A map layer updating method and system based on unmanned aerial vehicle orthographic images

Abstract

The application provides a map layer updating method and system based on unmanned aerial vehicle orthographic images, and belongs to the technical field of image processing, and specifically comprises the following steps: determining the identification result of the priority updating area based on the identification strategy, determining the early warning processing strategy based on the historical deviation situation of the orthographic images and the map layer of the power grid map in the preferred updating area and the similar situation of the electrical equipment in the preferred updating area and other updating target areas, determining the updating identification deviation data of the map layer based on the orthographic images based on the early warning processing strategy, and determining the map layer updating method of the updating target area except the priority updating area based on the updating identification deviation data, the historical deviation situation of the deviation position of the updating target area and the updating risk type, so that the accuracy of the layer updating processing is improved.

Description

Technical Field

[0001] This invention belongs to the field of image processing technology, and in particular relates to a method and system for updating map layers based on UAV orthophotos. Background Technology

[0002] Traditional maps (especially large-scale thematic maps such as "Power Grid Map", "Land Planning Map", and "Smart City Base Map") have long update cycles, high costs, and low efficiency, making it difficult to reflect rapid changes in the Earth's surface. UAV orthophotos, on the other hand, are characterized by high resolution, high timeliness, high flexibility, and low cost, providing a revolutionary technical means for high-frequency and accurate updates of map layers.

[0003] Specifically, the invention patent application CN202311352018.3, "A Method for Correcting Power Grid GIS Graphics Based on UAV Orthophotos," uses orthophotos and tower attributes collected during UAV inspections to correct the graphics and topology information of power grid equipment in the GIS system, thereby solving the problems of insufficient accuracy and low maintenance efficiency in traditional data maintenance models. However, the above technical solution has the following technical problems: When updating map layers, the differences in power equipment and external environment in different areas lead to variations in the number of locations requiring manual verification due to low similarity. Consequently, the reliability of update verification varies across different locations. Therefore, determining the update strategy for map layers in different areas based on the reliability of update verification has become an urgent technical problem to be solved.

[0004] To address the aforementioned technical problems, this application provides a method and system for updating map layers based on UAV orthophotos. Summary of the Invention

[0005] To achieve the objectives of this invention, the following technical solution is adopted: Specifically, this application provides a map layer update method based on UAV orthophotos, which includes: S1 uses historical update data of the power grid map as a basis to determine the similarity data between the orthophoto of the UAV in the update target area and the map layer of the power grid map. Using the similarity data and the location distribution data of the similarity degree not meeting the requirements, the update risk type of the update target area is determined. Based on the different update risk types of the update target areas and the distribution data of electrical equipment in the update target areas, the identification strategy of the priority update area in the update target area is determined. S2 determines the identification result of the priority update area based on the identification strategy, and determines the early warning processing strategy based on the historical deviation between the orthophoto of the priority update area and the map layer of the power grid map, and in combination with the similarity of electrical equipment in the priority update area and other update target areas. Based on the aforementioned early warning processing strategy, S3 determines the update identification deviation data of the map layer based on orthophotos. Based on the update identification deviation data, and combined with the historical deviation of the deviation location of the update target area and the update risk type, S3 determines the map layer update method for the update target area excluding the priority update area.

[0006] The beneficial effects of this invention are as follows: By utilizing similar data and location distribution data where the similarity does not meet the requirements, the update risk type of the target area is determined. This fully considers that if the target area has frequently triggered manual verification in similar historical updates, then the probability of that area triggering manual verification again in the current update is very high. The intervention of manual verification can directly discover and correct automated errors, thereby significantly reducing the error risk of the final layer. The targeted determination of the update risk type lays the foundation for further determination of the map layer update processing strategy.

[0007] Based on the updated identification deviation data, the historical deviation of the target area, and the type of update risk, this embodiment determines the map layer update method for the target area excluding priority update areas. It constructs an intelligent decision-making system based on a "dual-track analysis of facts and probability." Its ingenuity lies in employing differentiated data analysis strategies for the updated and unupdated areas: for the former, purely objective factual statistics are performed to determine the update deviation of the model's orthophoto; for the latter, in-depth probabilistic pattern deconstruction is conducted, achieving a comprehensive consideration of the current mismatch risk. This ensures the objectivity of the decision-making process while also endowing the system with the ability to detect hidden risks.

[0008] Furthermore, the historical update data of the power grid map is based on the historical update process of map layers in different areas of the power grid map and the similarity data between the orthophotos of UAVs and map layers of the power grid map in different historical update processes.

[0009] Furthermore, the target area for updating is determined based on the user's selection.

[0010] Furthermore, the similarity data between the orthophoto of the UAV and the map layer of the power grid map is determined based on the image similarity coefficient between the orthophoto and the map layer at different locations in the updated target area during different historical update processes.

[0011] Furthermore, the method for determining the update risk type of the target area is as follows: Based on the aforementioned similar data, determine the image similarity coefficients between orthophotos and map layers at different locations within the updated target area during the historical update process; Based on the image similarity coefficient, determine the historical update process of image changes at the location; Based on the historical update process data of image changes at different locations within the target update area, the update risk type of the target update area is determined.

[0012] Furthermore, the method for determining the map layer update method for the target update area excluding the priority update area is as follows: Based on the updated identification deviation data, the identification deviation position in the priority update region is determined; The target area for updating, excluding the priority update area, is taken as the update area. The probability of change of the deviation position is determined based on the historical deviation of the deviation position in the update area. Based on the identification deviation location, the probability of change of the deviation location, and the update risk type of the update target area that has not yet been updated, the map layer update method for the update area is determined.

[0013] In a second aspect, the present invention provides a computer system comprising: a memory and a processor connected in communication, and a computer program stored in the memory and capable of running on the processor, wherein the processor executes the above-described method for updating map layers based on UAV orthophotos when running the computer program.

[0014] Other features and advantages will be set forth in the description which follows, and will be apparent in part from the description, or may be learned by practicing the invention. The objects and other advantages of the invention are realized and obtained through the structures particularly pointed out in the description and the drawings.

[0015] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, preferred embodiments are described below in detail with reference to the accompanying drawings. Attached Figure Description

[0016] The above and other features and advantages of the present invention will become more apparent from a detailed description of exemplary embodiments thereof with reference to the accompanying drawings.

[0017] Figure 1 This is a flowchart of a map layer update method based on UAV orthophotos; Figure 2 This is a flowchart illustrating the method for determining the update risk type of the target area. Figure 3 This is a flowchart illustrating the method for determining the identification strategy of the priority update region in the target region. Detailed Implementation

[0018] Exemplary embodiments will now be described more fully with reference to the accompanying drawings. However, these exemplary embodiments can be implemented in many forms and should not be construed as limited to the embodiments set forth herein; rather, they are provided so that the invention will be thorough and complete, and the concept of the exemplary embodiments will be fully conveyed to those skilled in the art. The same reference numerals in the drawings denote the same or similar structures, and therefore their detailed description will be omitted.

[0019] The terms “a,” “one,” “the,” and “the” are used to indicate the existence of one or more elements / components / etc.; the terms “including” and “having” are used to indicate an open-ended meaning of inclusion and that other elements / components / etc. may exist in addition to the listed elements / components / etc.

[0020] Example 1 To solve the above problems, according to one aspect of the present invention, such as Figure 1 As shown, a method for updating map layers based on UAV orthophotos is provided, specifically including: S1 uses historical update data of the power grid map as a basis to determine the similarity data between the orthophoto of the UAV in the update target area and the map layer of the power grid map. Using the similarity data and the location distribution data of the similarity degree not meeting the requirements, the update risk type of the update target area is determined. Based on the different update risk types of the update target areas and the distribution data of electrical equipment in the update target areas, the identification strategy of the priority update area in the update target area is determined. S2 determines the identification result of the priority update area based on the identification strategy, and determines the early warning processing strategy based on the historical deviation between the orthophoto and the power grid map layer in the priority update area and the data of the priority update area. Based on the aforementioned early warning processing strategy, S3 determines the update identification deviation data of the map layer based on orthophotos. Based on the update identification deviation data, and combined with the historical deviation of the deviation location of the update target area and the update risk type, S3 determines the map layer update method for the update target area excluding the priority update area.

[0021] Furthermore, the historical update data of the power grid map is based on the historical update process of map layers in different areas of the power grid map and the similarity data between the orthophotos of UAVs and map layers of the power grid map in different historical update processes.

[0022] Furthermore, the target area for updating is determined based on the user's selection.

[0023] Furthermore, the similarity data between the orthophoto of the UAV and the map layer of the power grid map is determined based on the image similarity coefficient between the orthophoto and the map layer at different locations in the updated target area during different historical update processes.

[0024] It should be noted that the image similarity coefficient is determined based on the similarity of image features between the orthophoto and the map layer image, and its value ranges from 0 to 1.

[0025] Specifically, such as Figure 2 As shown, the method for determining the update risk type of the target area is as follows: The core decision-making objective of this embodiment is to establish a method for predicting the probability of manual verification in the current update task based on historical data, and to classify risk levels according to this probability. The core logic of this method is: for the upcoming layer update task, if the target area has frequently triggered manual verification in similar historical updates, then the probability of that area triggering manual verification again in the current update is high; and the intervention of manual verification will directly detect and correct automated errors, thereby significantly reducing the error risk of the final layer. Therefore, the higher the historical frequency of triggering verification, the higher the expected probability of the area undergoing verification in the current update, and the lower the risk level of its final layer data. This is a risk assessment method based on probability prediction and the effectiveness of quality control mechanisms.

