A method for automatically generating insulator defect reports

Abstract

The application belongs to the technical field of data processing, and provides a kind of insulator defect report automatic generation method, comprising: obtaining the original detection result of insulator defect, and the original detection result is preprocessed in multiple levels, and engineering level defect characteristic parameter set is obtained;According to engineering level defect characteristic parameter set, through structured semantic modeling and regularization generation, structured semantic description unit set is constructed;The corresponding inspection report template is matched for structured semantic description unit set, and based on structured semantic description unit set, chapter level mapping is carried out on inspection report template, and report skeleton data is generated;According to report skeleton data, through chapter level rendering and redundancy merging, report text content is generated, and insulator defect report is obtained.The scheme realizes the full automatic generation of insulator defect report through the closed loop process of multi-level preprocessing, structured semantic modeling, template matching mapping and chapter level rendering, and improves the report generation efficiency.

Description

Technical Field

[0001] This invention relates to the field of data processing technology, and in particular to a method for automatically generating insulator defect reports. Background Technology

[0002] With the intelligent upgrading of power systems, the inspection of insulators on transmission lines has widely adopted drone aerial photography and computer vision detection technologies, realizing automated identification and location of defects and significantly improving inspection coverage efficiency. However, in practical applications, there is still a significant technical gap between defect detection results and the generation of maintenance defect reports, making it difficult to meet the standardized and efficient management needs of power operation and maintenance.

[0003] In related technologies, insulator defect detection results are often output in fragmented, low-semantic forms such as defect category labels, image coordinates, and confidence scores, lacking unified structured standards. This makes it difficult to directly integrate and analyze data output by different detection models. Inspection report generation still heavily relies on manual intervention; maintenance personnel must manually extract detection data, organize defect information, and write report text. This report generation method is not only inefficient but also prone to inconsistencies in description, omission of key information, or logical inconsistencies due to differences in individual experience. Furthermore, different maintenance units and different inspection tasks have varying requirements for report format and content emphasis. Existing technologies often use fixed templates to generate reports, failing to adaptively adjust the report structure and content according to actual needs. This makes it difficult to adapt to complex inspection scenarios and fails to provide accurate engineering semantic support for maintenance decisions.

[0004] It is evident that traditional insulator defect detection report generation schemes suffer from technical problems such as low generation efficiency, poor standardization, and insufficient adaptability. Summary of the Invention

[0005] This invention provides an automatic method for generating insulator defect reports, which solves the problems of low generation efficiency, poor standardization, and insufficient adaptability of traditional insulator defect detection report generation schemes.

[0006] This invention provides a method for automatically generating insulator defect reports, comprising: The original detection results of insulator defects are obtained, and the original detection results are preprocessed in multiple levels to obtain a set of engineering-level defect feature parameters. Based on the set of engineering-level defect feature parameters, a set of structured semantic description units is constructed through structured semantic modeling and rule-based generation; Match the corresponding inspection report template to the set of structured semantic description units, and perform chapter-level mapping on the inspection report template based on the set of structured semantic description units to generate report skeleton data; Based on the report skeleton data, the report text content is generated through chapter-level rendering and redundancy merging to obtain the insulator defect report.

[0007] According to the automatic insulator defect report generation method provided by the present invention, the original detection results are subjected to multi-level preprocessing to obtain defect feature data, including: The original detection results are standardized to obtain a set of standardized defect description parameters; Based on the standardized defect description parameter set, defect information statistics and engineering-level feature modeling are performed to obtain an engineering-level defect feature parameter set.

[0008] According to the automatic insulator defect report generation method provided by the present invention, the original detection results are standardized to obtain a standardized defect description parameter set, including: For the defect category identifiers in the original detection results, perform defect category mapping processing to obtain standardized defect category information; Based on the standardized defect category information, the severity of the defect is determined, and the severity level of the defect is identified. The defect location information in the original detection results is subjected to location parameter standardization processing to obtain standardized defect location information; Based on the detection confidence information in the original detection results, confidence identification processing is performed to obtain confidence identification information; The key defect information, including the standardized defect category information, defect severity level, standardized defect location information, and credibility identification information, is encapsulated to generate a standardized defect description parameter set.

[0009] According to the automatic insulator defect report generation method provided by the present invention, defect category mapping processing is performed on the defect category identifier in the original detection results to obtain standardized defect category information, including: Construct a defect category mapping table, wherein the defect category mapping table is used to represent the correspondence between each defect type and a unique category code; Based on the defect category mapping table, defect categories with the same defect category identifier but different naming methods in the original detection results are uniformly mapped to their corresponding unique category codes to obtain standardized defect category information.

