plant's resource characteristics and application module records its use in mine wasteland remediation, its biological characteristics module records drought resistance and salt tolerance, and its image data displays its population cover status. The server first maps mine wasteland remediation to the use determination feature of ecological restoration grass, then maps drought resistance and salt tolerance to biological support features, and maps the population cover image to morphological support features. If the plant lacks remediation efficiency test records, the test evidence feature is empty and is not replaced by an image. Similarly, if a plant's data includes mowing and hay yield, the server maps it to the forage category; if only the plant has beautiful leaf color and long flowering period, this field is classified as ornamental grass, not forage grass.

[0060] When using it, the module formed in step one is packaged. Strength of Evidence The data is directly passed to step two. The matrix cell not only records whether there is a field, but also where the field comes from, the strength of the evidence, and whether there are empty spaces, thus providing continuous input for determining the primary and secondary uses in step three.

[0061] Complete Use Field Fit Strength Subsequently, the data processing server generates a use evidence matrix. In this matrix, each row corresponds to a plant ID, each column corresponds to a use category, and each cell stores direct use evidence, morphological supporting evidence, biological supporting evidence, genetic stability evidence, specific quality evidence, experimental evidence, approved material evidence, conflicting fields, and missing fields. To ensure that matrix cells are not misled by a single strong field, the server incorporates a material gating coefficient during the filling process. and conflict deduction factor , forming the application matrix filling value : Furthermore, in the engineering implementation, the use evidence matrix uses a ternary association index to maintain a fixed correspondence between plant ID, use category, and evidence field. The ternary association index includes a plant ID index, a use category index, and an evidence field index, and maintains the same field names, the same field sources, and the same evidence layer markers in two-dimensional tables, ternary tables, graph database edge relations, or key-value indexes.

[0062] When the data processing server reads the data required for the suitability of use, it first locates the plant number index, then the use category index, and finally reads the field values, source modules, data types, and evidence binding strength corresponding to the evidence field index.

[0063] In the formula: ... Indicates plant number In terms of usage category The available fill values ​​in the cell, ranging from 0 to 1, serve as input for reading the candidate category sequence in step three; their purpose is to assess the field's fit strength. Following the aforementioned definition, its function is to express the basic degree of fit between the field and its application category; material gating coefficient Indicates plant number Category of use The completeness of the necessary materials is rated from 0 to 1, with higher values ​​for more complete source reports, images, test data, and inspection data. Among them, the material gate coefficient and conflict deduction factor Generated using computable rules; material gating coefficient The specific form is:

[0064] Material gate coefficient Indicates plant number In terms of usage category The availability of the necessary materials, with a value ranging from 0 to 1. (Required materials set) Indicates the category of use Corresponding source reports, photographic materials, experimental data, testing data, or safety certification materials. Material weighting. Material Category of use The importance level, ranging from 0.1 to 1, is predetermined by the necessary evidence table. The material contains markers. Indicates plant number Does it have materials? It exists and can be associated with the plant ID. If the value is 1, then the value is 0; otherwise, the value is 0.

[0065] Among them, the conflict deduction coefficient The specific form is: Where: Conflict deduction coefficient Indicates plant number In terms of usage category The degree to which contradictory fields exist, ranging from 0 to 1. Set of contradictory fields. This represents a set of fields that contain contradictory or inconsistent records within the same usage category; conflict weight. Indicates conflicting fields Category of use The degree of impact, ranging from 0.1 to 1, is preset by the field mapping table. Conflict confirmation flag. Indicates whether the conflict has been identified and retained by the system; the value is 0 or 1. Field Reach Coefficient Strength of fit of application field The definition in [the document / reference].

[0066] Conflict deduction factor Indicates plant number In terms of usage category The degree to which conflicting fields exist is indicated, with a value ranging from 0 to 1. The higher value is used when the same field appears in contradictory records from different sources.

[0067] For example, a plant is described as both edible and possessing medicinal active ingredients. The server fills in evidence of use in the edible herb and medicinal herb columns respectively, but simultaneously writes missing fields for food safety testing, active ingredient testing, and description of medicinal parts into both cells; if safety materials are missing, the material gating coefficient is used. The ranking does not increase based on the presence of usage terms. For example, if a plant possesses attractive inflorescences, unique leaf color, and low-growing, mass-planting characteristics, the server fills in inflorescence, leaf color, ornamental period, and landscaping evidence for the ornamental grass category, and low-growing, dense, green period, and established state evidence for the turfgrass category, while writing back mowing tolerance and trampling tolerance as missing fields for the turfgrass category. For energy grasses, if a plant only has a high biomass description but lacks data on calorific value, cellulose content, or biomass conversion efficiency, the server retains the energy grass candidate cell but marks the specific quality characteristics as empty, not directly replacing the energy conversion field with the biomass field. During conflict isolation, the server does not overwrite the original fields but instead places conflicting fields side-by-side in the conflict field area, preserving their respective field origins and evidence binding strength.

