Illegal behavior early warning processing method and system based on image recognition

Abstract

The invention provides an illegal behavior early warning processing method and system based on image recognition, and belongs to the technical field of image processing. According to the configuration data of the behavior recording terminals of the maintenance construction site matched with the deviation type and the reference identification processing strategy, determining a control strategy for identifying all the behavior recording terminals in the maintenance construction site matched with the deviation type, and based on the switching data of the recognition processing target of the violation behavior of the behavior recording terminal and the control strategy, the early warning processing method of the behavior recording terminal of the maintenance construction site is determined, and the reliability of early warning processing of the violation behavior is improved.

Description

TECHNICAL FIELD

[0001] The present application belongs to the technical field of image processing, and particularly relates to a rule violation early warning processing method and system based on image recognition. BACKGROUND

[0002] TECHNICAL FIELD In order to realize the identification of rule violations in the process of power maintenance, a portable rule violation recorder is set up to facilitate the identification and processing of rule violations in the process of repair through the image mode. In the invention patent application CN201922076989.5 "On-site operation real-time supervision system based on data fusion processing", the scene recorder is set up to realize remote visual supervision of on-site operation, solve the problem of wide range of on-site operation and insufficient on-site supervision, and significantly improve the control efficiency of rule violations and operation risks. However, there are the following defects: When identifying and processing rule violations of different maintenance sites in the cloud platform, in order to improve the efficiency of rule violation identification and processing, in the case of more maintenance sites, only part of the behavior record terminals can be selected for rule violation identification and processing. Therefore, on the basis of using a fixed proportion to determine the behavior record terminals for rule violation identification and processing in different sites, how to regularly identify rule violations of all record terminals and specifically optimize the behavior record terminals for rule violation identification and processing to improve the reliability of rule violation identification and processing becomes a technical problem to be solved.

[0003] To solve the above technical problems, the present application provides a rule violation early warning processing method and system based on image recognition. SUMMARY

[0004] To achieve the purpose of the present application, the present application adopts the following technical solutions: Specifically, the present application provides a rule violation early warning processing method based on image recognition, which specifically includes: S1, based on the maintenance site data, determine the distribution data of the behavior record terminal, based on the distribution data of the behavior record terminal and the maintenance personnel data in different maintenance sites, determine the reference identification and processing strategy of the behavior record terminal in the maintenance site, based on the reference identification and processing strategy, determine the early warning processing matching type of the rule violation of the maintenance site, based on the maintenance personnel data in different early warning processing matching types, when the reliability of early warning processing does not meet the requirements, enter the next step; S2 when the pre-warning processing matching type of the violation behavior of the maintenance site is the matching deviation type, determining the control strategy of identifying processing for all the behavior record terminals in the maintenance site of the matching deviation type according to the configuration data of the behavior record terminals of the maintenance site of the matching deviation type and the reference identifying processing strategy; S3 determining the pre-warning processing method of the behavior record terminals of the maintenance site based on the switching data of the identifying processing target of the violation behavior of the behavior record terminals and the control strategy.

[0005] The present application has the following advantages: The control strategy of identifying processing for all the behavior record terminals in the maintenance site of the matching deviation type is determined according to the configuration data of the behavior record terminals of the maintenance site of the matching deviation type and the reference identifying processing strategy, so that the reliability of the identifying processing of the violation behavior caused by the reference identifying processing strategy is considered, and the idle condition of the behavior record terminals in the maintenance site is also considered, and then the control strategy of identifying processing for all the behavior record terminals is determined from the perspective of the idle condition and the reliability of the identifying processing of the violation behavior, so that the reliability of the identifying processing of the violation behavior is ensured, and the switching processing of the identifying processing target is laid a foundation.

[0006] The pre-warning processing method of the behavior record terminals of the maintenance site is determined based on the switching data of the identifying processing target of the violation behavior of the behavior record terminals and the control strategy, so that the difference of the deviation degree of the identifying processing of the violation behavior caused by the difference of the switching frequency of the identifying processing target is considered, and the reliability of the switching control of the behavior record terminals is evaluated in combination with the control strategy, and then the pre-warning processing method of the behavior record terminals of the maintenance site is determined from the perspective of the reliability of the switching control and the deviation degree of the identifying processing of the violation behavior, that is, the adjustment of the identifying processing target, so that the reliability of the identifying processing of the violation behavior is further improved.

[0007] Further, the distribution data of the behavior record terminals includes the number of the behavior record terminals configured in different maintenance sites.

[0008] Further, the maintenance personnel data in the maintenance site includes the number of the maintenance personnel in the maintenance site.

[0009] Further, the method for determining the reference identifying processing strategy of the behavior record terminals in the maintenance site is: determining the number of the behavior record terminals in different maintenance sites based on the distribution data of the behavior record terminals; Determine the number of maintenance personnel in different maintenance sites based on the maintenance personnel data in different maintenance sites; Determine the reference identification processing strategy of the behavior recording terminal in the maintenance site according to the number of behavior recording terminals and the number of maintenance personnel in different maintenance sites.

[0010] Further, the method for determining the control strategy of the identification processing of all behavior recording terminals in the maintenance site is: Determine the total number of behavior recording terminals in the matching deviation type of maintenance site minus the number of identified terminals as the number of idle terminals according to the configuration data of the behavior recording terminal in the matching deviation type of maintenance site; Determine the identification matching ratio of the matching deviation type of maintenance site according to the reference identification processing strategy; Determine the control strategy of the identification processing of all behavior recording terminals in the matching deviation type of maintenance site based on the number of idle terminals and the identification matching ratio in the matching deviation type of maintenance site.

[0011] On the other hand, the present application provides a computer system comprising a memory and a processor connected in communication, and a computer program stored on the memory and capable of running on the processor, wherein the processor executes the computer program to perform the above-mentioned image recognition-based illegal behavior early warning processing method.

[0012] Other features and advantages will be set forth in the following description, and in part will become apparent to those skilled in the art from the description, or can be learned by practice of the application. The objects and other advantages of the application will be realized and attained by the structure particularly pointed out in the description and the drawings.