[0010] According to the automatic insulator defect report generation method provided by the present invention, based on the standardized defect description parameter set, defect information statistics and engineering-level feature modeling are performed to obtain an engineering-level defect feature parameter set, including: Based on the insulator identification parameters of various defects in the standardized defect description parameter set, defect aggregation processing is performed to obtain multiple defect sets; For each defect set, based on the defect category information and defect quantity parameters in the standardized defect description parameter set, defect quantity and category distribution statistics are performed to obtain quantity feature parameters and category distribution feature parameters; For the same set of defects, based on the standardized defect location information in the standardized defect description parameter set, the degree of spatial concentration of defects is determined in an engineering manner, and defect spatial concentration feature parameters are generated. Based on the defect severity level, defect quantity parameter and defect spatial concentration characteristic parameter in the standardized defect description parameter set, a comprehensive operational risk level model is performed for each target insulator to generate a comprehensive risk level parameter. For each defect set, based on the credibility identifier information in the standardized defect description parameter set, credibility status aggregation is performed to obtain credibility status marking results; The quantitative characteristic parameters, category distribution characteristic parameters, defect spatial concentration characteristic parameters, comprehensive risk level parameters, and credibility status marking results are encapsulated to generate an engineering-level defect characteristic parameter set.

[0011] According to the automatic insulator defect report generation method provided by the present invention, based on the engineering-level defect feature parameter set, a set of structured semantic description units is constructed through structured semantic modeling and rule-based generation, including: Based on the pre-defined data structure, construct the initial description unit; For each defect category entry in the initial description unit, a rule engine-style slot filling is performed, and a data mapping from defect categories to technical terms and description templates is performed to obtain a semantic description unit; The semantic description units are subjected to multi-defect entry merging and redundancy removal processing to obtain a set of structured semantic description units.

[0012] According to the automatic insulator defect report generation method provided by the present invention, matching the structured semantic description unit set with the corresponding inspection report template includes: Construct a report structure template library, wherein the report structure template library contains multiple predefined report structure templates, and each report structure template adopts a parameterized chapter structure; The risk level parameter of each semantic description unit is extracted from the set of structured semantic description units, and the overall risk level of the report is determined based on the risk level parameters of all semantic description units. Based on the overall risk level of the report, the corresponding target report structure template is extracted from the report structure template library as the inspection report template.

[0013] According to the automatic insulator defect report generation method provided by the present invention, based on the set of structured semantic description units, the inspection report template is mapped at the chapter level to generate report skeleton data, including: The chapter organization method is determined based on the number of insulator identifiers in the set of structured semantic description units; According to the chapter organization method, the semantic description content in the set of structured semantic description units is mapped to each chapter in the inspection report template to obtain preliminary skeleton data; The chapter content in the preliminary skeleton data is conflict-resolved and merged to obtain the report skeleton data.

[0014] According to the automatic insulator defect report generation method provided by the present invention, based on the report skeleton data, report text content is generated through chapter-level rendering and redundancy merging to obtain an insulator defect report, including: Extract the chapter order parameters from the report skeleton data, and generate the corresponding chapter texts sequentially according to the chapter order parameters; Template rendering is performed on the semantic content entries contained in each chapter text to generate candidate text segments; Multiple candidate text segments within the same chapter are processed by sequential organization and redundancy merging to obtain the chapter content. By piecing together the contents of all chapters in order, a report text is generated, resulting in an insulator defect report.

[0015] The automatic insulator defect report generation method provided by the present invention further includes, before sequentially splicing all chapter contents: The consistency of defect terminology and risk level descriptions for the same insulator identifier across different sections must be verified; and / or, Determine whether the action level in the handling recommendations section is lower than the risk level reflected in the risk analysis section. If the determination result is yes, adjust the action level in the handling recommendations section to be consistent with the risk level.

[0016] The automatic insulator defect report generation method provided by this invention achieves fully automatic generation of insulator defect reports through a closed-loop process of multi-level preprocessing, structured semantic modeling, template matching mapping, and chapter-level rendering. It not only breaks down the semantic gap between raw inspection data and standardized reports, but also significantly improves report generation efficiency through engineering-level feature modeling and rule-based semantic generation. Furthermore, it can adaptively match report templates in different scenarios, ensuring the standardization, professionalism, and logical coherence of the reports. This avoids subjective differences and information omissions caused by manual operation, and provides accurate data support for power operation and maintenance decisions, effectively promoting the intelligent and efficient upgrading of the entire power inspection process. Attached Figure Description

[0017] To more clearly illustrate the technical solutions in this invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0018] Figure 1 This is a flowchart illustrating the automatic generation method for insulator defect reports provided in an embodiment of the present invention; Figure 2 This is a schematic diagram of the structure of the electronic device provided in an embodiment of the present invention. Detailed Implementation

[0019] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this invention. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without creative effort are within the scope of protection of this invention.

[0020] The following is combined Figure 1 and Figure 2 This invention describes the detailed scheme of the automatic generation method for insulator defect reports provided in embodiments of the present invention.

[0021] like Figure 1 As shown, the automatic insulator defect report generation method provided in this embodiment of the invention mainly includes the following steps: Step 110: Obtain the original detection results of insulator defects, and perform multi-level preprocessing on the original detection results to obtain a set of engineering-level defect characteristic parameters.