[0068] After the staff selects a confirmation value on the review interface, the server only updates the confirmation marker in the use evidence matrix; the original fields remain in the five-module plant feature package. This allows the use evidence matrix to be read in step three and also to be used to find the source of evidence in step one, preventing the review process from disrupting the data chain. In use, each cell simultaneously displays supporting, limiting, and supplementary information, preserving horizontal candidate space for plants with multiple uses and preventing the use category with insufficient data from being classified as a mature classification result.

[0069] Step 3: Calculate the use fit of each plant number relative to each use category based on the use evidence matrix, and output the primary use category, secondary use category, classification basis, and review status while preserving the space for multi-use applications. All of the above steps are completed by the data processing server; the manual review terminal only receives the candidate use category sequence, missing field list, and review prompts generated by the server.

[0070] Step two has mapped the five-module plant characteristic package into a use evidence matrix. However, the candidate labels in the matrix only indicate that a plant has the evidence position to enter a certain use category, and cannot directly indicate that the use should be the primary use category. In the approval of grass varieties, the same plant often has clues for ecological restoration, ornamental, lawn, forage, medicinal, spice, food, or energy use. If sorted only by the frequency of use words, it is easy to push promotional uses, species-level generalized descriptions, or uses lacking testing materials to the forefront. Therefore, Step three connects the field strength, evidence source, material gating, and conflict deduction in the use evidence matrix into a single chain, so that the use allocation is constrained by the evidence structure, rather than driven by a single word.

[0071] The data processing server first reads the plant number from the use evidence matrix. and category of use Each cell is then broken down into a semantic layer, a morphological support layer, a biological support layer, a genetic stability layer, a specific quality layer, a test evidence layer, and a safety compliance layer. The server then calls the usage matrix generated in step two to populate the values. As the base value of the cell, it reads conflicting fields, missing fields, and material gating status. If a certain use category only has direct use text without specific qualities or experimental evidence matching that use, the server retains the use category as a candidate but lowers its eligibility to enter the main use category; if a use category has less use text but has morphological support, biological support, and experimental evidence, the server enters it as a substantial candidate for sorting.

[0072] In terms of engineering implementation, the data processing server preferably uses a relational database to store plant IDs, usage categories, and field sources; a document database to store original text segments and image annotations; and a rule table to store the evidence layer names, layer weights, and thresholds for each usage category. The scoring program reads the usage evidence matrix row by row according to the plant ID and expands each cell into a fixed field object. The field object includes the field name, field value, field source, evidence type, missing status, and conflict status. If the deployment environment already has a variety approval business system, step three receives the usage evidence matrix via an interface file; if the deployment environment only has a local computer, step three uses a spreadsheet and folder path as input. Database devices with equivalent functionality can also execute the procedure.

[0073] To avoid a single field dominating the results, the server sets seven evidence layers for each use category, and each evidence layer only accepts fields with the same function. For example: the biological support layer for ecological restoration grasses accepts drought resistance, salt tolerance, cover, and biomass; the specific quality layer for forage grasses accepts fresh grass yield, hay yield, and nutritional value; and the safety compliance layer for medicinal grasses accepts test reports and descriptions of medicinal parts. The server does not directly convert "good growth" in an image into the experimental evidence layer, nor does it directly convert "edible" into the safety compliance layer. The fit is generated using the following formula. : In the formula: application suitability Indicates plant number Relative to the category of use The overall fit degree, with a value ranging from 0 to 1, serves to form a sequence of candidate use categories; plant numbering. This represents a single plant object generated in step one and used in step three; its purpose is categorized as follows. This indicates ecological restoration grass, pasture grass, ornamental grass, turfgrass, medicinal grass, spice grass, edible grass, or energy grass; the fill value in the usage matrix is... This indicates that the cells output in step two can be filled with values. Its purpose is to introduce the field matching, material gating, and conflict deduction from step two into step three. Evidence layer set Indicates the category of use Seven layers of evidence; layer weights Indicating the evidence layer Category of use The contribution value ranges from 0.05 to 0.35, and the sum of the layer weights for each usage category is 1; layer score Indicates plant number In terms of usage category Evidence layer The field support level formed in the middle takes a value from 0 to 1; layer completeness coefficient Indicating the evidence layer The required material availability level, with a value ranging from 0 to 1, is used to limit the suitability of missing materials for the intended use. The rise; In the coupled calculation of layered application suitability, the layer score is... It is generated jointly by the field reach status and field evidence fusion value within the evidence layer, and its specific form is as follows: In the formula: layer score Indicates plant number In terms of usage category Evidence layer The level of support for the fields formed in the middle is denoted by a value between 0 and 1. Layer field set Indicates the category of use Evidence layer The set of fields that need to be included in the scoring; the weight of fields within the layer. Representation field In terms of usage category and evidence layer The importance level, ranging from 0.1 to 1. Field reach coefficient. Representation field Whether there are available records; field evidence fusion value Representation field The overall strength of the evidence.