[0026] This method constructs a predictive framework for deriving the current risk level from historical verification trigger frequencies: Verification Trigger History Analysis: By analyzing historical records of manual verification triggered due to low image similarity coefficients after updates in various regions, a "verification trigger frequency" archive for each region is established.

[0027] Current verification probability prediction: Assuming that historical patterns are continuous, the frequency of historical verification triggers is used as the basis for predicting the probability that the region needs and will accept manual verification in the current update.

[0028] Risk transmission mechanism establishment: Establish a quantitative relationship of "verification probability → risk reduction" - the higher the verification probability, the greater the possibility that automated errors will be detected and corrected in a timely manner, and the lower the error risk of the final layer data.

[0029] Risk level mapping: Based on the predicted verification probability (and the resulting risk reduction effect), risk levels are divided, forming an assessment conclusion that the more frequently a region is historically triggered for verification, the lower its current updated risk level.

[0030] S11 Based on the similarity data, determine the image similarity coefficients between orthophotos and map layers at different locations in the target area during the historical update process; Determining the similarity coefficient of historical images—the original basis for verifying the trigger judgment: The image similarity coefficient is an automated processing result calculated by the system after each historical update and before manual verification begins, and is a consistency score between the system and the original orthophoto at each point. The value ranges from 0 to 1.

[0031] The similarity coefficient is a direct input to the objective threshold criterion for determining whether manual verification needs to be initiated. When the similarity coefficient of a certain location falls below the preset threshold, the workflow will trigger a manual verification task for that location. Therefore, the historical similarity coefficient sequence essentially records the decision-making basis data for "under what circumstances verification was triggered" in the past.

[0032] This step provides a quantified historical record, allowing us to backtrack and analyze whether the "verification trigger conditions" for each location and each update were met, thus establishing a data foundation for statistical verification trigger frequency.

[0033] For example, after three historical updates, the similarity coefficients calculated by the system for a certain key node are [0.94, 0.68, 0.91]. If the threshold is set at 0.75, then a similarity coefficient of 0.68 in the second update will trigger manual verification of that node.

[0034] S12 determines the historical update process of image changes at the location based on the image similarity coefficient; Identify the image change process at the location—verify the markers of the triggering events: The "historical update process of image changes" is a rule-based determination: if the image similarity coefficient of a certain location after a certain historical update is less than a preset "similarity coefficient threshold" (such as 0.75), then it is determined that the location has "changed" in that update.

[0035] The "Change" event is a synonym for "Triggering Manual Verification". This decision rule is set to transform consecutive similarity coefficient values ​​into a binary sequence of "Whether Verification is Triggered" events. This definition is directly related to the data quality status and the initiation of quality control actions (verification).

[0036] Significance: By using event-based tagging, a historical "verification trigger record" is generated for each location. This record forms the basis for calculating the "verification trigger frequency" (i.e., the "probability of change" in the current step S133), which will be directly used to predict the probability of triggering verification in the current update.

[0037] Specific example: Continuing from the previous example, the sequence of change events for this node is: [Not triggered, Triggered, Not triggered]. This indicates that manual verification was triggered once out of three historical updates.

[0038] S13 determines the update risk type of the update target area based on the historical update process data of image changes at different locations in the update target area.

[0039] Specifically, the historical update process of the image at the location where changes occur is the historical update process of the image similarity coefficient at the location being less than a preset similarity coefficient threshold.

[0040] Specifically, based on historical update process data showing image changes in the target area, the update risk type of the target area is determined, including: S131 uses the historical update process data of image changes in the updated target area to determine the number of positions in the historical update process that belong to image changes. The proportion of the number of positions in the historical update process that belong to image changes in the updated target area is used to determine the proportion of image change positions in the historical update process. It is determined whether there are any historical update processes where the proportion of image change positions is greater than a preset proportion threshold. If yes, proceed to step S132. If no, since the number of positions with image changes is not large, and generally only image changes will trigger manual verification, the reliability of manual verification in the model update process is not high. Therefore, the update risk type of the updated target area is determined to be a type 1 risk. Identify large-scale verification trigger scenarios: "Image change location ratio" refers to the proportion of locations marked as "changed" (i.e., triggering verification) out of the total number of monitored locations within the target area during a certain historical update. The "preset ratio threshold" (e.g., 60%) is used to identify historical update scenarios that trigger large-scale, systematic manual verification.

[0041] Analyzing large-scale verification trigger scenarios has two implications. First, these scenarios indicate that under certain specific conditions (such as extremely poor image quality), a full-scale verification will almost inevitably be triggered in the region. Second, and more importantly, as described later in S132, we consider these scenarios as data quality issues rather than valid "controllable verification" samples, and therefore exclude them from the subsequent analysis of routine verification trigger patterns in the region to ensure the accuracy of predictions.

[0042] Significance: This step identifies special cases where the verification mechanism is "fully activated" due to extreme conditions, preventing these abnormal data from interfering with the assessment of the region's verification trigger probability under normal operating conditions.

[0043] Specific logic and risk assessment: If the judgment is "No" (no large-scale verification triggers have occurred in all historical updates), then proceed to S133 to analyze the normal verification trigger mode. At this time, a risk assessment is not directly applied, because even without large-scale triggers, there may still be sporadic but high-frequency verification trigger points, which will also affect the risk of the current update.

[0044] Specific example: There are 100 points in the area. The number of points that triggered verification in four historical updates were 5, 8, 3, and 70, respectively. Assuming a threshold of 60%, the fourth update (70 points) is a large-scale verification trigger scenario and will be handled specially in subsequent steps.

[0045] S132 takes the historical update process in which the proportion of image change position is greater than the preset proportion threshold as a reliable identification process, and determines whether the number of reliable identification processes in the updated target area is greater than the preset reliable process number threshold. If yes, the update risk type of the updated target area is determined to be a three-type risk type. If no, proceed to step S133. Determine the effective historical sample size for probabilistic prediction: "Reliable identification process" refers to the historical update process that is determined in S131 to have an "image change position ratio greater than a preset ratio threshold". These processes represent updates completed under normal operating conditions, and their data reflect the verification triggering pattern of the region under normal circumstances. "Preset reliable process quantity threshold" is the minimum sample size required for effective statistical prediction.

[0046] Specific logic and risk rating: If the judgment is "yes" (i.e., the number of reliable processes), then the area is directly determined to be a Class III risk type (low risk). Decision implications: At this time, the number of reliable identification processes is relatively large, therefore its risk level is relatively low.

[0047] Specific example: A certain area has only one reliable identification process, which is less than the preset threshold of three reliable processes, so it is determined to proceed to the next step.

[0048] S133 uses historical update process data of image changes at different locations to determine the proportion of historical update processes of image changes at the location in all historical update processes, and uses it as the change probability to determine whether there is a location with a change probability greater than a preset probability threshold. If yes, proceed to step S134; otherwise, determine the update risk type of the update target area as a type of risk. Predict the probability of verification triggering at a specific location: "Probability of Change" is a metric calculated for a single geographic location. It equals the proportion of times that location "changes" (triggers verification) throughout all "historical identification processes." It directly predicts the historical probability that this location will trigger manual verification in the current regular update.

[0049] This is the core prediction step of our method. We assume that historical experience has continuity: a location that frequently triggers verification in past routine updates is also likely to trigger verification again in similar current routine updates. A high trigger probability means that the location is more likely to be manually checked, and its potential errors are more likely to be discovered.

[0050] Identifying these high-probability trigger points is crucial for assessing the overall risk of a region. If a large number of such points exist within a region, it means that many locations will likely be verified in the current update, thus significantly reducing the overall risk of layer errors.

[0051] Specific logic and risk rating: If the judgment is "no" (i.e., there are no points with a high probability of change), then the area is determined to be a Class III risk type (high risk). Decision implications: There are no "hot spots" in the area that have frequently triggered verification in the past. This means that in the current update, the probability of most locations triggering verification is low, and the overall expected probability of automated errors being discovered and corrected manually is also low. Therefore, the residual risk of the final layer data is high.

[0052] Specific example: Analysis shows that the probability of change for all points in the area is less than 0.3 (preset probability threshold), indicating that there are no points with a high probability of triggering verification. The expected verification coverage is low and the risk is high. Therefore, the update risk type of the target area is determined to be a type of risk.

[0053] S134 determines the verification reliability coefficient of the target area based on the probability of change at different locations and in combination with the proportion of image change locations in different historical update processes. Based on the verification reliability coefficient, the update risk type of the target area is determined.

[0054] Comprehensive assessment of expected verification coverage and risk level: Keyword Explanation: "Verification Reliability Coefficient" is a comprehensive evaluation index that combines the expected verification trigger probability (probability of change) of each point in the region with the verification trigger breadth (proportion of image change locations) of each regular update in history. It aims to quantify and predict "the overall probability or strength of quality assurance for the entire region through the manual verification mechanism in the current update".

[0055] Having only a few points with a high probability of triggering, or triggering a large number of points only in a few updates, is insufficient to guarantee a low overall risk. The verification reliability coefficient, by combining two aspects of information, predicts the overall reliability of the region's "coverage by the verification safety net." The higher the coefficient, the more comprehensive and reliable the predicted verification coverage, and the better the expected risk reduction effect.