[0022] In practical applications, the original detection results of insulator defects can be obtained from the insulator defect detection system. The defect detection system can be implemented based on image recognition or target detection algorithms, by processing the inspection images and outputting the detection results.

[0023] Understandably, the main purpose of multi-level preprocessing is to improve the data quality of the original detection results and provide accurate and effective data for subsequent data processing.

[0024] Step 120: Based on the set of engineering-level defect feature parameters, construct a set of structured semantic description units through structured semantic modeling and rule-based generation.

[0025] In this embodiment, a set of structured semantic description units can be obtained through semantic-level data processing operations, thereby providing a data foundation for determining the content of subsequent report texts.

[0026] Step 130: Match the corresponding inspection report template to the set of structured semantic description units, and perform chapter-level mapping on the inspection report template based on the set of structured semantic description units to generate report skeleton data.

[0027] It is understandable that this embodiment can solve the problem of poor scenario adaptability of a single template strategy by using template adaptive matching, thereby improving the flexibility and scenario adaptability of the report generation process.

[0028] Step 140: Based on the report skeleton data, generate the report text content through chapter-level rendering and redundancy merging to obtain the insulator defect report.

[0029] It should be noted that the solution provided by this invention automates, standardizes, makes interpretable and traceable the report generation process without increasing the complexity of the detection model, significantly reducing the cost of manual generation and subjective differences, and improving the quality of inspection reports and the ability to support operation and maintenance decisions.

[0030] In one embodiment, the original detection results are subjected to multi-level preprocessing to obtain defect feature data, specifically including: First, the original detection results are standardized to obtain a set of standardized defect description parameters.

[0031] In this embodiment, the original detection results specifically include one or more of the following: defect category identification information, the bounding box or pixel coordinate information of the defect in the inspection image, the detection confidence information corresponding to the defect, the image number to which the defect belongs, and the shooting time or inspection task identification. Table 1 below shows the original detection results under one scenario.

[0032] Table 1 Original test results

[0033] In one specific implementation, the original detection results are standardized to obtain a standardized set of defect description parameters, which specifically includes: On the one hand, defect category mapping processing is performed on the defect category identifiers in the original detection results to obtain standardized defect category information.

[0034] In this embodiment, defect category mapping processing is performed on the defect category identifier in the original detection result to obtain standardized defect category information, specifically including: The first step is to construct a defect category mapping table, which is used to represent the correspondence between each defect type and a unique category code.

[0035] Understandably, the defect category mapping table is primarily used to map defect category identifiers in the raw inspection results to unique category codes for predefined standard defects. Table 2 shows the correspondence between defect category identifiers and unique category codes for standard defects in the defect category mapping table.

[0036] Table 2 Defect Category Mapping Table

[0037] The second step is to map the defect categories with the same defect category identifier but different naming methods in the original detection results to their corresponding unique category codes based on the defect category mapping table, thereby obtaining the standardized defect category information.

[0038] It is understandable that when different detection models output inconsistent defect category names, the aforementioned defect category mapping table can unify them into the same unique category code, thereby achieving unified defect category mapping and coding.

[0039] On the other hand, based on the standardized defect category information, the severity of the defect is determined, and the severity level of the defect is identified.

[0040] In this embodiment, a correspondence rule between defect categories and defect severity levels can be pre-established. Subsequently, the corresponding defect severity level can be determined based on the standard defect category to which the defect belongs. In practical applications, the defect severity level can be set to two levels: high and low.

[0041] On the other hand, the defect location information in the original detection results is subjected to location parameter standardization processing to obtain standardized defect location information.

[0042] In this embodiment, the defect location information in the original detection result is subjected to location parameter standardization processing to obtain standardized defect location information, specifically including the following steps or any combination of steps: The pixel-level coordinates in the defect location information are converted into normalized coordinates corresponding to the inspection image size.

[0043] In practical applications, taking the pixel-level coordinates of the center point in the defect location information as (880, 430) as an example, the corresponding normalized coordinates are (0.46, 0.40).

[0044] Convert the bounding box information in the defect location information into defect center point location parameters.

[0045] Based on the correspondence between the pre-obtained inspection images and the target insulators, the number of the target insulator to which the defect belongs is determined.

[0046] It is understandable that the target insulator number can be a single insulator number or an insulator string number.

[0047] On the other hand, based on the detection confidence information in the original detection results, confidence labeling processing is performed to obtain confidence labeling information.

[0048] In practical applications, the detection confidence information in the original detection results can be compared with the preset confidence threshold. When the detection confidence information is lower than the preset confidence threshold, the corresponding defect can be manually reviewed.

[0049] In practical applications, the preset confidence threshold can be set to 0.7, but the specific value can be adjusted according to actual detection needs. Table 3 shows the confidence indicator information corresponding to different detection confidence levels.

[0050] Table 3. Confidence Identifiers for Different Detection Confidence Levels

[0051] Accordingly, key defect information, including standardized defect category information, defect severity level, standardized defect location information, and credibility identification information, is encapsulated to generate a standardized defect description parameter set.

[0052] In this embodiment, the standardized defect description parameter set includes at least the following fields: Standard defect category coding is used to characterize the engineering semantic type of defects.