[0013] In order to make the above-mentioned objects, features and advantages of the present application more obvious and easy to understand, the following preferred embodiments are specifically described in detail below, and the accompanying drawings are referred to. BRIEF DESCRIPTION OF DRAWINGS

[0014] The above-mentioned and other features and advantages of the present application will become more apparent by describing example embodiments thereof with reference to the accompanying drawings.

[0015] Figure 1 A flowchart of an image recognition-based illegal behavior early warning processing method; Figure 2 A flowchart of a method for determining the reference identification processing strategy of the behavior recording terminal in the maintenance site; Figure 3 A flowchart of a method for determining the matching type of the early warning processing of illegal behavior in the maintenance site. DETAILED DESCRIPTION

[0016] Example implementations are now described in greater detail with reference to the figures. Like reference numerals can be used to denote like elements throughout the figures. The example implementations can be implemented in numerous ways, and are not limited to the implementations described herein; instead, the implementations described herein provide a thorough teaching of the example implementations to enable persons skilled in the art to employ the example implementations.

[0017] The terms "a," "an," "the" and "said" are used to refer to one or more than one (i.e., to "at least one") of the referenced elements / constituents / etc.; the terms "comprising," "having" and "including" are used to mean "including but not limited to;" and the term "consisting essentially of" means including an element or ingredient in addition to those recited.

[0018] Example 1 To solve the above problems, according to one aspect of the present application, as shown in Figure 1 An image recognition-based rule violation early warning processing method is provided, and specifically includes: S1, based on the maintenance site data, determining the distribution data of the behavior recording terminal, based on the distribution data of the behavior recording terminal and the maintenance personnel data in different maintenance sites, determining the reference identification processing strategy of the behavior recording terminal in the maintenance site, based on the reference identification processing strategy, determining the early warning processing matching type of the rule violation in the maintenance site, when the reliability of the early warning processing does not meet the requirements based on the maintenance personnel data in different early warning processing matching types, going to the next step; S2, when the early warning processing matching type of the rule violation in the maintenance site is the matching deviation type, according to the configuration data of the behavior recording terminal of the matching deviation type in the maintenance site and the reference identification processing strategy, determining the control strategy of identifying all behavior recording terminals in the matching deviation type in the maintenance site; S3, based on the switching data of the identification processing target of the rule violation of the behavior recording terminal and the control strategy, determining the early warning processing method of the behavior recording terminal in the maintenance site.

[0019] Further, the distribution data of the behavior recording terminal includes the number of behavior recording terminals configured in different maintenance sites.

[0020] Further, the maintenance personnel data in the maintenance site includes the number of maintenance personnel in the maintenance site.

[0021] Specifically, as shown in Figure 2 The method for determining the reference identification processing strategy of the behavior recording terminal in the maintenance site is: The root problem to be solved by the present solution is: how to determine the number of image / video returned by the "behavior record terminal" for real-time and high-load AI illegal behavior identification under the condition of limited computing resources (such as server computing power and network bandwidth)? The core idea is: according to the ratio of "personnel-equipment", to determine the intensity of monitoring, so as to realize the determination of the identification mode of the benchmark illegal behavior in different maintenance sites.

[0022] Core logic: count the number of devices and personnel in each site → calculate the matching factor → determine the number of devices for illegal behavior identification according to the overall matching level.

[0023] S21 determines the number of behavior record terminals in different maintenance sites based on the distribution data of the behavior record terminals. In the above step, the number of behavior record terminals in different maintenance sites is determined. Behavior record terminal refers to intelligent safety helmet, monitoring ball, fixed monitoring and other devices that can collect on-site video / image.

[0024] Distribution data of behavior record terminals: refers to the number distribution of these terminals in different sites.

[0025] S22 determines the number of maintenance personnel in different maintenance sites based on the maintenance personnel data in different maintenance sites. In the above step, the number of maintenance personnel in different maintenance sites is determined, which is a demand assessment. The number of personnel represents the potential risk points and the scale of objects that need to be monitored. The more intensive the personnel, the higher the possibility and severity of illegal behavior.

[0026] Maintenance personnel data: refers to the total number of construction personnel, technical personnel and other personnel working in the maintenance site.

[0027] S23 determines the benchmark identification processing strategy of the behavior record terminal in the maintenance site according to the number of behavior record terminals and the number of maintenance personnel in different maintenance sites.

[0028] It can be understood that according to the number of behavior record terminals and the number of maintenance personnel in different maintenance sites, the benchmark identification processing strategy of the behavior record terminal in the maintenance site is determined, which specifically includes: Based on the ratio of the number of behavior record terminals and the number of maintenance personnel in different maintenance sites, determine the record matching factor in different maintenance sites; Based on the record matching factor, determine the benchmark identification processing strategy of the behavior record terminal in the maintenance site.

[0029] In the above steps, a standardized, quantifiable indicator is needed to objectively compare the monitoring resource adequacy of different construction sites. Directly using absolute numbers is not comparable, while the ratio (number of terminals / number of personnel) can well reflect "how many terminals can be covered on average per person".

[0030] Record matching factor: for a particular maintenance site, the ratio of the number of behavior recording terminals to the number of maintenance personnel, formula: record matching factor = number of terminals / number of personnel Example: a construction site has 10 terminals and 20 workers, its record matching factor = 10 / 20 = 0.5.

[0031] It can be understood that, based on the record matching factor, the reference identification processing strategy of the behavior recording terminal in the maintenance site is determined, specifically including: Case 1: when the average value of the record matching factor in different maintenance sites is greater than the preset matching factor threshold, the reference identification processing strategy of the behavior recording terminal in the maintenance site is determined as using a preset proportion of behavior recording terminals in different maintenance sites for early warning identification processing of violation behaviors; Case 1: overall resource adequacy (average value > threshold), logic: when the average record matching factor is high, it means that on average, each person can be covered by more terminals. The monitoring resources are relatively adequate.