[0074] Layer completeness coefficient From the evidence layer The corresponding necessary fields and materials are jointly determined. When the evidence layer... When the necessary fields have available records and the corresponding materials exist, the layer completeness coefficient is... Take the higher value; when necessary fields are missing or the corresponding material cannot be associated with the plant number. At that time, the layer completeness coefficient Take the lower value. This will improve the suitability for the application. Each score can be traced back to the fields and evidence in the use evidence matrix, rather than being scored directly by humans.

[0075] Furthermore, the layer completeness coefficient It is generated using the necessary item completeness rule, and its specific form is as follows: Where: layer completeness coefficient Indicates plant number In terms of usage category Evidence layer The completeness of necessary fields and necessary materials is indicated by a value between 0 and 1; the set of necessary items. Indicates the category of use Evidence layer Fields and materials that must be verified; weight of necessary items Indicates the sequence number of the necessary items The importance of this layer of evidence is determined by a value ranging from 0.1 to 1, preset by the usage classification index library; necessary items are marked. Indicates plant number Does it have a required item number? The corresponding necessary item is set to 1 if it exists and can be associated with the plant number, otherwise it is set to 0; necessary item number Represents the set of necessary items One of the necessary fields or necessary materials. This set of symbols corresponds to the determination of the completeness of necessary fields and necessary materials.

[0076] For example, when a data clerk opens a plant ID on the server interface, the ecological restoration grass list displays information on mine wasteland remediation, drought resistance records, cover photos, and regional trial summaries. The server then sends this information to the use semantic layer, biological support layer, morphological support layer, and trial evidence layer, respectively. Even if the same plant only has an inflorescence photo and landscape configuration text in the ornamental grass list, the server still calculates the use suitability of the ornamental grass. However, due to the lack of viewing period records and control materials, its stratum completeness coefficient is... The data is not supplemented. Consequently, multi-purpose plants can be scored in parallel, and uses with insufficient data will not be included in the high-order sequence due to a single description.

[0077] Obtain the application fit of each application category Then, the server sorts all use categories under the same plant ID in descending order, resulting in the first use category, the second use category, and the remaining use categories. To avoid mechanically grouping two similar uses into a single category, the server introduces a use separation factor. : In the formula: (Application / Separation Degree) Indicates plant number The degree of separability between the primary and secondary use categories, with a value ranging from 0 to 1. Primary fit Indicates plant number The first priority is suitability for the intended use. The second priority is suitability for the intended use. Indicates plant number The second most important factor is its suitability for the intended use. It serves as a stable safety net. Indicates the use of avoiding the first fit. A baseline parameter, ranging from 0.01 to 0.10, is used to prevent plants with insufficient data from being forcibly classified due to extremely small differences.

[0078] When the first fit Reaching the preset classification threshold and application separation When the preset separation threshold is reached, the server writes the first purpose category into the primary purpose category field and retains the second purpose category as the reference category. When the first fit... The separation degree of application reaches the preset classification threshold. If the preset separation threshold is not reached, the server writes the first purpose category into the primary purpose category field, writes the second purpose category into the secondary purpose category field, and generates a review status.

[0079] If all uses are suitable None of the data met the preset classification threshold, so the server outputs insufficient data and passes the list of missing fields to step four.

[0080] For example, if a plant simultaneously possesses records of prominent inflorescences, distinctive leaf color, low and dense growth, and a long green period, the server will sort ornamental grasses and turfgrass side by side. The manual review terminal will display two columns of evidence sources, allowing reviewers to directly see that the ornamental evidence comes from the inflorescence and leaf color fields, while the turfgrass evidence comes from the low and dense growth and green period fields. Furthermore, the primary and secondary uses are not simply a matter of choosing one over the other, but rather a clear boundary is formed between the differences in use, the completeness of evidence, and the review status.

[0081] Classification threshold Sum and difference thresholds All are stored in the threshold table of the usage classification index library. Classification thresholds Indicates the category of use The minimum use fit score for inclusion in the primary use determination ranges from 0.5 to 0.8; for medicinal herbs, aromatic herbs, and edible herbs, due to safety and compliance considerations, the classification threshold is... Use the high end of this range. Difference threshold. This indicates whether the separation boundary of the auxiliary use category is retained between the first use category and the second use category, with a value ranging from 0.05 to 0.20.

[0082] Furthermore, the aforementioned preset separation thresholds are uniformly represented by difference thresholds, which are written into the threshold table according to the usage category. Each time the data processing server forms a candidate usage category sequence, it reads the classification threshold and difference threshold of the corresponding usage category from the threshold table, without temporarily changing the meaning of the thresholds during the scoring process; when a certain usage category is not enabled in the threshold table, the data processing server only outputs the candidate evidence field and the list of missing fields for that usage category, without writing the main usage category field.