[0056] Significance: This is the final risk quantification and adjudication step, which translates the predicted strength of verification coverage into a clear risk level.

[0057] Final risk rating: Verification reliability coefficient > preset reliability coefficient threshold: determined to be a low-risk category (Class III). Meaning: The predictive model indicates that there is a high probability that extensive and effective manual verification will be initiated in this area during the current update. High-probability verification intervention will significantly reduce the risk of automated error retention, thus resulting in a low expected risk level for the final layer data.

[0058] Verification reliability coefficient ≤ preset reliability coefficient threshold: classified as a Class II risk type (medium risk). Meaning: the predicted verification coverage is of moderate strength, providing some risk reduction protection, but not as sufficient as in low-risk areas.

[0059] It should be noted that the verification reliability coefficient is determined based on the average of the change probability of different locations and the proportion of image change locations in different historical update processes. For example, if the verification reliability coefficient in a certain area is greater than 0.5, then the update risk type of the target area is determined to be a type three risk.

[0060] It is understood that if the verification reliability coefficient is greater than the preset reliability coefficient threshold, the update risk type of the target area is determined to be a type three risk; if the verification reliability coefficient is not greater than the preset reliability coefficient threshold, the update risk type of the target area is determined to be a type two risk.

[0061] It should be noted that the first type of risk is greater than the second type of risk, and the second type of risk is greater than the third type of risk.

[0062] The core innovation of this embodiment lies in transforming risk assessment from a static description of the current situation into a dynamic process assessment based on probability prediction.

[0063] From "Historical Results" to "Future Probability": Traditional methods look at historical error rates, while this method looks at historical verification trigger frequencies and uses this to predict the current verification probability. This is more in line with the actual workflow of "verification as a proactive error correction mechanism".

[0064] A "probability-risk" transmission model was established: a logical chain of "high verification trigger probability → high error detection probability → low final error risk" was clearly constructed, so that the classification of risk levels has a clear and interpretable causal basis.

[0065] Supports differentiated quality trust decisions: Output results directly guide management in assigning different levels of trust to data in different regional layers. Data from low-risk areas can be used for business operations requiring higher confidence (such as automated planning), while data from high-risk areas may require additional verification steps before use.

[0066] Optimize quality control resource expectations: Although resources are not directly allocated, this method allows managers to anticipate which areas are more likely to consume verification resources in the current update, which helps to manage workload and time expectations.

[0067] Specifically, such as Figure 3 As shown, the method for determining the identification strategy of the priority update region in the target update region is as follows: The core decision-making objective of this embodiment is to formulate an intelligent "priority update area" identification strategy after completing the risk level assessment of the power grid layer update (obtaining risk areas of categories I, II, and III). This strategy aims to address how to scientifically select the most urgently needed and worthy-of-priority resource subset from all target areas to be updated, under constraints of limited update resources (such as computing power, time, and manual verification capabilities). Its decision-making logic integrates two major principles: "risk-driven" and "asset commonality-driven," focusing on both the urgency of high-risk areas and considering maximizing experience reuse and value spillover effects by updating typical equipment areas.

[0068] This method constructs a hierarchical, condition-triggered decision tree to dynamically determine the optimal priority region selection ratio and selection rules: Risk area identification: Based on the results of the previous risk level assessment, high-risk (Category I) areas are identified as the primary candidate pool for priority updates.

[0069] Equipment commonality analysis: Calculate the model similarity of electrical equipment in each region to identify regions where equipment types are widely representative. Updating these regions allows for better validation and optimization of the identification algorithm for common equipment.

[0070] Dynamic strategy decision-making: Based on key indicators such as the number of high-risk areas, the total risk weight, and the similarity of equipment between areas, the system dynamically selects the most appropriate priority update area selection ratio (high, medium, and low) and specific selection rules through multi-level condition judgment.

[0071] Priority region generation: Based on the final determined strategy and ratio, and combined with device similarity ranking, a specific list of priority update regions is determined from the candidate regions.

[0072] S21 determines an update target region of one risk type based on different update risk types of the update target regions; Identify key high-risk areas: The target area for updating a type of risk refers to all areas that were identified as "Type 1 Risk" (highest numerical value, highest risk) in the preceding risk assessment process (S11-S134). These areas typically have characteristics of extremely unstable historical update quality and poor reliability of automated processing.

[0073] Using a high-risk area as the starting point for strategy formulation reflects the core management principle of "risk first." If these areas are not addressed first, the poor timeliness and low accuracy of their layered data will pose the greatest potential threat to power grid operation, planning, and maintenance decisions.

[0074] Significance: This step outlines the primary areas for prioritizing resource updates, ensuring that the most urgent and critical problem areas are given priority.

[0075] One update task covered 15 distribution grids (numbered Grid-01 to Grid-15). Based on previous assessments, grids Grid-03, Grid-07, and Grid-12 were classified as "Class 1 Risk" due to extremely unstable historical image recognition and frequent tower coordinate deviations. The output of step S21 is this set of high-risk areas: {Grid-03, Grid-07, Grid-12}.

[0076] S22 determines the equipment similarity coefficient between the electrical equipment in the updated target area and other updated target areas based on the distribution data of electrical equipment in the updated target area; Commonalities in equipment assets across different regions: The equipment similarity coefficient is an indicator that measures the degree of similarity in model of electrical equipment between any two target update regions. The specific calculation method is as follows: For region X and region Y, count the number of devices of the same model that also exist in region Y in region X, divide this number by the total number of devices in region X (or a calculation of the total number of devices in both regions), and the resulting ratio is the equipment similarity coefficient of X to Y (or a two-way average).

[0077] The universality of power grid equipment models is key to algorithm optimization and experience reuse. If multiple regions share a large number of transformers, circuit breakers, towers, etc., of the same model, then a recognition model or parameters that have been validated in one region can be more reliably transferred to other regions. Updating regions with high equipment similarity yields results with greater spillover value.

[0078] Significance: By introducing the equipment dimension, technical feasibility and update efficiency are incorporated into decision-making. This avoids the inefficiency that may result from making decisions solely based on risk, such as "prioritizing the update of a unique, high-risk isolated piece of equipment where experience cannot be reused."

[0079] Specific examples: Grid-04 mainly includes 5 types of equipment: [Transformer T-1000kVA, Circuit Breaker CB-10kV-A type, Disconnect Switch DS-10kV-I type, Surge Arrester LA-10kV, Voltage Transformer PT-10kV]. Grid-08 mainly includes: [Transformer T-800kVA, Circuit Breaker CB-10kV-A type, Disconnect Switch DS-10kV-II type, Surge Arrester LA-10kV, Coupling Capacitor CC-10kV]. Both share the equipment types "Circuit Breaker CB-10kV-A type" and "Surge Arrester LA-10kV". Assuming a simple calculation, the equipment similarity coefficient between Grid-04 and Grid-08 can be calculated as the number of shared models (2) divided by the total number of models in Grid-04 (5), which is 0.4. The system will calculate this similarity coefficient for each pair of areas.

[0080] S23 determines the identification strategy for priority update areas in the update target areas based on a risk type of update target areas and the equipment similarity coefficient between the electrical equipment in the update target areas and other update target areas.

[0081] It should be noted that the equipment similarity coefficient is determined based on the proportion of electrical equipment of the same model that exists in other target areas.

[0082] It is understandable that in the above steps, the number of update target areas for a certain type of risk is obtained. If the number of update target areas for a certain type of risk is greater than the preset target area number threshold, then the identification strategy for the priority update area among the update target areas is to select the priority update area with the target proportion.

[0083] Level 1 Judgment – ​​Strategic Decision-Making Based on the Number of High-Risk Areas: The "Preset Target Area Quantity Threshold" is an integer threshold set by management strategy, representing the minimum number of core high-risk areas that management deems "worth launching a high-intensity, targeted priority update program." When the number of identified "Type 1 Risk" areas exceeds this threshold, it indicates that the high-risk issue has become widespread, requiring broader priority update actions.

[0084] This judgment marks the first branch point in the decision tree, embodying the principle of "focusing on the principal contradiction." If a category of high-risk areas is already numerous (exceeding the threshold), it indicates that the current update task faces widespread and severe quality risk challenges. In this case, the decision logic should be simplified: a large-scale priority update plan must be immediately initiated to ensure sufficient resources are allocated to these high-risk areas and to curb the spread of risk. At this point, excessively complex analysis is unnecessary; the most proactive strategy (i.e., prioritizing updates for areas with the highest "target proportion") should be adopted directly to quickly respond to and cover the main risks.

[0085] Significance: This level of judgment provides decision-making agility. When risk signals are clear and focused, the system can quickly make decisive decisions to "increase investment and respond comprehensively," avoiding delays caused by getting bogged down in lengthy detailed analysis in critical situations. It ensures that the strategy has sufficient responsiveness during periods of high risk.

[0086] Specific logic and example: Continuing from example S21, a risk area category is identified as {Grid-03, Grid-07, Grid-12}, with a quantity of 3. Assume the "preset target area quantity threshold" is 2. Since 3 > 2, the condition is met. At this point, the system skips subsequent complex judgments and determines the update strategy as: "Select the priority update areas based on the target proportion (let's call it P1, e.g., 20%)". The decision-making process ends here.