[0032] Strategy: a preset proportion of behavior recording terminals are used in different maintenance sites for early warning identification processing of violation behaviors.

[0033] Because there are many resources, even if only a part of the terminals (for example, the preset proportion = 30%) are enabled, effective monitoring coverage can be basically formed. This can save a lot of computing resources and energy, and concentrate the computing power on part of the terminals to ensure their identification accuracy and speed. This is a "optimize resource use efficiency" strategy.

[0034] Case 2: when the average value of the record matching factor in different maintenance sites is not greater than the preset matching factor threshold, the reference identification processing strategy of the behavior recording terminal in the maintenance site is determined as using a second preset proportion of behavior recording terminals in different maintenance sites for early warning identification processing of violation behaviors.

[0035] It can be understood that the second preset proportion is greater than the preset proportion.

[0036] Case 2: overall resource shortage (average value ≤ threshold), logic: when the average record matching factor is low, it means that the number of terminals is relatively insufficient compared to the number of personnel, and the monitoring coverage itself has the risk of blind area.

[0037] Strategy: A second preset ratio of behavior recording terminals will be used at various maintenance sites to identify and issue warnings for violations. Furthermore, the second preset ratio must be greater than the preset ratio.

[0038] Due to resource constraints, in order to maximize security and reduce false negatives, a higher proportion of terminals (e.g., the second preset proportion = 50%) must be used for AI identification. This is a strategy of "prioritizing the effectiveness of security monitoring," even though it consumes more computing resources.

[0039] Specifically, a behavior recording terminal with a second preset ratio is used to identify and process violations, including: The number of behavior recording terminals that need to be identified and processed for violations is determined by multiplying the second preset ratio by the number of behavior recording terminals in the maintenance site. Select a number of behavior recording terminals from the maintenance site that require identification and processing of violations, and perform violation identification and processing.

[0040] Specifically, such as Figure 3 As shown, the method for determining the matching type of the early warning processing for violations at the maintenance site is as follows: Previous process ( Figure 2 This establishes a baseline strategy from a global perspective, while this process ( Figure 3 The approach involves going deep into each specific maintenance site, "diagnosing" it based on its own situation, and "labeling" it for more refined management in the future.

[0041] Objective: To assess the potential reliability of AI-based violation detection for each construction site under its current configuration.

[0042] Method: The monitoring density is determined by comparing the number of terminals used for identification with the number of personnel to be monitored.

[0043] Output: The construction sites are divided into two types: "reliable identification type" and "mismatch type", providing a basis for subsequent differential processing.

[0044] Core logic: Obtain the number of recognition terminals at a single construction site → Calculate the ratio of the number of terminals to the number of personnel → Determine the reliability of AI recognition at the construction site based on the ratio.

[0045] S31 takes the number of behavior recording terminals in the maintenance site that need to be identified and processed for violations as the number of identification terminals. In the steps described above, determining the number of identification terminals is an assessment of the actual "monitoring force" deployed at the construction site. This number may come from the allocation results of the previous process (e.g., only 50% of the terminals were allocated for identification), or it may be all the terminals at the construction site.

[0046] Number of Identification Terminals: The number of behavior recording terminals actually activated and undergoing AI-based violation identification and analysis at this maintenance site. This represents the actual monitoring resources in effect.

[0047] S32 uses the ratio of the number of identification terminals to the number of maintenance personnel at the maintenance site as the identification matching ratio. Specifically, calculating the matching ratio requires a key indicator to quantify the degree of matching between "monitoring resources" and "monitoring tasks." Looking at the number of terminals alone is meaningless; it must be compared to the scale of the personnel it aims to cover.

[0048] Definition of the matching ratio: This refers to the number of AI identification terminals divided by the number of maintenance personnel. A higher ratio means that on average, more AI identification terminals are deployed per person, theoretically resulting in denser monitoring coverage and more reliable identification results. A lower ratio means that more personnel need to be covered by the terminals, potentially leading to blind spots and an increased risk of missed detections.

[0049] S33 determines the early warning processing matching type for the violation at the maintenance site based on the identification matching ratio.

[0050] It is understandable that when the identification matching ratio of the maintenance site is greater than the preset identification matching threshold, the early warning processing matching type of the violation at the maintenance site is determined to be the reliable identification type. However, if the identification matching ratio of the maintenance site is not greater than the preset identification matching threshold, the number of identification terminals is relatively small, which may prevent reliable monitoring of violations at the maintenance site. Therefore, the early warning processing matching type of the violation at the maintenance site is determined to be the matching deviation type.

[0051] Specifically, determining the matching type for early warning processing is the decision-making step in this process. By comparing the quantified "identification matching ratio" with a standard (preset identification matching threshold), a qualitative judgment is made on the monitoring status of the construction site, thereby providing clear instructions for subsequent actions.

[0052] Preset identification and matching threshold: A minimum density standard set by management to ensure effective monitoring. For example, it can be set to 0.5, meaning that at least one AI identification terminal is needed for every two workers on average to guarantee basic reliability. Matching type for early warning of violations: A classification label for the reliability of the current AI monitoring capabilities at the construction site.

[0053] Specifically, in scenario one: construction site A (with sufficient resources); S31: Construction site A has 10 behavior recording terminals. According to the global policy, 8 of them are used for AI recognition. Therefore, the number of recognition terminals = 8.

[0054] S32: The number of maintenance personnel at site A is 10.

[0055] The matching ratio is identified as 8 / 10 = 0.8, S33: 0.8 > 0.5 (preset threshold), determine the type: identify reliable type. Analysis: On average, there is one AI recognition terminal covering every 1.25 workers at this construction site, indicating a high monitoring density and a high degree of confidence in the system's recognition results.

[0056] Scenario 2: Construction Site B (Resources are scarce); S31: Construction site B has 5 behavior recording terminals, all of which are used for AI recognition. Therefore, the number of recognition terminals = 5.