[0053] The insulator identification parameters to which the defect belongs, including the insulator number or insulator string number, are used to characterize the engineering object to which the defect belongs.

[0054] Defect spatial distribution parameters are used to characterize the location characteristics of defects in the corresponding insulators or insulator strings.

[0055] The defect quantity parameter is used to characterize the quantity of defects in the same insulator or insulator string.

[0056] The defect severity level parameter characterizes the degree to which a defect affects the operational safety of the equipment. And, The defect reliability indicator parameter is used to characterize the reliability of defect detection results.

[0057] In a specific example, Table 4 shows the data information in the standardized defect description parameter set.

[0058] Table 4. Data information in the standardized defect description parameter set

[0059] Then, based on the standardized defect description parameter set, defect information statistics and engineering-level feature modeling are performed to obtain the engineering-level defect feature parameter set.

[0060] In a specific implementation, based on a standardized set of defect description parameters, defect information statistics and engineering-level feature modeling are performed to obtain a set of engineering-level defect feature parameters, which specifically includes: First, based on the insulator identification parameters of various defects in the standardized defect description parameter set, defect aggregation processing is performed to obtain multiple defect sets.

[0061] In the aggregation process, defects can be grouped by insulator number or insulator string number. Defects belonging to the same insulator or the same insulator string can be merged into the same defect set, while defect sets of different insulators or different insulator strings can be stored separately.

[0062] Through the above processing, a defect data structure can be established with insulators or insulator strings as the smallest engineering analysis unit.

[0063] On the one hand, for each defect set, based on the defect category information and defect quantity parameters in the standardized defect description parameter set, defect quantity and category distribution statistics are performed to obtain quantity feature parameters and category distribution feature parameters.

[0064] In practical applications, the total number of defects in the corresponding insulator or insulator string can be counted, the number of defects corresponding to each standard defect category can be counted, and the proportion of each defect category to the total number of defects can be calculated.

[0065] The above statistics can be used to generate quantitative characteristic parameters and category distribution characteristic parameters that reflect the scale and type of defects.

[0066] On the other hand, for the same set of defects, based on the standardized defect location information in the standardized defect description parameter set, the degree of spatial concentration of defects is determined in an engineering manner, and defect spatial concentration feature parameters are generated.

[0067] In practical applications, the spatial distance between any two defects in the defect set can be calculated first. This spatial distance can be calculated based on normalized coordinates or defect center point location parameters. Then, this spatial distance can be compared with a preset judgment threshold for the spatial set. This judgment threshold is used to characterize the maximum spatial distance that can be regarded as the same area on the insulator or insulator string structure. Next, the number of defect pairs with a spatial distance less than the judgment threshold or the number of defects participating in the spatial set distribution can be counted to obtain statistical results. Subsequently, the statistical results can be compared with a preset quantity threshold or proportion threshold.

[0068] Specifically, when the statistical results meet a preset threshold condition, the defect set is determined to be spatially concentrated; when the statistical results do not meet the preset threshold condition, it is determined to be spatially discrete. Based on the above determination results, defect spatial concentration characteristic parameters can be generated.

[0069] On the other hand, based on the defect severity level, defect quantity parameters, and defect spatial concentration characteristic parameters in the standardized defect description parameter set, a comprehensive operational risk level model is performed for each target insulator to generate comprehensive risk level parameters.

[0070] In this embodiment, the process of performing comprehensive operational risk level modeling specifically includes: First, the highest defect severity level can be selected from the defect set as the basic risk level for the insulator or insulator string.

[0071] Then, the total number of defects is compared with a preset threshold. When the total number of defects is less than the first threshold, the basic risk level remains unchanged; when the total number of defects is greater than or equal to the first threshold, the basic risk level is increased by one level.

[0072] Subsequently, rule-based corrections can be made based on spatial concentration. When the defect spatial concentration characteristic parameter indicates that the defects are concentrated, the basic risk level is increased by one level; when the defect spatial concentration characteristic parameter indicates that the defects are discrete, the basic risk level is not increased.

[0073] Ultimately, a comprehensive risk level parameter can be generated.

[0074] On the other hand, for each defect set, based on the credibility identification information in the standardized defect description parameter set, credibility status is summarized to obtain credibility status marking results.

[0075] In this embodiment, during the process of summarizing the credibility status, it can first be determined whether there are defects marked as requiring manual review in the defect set; when there is at least one manually reviewed defect, the credibility status of the corresponding insulator or insulator string is marked as requiring review; when there is no manually reviewed defect, the credibility status is marked as reliable.

[0076] Based on this, the quantitative characteristic parameters, category distribution characteristic parameters, defect spatial concentration characteristic parameters, comprehensive risk level parameters, and credibility status marking results are encapsulated to generate an engineering-level defect characteristic parameter set.

[0077] In this embodiment, the set of engineering-level defect characteristic parameters includes at least: insulator or insulator string identifier, total number of defects, number of each defect category, defect category distribution parameters, defect spatial concentration characteristic parameters, comprehensive risk level parameters, and credibility status marking results.