[0087] Additionally, it is understandable that if the number of target regions for a certain type of risk is not greater than a preset threshold for the number of target regions, the following content is also included: S231 determines the update risk weight coefficient of different update target areas based on the update risk type of different update target areas, and judges whether the sum of the update risk weight coefficients of different update target areas is greater than the preset weight coefficient threshold. If so, the identification strategy of the priority update area in the update target area is to select the priority update area of ​​the target proportion. If not, proceed to step S232. Second-level judgment – ​​Strategy decision based on global risk weights: The "update risk weight coefficient" is a quantified weight value assigned to each risk level, typically designed as follows: the highest weight coefficient for Category I risk (e.g., 1.0), followed by Category II risk (e.g., 0.6), and the lowest for Category III risk (e.g., 0.3). The "sum of update risk weight coefficients for different update target areas," or the global risk weight sum, is the sum of the weights corresponding to the risk levels of all areas to be updated, reflecting the overall risk load faced by this update task. The "preset weight coefficient threshold" is another criterion for determining whether the overall risk load warrants a high-intensity priority update strategy.

[0088] When the number of high-risk areas is small (not exceeding the first-level threshold), it does not necessarily mean the overall situation is optimistic. There may be a large number of "second-level risk" areas; while not the highest-risk, their accumulated risk pressure should not be underestimated. This assessment aims to capture this pattern of "dispersed but large-scale risk." By assigning weights to different levels of risk and summing them, risks of different qualities and levels can be unified onto a comparable quantitative scale, thus providing a more comprehensive assessment of overall pressure. If the total load exceeds the preset threshold, it indicates that even without prominent "peak" risks, widespread "plateau" risks still require systematic attention, thus necessitating a more intensive priority update strategy.

[0089] This level of assessment fills the gaps that may exist in decision-making based solely on the "number of risk areas of a certain type," making strategy formulation more holistic and systematic. It prevents situations where a few extreme problems are focused on while neglecting generalized medium risks, ensuring the coverage and depth of risk management.

[0090] Specific logic and example: Assume there is 1 risk area of ​​type I (not exceeding the threshold of 2), but 10 of the 15 areas are type II risks and 4 are type III risks. Calculate the global risk weight sum: 1*1.0 (type I) + 10*0.6 (type II) + 4*0.3 (type III) = 1 + 6 + 1.2 = 8.2. Set the "preset weight coefficient threshold" to 7.0. Since 8.2 > 7.0, the condition is met. At this point, the system decision is still: "Select the priority update area of ​​the target proportion (P1)".

[0091] S232 determines whether the sum of the update risk weight coefficients of different update target regions is less than a preset coefficient threshold (less than a preset weight coefficient threshold). If so, the identification strategy of the priority update region in the update target region is to select the priority update region of the third target ratio. If not, proceed to step S233. Level 3 Judgment – ​​Identifying Low-Risk Load Situations: The "preset coefficient threshold (less than the preset weight coefficient threshold)" is a numerical threshold lower than the threshold in S231, used to clearly define the "low-risk load" scenario. When the global risk weight is lower than this low threshold, it means that the overall risk level of this update task is very low.

[0092] In project management, resource allocation should be commensurate with the severity of the problem. If, after the first two steps of assessment, there are not enough Class 1 risk areas and the overall risk load is not high, it may mean that the current update task is relatively easy. In this case, allocating a high proportion of priority update resources may be uneconomical. Setting this low threshold is to identify this "benign" situation, thereby triggering a more resource-efficient "light" priority update strategy (i.e., prioritizing updates to areas with the lowest "third target ratio"), reserving the main resources for more complex future tasks or other work.

[0093] Significance: This judgment reflects the principles of refined decision-making and cost-effectiveness. It avoids mechanically adopting high-input strategies in low-risk situations, achieving flexible allocation and efficient utilization of resources. This aligns with the concept of modern lean management, which aims to optimize resource allocation as much as possible while ensuring basic quality objectives.

[0094] Specific logic and example: Continuing from the previous example, if there are 0 Class I risk areas, 5 Class II risk areas, and 10 Class III risk areas, the global risk weight sum = 0*1.0 + 5*0.6 + 10*0.3 = 0 + 3 + 3 = 6.0. Let the "preset weight coefficient threshold" be 7.0 (high threshold) and the "preset coefficient threshold (low threshold)" be 5.0. Since 6.0 is neither greater than 7.0 (high threshold) nor less than 5.0 (low threshold), it does not enter this branch. If the global weight sum is 4.0 (<5.0), the condition is met, and the strategy is defined as: "Select the priority update area of ​​the third target proportion (let's call it P3, for example, 5%)".

[0095] S233 Based on the equipment similarity coefficient between the electrical equipment in the target area and other target areas, determine the average value of the equipment similarity coefficients between different target areas, and determine whether the average value of the equipment similarity coefficients is less than a preset similarity coefficient threshold. If so, determine that the identification strategy for the priority update area in the target area is to select the priority update area with the target proportion. If not, proceed to step S234. Level 4 Judgment – ​​Analyzing the homogeneity of equipment composition in the analysis area: The "average of device similarity coefficients among different update target areas" refers to calculating the device similarity coefficients between a specific update target area and all other update target areas, and then averaging these coefficients. This average reflects the "prevalence" or "uniqueness" of device models in that area—a high average indicates that the device models in that area are similar to most other areas; a low average indicates that the device composition in that area is relatively unique. The "preset similarity coefficient threshold" is the criterion for judging whether the devices in a region are sufficiently "unique" or "prevalent".

[0096] When the overall risk level is in a neutral, "neither high nor low" state, a deeper decision-making dimension needs to be introduced. The homogeneity (or heterogeneity) of equipment composition directly affects the technical strategies and knowledge accumulation efficiency of the update work. If the similarity of equipment in all regions is high (high average similarity coefficient), it means that the update work is highly repeatable and has a large space for strategy selection. If the differences between equipment in different regions are large (low average similarity coefficient), it means that the update in each region may face unique technical challenges, and experience is difficult to reuse. In this case, the strategy should be more inclined to "cast a wide net" (select a higher proportion) in order to accumulate experience in diverse scenarios.

[0097] Significance: This level of judgment incorporates technical feasibility and long-term learning effects into real-time decision-making. It guides the system to proactively select update strategies that maximize knowledge accumulation and technical verification effects, within the limits of risk level, making each update task a "training ground" for improving overall automated processing capabilities.

[0098] Calculate the average device similarity coefficient of each region to all other regions. Assume that the average similarity coefficient of most regions is found to be above 0.5, but the preset "preset similarity coefficient threshold" is 0.4. Since the average (e.g., 0.5) is not less than the threshold 0.4, the condition is not met, and proceed to the next sub-step S234. If the average is less than 0.4, the strategy reverts to "selecting the priority update region of the target proportion (P1)".

[0099] S234 identifies update target regions whose average device similarity coefficient with other update target regions is less than a preset similarity coefficient threshold as risk target regions, and determines the identification strategy of priority update regions among the update target regions based on the sum of update risk weight coefficients of the risk target regions.

[0100] Level 5 Assessment – ​​Focusing on Risk Assessment in Unique Regions: "Risk target area" is a subset defined based on the S233 judgment, specifically referring to those update target areas whose "average device similarity coefficient with other update target areas is less than a preset similarity coefficient threshold". These areas are "outlier" areas with unique device model composition. "Sum of update risk weight coefficients of risk target areas" is calculated only for this subset of "outlier areas", summing the update risk weights (based on their respective risk levels) of all areas within it.

[0101] This is the final and most refined level of decision-making. When the overall similarity of devices between regions is not low (not meeting the S233 condition), we need to pay special attention to those few unique "outlier" regions. Because the technical experience gained from updating them has limited benefit to other regions, their value lies primarily in solving their own problems. Therefore, the key decision becomes: how significant are the risks posed by these "outlier" regions themselves? If their risk weights are high, it indicates that these unique regions are also high-risk areas with serious problems, requiring priority handling (adopting a high-intensity strategy). If their risk weights are not high, it indicates that these regions are both unique and relatively stable, and their priority can be appropriately reduced (adopting a medium-intensity strategy).

[0102] This judgment enables precise management of "specificity." It ensures that the strategy for unique areas of equipment is not a one-size-fits-all approach, but rather differentiated based on their actual risk levels. This avoids ignoring unique high-risk areas while also preventing over-focusing on unique but low-risk areas, achieving an optimal balance between risk and cost at the micro level.

[0103] Specific logic and example: Assume Grid-05 and Grid-10 are identified as "risk target areas" (i.e., unique areas of the device). Grid-05 is a Class II risk (weight 0.6), and Grid-10 is a Class III risk (weight 0.3). Their combined risk weights are 0.9. Let the "preset coefficient threshold" be 1.0. Since 0.9 is not greater than 1.0, the condition is not met. Therefore, the final strategy is: "Select the priority update area based on the second target proportion (let's call it P2, for example, 10%)."

[0104] It should be noted that when the sum of the updated risk weight coefficients of different risk target areas is greater than the preset coefficient threshold, the identification strategy for the priority update area in the update target area is determined to be the priority update area of ​​the target ratio; otherwise, the identification strategy for the priority update area in the update target area is determined to be the priority update area of ​​the second target ratio.

[0105] It should be noted that the target ratio is greater than the second target ratio, and the second target ratio is greater than the third target ratio. Specifically, if the identification strategy for the priority update region in the updated target region is to select the priority update region of the target ratio, it includes the following: The target update regions with the target proportion among the three risk types are selected as priority update regions. Specifically, the number of target update regions with the target proportion among the three risk types is used as a constraint, and the priority update regions are determined from large to small based on the average value of the equipment similarity coefficient between them and other target update regions.