[0083] When the first fit Greater than or equal to the classification threshold And application separation Greater than or equal to the difference threshold At that time, the data processing server outputs the primary use category and writes the secondary use category into the reference category field; when the first fit... Greater than or equal to the classification threshold And application separation Less than the difference threshold At that time, the data processing server outputs the primary use category, secondary use category, and review status; when all uses are compatible... All are less than the classification threshold At that time, the data processing server outputs a list of insufficient data status and missing fields.

[0084] To enable reviewers to reconstruct the reasons for the assignment, the server generates a classification basis fragment after each sorting. This fragment does not change the original fields but extracts the field name, source material, and the corresponding evidence layer. For example, when a plant's ecological restoration grass is listed as the first use category, the interface simultaneously displays information from the "Damaged Habitat" module (resource characteristics and application module), "Drought Resistance" (biological characteristics module), and "Coverage Status" (image evidence). Reviewers use this information to examine the original materials and confirm whether to accept the assignment result.

[0085] For medicinal herbs, spice herbs, edible herbs, and energy herbs, the server's suitability for different uses is... In addition, pre-construction safety and specialized material checks are performed. This check does not negate candidate use categories, but rather transforms gaps in the safety compliance layer, specialized quality layer, and experimental evidence layer into a list of missing fields. If the data states it is edible but lacks edible safety testing, the server retains the edible grass candidate category and adds edible safety testing, nutritional quality, and processing methods to the list of missing fields. If the data states it has medicinal active ingredients but lacks information on the content of effective ingredients and the appearance of medicinal parts, the server retains the medicinal grass candidate category and lists the corresponding test reports as verification fields. If the energy grass candidate only has high biomass information, the server adds calorific value, cellulose content, and biomass conversion efficiency to the list of missing fields.

[0086] Safety Pre-judgment Model Suitable for medicinal herbs, aromatic herbs, edible herbs, and energy herbs. (Regarding the category of use...) For data involving food safety, medicinal components, aromatic volatile oils, calorific value, cellulose content, or biomass conversion efficiency, the data processing server first reads the necessary set of materials. And calculate the material gate coefficient. When the material gate coefficient When the material threshold in the application classification index library is not met, the data processing server retains candidate evidence for that application category, but prohibits that application category from being directly written into the main application category field, and writes the missing materials into the missing field list.

[0087] Furthermore, the safety pre-judgment model takes the necessary material set and material gating coefficients as inputs, and outputs candidate retention status, prohibited main use status, and a list of missing fields. For medicinal herbs, the necessary material set includes at least a description of the medicinal parts, data on the detection of active ingredients, and photographs of the appearance of the medicinal parts; for aromatic herbs, the necessary material set includes at least data on the detection of aromatic volatile oils, a description of processing characteristics, and safety information; for edible herbs, the necessary material set includes at least data on edible safety, nutritional quality information, and a description of processing and utilization; for energy herbs, the necessary material set includes at least calorific value data, cellulose content data, and biomass conversion efficiency data.

[0088] For example, a staff member uploads a description of the plant's uses, which states that the tender stems and leaves are edible and the roots are used medicinally. The server reads the text indicating edible use in the edible herbs category and the text indicating medicinal parts in the medicinal herbs category, then searches for testing data and image data. If the system cannot find edible safety testing results or photos of the medicinal parts, the interface will not directly display the final conclusion on whether it is an edible or medicinal herb. Instead, it will display edible herb candidates, medicinal herb candidates, and two sets of materials to be supplemented.

[0089] Furthermore, the use of insufficient safety or special materials is still retained as a technical fact, but is not confirmed as a mature classification result in advance; at the same time, the list of missing fields can directly enter the material list generation stage in step four, reducing repeated manual verification.

[0090] The results of step three include plant ID, candidate use category sequence, and use suitability. Application Separation The process involves three steps: primary use category, secondary use category, verification status, classification criteria, and a list of missing fields. For ecological restoration grasses, the classification criteria record damaged habitat, stress resistance, cover status, restoration efficiency, and experimental source. For forage grasses, the classification criteria record feeding scenarios, yield, nutritional value, and palatability. For ornamental grasses and turfgrasses, the classification criteria record inflorescence, leaf color, ornamental period, low dwarfness, dense growth, green period, and turf establishment status, respectively. For medicinal grasses, spice grasses, edible grasses, and energy grasses, the classification criteria record medicinal parts, volatile oils, food safety, calorific value, cellulose content, and conversion efficiency, respectively. Thus, step three transforms the use evidence matrix into classification results that can be accessed in step four.