[0078] In one embodiment, based on a set of engineering-level defect feature parameters, a set of structured semantic description units is constructed through structured semantic modeling and rule-based generation, specifically including: The first step is to construct the initial description unit based on the pre-defined data structure.

[0079] Specifically, the data structure of the initial description unit should contain at least the following fields: ObjectID represents the identifier parameter for an insulator or insulator string; DefectSummary represents the defect summary field, such as category distribution and total number; RiskLevel represents the overall risk level parameter; ConfidenceState represents the defect confidence state parameter, i.e., the confidence state labeling result; SpatialPattern represents the spatial concentration characteristic parameter; PhenomenonSlots represents the slots for describing phenomena. ImpactSlots represents the impact analysis slots; SuggestionSlots indicates the slots for suggested solutions. EvidenceRefs represents evidence references, such as image numbers, timestamps, and defect locations.

[0080] The second step is to perform rule engine-style slot filling for each defect category entry in the initial description unit, and to perform data mapping from defect categories to technical terms and description templates to obtain semantic description units.

[0081] Specifically, for each defect category entry in the initial description unit, a rule engine-style slot filling can be performed. In this embodiment, the rule engine uses if-then conditional triggering, and each rule needs to explicitly define its input, condition, and output.

[0082] The spatial distribution slot filling rule can be set as follows: if SpatialPattern is centralized, then the output SpatialPatternText is centralized; if SpatialPattern is discrete, then the output SpatialPatternText is discrete.

[0083] The mapping rule from risk level to action level can be set as follows: if the RiskLevel is high, the action level is immediate inspection or replacement; if the RiskLevel is medium, the action level is to arrange a re-inspection and formulate a treatment plan as soon as possible; if the RiskLevel is low, the action level is to continuously observe and include it in periodic maintenance.

[0084] The mapping rule for confidence status to review prompts can be set as follows: if ConfidenceState is "requires review", the load prompt message is "the confidence of the test result is low, and manual review and confirmation are recommended"; if ConfidenceState is "confident", the review prompt message is "the confidence of the test result is high".

[0085] In the mapping process from defect categories to technical terms and description templates, a defect terminology and template library can be established. The terminology mapping table contains keys representing standard defect category codes and values ​​covering term names, synonyms, and typical phenomenon phrases. The template mapping table contains keys representing standard defect category codes and values ​​covering phenomenon templates, impact templates, and suggested templates.

[0086] For example, the phenomenon template could state that IS-01 detected two concentrated defects; the impact template could state that such defects may lead to a decrease in insulation performance and pose a high risk; and the recommendation template could state that immediate repair or replacement is recommended, as the test results have low reliability and manual verification is advised.

[0087] In the application, for each insulator identifier, the number parameters of each defect category can be traversed. For each type of defect, the difference category can be obtained from the terminology mapping table, and three types of templates can be obtained from the template mapping table. Then, the entries formed by combining the templates and slots are written into the Semantic Description Unit (SDU), but the final sentence is not formed yet.

[0088] The third step is to merge multiple defect entries and remove redundancy from the semantic description units to obtain a set of structured semantic description units.

[0089] In this embodiment, when multiple types of defects exist for the same insulator identifier, a general summary and specific descriptions are required to avoid repetitive and lengthy reports. In practical applications, a priority-based merging algorithm can be used. First, a defect priority table is defined, for example, the priorities of damage, flashover marks, hardware corrosion, and dirt decrease in that order.

[0090] Next, a merging algorithm can be executed. In this process, all defect category entries can be retrieved first and sorted in descending order of defect priority. Then, a summary sentence slot can be generated, specifically taking the total number of defects, the main defect categories, and the risk level to form a summary template. Next, a detailed description slot can be generated for each of the sorted entries, and a redundancy removal rule can be executed. If two entries have the same risk level and concern level, it is recommended that the paragraph be retained only once, and subsequent entries should only retain the phenomenon description. Finally, a set of structured semantic description units that have been merged and deredundant can be output.

[0091] In this embodiment, the structured semantic description unit set specifically includes: insulator or insulator string identification parameters, general description slots, phenomenon description slots, impact analysis slots, handling suggestion slots, risk level, defect credibility status parameters, evidence citations, and other information.

[0092] In a specific instance, Table 5 shows the specific information contained in the set of structured semantic description units.

[0093] Table 5. Detailed information about the set of structured semantic description units.

[0094] In one embodiment, matching the corresponding inspection report template to the set of structured semantic description units specifically includes: First, a report structure template library is built, which contains multiple predefined report structure templates, each of which adopts a parameterized chapter structure.

[0095] In this embodiment, each report structure template is defined in the form of a parameterized chapter structure, specifically including: template identifier parameters, template applicable condition parameters, and template chapter set.

[0096] Each chapter in the report structure template must define at least the following attribute information: chapter identifier, chapter type (such as overview, phenomenon description, impact analysis, handling suggestions, credibility statement, etc.), chapter sorting parameters, and chapter content source rules. In practical applications, the report structure template library can be stored in the system as a configuration file or a data table.