[0106] When the strategy is ultimately determined to be "selecting the priority update region of the target ratio (P1)," the specific execution rules are as follows: The "Updated Target Area for the Target Proportion in the Three Risk Types" specifies the specific source pool for the priority areas: selected proportionally from all areas assessed as "Three Risk Types" (i.e., the lowest risk). The "Average Equipment Similarity Coefficient" serves as the sorting criterion, referring to the average equipment similarity of each area relative to all other areas, calculated in S22.

[0107] This is a highly strategic design. When resources are relatively abundant (using a high proportion of P1 strategies), prioritizing updates to low-risk (Category III) areas achieves the tactical goal of "rapid advancement and improved overall readiness." This is because low-risk areas have a high success rate of automated processing, require minimal manual verification, and can be completed efficiently. Simultaneously, within these low-risk areas, prioritizing areas with the most common equipment models (highest average similarity coefficients) ensures that the technological achievements (such as optimized algorithms) of the first updated areas can be reused to the greatest extent possible in subsequent areas, clearing common technical obstacles and accumulating reliable parameters for subsequent updates (including high-risk areas). This follows the excellent engineering practice principle of "starting with the easy and progressing to the difficult, using specific examples to drive broader implementation, and reusing experience."

[0108] This rule translates macro-level strategies into actionable selection steps. It ensures that the selection of priority update areas is not only fair and rule-based, but also intelligent and forward-looking. By prioritizing the resolution of "easy-to-solve and representative" issues, it builds up the technical expertise and confidence needed to tackle "difficult-to-solve and special" problems later, thereby systematically improving the efficiency and final quality of the entire update project.

[0109] Specific example: The final strategy is set as "selecting priority update areas based on the target proportion P1 (20%)". There are a total of 15 areas, 8 of which are Category III risk areas. The number of priority areas to be selected is 15 * 20% = 3. The rule requires selecting 3 from the 8 Category III risk areas. At this point, it's possible to supplement the selection according to the rules, for example, by selecting 3 from the Category III risk areas based on device similarity from high to low, thus forming a list of 3 priority update areas.

[0110] Specifically, the method for determining the early warning processing strategy is as follows: The core decision-making objective of this embodiment is to develop a refined and differentiated "early warning processing strategy" after identifying the "priority update areas." This strategy aims to solve a core problem: when the system is about to perform automated update processing on the priority update areas, what "similarity coefficient threshold" should be set to trigger a manual verification early warning? Setting a higher threshold (such as the first coefficient threshold) means more sensitive early warnings and stricter verification, but also a greater workload; setting a lower threshold (such as the second coefficient threshold) results in more lenient early warnings and less workload, but may lead to missed detections. This method intelligently selects the most suitable early warning sensitivity level by comprehensively analyzing the historical risk characteristics of the priority update areas themselves, their proportion in the overall task, and their equipment representativeness, thereby seeking the optimal balance between ensuring update quality and optimizing verification costs.

[0111] This method constructs a framework for data-driven early warning strategy decision-making: Input integration: Gathering three major categories of key input data: historical risk frequency (probability of change) of each location within the priority update area; the proportion of the priority update area set relative to the overall update task; and the device representativeness of the priority update area (similarity coefficient with other areas).

[0112] Initial assessment of strategy strength: Preliminary decisions are made based on the proportion of priority update areas. If the proportion is very high, it indicates that the focus of this update is highly concentrated on this area, requiring the highest level of alert strategy to ensure its quality.

[0113] In-depth conditional decision-making: When the scope coverage is relatively low, a multi-level refined judgment is initiated. This involves examining, in turn, the representativeness of equipment in the priority update area, its alignment with the overall mission, the concentration of high-risk points within the area, and finally, a weighted comprehensive evaluation. Each level of judgment attempts to prove or disprove the necessity of adopting a high-level early warning strategy from different dimensions.

[0114] Strategy Output: The final decision output is one of two preset warning strategies: "Preset Warning Strategy" (using a higher first coefficient threshold to trigger more frequent manual checks) or "Second Preset Warning Strategy" (using a lower second coefficient threshold to trigger relatively lenient manual checks).

[0115] S31 determines the probability of change at different locations in the priority update area based on the historical deviation between the orthophoto and the map layer of the power grid map in the priority update area; The inherent risks of prioritizing updates to the historical data of the quantified regions: "Probability of Change" here specifically refers to the frequency with which each geographical location (such as a pole point or equipment center point) within the priority update area experiences "image change" (i.e., the image similarity coefficient falls below the preset threshold at that time, triggering verification) during historical updates. The calculation method is as follows: count the number of times the location was marked as "changed" in the historical identification process (refer to the previous embodiment), and divide by the total number of reliable historical processes. This probability value reflects the historical tendency of unstable automated identification results for that location.

[0116] The formulation of early warning strategies must be based on a deep understanding of the risk characteristics of the target area. If a region has a large number of persistent "problem points" with high historical change probabilities, it means that automated identification of that region faces continuous challenges, and the likelihood of these points recurring in future updates is extremely high. Therefore, a more sensitive early warning mechanism (with a higher similarity coefficient threshold) is needed to capture these potential problems earlier and more rigorously, preventing them from slipping through the cracks. Calculating the change probability of each point provides the most fundamental micro-data for subsequent assessment of the overall risk concentration (S333) and comprehensive risk level (S334) of the region.

[0117] Significance: This step moves the regional-level conclusions from the preceding risk assessment down to a more refined location-level risk frequency analysis. This allows for more targeted early warning strategies, focusing on specific locations that have historically experienced recurring problems and require the most human monitoring, thereby improving the accuracy and effectiveness of early warnings.

[0118] Specific example: In the selected priority update area "Grid-08", there are 100 monitoring points. The system retrieves records from the past 5 reliable updates for each point. For example, point P_08_45 is marked as "changed" in 3 of these updates due to low similarity coefficients, so its probability of change is 3 / 5 = 0.6. The system calculates the probability of change for each of the 100 points in the area, forming a risk probability distribution map.

[0119] S32 determines the proportion of the priority update region in the update target region based on the priority update region data; Assess the breadth and scope of priority update tasks: The "proportion of priority update areas among the target update areas" is a macro-structural indicator, calculated as: (Number of areas selected as priority update / Total number of target areas planned for this task) * 100%. This proportion directly reflects the scope of the "focus" of resource allocation in this update work.

[0120] This is a high-level strategy switch. If the proportion of priority update areas is high (e.g., over 70%), it means that the main part of this update adopts the priority strategy, and the success of the entire task depends almost entirely on the update quality of these areas. In this case, the strictest and safest quality control measures must be taken, that is, the highest early warning sensitivity (preset early warning strategy) must be adopted to ensure that the main work is flawless. Conversely, if the proportion is low, it means that the priority update is only a part of the overall task or a pilot project, and there is room for more refined strategy selection based on its specific characteristics to optimize the overall resource utilization.

[0121] Significance: This step achieves the top-level decision-making principle of "strategy determines tactics." From a macro-level project management perspective, it determines the intensity of quality control based on the weight of priority tasks, avoiding quality risks due to excessive cost considerations in the main project, and aligning with the management principle of prioritizing key tasks.

[0122] Specific example: The target area for this update consists of 20 grids (Grid-01 to Grid-20). Based on the prioritization strategy, 3 grids were ultimately identified as priority update areas. Therefore, the proportion of priority update areas is 3 / 20 = 15%. Assuming the "preset proportion threshold" is 60%, since 15% is not greater than 60%, the preset warning strategy is not directly adopted. Instead, the subsequent refined judgment process (starting from S331) is initiated.

[0123] S33 determines the early warning processing strategy based on the change probability of different locations in the priority update area, the proportion of the priority update area in the update target area, and the device similarity coefficient between the priority update area and other update target areas.

[0124] It is understood that if the proportion of the priority update area in the update target area is greater than a preset proportion threshold, then the early warning processing strategy is determined to be a preset early warning strategy.

[0125] Additionally, it should be noted that if the proportion of the priority update region in the target update region is not greater than a preset proportion threshold, the following applies: S331 takes the average of the device similarity coefficients between the priority update area and other update target areas as the average similarity coefficient of the priority update area, and determines whether there is a priority update area with an average similarity coefficient greater than the preset similarity coefficient value. If yes, proceed to step S332; otherwise, determine that the early warning processing strategy is the second preset early warning strategy. Assess the representativeness (breadth) of equipment in priority update areas: The "mean similarity coefficient" is an indicator calculated for a single priority update region. It equals the average of the device similarity coefficients between this priority update region and all other update target regions (including both priority and non-priority regions). This mean measures the "typicality" or "representativeness" of this priority region in terms of device model composition for the entire update task batch. The "preset similarity coefficient value" is a threshold for judging whether the representativeness is strong enough.

[0126] The selection of early warning strategies needs to consider the spillover value of update work. If the areas prioritized for updates are highly representative in terms of equipment type (high average similarity coefficient), then investing rigorous early warning and verification resources in them will yield "high-quality results" (such as rigorously verified and confirmed correct equipment identification rules and parameters) that can be reliably applied to a large number of other areas subsequently. This is equivalent to using a high-cost, in-depth quality control to pave the way for a large amount of subsequent work, resulting in high overall benefits. Therefore, identifying such highly representative areas is one of the important reasons for adopting a high-level early warning strategy.