[0091] Step 4: Based on the primary use category and secondary use category, call the approval template to generate classification results, missing materials list and approval application material framework, and fix each output to the plant number, evidence field and material source.

[0092] Step three has already provided the primary use category, secondary use category, and review status. However, in the actual application process, staff still need to compile the plant name, family, species, source type, morphological photos, test reports, quality testing, suitable promotion area, cultivation techniques, and authenticity commitment letter into the review materials. If only a single use label is output, the review template cannot determine which fields have existing evidence, which fields are merely candidate clues, and which fields require supplementary testing. Therefore, step four links the use classification results to template slots, allowing ecological restoration grasses, pasture grasses, turfgrass, ornamental grasses, medicinal grasses, spice grasses, edible grasses, and energy grasses to enter their respective material paths. Simultaneously, supplementary slots for secondary use categories are retained to prevent multi-purpose plants from being truncated by a single template.

[0093] The following actions are all performed by the data processing server; the material verification terminal only displays the classification report and material package. The data processing server first reads the plant number. The corresponding primary use category is used as the primary template index; then, the secondary use category is read and used as the secondary template index; subsequently, the direct use evidence, morphological supporting evidence, biological supporting evidence, genetic stability evidence, specific quality evidence, experimental evidence, and approved material evidence corresponding to that category are extracted from the use evidence matrix and written into the template slots. If the review status is insufficient data, the data processing server does not generate a final use conclusion page, but instead generates a supplementary materials page and a review prompt page; if the review status is a combined use, the primary template retains the complete indicator table, and the secondary template only opens the indicator columns and experimental columns directly related to the secondary use.

[0094] In the preferred embodiment, staff first select the batch to be processed at the material verification terminal. The data processing server reads the output files from step three one by one according to the plant number, and then enters the template library according to the main purpose category. The template library is not a collection of blank documents, but a structured file composed of slot names, field sources, allowed evidence types, prompts, and output page segments. Each slot only receives the corresponding source field. For example, the source report slot receives resource source, breeding generation, or introduction certificate; the image slot receives photos of single plants, populations, and organs with completed part annotations; and the inspection slot receives quality inspection reports. Staff only need to confirm whether the slot accepts the field on the interface, and the server then generates the report text, table of contents, and attachment list. The output results can be printed as paper materials or exported as document files.

[0095] As an alternative implementation, the template library can be deployed on the provincial approval server or on the applicant's local computer; the field index can be a relational table or a document directory with plant IDs and slot numbers. These alternative methods do not change the processing order of step four, only the storage location and file format, thus covering three scenarios: centralized application, batch initial screening, and single-variety correction.

[0096] The data processing server pre-sets fixed slots for each use category in the template library. These fixed slots include a plant basics slot, a use conclusions slot, a trait index slot, an experimental evidence slot, a materials catalog slot, and a supplementary suggestions slot. The plant basics slot receives the basic information module and genetic information module formed in step one; the trait index slot receives the botanical characteristics module, biological characteristics module, and resource characteristics and application module; the experimental evidence slot receives the experimental supporting evidence from the use evidence matrix in step two; and the supplementary suggestions slot receives the list of missing fields from step three. When a template is called, the data processing server does not simply concatenate templates from different uses. Instead, it first writes the main use category into the main template, and then writes the auxiliary use categories into the auxiliary use column of the same report. Only when the auxiliary use category has independent specific quality evidence is an additional template index page added.

[0097] Template slot matching operator approved Taking the primary use category, secondary use category, and use evidence matrix as input, the output is a set of template slots. The template library includes template number, purpose category, slot number, slot name, source field, allowed evidence type, and missing information message. The main purpose category calls the complete main template, while auxiliary purpose categories call the supplementary slots. Each slot only receives evidence consistent with its source field; unfilled slots are added to the missing materials list.

[0098] For example, when the primary use category is forage, the template slot set The fresh grass yield slot only accepts the yield test field or production test field, the nutritional value slot only accepts the nutritional testing field, and the mowing or grazing method slot only accepts the utilization method field from the resource characteristics and application module. If a slot only has the usage text without corresponding test or testing materials, the slot remains unfilled.

[0099] For example, a plant's primary use category is "ecological restoration grass," and its secondary use category is "ornamental grass." The data processing server opens the ecological restoration grass approval template and writes the damaged habitat type, restoration target, cover change, drought resistance, salt tolerance, overwintering rate, summering rate, aboveground biomass, belowground biomass, and control varieties into the ecological restoration grass index table. Simultaneously, it records inflorescence, leaf color, ornamental period, and landscape configuration in the secondary use column, but does not forcibly incorporate the entire catalog of ornamental grass materials into the main report. On-site staff see a report focused on ecological restoration grass at the material verification terminal, with evidence sources and supplementary materials for ornamental use displayed alongside. This ensures the primary use indicators are complete, secondary use indicators are not lost, and reduces the risk of duplicate entry when materials for combined uses are split into multiple reports.