[0097] Then, the risk level parameter of each semantic description unit is extracted from the set of structured semantic description units, and the overall risk level of the report is determined based on the risk level parameters of all semantic description units.

[0098] In practical applications, based on the risk level parameters within the set of structured semantic description units generated for the current inspection task, report structure template matching can be performed. The specific process is as follows: The first step is to extract the risk level parameters of all semantic description units from the set of structured semantic description units.

[0099] The second step is to determine the overall risk level of the report based on preset rules, including: If any semantic description unit has a risk level parameter of "high", then the overall risk level of the report is determined to be "high"; if any semantic description unit has a risk level parameter of "medium", then the overall risk level of the report is determined to be "medium"; otherwise, the overall risk level of the report is determined to be "low".

[0100] Finally, based on the overall risk level of the report, the corresponding target report structure template is extracted from the report structure template library as the inspection report template.

[0101] Specifically, the overall risk level of the report can be matched with the applicable parameters of the templates in the report structure template library, and the report structure template that matches the overall risk level of the report can be selected as the structure template for the current inspection report. This rule ensures that high-risk inspection tasks automatically adopt a report structure template that includes sections for key analysis and urgent recommendations.

[0102] In one embodiment, based on a set of structured semantic description units, the inspection report template is mapped at the chapter level to generate report skeleton data, specifically including: The first step is to determine the chapter organization method based on the number of insulator identifiers in the structured semantic description unit set.

[0103] In this embodiment, the process of determining the chapter organization method specifically includes: When the set of structured semantic description units contains only one insulator identifier, a single-object chapter organization method can be adopted to specifically describe all phenomena, impacts and suggestions in one go.

[0104] When the set of structured semantic description units contains multiple insulator identifiers, a multi-object chapter organization method is adopted, specifically, each insulator identifier independently generates a corresponding sub-chapter; the sub-chapter is arranged in descending order of insulator identifier order or risk level.

[0105] Understandably, the above-mentioned chapter organization method ensures that the report structure maintains a clear hierarchy even when the number of objects changes.

[0106] The second step is to map the semantic description content in the structured semantic description unit set to each chapter in the inspection report template according to the chapter organization method, so as to obtain the preliminary skeleton data.

[0107] In this embodiment, for chapters of the general overview type, the general overview sentence slot of each semantic description unit is read and merged or arranged sequentially according to the chapter organization rules; for chapters of the phenomenon description type, the phenomenon description slot list in the semantic description unit is read, arranged in order of defect priority, and then filled into the chapter; for chapters of the impact analysis type, the impact analysis slot in the semantic description unit is read, and when the chapter is in multi-object mode, each insulator identifier is divided into a separate segment; for chapters of the handling suggestion type, the handling suggestion slot in the semantic description unit is read, sorted in order of comprehensive risk level from high to low, and then filled into the chapter; for chapters of the credibility description type, the defect credibility status parameter in the semantic description unit is read; when there is a status that needs to be reviewed, a corresponding description entry is generated.

[0108] In a specific example, the preliminary skeleton data after chapter mapping can be seen in Table 6.

[0109] Table 6 Preliminary Skeleton Data

[0110] The third step is to resolve and merge the chapter content in the preliminary skeleton data to obtain the report skeleton data.

[0111] In this embodiment, in scenarios with multiple semantic description units or multiple defects, content duplication or logical conflicts may occur. Therefore, the following specific rules need to be executed: On the one hand, when multiple semantic description units have the same processing suggestion text content and the same level of handling action, only one suggestion description is retained.

[0112] On the other hand, when different semantic description units have different levels of action, suggestions with higher action levels are retained first.

[0113] On the other hand, items that differ only in quantity in the description of a phenomenon are merged into one item and described using a range description method.

[0114] In practical applications, by encapsulating the above matching, organization, and mapping results, report skeleton data can be generated. The report skeleton data includes at least: report template identifier parameters, chapter order parameters, a list of semantic content corresponding to each chapter, and the correspondence between chapters and insulator identifiers.

[0115] In one embodiment, based on the report skeleton data, the report text content is generated through chapter-level rendering and redundancy merging to obtain an insulator defect report, specifically including: First, extract the chapter order parameters from the report skeleton data, and then generate the corresponding chapter texts in sequence according to the chapter order parameters.

[0116] In this embodiment, the report skeleton data is used as input, the chapter order parameters are read, and the corresponding chapter text is generated sequentially according to the chapter order parameters. The specific steps include: The first step is to sort the chapter set according to the chapter order parameter.

[0117] The second step is to perform text generation processing on each of the sorted chapters.

[0118] The third step is to cache the already generated chapter text before generating the next chapter text for overall format control.

[0119] By using the methods described above, it can be ensured that the final generated report text is structurally consistent with the selected report template.

[0120] Then, template rendering is performed on the semantic content entries contained in each chapter text to generate candidate text segments.