[0127] Significance: This judgment extends the concept of "equipment commonality" from a strategic influencing factor in the priority identification stage (previous implementation) to a value consideration factor in formulating early warning strategies at this stage. It guides the system to consider not only the regional risk itself but also the potential for reuse of its results when determining the intensity of early warnings, reflecting a long-term perspective and a focus on maximizing benefits.

[0128] Specific logic and example: Calculate the average similarity coefficient of each of the three priority update regions. Suppose that the average similarity coefficients of regions Grid-08, Grid-12, and Grid-15 are 0.82, 0.78, and 0.85 respectively, all higher than the "preset similarity coefficient value" (e.g., 0.70). Since such regions exist (number > 0), the condition is met, and further integration and evaluation proceeds to S332. If the average similarity coefficient of all priority regions is lower than 0.70, it indicates that they are relatively unique on their respective devices, and a more lenient "second preset warning strategy" is directly adopted.

[0129] S332 uses the average similarity coefficient of different priority update regions and the proportion of similar update regions in the update target region to determine the verification matching coefficient. It then determines whether the verification matching coefficient is greater than the preset matching coefficient threshold. If so, the warning processing strategy is determined to be the preset warning strategy. If not, the process proceeds to step S333. Assess the coverage of highly representative regions (a combination of breadth and weight): The "Verification Matching Coefficient" is a comprehensive indicator designed to quantify "the extent to which highly representative priority update regions cover the main body of this priority update task." A simplified calculation example could be: Verification Matching Coefficient = Average of the similarity coefficients of different priority update regions and the average proportion of similar update regions in the update target region. It measures the dominance of high-value (highly representative) regions in the priority set. The "Preset Matching Coefficient Threshold" is the standard for judging whether this dominance is high enough.

[0130] The mere existence of a few highly representative areas is insufficient. If these areas constitute only a small portion of the priority update set (e.g., only one out of three priority areas), then applying a high-level warning strategy to all three areas for that one area may be uneconomical. The purpose of this step is to determine whether highly representative areas constitute the "main force" of priority updates. If they constitute the majority of the priority update work (high verification matching coefficient), then applying a high-level warning strategy to the entire priority set is worthwhile, because the main body of the quality deliverables being ensured has high reusability.

[0131] Significance: This judgment is a deepening and quantification of S331, preventing the characteristics of individual regions from affecting the overall strategy. It requires that high representativeness must achieve a certain scale effect to prove that the investment in high-level early warning strategies has sufficient returns on scale.

[0132] Specific logic and example: Continuing from the previous example, assume that the highly representative area of ​​Grid-08 accounts for 33% of all three priority areas. The "verification matching coefficient" is calculated to be 0.72 using a specific formula. Let the "preset matching coefficient threshold" be 0.525. Since 0.525 > 0.3, the condition is met, so the "preset early warning strategy" is directly adopted. If the condition is not met, proceed to S333 to examine the risk concentration.

[0133] S333: Based on the change probability of different positions in the priority update area, determine the positions in the priority update area whose change probability is greater than a preset probability threshold, and determine whether the average proportion of the proportions of positions in different priority update areas whose change probability is greater than the preset probability threshold is greater than a preset proportion value. If yes, determine that the warning processing strategy is the second preset warning strategy; otherwise, proceed to step S334. This step first identifies locations within each priority update area where the "probability of change is greater than a preset probability threshold," these can be termed "high-risk locations." Then, the proportion of high-risk locations within each priority area is calculated. Finally, the average of this proportion across all priority areas is calculated, i.e., "the average proportion of locations with a probability of change greater than the preset probability threshold across different priority update areas." This average reflects the prevalence and concentration of high-risk locations within the set of priority update areas. The "preset proportion value" is the threshold used to determine whether this risk concentration is high.

[0134] The core purpose of early warning is to detect risks. If a large number of historically high-risk locations (a high average proportion of high-probability-of-change points) are found within the priority update area, it indicates that the early warning reliability for these areas is high. In this case, even if the equipment is not highly representative (S331 failed) or has not been scaled up (S332 failed), a more lenient approach can be used for early warning, thereby reducing the difficulty of data processing while ensuring the reliability of identification.

[0135] Specific logic and example: For three priority areas, calculate the proportion of points within each area with a change probability > 0.5 (assuming a preset probability threshold). Assume the three proportions are [0.1, 0.8, 0.15]. Their average is (0.1 + 0.8 + 0.15) / 3 ≈ 0.35. Let the "preset proportion value" be 0.30. Since 0.35 > 0.30, the condition is met, therefore the "second preset early warning strategy" is adopted. If not, proceed to the final S334 comprehensive evaluation.

[0136] S334 determines a comprehensive matching coefficient based on the average similarity coefficient of different priority update regions and the proportion of positions in different priority update regions with a change probability greater than a preset probability threshold, and determines the early warning processing strategy based on the comprehensive matching coefficient.

[0137] The "weight coefficient of priority update areas" is a coefficient assigned based on the average similarity coefficient (equipment representativeness) of the area. Generally, the higher the representativeness, the larger the weight coefficient, reflecting the potential value of the results. The "comprehensive matching coefficient" is a final composite indicator, calculated as follows: Comprehensive matching coefficient = Σ [(weight coefficient of a priority area) * (1 - proportion of high-risk locations within that area)] / number of priority update areas. It simultaneously integrates the area's "value weight" (equipment representativeness) and the "degree of deviation in the number of locations with high warning probability."

[0138] This is the final and most refined layer of decision-making logic. It no longer views "value" and "risk" in isolation, but assesses their spatial correlation, achieving a true cost-benefit analysis. It ensures that the strategy choice remains optimal even in marginal situations.

[0139] Specific logic and example: Calculate a weight coefficient for each priority region (e.g., the mean similarity coefficient is directly used as the weight). Calculate the proportion of high-risk locations in each region (same as S333). Then calculate the weighted sum. The mean values ​​of the three regions Grid-08, Grid-12, and Grid-15 are 0.82, 0.78, and 0.85, respectively, with proportions of [0.1, 0.8, 0.15]. Therefore, the calculated "comprehensive matching coefficient" is 0.8 multiplied by 0.1. Set the "preset comprehensive matching coefficient threshold" to 0.54. Since 0.54 ≤ 0.60, the condition is not met, so the "second preset early warning strategy" is ultimately adopted.

[0140] It should be noted that the value of the comprehensive matching coefficient is between 0 and 1. The smaller the average similarity coefficient of different priority update regions and the greater the proportion of positions with a change probability greater than the preset probability threshold in different priority update regions, the smaller the comprehensive matching coefficient will be.

[0141] It is understood that when the comprehensive matching coefficient is greater than the preset comprehensive matching coefficient threshold, the early warning processing strategy is determined to be the preset early warning strategy; when the comprehensive matching coefficient is not greater than the preset comprehensive matching coefficient threshold, the early warning processing strategy is determined to be the second preset early warning strategy.

[0142] It should be noted that the preset early warning strategy is to issue an early warning for positions with a similarity coefficient less than the first coefficient threshold, thereby requiring manual verification. The second early warning strategy is to issue an early warning for positions with a similarity coefficient less than the second coefficient threshold, thereby requiring manual verification.

[0143] It should be noted that the first coefficient threshold is greater than the second coefficient threshold.

[0144] The core difference between the "preset early warning strategy" and the "second preset early warning strategy" lies in the different "similar coefficient thresholds" used to trigger manual verification. The "first coefficient threshold" is higher (e.g., 0.80), while the "second coefficient threshold" is lower (e.g., 0.70). When the system performs automated update processing on a priority update area and generates a new layer, it immediately calculates the similarity coefficient between the new layer and the latest orthophoto. If the similarity coefficient at a certain location is lower than the threshold specified by the current strategy (either the first or second coefficient threshold), the system will automatically generate an early warning, mark that location as "requiring manual verification," and push it to quality control personnel.

[0145] Setting different threshold levels essentially controls the "sensitivity" and "workload" of manual verification. A high threshold (pre-set early warning strategy) implies a more conservative quality standard, issuing an alert for even a slight decrease in similarity, resulting in more frequent verifications and the strongest quality assurance, but also the highest labor costs. A low threshold (second pre-set early warning strategy) implies a more lenient standard, issuing an alert only when there is a significant deviation in similarity, resulting in less verification work and saving manpower within an acceptable risk range. Dynamically selecting thresholds through the aforementioned series of analyses finds an optimal solution for the trade-off between "quality" and "efficiency" in specific situations.

[0146] Significance: This step transforms abstract decision conclusions into specific, actionable quality control parameters. It enables a closed loop in the entire intelligent decision-making chain: from historical data analysis to priority identification, then to early warning strategy formulation, and finally to automated operational rules. This significantly enhances the intelligence and adaptability of the power grid layer update process.

[0147] Following the aforementioned decision-making process, the "second preset early warning strategy" was ultimately adopted for the three areas prioritized for this update, setting a similarity coefficient threshold of 0.70. After the automated update process was completed, the system scanned all points within these three areas. For example, if the new similarity coefficient of point P_08_45 in Grid-08 was found to be 0.65, which is below the threshold of 0.70, the system automatically highlighted the point as a red early warning on the GIS platform and generated a verification work order containing location information and the degree of deviation, which was then sent to the quality control engineer responsible for that area.