[0057] S32: The number of maintenance personnel at construction site B is 15, and the identification matching ratio is approximately 5 / 15 ≈ 0.33; S33: 0.33 ≤ 0.5 (preset threshold); Type determination: Matching deviation type; Analysis: On average, there is only one AI recognition terminal for every three workers at this construction site, resulting in large blind spots in the monitoring. The system has determined that the reliability of its AI recognition is questionable and that other monitoring methods are needed to supplement it.

[0058] This process acts like a "health diagnostic tool for the monitoring system." Through a simple calculation, it labels each construction site as either "healthy" (reliable identification) or "sub-healthy" (mismatch). This allows safety management to move from a broad-based to a more refined approach, enabling: Identify weaknesses: Quickly locate construction sites with insufficient monitoring resources; Differentiated management: Adopt different response strategies for construction sites in different states to maximize resource efficiency; Risk warning: Be aware in advance of which construction sites may have incomplete monitoring data, thereby avoiding misjudgments of the overall safety situation.

[0059] Specifically, determining that the reliability of early warning processing does not meet the requirements includes: The objective of the following steps is to assess the reliability of the entire monitoring system from a macro perspective.

[0060] Method: We not only look at the number of "problem sites" (mismatched sites), but also the size (number of personnel) of these problem sites and the severity of the problems (the degree of poorness of the match ratio), and conduct a weighted comprehensive assessment.

[0061] Output: A final judgment – ​​whether the reliability of the early warning processing meets the requirements.

[0062] Core logic: Identify the problematic construction site → Assess its overall impact from multiple dimensions (quantity, scale, severity, weighted impact) → Make a global reliability judgment.

[0063] S41 uses data on different early warning processing matching types to identify maintenance sites whose early warning processing matching type is a matching deviation type, and then uses them as matching deviation sites. In the steps described above, identifying the sites with mismatches is the starting point for the assessment. First, all sites diagnosed as "unhealthy" need to be screened out as the subjects of this assessment.

[0064] Matching Deviation Sites: These refer to maintenance sites identified as "Matching Deviation Type" in Process Three. These sites share the common characteristic of having an insufficient number of AI recognition terminals relative to the number of personnel, resulting in monitoring blind spots.

[0065] S42 determines the number of maintenance personnel in different matching deviation sites based on the maintenance personnel data in the matching deviation sites; In the steps above, the number of maintenance personnel at the mismatch sites is determined to understand the "scale" of each problem site. A problem site with only 5 people and a problem site with 50 people have completely different risk levels. The number of personnel represents the potential risk impact range of the site.

[0066] S43 determines whether the reliability of the early warning processing meets the requirements based on the number of maintenance personnel at different matching deviation sites and the identification matching ratio.

[0067] Understandably, the reliability of the early warning system is determined based on the number of maintenance personnel at different matching deviation sites and the matching ratio, specifically including: S431 obtains the number of the matching deviation construction sites, and determines whether the number of the matching deviation construction sites is greater than the preset matching deviation construction site number threshold. If yes, it is determined that the reliability of the early warning processing does not meet the requirements. If no, proceed to the next step. In the above steps, the judgment is based on the number of problematic construction sites. The logic is: if the absolute number of "mismatched construction sites" is too large, it means that the problem is very common, and there is no need to look at other details. It can be directly determined that the system is unreliable.

[0068] Preset threshold for the number of construction sites with matching deviation: a critical value used to determine whether the problem is "abundant" based on quantity. For example, if there are a total of 100 construction sites, the threshold is set to 20. If more than 20 of them are problematic construction sites, the system is directly judged to be unreliable.

[0069] S432 determines whether the total number of maintenance personnel in all matching deviation sites is greater than the preset maintenance personnel number threshold based on the number of maintenance personnel in different matching deviation sites. If so, it is determined that the reliability of the early warning processing does not meet the requirements. If not, it proceeds to the next step. Specifically, based on the total number of affected personnel, even if the number of problematic construction sites is not large, if these problematic construction sites are all large-scale construction sites and the total number of affected personnel is very large, then the entire system still faces a high risk and is unreliable.

[0070] Preset maintenance personnel number threshold: a critical value used to judge whether the overall impact is too large based on the number of personnel. For example, if there are a total of 1000 workers, the threshold is set to 200. If the total number of workers at all problem sites exceeds 200, the system is considered unreliable.

[0071] S433 determines whether there are any construction sites with matching deviations whose matching ratio is less than a preset threshold based on the identification matching ratio of different construction sites with matching deviations. If so, it proceeds to the next step; otherwise, it determines that the reliability of the early warning processing meets the requirements. S434 identifies matching deviation construction sites with matching ratios less than a preset ratio threshold as target deviation construction sites, and determines whether the number of target deviation construction sites is greater than the preset deviation construction site threshold. If so, it determines that the reliability of the early warning processing does not meet the requirements; otherwise, it proceeds to the next step. In the above steps, based on the severity of the problem, the logic is as follows: the first two steps look at the "surface," while this step delves into the severity of the "points." The system needs to identify construction sites with particularly severe problems (i.e., "target deviation sites" with extremely low monitoring density). If there are many such "severe" construction sites, it also indicates that the system is unreliable.

[0072] Preset ratio threshold: A lower and more stringent ratio than the "match deviation" criterion (preset matching threshold). Used to identify construction sites with extremely limited monitoring resources. For example, the threshold for "match deviation" is 0.5, while the threshold for "target deviation construction site" can be set to 0.2.

[0073] Target Deviation Sites: Among the "Matching Deviation Sites", a subset of sites with particularly serious problems (identifying matching ratios < preset ratio thresholds). Preset Deviation Site Threshold: A critical value used to determine whether the number of "severe" sites is excessive.

[0074] S435 determines the influence weighting factor of different matching deviation sites based on the number of maintenance personnel and the identification matching ratio at different matching deviation sites, and determines whether the reliability of the early warning processing meets the requirements based on the sum of the influence weighting factors of different matching deviation sites.