[0121] In this embodiment, template rendering processing is performed on the semantic content entries contained in each chapter, specifically including: The first step is to read the corresponding semantic template from the chapter content entries.

[0122] The second step is to read the parameter values ​​corresponding to the template slots from the semantic description unit.

[0123] The third step is to replace the parameter values ​​in the semantic template according to the slot identifier to generate candidate text segments.

[0124] Next, multiple candidate text segments in the same chapter are processed by sequential organization and redundancy merging to obtain the chapter content.

[0125] When there are multiple candidate text segments within the same chapter, they can be organized according to the following specific rules: The first step is to sort the candidate segments according to the insulator identification or risk level.

[0126] The second step is to merge semantically repetitive and parameter-consistent content in consecutive sentence segments.

[0127] The third step is to merge sentences when the differences lie only in quantity or degree parameters, using parallel or range-based expressions.

[0128] Finally, all chapter contents are pieced together in order to generate the report text, resulting in the insulator defect report.

[0129] In this embodiment, by sequentially splicing together the formatted text of each section, a complete report text can be generated, thereby obtaining an insulator defect report.

[0130] In one embodiment, before sequentially concatenating all chapter contents, the following may be included: The consistency of defect terminology and risk level descriptions for the same insulator identifier across different sections must be verified; and / or, Determine whether the action level in the handling recommendations section is lower than the risk level reflected in the risk analysis section. If the determination result is yes, adjust the action level in the handling recommendations section to be consistent with the risk level.

[0131] In practical applications, after the initial text generation of all chapters is completed, cross-chapter consistency checks and conflict resolution can be performed, specifically including: On the one hand, it verifies whether the defect terminology and risk level descriptions used for the same insulator label are consistent across different chapters.

[0132] On the other hand, when it is found that the level of the action to be taken in the handling recommendations section is lower than the level of risk reflected in the risk analysis section, the handling recommendations can be automatically adjusted to be consistent with the higher risk level.

[0133] It is evident that the automatic insulator defect report generation method provided in this embodiment of the invention, by completing category mapping, location standardization, and severity determination based on defect category in the early stages, and by engineering-modeling the number of defects, category distribution, spatial concentration, and comprehensive risk level, can stably output closed-loop suggestions of risk level—action level—review prompt. Higher risk triggers higher-level actions, while low confidence only triggers a review indicator without confusing risk assessment. This mechanism avoids logical errors such as mismatch between suggestions and risks or substituting confidence level for severity. Simultaneously, the report structure is configurable and transferable through a template library and chapter mapping rules; different units / tasks only need to adjust the templates and rules to adapt to new formats. Furthermore, chapter merging and cross-chapter conflict resolution reduce redundancy and contradictions in multi-defect scenarios, ensuring consistent report expression.

[0134] In summary, this invention achieves automated, standardized, interpretable, and traceable report generation without increasing the complexity of the detection model, significantly reducing labor costs and subjective differences, and improving the quality of inspection reports and the ability to support operation and maintenance decisions.

[0135] Based on the same general inventive concept, this invention also protects an electronic device. The electronic device provided by this invention will be described below. The electronic device described below and the automatic generation method for insulator defect reports described above can be referred to and correspond to each other.

[0136] like Figure 2 As shown, the electronic device may include a processor 210, a communications interface 220, a memory 230, and a communication bus 240, wherein the processor 210, the communications interface 220, and the memory 230 communicate with each other via the communication bus 240. The processor 210 can call logical instructions in the memory 230 to execute the automatic insulator defect report generation method provided in the above embodiments.

[0137] Furthermore, the logical instructions in the aforementioned memory 230 can be implemented as software functional units and, when sold or used as independent products, can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0138] On the other hand, the present invention also provides a computer program product, which includes a computer program that can be stored on a non-transitory computer-readable storage medium. When the computer program is executed by a processor, the computer is able to execute the automatic generation method for insulator defect reports provided in the above embodiments.

[0139] In another aspect, the present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the automatic generation method for insulator defect reports provided in the above embodiments.

[0140] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.

[0141] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.

[0142] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for automatically generating insulator defect reports, characterized in that, include: The original detection results of insulator defects are obtained, and the original detection results are preprocessed in multiple levels to obtain a set of engineering-level defect feature parameters. Based on the set of engineering-level defect feature parameters, a set of structured semantic description units is constructed through structured semantic modeling and rule-based generation; Match the corresponding inspection report template to the set of structured semantic description units, and perform chapter-level mapping on the inspection report template based on the set of structured semantic description units to generate report skeleton data; Based on the report skeleton data, the report text content is generated through chapter-level rendering and redundancy merging to obtain the insulator defect report.

2. The method for automatically generating insulator defect reports according to claim 1, characterized in that, The original detection results are preprocessed at multiple levels to obtain defect feature data, including: The original detection results are standardized to obtain a set of standardized defect description parameters; Based on the standardized defect description parameter set, defect information statistics and engineering-level feature modeling are performed to obtain an engineering-level defect feature parameter set.