[0148] In this embodiment, quality control is made more refined and dynamic: the "one-size-fits-all" early warning standard is changed, and the intensity of quality control can be dynamically adjusted according to the specific composition and characteristics of the updated task, thus realizing the leap from extensive management to lean management.

[0149] Optimize the allocation of manual verification resources: By scientifically selecting early warning thresholds, the number of manual verification work orders is directly affected. While ensuring core quality, unnecessary waste of manpower is avoided, or valuable manpower is precisely directed to the verification points with the highest risk and greatest value, significantly improving the efficiency of human resource utilization.

[0150] Supporting the optimal global balance between risk and benefit: The decision-making process is consistently guided by the dual considerations of "risk prevention and control" and "updating benefits." Whether focusing on equipment representativeness (improving future efficiency), the reliability of current early warnings, or the final comprehensive matching assessment, the goal is to find the balance point that maximizes the overall benefit (including current quality assurance and future efficiency improvement) of limited resources within a given task context.

[0151] Enhancing the credibility and acceptance of automated systems: By employing intelligent and interpretable early warning strategies, a dynamic and resilient "smart fuse" is added to the results of automated processing. This prevents automated errors from flowing into the production system and avoids "alarm fatigue" among maintenance personnel due to overly frequent warnings, thereby increasing trust and reliance on the entire automated update system.

[0152] Furthermore, the updated identification deviation data includes locations where the position corresponding to the UAV's orthophoto is inconsistent with the map layer, that is, data that identifies orthophotos of other locations outside the stated location as the stated location.

[0153] Specifically, the method for determining the map layer update method for the target update area, excluding the priority update area, is as follows: The core decision-making objective of this embodiment is to construct a dynamic decision-making system based on the combination of fact-checking results of updated areas and probabilistic risk characteristics of areas to be updated. This system employs a hierarchical data strategy: for completed areas, it conducts objective analysis based strictly on manually verified errors ("identification deviation locations"); for "updated areas" to be processed, it focuses on analyzing the probabilistic characteristics in their historical risk files, paying particular attention to the specific high risks indicated by "low-probability deviation locations." By integrating these two types of data, the system intelligently determines the layer update method for the updated areas, achieving an organic combination of fact-based prudent decision-making and probability-based forward-looking early warning.

[0154] This method constructs a decision-making framework driven by both "facts" and "probability-based early warning": Fact-level analysis: Statistically analyze the objective distribution and quantity characteristics of genuine errors (identification deviation locations) confirmed by manual verification in all updated areas (including priority areas and other completed areas).

[0155] Probabilistic layer analysis: For the "updated area" to be processed, analyze the probability distribution of changes of all deviation positions in its historical archives, identify the composition ratio of "low probability deviation positions", and assess the hidden special risks of the deviation positions with low probability of change, where the orthophotos of the UAV are more likely to be mismatched.

[0156] Integrated decision-making: Based on the facts of the prevalence / severity of errors in the updated area, the current early warning mechanism's ability to detect low-probability risks (implied in the factual data), and the specific probability risk structure of the updated area, multi-level conditional judgments are made.

[0157] Strategy Output and Closed Loop: Output "Strict Mode" or "Standard Mode" decision and feed the results of this round of execution back to the fact database.

[0158] S41 determines the location of the identification deviation in the priority update area based on the updated identification deviation data; It should be noted that the identified deviation location refers to the location where the orthophoto of the UAV in the priority update area is inconsistent with the map layer.

[0159] "Identifying deviation locations" here specifically refers to, and only to, the locations in all target areas (including priority update areas and other updated areas) that have completed automated updates in this round or historical batches and undergone manual verification, where the location corresponding to the UAV's orthophoto is inconsistent with the map layer, as ultimately determined and confirmed by a manual verifier. The system records the coordinates, region, and error type of these locations, but in this step, it does not retrieve or analyze the historical change probability data of these locations.

[0160] The core of this step is establishing a "factual basis" for decision-making. Decisions will be made strictly based on the objective existence, quantity, and distribution of errors that have occurred. Probability data is excluded to avoid introducing historical speculative information during the factual analysis phase, ensuring that judgments about the current operational quality are entirely based on "verified and confirmed realities." This guarantees the purity and objectivity of the first stage of decision-making (factual judgment).

[0161] Significance: This step establishes an "empirical" starting point for the decision-making process. It begins with the most reliable results from manual verification and uses statistical factual data (whether errors are widespread or seriously excessive) to form a first-hand, unbiased assessment of the quality status of the current updated batch.

[0162] Specific examples: In this round and recent updates, the areas that have been completed and verified include: priority areas Grid-08, Grid-12, and Grid-15, as well as the earlier completed Grid-01 and Grid-03. The verification report summary shows: Grid-08 has 5 verification errors, Grid-12 has 1, Grid-15 has 0, Grid-01 has 2, and Grid-03 has 4. Step S41 collects information on these 12 "identification deviation locations" only for subsequent factual statistics (such as the number of area errors, whether any area exceeds the limit).

[0163] S42 takes the target area for updating, excluding the priority update area, as the update area, and determines the probability of change of the deviation position based on the historical deviation of the deviation position in the update area. Quantitatively update the historical risk characteristics of the region, focusing on the probability structure (probability analysis): "Updated Area" refers to the non-priority target area to be processed next. The definitions of "deviation location" and "probability of change" are the same as before.

[0164] This step involves a "risk pre-diagnosis" of the "updated area." The system retrieves its historical records, not only counting the total number of deviation locations but also analyzing the distribution structure of their probability of change. The core focus is identifying the number and proportion of "low-probability deviation locations" (e.g., probability < 0.2). These locations are "risk-sensitive points": they rarely make mistakes under normal historical conditions, therefore their status as "deviation locations" implies a potential mismatch risk. If an updated area has a large number of such points or a high proportion, even if its total historical error count is low, it indicates a high and widespread mismatch risk in the current operational environment where special changes may occur.

[0165] This step elevates risk assessment from simple "counting" to "deconstruction." By analyzing probability distributions, it reveals the "texture" of regional risks rather than just their "quantity," enabling early identification and special attention to potential systemic risks.

[0166] A specific example: The "update area" to be processed is Grid-02. Its historical records show a total of 25 deviation locations. The distribution of change probability is as follows: 15 points belong to "old problem points" with a high change probability (>0.6); 10 points belong to "sensitive points" with a low change probability (<0.3), i.e., locations with a high risk of mismatch. These 10 "sensitive points" are the focus of this assessment.

[0167] S43 determines the map layer update method for the update area based on the identified deviation location, the probability of change of the deviation location, and the update risk type of the update target area that has not yet been updated in the priority update area.

[0168] Furthermore, if there are identification deviation positions in all of the priority update areas, the map layer update method for the update area is to manually review and update all positions with similarity coefficients less than the first coefficient threshold.

[0169] Are there genuine errors prevalent in the updated areas? Logic: Check all priority update regions summarized in S41 to see if each region has at least one confirmed real error (identify the deviation location).

[0170] This is a "systemic risk" detection based on the broadest possible factual basis. If genuine errors are found in all priority update areas, this provides extremely strong evidence of widespread deficiencies in current operating conditions, indicating that risk is omnipresent. The most stringent measures must be taken in subsequent areas.

[0171] Result: If satisfied, the "strict mode" (first coefficient threshold) is applied to the updated region.

[0172] Specific example: The updated area Grid-15 has no errors. Therefore, the condition is not met, and we proceed to S431.

[0173] It should be noted that the deviation position refers to the position where, during the update process, the similarity coefficient between the orthophoto and the map layer of the power grid map is less than a preset similarity coefficient threshold.

[0174] Furthermore, if there are not uniformly identified deviation locations in the priority update region, the following is also included: S431 Based on the deviation identification positions in the target area of ​​the update process that was updated before the update area, determine the number of deviation identification positions in the target area of ​​the update process that was updated before the update area, and determine whether there are any target areas of the update process that were updated before the update area where the number of deviation identification positions does not meet the requirements. If so, determine that the map layer update method of the update area is to manually review all positions with similarity coefficients less than the first coefficient threshold before updating. If not, proceed to step S432. Is there a serious over-limit of errors in the updated regions? Logic: Check if any of the updated regions (within the same scope as S43) has a number of actual errors exceeding the tolerance limit set for that type of region.

[0175] When general risks have not yet materialized, focus on "outliers." A surge of errors in a particular area is a strong signal of localized risk, potentially indicating a serious conflict between the characteristics of that area and current operating conditions. Based on this, it is necessary to strengthen precautions for subsequent updates to areas that may exhibit similar characteristics.

[0176] Result: If satisfied, then the "strict mode" is adopted.

[0177] Specific example: Suppose the maximum tolerance for the number of errors in a region is 4. Grid-08 has 5 errors (>4), so the condition is met, and strict mode is used. If not, proceed to S432.

[0178] S432 determines the number of deviation positions in the updated area based on the probability of change of the deviation positions in the updated area, and determines whether the number of deviation positions in the updated area is greater than a preset threshold for the number of deviation positions. If so, the map layer update method for the updated area is to manually review all positions with similarity coefficients less than the second coefficient threshold before updating. If not, proceed to step S433. Is the total historical deviation of the updated region too large? Logic: Count the total number of all deviation positions in the current "updated area".

[0179] This is a basic historical risk load indicator. An excessively large number indicates that there are many known identification challenges in the area, in which case the reliability of manual verification is higher, and a more lenient approach can be adopted.