[0100] To avoid situations where a template is invoked but the slot lacks a traceable field, the data processing server calculates the template slot completion rate for each plant number. : In the formula: template slot completion rate Indicates plant number The report slot is filled with valid evidence to a degree ranging from 0 to 1, and its function is to determine whether the report can form a classification results page; plant number. Represents a single plant object throughout the four steps; a collection of slots. Indicates plant number All report slots that need to be filled in the main template and the supplementary template; Slot number Represents a report slot; slot weight Indicates slot number The importance of the corresponding template is assigned a value ranging from 0.1 to 1; slot fill mark. Indicates plant number Slot number Does it have available fields, with a value of 0 or 1; evidence set Indicates the supported slot number A collection of textual, pictorial, experimental, or testing evidence; strength of evidence binding. Following the meaning of step one, the evidence number is indicated. The degree of association with the target field, with a value ranging from 0 to 1; Maximum selection function Indicates from the set of evidence Selecting evidence binding strength The highest priority is to avoid low-reliability sources covering proofread material; When the same reporting slot contains multiple pieces of evidence, the data processing server first generates a slot evidence fusion value. Then calculate the template slot completion rate. Slot evidence fusion value Field evidence fusion values ​​can be reused The fusion logic is based on the support slot sequence number. The evidence set is generated accordingly. Correspondingly, the template slot completion rate... Calculate using the following formula: The specific form of the slot evidence fusion value is as follows: Where: slot evidence fusion value Indicates plant number Slot number The overall strength of evidence for the corresponding report slot is rated from 0 to 1; evidence set. Indicates the supported slot number A collection of textual, visual, experimental, or testing evidence. Evidence binding strength. Similar to step one, this refers to obtaining the evidence sequence number. The degree of association between the data and the target field ranges from 0 to 1. This formula allows multiple pieces of evidence from a specific report slot to participate in the completion rate calculation.

[0101] In the formula: template slot completion rate Indicates plant number The report slots are filled with valid evidence to a degree ranging from 0 to 1. Slot set Indicates plant number All report slots that need to be filled in the main template and supplementary templates. Slot weight. Indicates slot number The importance of the corresponding template ranges from 0.1 to 1. Slot fill marker. Indicates plant number Slot number Does it have an available field? The value can be 0 or 1. Slot evidence fusion value. Indicates the supported slot number The fusion strength of textual, image, experimental, or testing evidence, with a value ranging from 0 to 1.

[0102] For example, a candidate forage grass already has fields for mowing method, fresh grass yield, hay yield, and nutritional value, and a variety comparison test report has been uploaded. The data processing server fills these fields into the forage grass approval template and writes the corresponding report page number and image number in the slot source field. If palatability description is missing, the palatability slot remains unfilled and is not replaced with livestock feeding terms. For the turfgrass template, density, uniformity, color, texture, green period, and turf establishment speed become different slots. If there is only a turfgrass photo without trampling resistance evaluation, the trampling resistance slot is added to the missing materials list. The effect of this processing is that the report content corresponds one-to-one with the evidence fields, and the examiners can directly see the filled and unfilled slots, reducing field mismatches caused by supplementing materials based on usage labels.

[0103] Furthermore, the success rate of template slots After generation, the data processing server calculates the material priority value for the unfilled slots. : In the formula: material preference value Indicates plant number Slot number The priority order in the supplementary materials list, with a value not less than 0, determines whether to supplement test reports, inspection reports, or image materials first; slot filling mark. Following the aforementioned meaning, the value is either 0 or 1; slot weight. Following the aforementioned meaning, the value ranges from 0.1 to 1; application suitability. Using the meaning from step three, we can represent the plant number. Category of use The overall compatibility level, with a value ranging from 0 to 1; Application Category This indicates that step four is calling the template category; the main template uses the primary purpose category, and the supplementary template uses the secondary purpose category; [This refers to] the compensation item for composite purposes. express The value ranges from 0 to 1, and its function is to improve the separation of applications. Increase the arrangement position of missing slots related to auxiliary uses when the position is low; The specific form of the compensation item for multi-purpose use is as follows: In the formula: Compensation item for multi-purpose use Indicates plant number The degree of compensation for missing slots when the primary and secondary use categories are similar, with a value ranging from 0 to 1; Use separation degree. Following the meaning of step three, this represents the degree of separability between the first and second use categories, with a value ranging from 0 to 1. When the use separation degree... At a lower level, the compensation item for multi-purpose use The missing slots corresponding to the auxiliary use category are moved to the front of the bill of materials.