3. The method for automatically generating insulator defect reports according to claim 2, characterized in that, The original detection results are standardized to obtain a set of standardized defect description parameters, including: For the defect category identifiers in the original detection results, perform defect category mapping processing to obtain standardized defect category information; Based on the standardized defect category information, the severity of the defect is determined, and the severity level of the defect is identified. The defect location information in the original detection results is subjected to location parameter standardization processing to obtain standardized defect location information; Based on the detection confidence information in the original detection results, confidence identification processing is performed to obtain confidence identification information; The key defect information, including the standardized defect category information, defect severity level, standardized defect location information, and credibility identification information, is encapsulated to generate a standardized defect description parameter set.

4. The method for automatically generating insulator defect reports according to claim 3, characterized in that, For the defect category identifiers in the original detection results, defect category mapping processing is performed to obtain standardized defect category information, including: Construct a defect category mapping table, wherein the defect category mapping table is used to represent the correspondence between each defect type and a unique category code; Based on the defect category mapping table, defect categories with the same defect category identifier but different naming methods in the original detection results are uniformly mapped to their corresponding unique category codes to obtain standardized defect category information.

5. The method for automatically generating insulator defect reports according to claim 2, characterized in that, Based on the standardized defect description parameter set, defect information statistics and engineering-level feature modeling are performed to obtain an engineering-level defect feature parameter set, including: Based on the insulator identification parameters of various defects in the standardized defect description parameter set, defect aggregation processing is performed to obtain multiple defect sets; For each defect set, based on the defect category information and defect quantity parameters in the standardized defect description parameter set, defect quantity and category distribution statistics are performed to obtain quantity feature parameters and category distribution feature parameters; For the same set of defects, based on the standardized defect location information in the standardized defect description parameter set, the degree of spatial concentration of defects is determined in an engineering manner, and defect spatial concentration feature parameters are generated. Based on the defect severity level, defect quantity parameter and defect spatial concentration characteristic parameter in the standardized defect description parameter set, a comprehensive operational risk level model is performed for each target insulator to generate a comprehensive risk level parameter. For each defect set, based on the credibility identifier information in the standardized defect description parameter set, credibility status aggregation is performed to obtain credibility status marking results; The quantitative characteristic parameters, category distribution characteristic parameters, defect spatial concentration characteristic parameters, comprehensive risk level parameters, and credibility status marking results are encapsulated to generate an engineering-level defect characteristic parameter set.

6. The method for automatically generating insulator defect reports according to claim 1, characterized in that, Based on the aforementioned set of engineering-level defect feature parameters, a set of structured semantic description units is constructed through structured semantic modeling and rule-based generation, including: Based on the pre-defined data structure, construct the initial description unit; For each defect category entry in the initial description unit, a rule engine-style slot filling is performed, and a data mapping from defect categories to technical terms and description templates is performed to obtain a semantic description unit; The semantic description units are subjected to multi-defect entry merging and redundancy removal processing to obtain a set of structured semantic description units.

7. The method for automatically generating insulator defect reports according to claim 1, characterized in that, Matching the corresponding inspection report template to the set of structured semantic description units includes: Construct a report structure template library, wherein the report structure template library contains multiple predefined report structure templates, and each report structure template adopts a parameterized chapter structure; The risk level parameter of each semantic description unit is extracted from the set of structured semantic description units, and the overall risk level of the report is determined based on the risk level parameters of all semantic description units. Based on the overall risk level of the report, the corresponding target report structure template is extracted from the report structure template library as the inspection report template.

8. The method for automatically generating insulator defect reports according to claim 1, characterized in that, Based on the set of structured semantic description units, the inspection report template is mapped at the chapter level to generate report skeleton data, including: The chapter organization method is determined based on the number of insulator identifiers in the set of structured semantic description units; According to the chapter organization method, the semantic description content in the set of structured semantic description units is mapped to each chapter in the inspection report template to obtain preliminary skeleton data; The chapter content in the preliminary skeleton data is conflict-resolved and merged to obtain the report skeleton data.

9. The method for automatically generating insulator defect reports according to any one of claims 1 to 8, characterized in that, Based on the aforementioned report skeleton data, the report text content is generated through chapter-level rendering and redundancy merging, resulting in an insulator defect report, including: Extract the chapter order parameters from the report skeleton data, and generate the corresponding chapter texts sequentially according to the chapter order parameters; Template rendering is performed on the semantic content entries contained in each chapter text to generate candidate text segments; Multiple candidate text segments within the same chapter are processed by sequential organization and redundancy merging to obtain the chapter content. By piecing together the contents of all chapters in order, a report text is generated, resulting in an insulator defect report.

10. The method for automatically generating insulator defect reports according to claim 9, characterized in that, Before piecing together all the chapter contents in order, it also includes: The consistency of defect terminology and risk level descriptions for the same insulator identifier across different sections must be verified; and / or, Determine whether the action level in the handling recommendations section is lower than the risk level reflected in the risk analysis section. If the determination result is yes, adjust the action level in the handling recommendations section to be consistent with the risk level.

Images