[0075] Of the steps described above, the weighted comprehensive impact assessment is the most refined. If the system passes the first four steps of screening, it means the problem is neither very widespread nor has a huge impact, and there are not many severely affected construction sites. However, to make a final judgment, the impact of each problematic construction site needs to be comprehensively considered. A construction site with many workers and an extremely low identification ratio has a greater impact than a construction site with fewer workers and a slightly lower identification ratio.

[0076] Definition: Influence weighting factor: An indicator that quantifies the negative impact of each "matching deviation site" on the overall reliability of the system.

[0077] Calculation logic: The influence weight factor is directly proportional to the number of maintenance personnel and inversely proportional to the identification matching ratio. The more personnel and the lower the identification ratio, the larger the weight factor. Example formula: Influence weight factor = Number of maintenance personnel * (1 - Identification matching ratio). Example: Site A has 10 people and an identification matching ratio of 0.3; Site B has 20 people and an identification matching ratio of 0.4. The weight of Site A = 10 * (1 - 0.3) multiplied by 0.1 = 0.7, and the weight of Site B = 20 * (1 - 0.4) multiplied by 0.1 = 1.2. Preset weighting factor threshold: The upper limit of tolerance for the sum of the weighted impact of all problematic construction sites. If the total weight exceeds this threshold, it indicates that the overall risk of these construction sites is unacceptably high, and the system is judged as unreliable.

[0078] It is understandable that when the sum of the influence weight factors of different matching deviation construction sites is greater than the preset weight factor threshold, for example, greater than 5, then the reliability of the early warning processing is determined to be unsatisfactory.

[0079] Specifically, the method for determining the control strategy for identifying and processing all behavior recording terminals is as follows: For construction sites that have been identified as "matching deviation type" (monitoring resources are scarce), how can we intelligently identify violations by all terminals during certain time periods under limited computing resources, so that we can periodically switch between different behavior recording terminals, that is, determine the behavior recording terminal for violation identification and processing, thereby ensuring the reliability of violation identification and processing.

[0080] Scenario setting: A maintenance site for a large substation (matching deviation site) Site Overview: A large substation maintenance site with complex conditions and high risks. Number of maintenance personnel: 20. Total number of behavior recording terminals: 10 (e.g., 10 smart safety helmets). Current baseline recognition and processing strategy: To save computing power, the company stipulates that only 30% of the terminals are used for real-time AI recognition at this site.

[0081] Therefore: The current number of recognition terminals is 10 (total) * 30% = 3 (only 3 terminals are used for AI recognition), the current recognition matching ratio is 3 / 20 = 0.15, and the matching factor (total number of terminals / total number of personnel) is 10 / 20 = 0.5. The decision thresholds set by management (remain unchanged): preset identification matching ratio threshold: 0.3, preset matching factor threshold: 0.6, preset idle identification ratio threshold: 1.0.

[0082] S51 uses the configuration data of the behavior recording terminals at the maintenance site of the matching deviation type to determine the total number of behavior recording terminals in the maintenance site of the matching deviation type minus the number of identification terminals, and uses this as the number of idle terminals. In the above steps, the number of idle terminals is calculated. The action is: Number of idle terminals = Total number of behavior recording terminals (10) - Number of identification terminals (3) = 7. First, it is necessary to determine how much "backup power" is available. When the proportion of activated terminals is low (30%), there will be more idle resources, and the flexibility and potential of system resource calls will be greater.

[0083] Number of idle terminals: refers to the number of terminals deployed on the construction site but not used for AI recognition. These are reserve resources for improving monitoring levels.

[0084] S52 determines the identification matching ratio of the maintenance site of the matching deviation type according to the benchmark identification processing strategy; Step S52: Obtain the current identification matching ratio. Action: Confirm that the current identification matching ratio of the construction site is 0.15. The identification matching ratio has become very low (0.15), which greatly increases the monitoring blind spots and the risk of missed reports. This data is a key signal to trigger a high-intensity monitoring strategy.

[0085] S53 determines a control strategy for identifying and processing all behavior recording terminals in the maintenance site of the matching deviation type based on the number of idle terminals and the identification matching ratio.

[0086] Specifically, based on the number of idle terminals and the identification matching ratio in the maintenance site of the matching deviation type, a control strategy for identifying and processing all behavior recording terminals in the maintenance site of the matching deviation type is determined, including: S531 obtains the identification matching ratio of the maintenance site of the matching deviation type, and determines whether the identification matching ratio of the maintenance site of the matching deviation type is less than the preset identification matching ratio threshold. If so, the preset time period is used to identify all behavior recording terminals in the maintenance site of the matching deviation type, that is, to identify and process the violation behavior in all behavior distance terminals. If not, proceed to the next step. In the above steps, the first level judgment - based on the identification matching ratio (risk urgency judgment), action: judge whether the identification matching ratio (0.15) < the preset identification matching ratio threshold (0.3)? Yes.

[0087] Decision: Immediately adopt a preset time period (e.g., every 10 minutes, identify and process violations of all behavior recording terminals at the maintenance site for 1 minute), and identify and process all (10) behavior recording terminals at the site.

[0088] The current monitoring density (0.15) is far below the safety threshold (0.3), and the site is in a highly dangerous state. The system must immediately activate the highest level of response, disregarding cost, to ensure safety at all costs. This is a "red alert" emergency response.

[0089] Preset time period: After each preset time period, the system will automatically upgrade all terminals to full power recognition mode and continue to run for a fixed period of time.

[0090] S532 determines whether the matching factor of the records in the maintenance site of the matching deviation type is not greater than the preset matching factor threshold based on the matching factor of the records in the maintenance site of the matching deviation type. If so, the behavior recording terminals in the maintenance site of the matching deviation type are identified and processed using a preset time period, that is, the violation behavior is identified and processed in all behavior distance terminals. If not, proceed to the next step. Hypothetical deduction: If the first-level judgment is not triggered (e.g., the threshold is 0.1), proceed to S532. In this step, the second-level judgment - based on the record matching factor (resource inherent sufficiency judgment) - action: Determine if the record matching factor (0.5) ≤ the preset matching factor threshold (0.6)? Yes.