[0104] Application Separation Following the meaning of step three, this indicates the degree of separability between the first use category and the second use category, with a value ranging from 0 to 1; For example, a plant is identified as a medicinal herb with secondary uses as an edible herb. The data processing server generates a list of medicinal parts, fresh and dried yields, appearance characteristics, extract content, active ingredient content, and test reports in the medicinal herb template, and generates nutritional quality, protein content, taste evaluation, food safety, and processing methods in the edible herb auxiliary section. If both active ingredient and food safety tests are missing, the material priority value is used. The two types of tests are prioritized before regular image supplementation. If a close-up of the medicinal part is missing from the image, the system will prompt the material verification terminal to supplement with a clear photo of the root, stem, leaf, flower, fruit, or seed corresponding to the medicinal part. For energy grasses, if calorific value, cellulose content, and biomass conversion efficiency are missing, the report will still retain the energy grass candidate page, but the classification results page will be marked as pending completion.

[0105] This ensures that the missing materials list is no longer a generalized list, but rather an orderly path to complete the list based on the primary use category, secondary use category, and slot weight.

[0106] Step four: Final packaging of the classification report and materials package. The classification report should include at least the plant number, plant name, family, genus and species information, primary use category, secondary use category, and use suitability. Application Separation The application package includes supporting evidence fields, evidence sources, missing evidence fields, recommended supplementary trials, recommended control varieties, suitable promotion areas, key cultivation techniques, and a list of application materials. The materials are arranged in the following order: application form, source report, botanical characteristic images, variety comparison test report, regional trial report, production trial report, resistance identification report, quality testing report, description of propagation materials, and a commitment to authenticity.

[0107] Preferably, the server uses a relational database to store the template slots and field indexes, and a document storage system to store the report text and image paths; equivalent local file systems, database devices, and document generation tools can also be used. As long as there is a fixed correspondence between the plant number, slot number, and evidence source, it is acceptable.

[0108] Recommended Operator for Comparison Varieties Plant numbering The input fields include: main use category, suitable promotion area, family, genus and species information, and evidence of use. The data processing server filters from the control variety database for plants with the correct plant number. Control varieties with the same or similar family, genus, species, main use category, and suitable promotion area are then selected, and a recommended list is generated in the order of comparable trait fields, complete experimental data, and complete source materials. If no control variety meets the criteria in the control variety library, the data processing server outputs a control variety to be manually specified, and the reason for the missing variety is written into the material list.

[0109] Furthermore, the control variety database should at least store the control variety name, family, genus, species, application category, suitable area, main traits, trial year, and source material. The data processing server first filters control varieties by family, genus, species, and main application category, then filters comparable objects by suitable promotion area, and subsequently verifies whether the trait fields correspond to the indicator tables in the classification report. If no control variety that meets the criteria is found in the control variety database, the data processing server outputs a control variety to be manually specified and writes the reason for the missing variety into the material list.

[0110] In this embodiment, the data processing server pre-establishes a weight configuration table, a usage classification index library, a threshold table, a template slot table, and a control variety library. The weight configuration table stores the field weights. Field weights Intra-layer field weights Layer weights Material weight and slot weight The usage classification index library stores usage categories, evidence layers, field names, source modules, and candidate rules; the threshold table stores classification thresholds. Sum and difference thresholds The template slot table stores the template number, slot number, slot name, source field, and missing information; the control variety library stores the control variety name, family, genus, species, use category, suitable area, main traits, and experimental data.

[0111] Furthermore, the field weights in the weight configuration table correspond to the field weights in module package degree, the field weights in purpose field fit strength, and the intra-layer field weights in evidence layer score, respectively. These three are stored separately in the weight configuration table using the weight object field. The purpose classification index library stores purpose category, evidence layer, field name, source module, candidate rules, and necessary material set; the threshold table stores classification threshold and difference threshold; the template slot table stores template number, slot number, slot name, source field, allowed evidence type, and missing message. The above configuration tables are loaded before the data processing server starts the classification task and are only called during operation, without changing the parameter names and meanings.

[0112] When using this template, the primary use category directly drives the main template, while secondary use categories are reserved in the supplementary template fields to avoid omitting plants with multiple uses. Template slot completion rate. Make report generation subject to evidence fields and material priority values. This ensures that supplementary materials have a clear order, and that classification results, lists of missing materials, and application frameworks can return to previous steps along the same plant number.

[0113] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A method for classifying and approving grass varieties based on their intended use, characterized by: The process is executed by the data processing server, including receiving the dataset of herbaceous plants to be classified, generating a plant ID for each plant, attaching text, images, experimental, detection, genetic and application data to the plant ID, and extracting fields to form a five-module plant feature package. Call the pre-set use classification index library and map the five-module plant feature package into a use evidence matrix with plant number as the row, use category as the column, and evidence field as the cell content according to use category and evidence layer; The use fit of each use category is calculated and ranked based on the use evidence matrix; When the highest-level suitability reaches the classification threshold and the difference between the highest-level suitability and the second-highest-level suitability reaches the difference threshold, the primary-level suitability category is output; when the highest-level suitability reaches the classification threshold but the difference does not reach the difference threshold, the primary-level suitability category, secondary-level suitability category, and review status are output; when the highest-level suitability does not reach the classification threshold, the data insufficiency status and missing field list are output. Use the corresponding approved template to generate a classification report.