[0180] Result: If the total number exceeds the threshold, the "standard strict mode" (second coefficient threshold) is adopted.

[0181] Specific example: For Grid-02, the total number of deviation locations is 25. Let the threshold be 20. Since 25 > 20, the second coefficient threshold is used for verification. If the condition is not met, proceed to S433.

[0182] S433 Based on the change probability of the deviation position, determine whether the average change probability of the deviation position in the updated area is less than a preset change probability threshold. If yes, proceed to step S434. If no, determine that the map layer update method of the updated area is to manually review all positions with similarity coefficients less than the second coefficient threshold before updating. Does the deviation location in the updated region primarily consist of points with low probability of change? Key logic: Calculate the average of the probability of positional changes of all deviations in the current "updated area" and compare it with the "preset probability threshold".

[0183] Why is it set up this way? This judgment is the core link between "facts" and "probabilities." An average value less than the threshold means that the deviation locations in this area are mainly composed of "sensitive points with low probability of change." Although these points have historically had a low total number of errors, they are currently deviation locations, indicating a high risk of mismatch. Therefore, the risk in this type of area has the characteristics of "low historical frequency and high current sensitivity," with high uncertainty, requiring more prudent decision-making (go to S434).

[0184] Results: If the average value is less than the threshold, proceed to S434; if the average value is greater than or equal to the threshold, it indicates that the problem is mainly a high-frequency persistent issue, and the "standard strict mode" can be used.

[0185] For example, the average probability of change for the 25 deviation locations in Grid-02 is 0.28 (less than the threshold of 0.6). This indicates that the deviation locations are mostly "sensitive points" with a special risk pattern, requiring final decision by S434.

[0186] S434 determines the number of update target areas of a certain risk type in the update target areas that have not yet been updated, based on the update risk type of the update target areas that have not yet been updated. It then determines whether the number of update target areas of a certain risk type is greater than a preset area quantity threshold. If so, it determines that the map layer update method for the update area is to manually review and update all positions with similarity coefficients less than a first coefficient threshold. If not, it determines that the map layer update method for the update area is to manually review and update all positions with similarity coefficients less than a second coefficient threshold.

[0187] The final trade-off between comprehensive factual feedback and specific risk models: If the current updated area itself is dominated by low-probability points and the project has many remaining Class I risk areas (high pressure to overcome), then a "strict mode (first coefficient threshold)" should be adopted for the current updated area. This is because the facts show that this type of risk is emerging and the current area is susceptible. At this time, by improving the reliability of manual verification, the foundation is laid for optimizing the manual verification strategy in the later stage.

[0188] If the proportion of "low-probability error points" in the updated area is high but the project has few remaining Class I risk areas (low pressure), the "second coefficient threshold" can also be considered, as the need for subsequent verification is not high.

[0189] This is the highest level of decision-making wisdom. It secretly utilizes the probabilistic information behind factual data, assesses the ratio of new to old risk patterns, and combines this with the overall allocation of resources to make the most targeted decisions.

[0190] Specific example: Currently, Grid-02 is dominated by points with low probability of change (S433 conclusion). There are 4 remaining risk areas (assuming a threshold of 3, high pressure). Based on rule 1, "strict mode" is adopted for Grid-02.

[0191] It should be noted that when deviations are found in the updated area during the manual review process, all locations will be updated after manual review.

[0192] This embodiment constructs an intelligent decision-making system based on a "dual-track analysis of facts and probabilities." Its ingenuity lies in employing differentiated data analysis strategies for updated and unupdated regions: for the former, purely objective factual statistics are performed to ensure a solid foundation; for the latter, in-depth probabilistic pattern deconstruction is conducted to achieve risk foresight, thus guaranteeing the objectivity of decision-making while endowing the system with the ability to perceive hidden risks.

[0193] The decision-making basis is objective and reliable: it takes the facts verified by humans as the starting point, avoiding the impact of the uncertainty of the prediction model on the core judgment.

[0194] Forward-looking risk identification: By analyzing the probability structure of "updated areas", it is possible to identify areas that are sensitive to changes in the current environment in advance and prevent problems before they occur.

[0195] Precise strategy generation and adaptation: By integrating "what has happened" (facts) and "what might happen" (probability), and making differentiated responses to different risk patterns (old problems / new risks), the strategy is highly matched with the real risk landscape.

[0196] The system possesses implicit learning capabilities: by associating facts with historical probabilities in the background, the system can silently evaluate the detection effectiveness of the current early warning mechanism for different risk categories and use it to optimize subsequent decisions, forming an implicit learning loop.

[0197] Example 2 In a second aspect, the present invention provides a computer system comprising: a memory and a processor connected in communication, and a computer program stored in the memory and capable of running on the processor, wherein the processor executes the above-described method for updating map layers based on UAV orthophotos when running the computer program.

[0198] The various embodiments in this specification are described in a progressive manner. Similar or identical parts between embodiments can be referred to mutually. Each embodiment focuses on describing the differences from other embodiments. In particular, the embodiments of apparatus, devices, and non-volatile computer storage media are basically similar to the method embodiments, so the descriptions are relatively simple; relevant parts can be referred to the descriptions of the method embodiments.

[0199] The foregoing has described specific embodiments of this specification. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps recited in the claims may be performed in a different order than that shown in the embodiments and may still achieve the desired result. Furthermore, the processes depicted in the drawings do not necessarily require the specific or sequential order shown to achieve the desired result. In some embodiments, multitasking and parallel processing are possible or may be advantageous.

[0200] The above description is merely one or more embodiments of this specification and is not intended to limit this specification. Various modifications and variations can be made to the one or more embodiments of this specification by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principle of one or more embodiments of this specification should be included within the scope of the claims of this specification.

Claims

1. A method for updating map layers based on UAV orthophotos, characterized in that, Specifically, it includes: Based on historical update data of the power grid map, similarity data between the orthophoto of the UAV in the update target area and the map layer of the power grid map is determined. Using the similarity data and the location distribution data of the similarity degree not meeting the requirements, the update risk type of the update target area is determined. Based on the different update risk types of the update target areas and the distribution data of electrical equipment in the update target areas, the identification strategy of priority update areas in the update target areas is determined. Based on the identification strategy, the identification result of the priority update area is determined. Based on the historical deviation between the orthophoto and the map layer of the power grid map in the priority update area, and combined with the similarity of electrical equipment in the priority update area and other update target areas, an early warning processing strategy is determined. Based on the aforementioned early warning processing strategy, update identification deviation data of the map layer based on orthophotos is determined. Based on the update identification deviation data, and combined with the historical deviation of the deviation location of the update target area and the update risk type, a map layer update method for the update target area, excluding the priority update area, is determined.

2. The map layer update method based on UAV orthophotos as described in claim 1, characterized in that, The historical update data of the power grid map is based on the historical update process of map layers in different areas of the power grid map and the similarity data between the orthophotos of UAVs and the map layers of the power grid map in different historical update processes.

3. The map layer update method based on UAV orthophotos as described in claim 1, characterized in that, The target area for updating is determined based on the user's selection.

4. The map layer update method based on UAV orthophotos as described in claim 1, characterized in that, The similarity data between the orthophoto of the UAV and the map layer of the power grid map is determined based on the image similarity coefficient between the orthophoto and the map layer at different locations in the updated target area during different historical update processes.

5. The map layer update method based on UAV orthophotos as described in claim 1, characterized in that, The method for determining the update risk type of the target update area is as follows: Based on the aforementioned similar data, determine the image similarity coefficients between orthophotos and map layers at different locations within the updated target area during the historical update process; Based on the image similarity coefficient, determine the historical update process of image changes at the location; Based on the historical update process data of image changes at different locations within the target update area, the update risk type of the target update area is determined.

6. The map layer update method based on UAV orthophotos as described in claim 5, characterized in that, The historical update process of the image at the specified location being changed is the historical update process of the image at the specified location having a similarity coefficient less than a preset similarity coefficient threshold.

7. The map layer update method based on UAV orthophotos as described in claim 1, characterized in that, The method for determining the early warning processing strategy is as follows: Based on the historical deviation between the orthophoto and the map layer of the power grid map in the priority update area, the probability of change at different locations in the priority update area is determined; Based on the priority update region data, determine the proportion of the priority update region in the target update region; The early warning processing strategy is determined based on the probability of change at different locations in the priority update area, the proportion of the priority update area in the update target area, and the device similarity coefficient between the priority update area and other update target areas.

8. The map layer update method based on UAV orthophotos as described in claim 7, characterized in that, If the proportion of the priority update area in the target update area is greater than a preset proportion threshold, then the early warning processing strategy is determined to be the preset early warning strategy.

9. The map layer update method based on UAV orthophotos as described in claim 1, characterized in that, The method for determining the map layer update method for the target update area excluding the priority update area is as follows: Based on the updated identification deviation data, the identification deviation position in the priority update region is determined; The target area for updating, excluding the priority update area, is taken as the update area. The probability of change of the deviation position is determined based on the historical deviation of the deviation position in the update area. Based on the identification deviation location, the probability of change of the deviation location, and the update risk type of the update target area that has not yet been updated, the map layer update method for the update area is determined.

10. A computer system, comprising: A memory and processor connected in communication, and a computer program stored in the memory and capable of running on the processor, characterized in that, when the processor runs the computer program, it executes a map layer update method based on UAV orthophotos as described in any one of claims 1-9.

Images