[0091] Decision: Using the same preset time period... all behavior recording terminals are identified and processed, indicating that the construction site is inherently "resource-scarce". Even with 100% deployment of all terminals, its resource level (0.5) is still below the health standard (0.6). Therefore, it needs to be regularly scanned comprehensively to compensate for its inherent deficiencies.

[0092] S533 determines a control strategy for identifying and processing all behavior recording terminals in the maintenance site of the matching deviation type based on the number of idle terminals in the maintenance site of the matching deviation type.

[0093] It is understood that, based on the number of idle terminals in the maintenance sites of the aforementioned matching deviation type, a control strategy for identifying and processing all behavior recording terminals in the aforementioned matching deviation type of maintenance sites is determined, specifically including: Based on the ratio of the number of idle terminals to the number of identification terminals in the maintenance site of the matching deviation type, an idle identification ratio is determined. When the idle identification ratio is less than a preset threshold, the control strategy for identifying all behavior recording terminals in the maintenance site of the matching deviation type is determined to be that if any identification terminal has a violation, all behavior recording terminals are identified, that is, the violation is identified among all behavior distance terminals. Otherwise, a preset time period is used to identify all behavior recording terminals in the maintenance site of the matching deviation type, that is, the violation is identified among all behavior distance terminals.

[0094] Hypothetical scenario: If the second-level judgment is not triggered (e.g., the preset matching factor threshold is 0.4), the third-level judgment - based on the idle identification ratio (resource call cost-effectiveness judgment), action: calculate the idle identification ratio = number of idle terminals (7) / number of identified terminals (3) ≈ 2.33, determine if the idle identification ratio (2.33) < preset threshold (1.0)? No, decision: use the preset time period... to identify all behavior recording terminals. (i.e., use periodic full-power scanning).

[0095] In this example, the idle detection ratio (2.33) is much greater than the threshold (1.0). This means that there are abundant available idle resources, and the monitoring equipment is idle at this time. Therefore, in order to improve the reliability of violation detection and processing, it is necessary to perform a full-range scan regularly.

[0096] In contrast, if the idle identification ratio is less than 1.0 (for example, 2 idle devices and 3 in use, with a ratio of 0.67), the system will be more stingy and will only start full-power identification when an event is triggered (such as when a violation is detected). Since the number of monitoring devices in an idle state is relatively small, in order to improve the reliability of violation identification and processing, the system can only perform violation identification and processing of all behavior recording terminals when a violation exists.

[0097] Idle Identification Ratio: Number of idle terminals / Number of identification terminals. The higher this ratio, the lower the marginal cost of upgrading the monitoring level and the more feasible it is.

[0098] It is understandable that, after identifying and processing violations among all the behavioral distance terminals, the behavioral recording terminal with the largest number of people exhibiting violations, and the number of such terminals that require violation identification and processing, is selected as the behavioral recording terminal for violation identification and processing.

[0099] In another possible implementation: After 30 minutes of full-power identification, the system analyzes the data. It finds that among the 10 terminals, three terminals with the highest number of violations and the most concentrated high-risk personnel coverage can be selected as the new terminals for violation identification. Additionally, two terminals (Terminal A and Terminal B) can also have the highest number of violations and the most concentrated high-risk personnel coverage. The decision is: the system will not randomly restore the three terminals for identification, but will prioritize retaining these two "highest value" terminals (A and B), and then select one more from the remaining terminals (e.g., Terminal C with the widest field of view) to form a new, more efficient three-terminal identification combination.

[0100] This is data-driven, precise resource allocation. The system conducted a comprehensive survey through full-power identification, learning which monitoring points were most efficient. After resuming regular monitoring, it can precisely allocate its limited computing power (slots for 3 terminals) to the most efficient points, maximizing monitoring effectiveness.

[0101] This process, as a closed-loop optimization process, enables the system's resource allocation strategy to continuously evolve based on actual monitoring results, thereby achieving continuous performance improvement.

[0102] Specifically, the method for determining the early warning processing method of the behavior recording terminal at the maintenance site is as follows: In this embodiment, the system's own adjustment history is analyzed to determine whether the current monitoring layout is stable, and the final monitoring scheme is determined accordingly, thus determining the final "early warning handling method" for a maintenance site.

[0103] Application Scenarios and Background Information Continue to monitor that substation maintenance site. In the previous process: because it was a "matching deviation type" and the identification matching ratio was very low, the system control strategy was set to: "use a preset time period (e.g., 10 minutes) to identify and process all behavior recording terminals" (i.e., full-power identification triggered by S531). After the full-power identification was completed, the system performed optimization, switched the identification target, and selected the three most efficient terminals (let's assume they are terminals A, B, and C) as the new regular identification terminals.

[0104] Now, the system needs to develop a long-term, stable early warning and handling method based on the frequency and content of such "switching" in recent times.

[0105] S61 determines the number of switching processes within the most recent preset time period based on the switching data of the target of the identification and processing of violations by the behavior recording terminal. Specifically, analyzing the switching frequency and actions: the system query found that in the most recent 2 hours (preset duration), due to changes in the on-site work area and the continuous learning of AI, the system actively adjusted (switched) the identified target combination for the construction site 3 times.

[0106] Switching frequency is a primary indicator of the stability of a monitoring environment. Frequent switching indicates rapid changes in the field, and fixed monitoring points may not remain effective indefinitely.

[0107] Switching process: refers to the decision-making and execution process by which the system reselects and determines which behavior recording terminals will perform AI recognition tasks in order to optimize resources.

[0108] Preset duration: The length of the historical time period used for backtracking and analysis, such as 2 hours.