2. The method for classifying and approving grass varieties based on their usage characteristics according to claim 1, characterized in that: The data processing server extracts the classification field group, source field group, source proof field group, and batch field from the herbaceous plant dataset to be classified; when the classification field group, source field group, and source proof field group are all consistent, the corresponding data are merged into the same plant number; When at least one field group is inconsistent, generate different plant IDs for the corresponding data; and write the plant IDs into each field record of the five-module plant feature package.

3. The method for classifying and approving grass varieties based on their usage characteristics according to claim 1, characterized in that: When the data processing server generates the five-module plant feature package, it records the source module, data type, original page segment, manual confirmation mark, and species-level generalization mark for each field. When a field record points to the plant itself corresponding to the plant number, the field record is written to the corresponding module. When a field record only points to the general characteristics of the plant species to which it belongs, the field record is retained as a species-level generalization field and excluded from the mapping input of the use evidence matrix; and the module packaging degree of each module is generated according to the field weight, the effective field label and the source credibility coefficient.

4. The method for classifying and approving grass varieties based on their usage characteristics according to claim 3, characterized in that: Each cell in the use evidence matrix is ​​associated with a plant number and a use category; Each cell includes evidence of direct use, morphological support, biological support, genetic stability, specific quality, experimental evidence, and approval materials. Each piece of evidence records the field name, field value, source module, data type, original page segment, and manual confirmation mark; the purpose suitability only calls the evidence fields that have been written in the corresponding cell.

5. The method for classifying and approving grass varieties based on their usage characteristics according to claim 1, characterized in that: The use evidence matrix uses a ternary association index to store a fixed correspondence between plant ID, use category, and evidence field; the ternary association index includes a plant ID index, a use category index, and an evidence field index. A ternary join index maintains the same field names, field sources, and evidence layer markers in at least one of the following data structures: a two-dimensional table, a ternary table, a graph database edge relation, and a key-value index; the data processing server reads the evidence fields used to calculate the suitability of use based on the ternary join index.

6. The method for classifying and approving grass varieties based on their usage characteristics according to claim 5, characterized in that: When the data processing server calls the application classification index library, it reads the application trigger field, supporting field, and material field for each application category in sequence; when all three fields exist, it generates candidate application categories. When the application trigger field is missing but the supporting field and material field exist, a candidate application category to be reviewed is generated; When only the use trigger field exists and the supporting and material fields are missing, use clues are generated and writing to the candidate use category sequence is prohibited; when other field combinations appear, a corresponding list of missing fields is generated and the use category remains in an undetermined state.

7. The method for classifying and approving grass varieties based on their usage characteristics according to claim 6, characterized in that: When the candidate use category belongs to at least one of medicinal herbs, spice herbs, edible herbs, and energy herbs, the data processing server reads the security compliance field and the special quality field. When both the safety compliance field and the special quality field exist, the candidate use category is allowed to enter the use suitability ranking; If at least one field is missing, retain the evidence field for that candidate use category, output the review status and a list of missing fields, and prohibit that candidate use category from being directly written into the main use category.

8. The method for classifying and approving grass varieties based on their usage characteristics according to claim 7, characterized in that: The data processing server generates a classification results page when generating a classification report; the classification results page displays the plant number, candidate use category sequence, main use category, auxiliary use category, review status, data insufficiency status, missing field list, and approval template number; When an auxiliary use category exists, the classification results page displays the auxiliary use column; when no auxiliary use category exists, the classification results page hides the auxiliary use column and retains the primary use category column.

9. The method for classifying and approving grass varieties based on their usage characteristics according to claim 8, characterized in that: The data processing server provides a classification result interface; the request message of the classification result interface includes a plant number field, a batch field, and a template number field. The response message from the classification results interface includes fields for usage sorting, primary usage, secondary usage, review status, insufficient data, missing data, and report address. When the data is insufficient, the primary purpose field is set to null and the missing field is written to the field name set; when the data is insufficient, the primary purpose field is written to the primary purpose category and the report address field is written to the storage address of the category report.

10. The method for classifying and approving grass varieties based on their use characteristics according to claim 9, characterized in that: The data processing server generates a classification report file, a bill of materials file, and an evidence index file at the location corresponding to the report address field. The classification report file should include the plant number, primary use category, secondary use category, and review status; The bill of materials file is filled with filled material items and missing material items according to the slots of the approved template; the evidence index file is filled with evidence fields, source data names, original page segments and image numbers; after each call to the classification result interface, an interface log containing request time, plant number field, response status and report address field is generated.

Images