[0109] S62 determines the target for identifying and processing violations that has changed between the previous handover process based on the change data of the target for identifying and processing violations between different handover processes, and uses it as the target for identifying changes. In one possible embodiment, the changes and actions are identified as follows: The system analyzes these three handovers. Findings: Handover process 1: Terminals [A, B, D] are changed to [A, B, C]. (Change identification targets: C and D); Handover process 2: Terminals [A, B, C] are changed to [B, C, E]. (Change identification targets: A and E); Handover process 3: Terminals [B, C, E] are changed to [A, C, F]. (Change identification targets: A, B, E, F) It's not just about the number of switches, but also the "operational scope" of each switch. Significant changes in a large number of targets are more telling than minor adjustments.

[0110] Change identification target: Terminals newly added or removed from the identification task list during two adjacent handovers. It reflects the magnitude of changes in the monitoring layout.

[0111] S63 identifies the target based on the number of switching processes within the most recent preset time period and the changes in different switching processes, and determines the early warning processing method for the behavior recording terminal of the maintenance site in conjunction with the control strategy of the maintenance site.

[0112] Furthermore, based on the number of switching processes within the most recent preset time period and the target identification of changes in different switching processes, and in conjunction with the control strategy of the maintenance site, a warning processing method for the behavior recording terminal of the maintenance site is determined, specifically including: S631 obtains the number of switching processes within the most recent preset time period, and determines whether the number of switching processes within the most recent preset time period is greater than the preset switching process number threshold. If so, the switching is too frequent. In order to ensure the reliability of the identification and processing, the early warning processing method of the behavior recording terminal at the maintenance site is determined to identify and warn of violations of all behavior recording terminals. If not, proceed to the next step. Specifically, a system disturbance assessment was conducted—was the switching too frequent? The system reviewed the historical records of the substation construction site over the past two hours. In order to continuously track the optimal monitoring point, the AI ​​system proactively switched the "identification and processing target" three times.

[0113] Judgment: The number of switching processes (3) is not greater than the preset threshold for the number of switching processes (assumed to be 4). This is a macro-frequency stability assessment of the system's "adjustment behavior". Switching itself implies that the strategy is unstable, but a reasonable system must allow for a certain amount of adjustment space. This step is to determine whether the adjustment has "gone out of control".

[0114] A threshold of 4 means the system allows no more than 4 strategy optimizations per day. Three optimizations are within the allowed range, indicating that while the system is adjusting, it hasn't reached the level of "frequent fluctuations," therefore the changes in violations are not drastic, and the system needs to proceed to the next step.

[0115] S632 takes the switching process where the number of changing targets exceeds the preset threshold for the number of changing targets as the changing switching process, and determines whether the number of changing switching processes exceeds the preset threshold for the number of changing processes. If so, the changing of the target for identifying and processing violations is too frequent. Therefore, in order to ensure the reliability of the identification and processing, the early warning processing method of the behavior recording terminal at the maintenance site is determined to identify and issue early warnings for violations on all behavior recording terminals. If not, proceed to the next step. S632: Persistence Assessment of Changes - Has Change Become the Norm? Based on the selected "effective switching", that is, the number of switching processes with no less than 2 change switching targets, which is 3 at this time, the system counts the number of such switching processes, that is, the "number of change switching processes", and judges: the number of change switching processes (3) > the preset threshold for the number of change processes (assuming it is 2)? Yes.

[0116] This is a quantitative frequency assessment, while S632 determines whether "there are too many changes". Even if the total number of switching (S631) is acceptable, if the vast majority of them are "valid changes", it means that the system's monitoring strategy cannot be fixed and is always in a state of searching and adjustment.

[0117] This reveals a deep-seated instability in the system. It indicates that the dynamic complexity of the construction site means that any locally optimized monitoring strategy can only be effective temporarily, and the system cannot find a consistently stable monitoring solution. In this situation, continuing to rely on unstable local monitoring will lead to a significant risk of missed detections.

[0118] Decision: Condition 3 > 2 is met. The system determines that the current monitoring strategy has changed too frequently and has lost its reliability. To ensure security, the most conservative but comprehensive solution must be adopted. Therefore, the final warning processing method is determined to be: identify and issue warnings for violations on all behavior recording terminals.

[0119] S633 determines the target change coefficient of the maintenance site based on the number of change switching processes and the number of change identification targets of different change switching processes within the most recent preset time period, and determines whether the target change coefficient of the maintenance site is greater than the preset change coefficient threshold. If yes, proceed to the next step; if no, determine that the warning processing method of the behavior recording terminal of the maintenance site is to use the benchmark identification processing strategy to identify and warn of violations. In another embodiment, assume that the system has performed 3 handovers in the past 3 hours. Handover details are as follows: Switch 1: Target changed from [Terminal A, B, C] to [A, B, D]. (Only one target changed). Switch 2: Target changed from [A, B, D] to [A, B, C]. (Only one target changed; actually, it switched back). Switch 3: Target changed from [A, B, C] to [A, C, D]. (One target changed). In the steps described above, the severity of change is assessed—monitoring the fluctuation range of the layout. The system is no longer satisfied with simply determining "whether there has been a change," but begins to measure the severity of each change. It calculates a new metric: the target change coefficient.

[0120] Calculation method: Target change coefficient = (Sum of the number of changed targets identified during all change switching processes) / (Number of change switching processes × Number of identification terminals); In this example, the calculations are as follows: Number of targets changed in Switch 1: 1 (D replaces C), Number of targets changed in Switch 2: 1 (C replaces D), Number of targets changed in Switch 3: 1 (D replaces B), Total = 1 + 1 + 1 = 3; Number of identification terminals = 3, target variation coefficient = 3 / (3 × 3) = 0.33; Judgment: Is the target variation coefficient (0.33) greater than the preset variation coefficient threshold (assumed to be 0.5)? No.

[0121] This is a quantitative assessment of the scope of impact. S633 indicates frequent changes, while S634 indicates that each change is minor. In this example, a coefficient of 0.33 means that on average, only about 1 / 3 of the monitoring points are replaced each time, which is considered a "fine-tuning" rather than a "reconstruction." The core of the system's monitoring layout (Terminal A) is stable.

[0122] Significance: This step can identify "frequent but mild" adjustment patterns. This pattern may be optimizing around a relatively stable core, and the reliability of the overall monitoring strategy is still acceptable, without necessarily needing to escalate to a "full coverage" ultimate solution.

[0123] Decision: Condition not met (0.33 < 0.5). The system determines the change is within acceptable limits. Therefore, the warning handling method is determined to be: using a baseline identification and handling strategy (i.e., 30% ratio) to identify and issue warnings for violations. The process ends here.

[0124] S634 acquires the control strategy of the maintenance site, and determines whether the control strategy of the maintenance site is to identify and process all behavior recording terminals in the maintenance site using a preset time period. If yes, then the behavior recording terminals in the maintenance site using a third preset proportion are used for early warning identification processing of violations. If no, then in order to ensure the reliability of the identification processing, the early warning processing method of the behavior recording terminals in the maintenance site is determined to be to identify and issue early warnings for violations of all behavior recording terminals.

[0125] (Another deduction path) Assuming that the calculated target variation coefficient in S634 is 0.67 (greater than the threshold of 0.5), then proceed to the final step S635.

[0126] Strategy Coordination Assessment – ​​In conjunction with control strategies, the system reads the currently effective control strategies for the construction site.

[0127] Judgment: Determine whether the control strategy of the maintenance site is "to identify and process all behavior recording terminals in the maintenance site using a preset time period"? This represents the ultimate synergy between historical dynamic analysis and current static configuration. If the system history exhibits high instability (triggered by S634), but the system has already configured periodic "full power" identification for the construction site as a safety fallback (control strategy), it indicates that the system has recognized the problem and taken compensatory measures.

[0128] In this situation, a compromise can be chosen for the final early warning handling method. Neither extremely costly continuous full coverage (the early warning handling method) nor high-risk low-proportion monitoring (the baseline strategy) can be adopted; instead, a high-proportion continuous monitoring approach can be selected as the new early warning solution. This reflects the hierarchical and economical nature of system decision-making.

[0129] Furthermore, if the third preset ratio is greater than the second preset ratio, and if so, the final warning processing method is to "use the behavior recording terminals of the third preset ratio (e.g., 60%) to perform warning identification processing for violations." This is an optimized balance between ensuring security and controlling costs. If not, it indicates that the system has a history of instability and currently lacks a security fallback plan, posing the highest risk. Therefore, it is necessary to "identify and issue warnings for violations on all behavior recording terminals."

[0130] Example 2 On the other hand, 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 image recognition-based violation warning processing method when running the computer program.

[0131] 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.

[0132] 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.

[0133] 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. An image recognition-based violation pre-warning processing method, characterized in that, Specifically comprising: Determine the distribution data of the behavior recording terminal based on the maintenance site data, determine the reference identification processing strategy of the behavior recording terminal in the maintenance site based on the distribution data of the behavior recording terminal and the maintenance personnel data in different maintenance sites, determine the early warning processing matching type of the violation behavior of the maintenance site based on the reference identification processing strategy, and when the reliability of the early warning processing does not meet the requirements based on the maintenance personnel data in different early warning processing matching types, proceed to the next step; When the early warning processing matching type of the violation behavior of the maintenance site is the matching deviation type, determine the control strategy of identifying all behavior recording terminals in the maintenance site of the matching deviation type according to the configuration data of the behavior recording terminal of the matching deviation type of the maintenance site and the reference identification processing strategy; Determine the early warning processing method of the behavior recording terminal of the maintenance site based on the switching data of the identification processing target of the violation behavior of the behavior recording terminal and the control strategy.

2. The image recognition-based violation pre-warning processing method of claim 1, wherein, The distribution data of the behavior recording terminal includes the number of behavior recording terminals configured in different maintenance sites.

3. The image recognition-based violation pre-warning processing method of claim 1, wherein, The maintenance personnel data in the maintenance site includes the number of maintenance personnel in the maintenance site.

4. The image recognition-based violation pre-warning processing method of claim 1, wherein, The method for determining the reference identification processing strategy of the behavior recording terminal in the maintenance site is: Determine the number of behavior recording terminals in different maintenance sites based on the distribution data of the behavior recording terminal; Determine the number of maintenance personnel in different maintenance sites based on the maintenance personnel data in different maintenance sites; Determine the reference identification processing strategy of the behavior recording terminal in the maintenance site according to the number of behavior recording terminals and the number of maintenance personnel in different maintenance sites.

5. The image recognition-based violation pre-warning processing method of claim 4, wherein, Determine the reference identification processing strategy of the behavior recording terminal in the maintenance site according to the number of behavior recording terminals and the number of maintenance personnel in different maintenance sites, specifically comprising: Determine the record matching factor in different maintenance sites based on the ratio of the number of behavior recording terminals to the number of maintenance personnel in different maintenance sites; Determine the reference identification processing strategy of the behavior recording terminal in the maintenance site based on the record matching factor.

6. The image recognition-based violation pre-warning processing method of claim 1, wherein, The method for determining the early warning processing matching type of the violation behavior of the maintenance site is: Take the number of behavior recording terminals in the maintenance site that need to be identified as the identification terminal number; Take the ratio of the identification terminal number to the number of maintenance personnel in the maintenance site as the identification matching ratio; Determine the early warning processing matching type of the violation behavior of the maintenance site based on the identification matching ratio.

7. The image recognition-based violation pre-warning processing method of claim 6, wherein, When the identification matching ratio of the maintenance site is greater than a preset identification matching threshold, it is determined that the early warning processing matching type of the violation behavior of the maintenance site is the identification reliable type.

8. The image recognition-based violation pre-warning processing method of claim 1, wherein, The method for determining the early warning processing method of the behavior recording terminal of the maintenance site is: Determine the number of switching processes in the recent preset time period based on the switching data of the identification processing target of the violation behavior of the behavior recording terminal; According to the change data of the identification processing target of the violation behavior between different switching processes, determine the identification processing target of the violation behavior changed between the last switching process, and take it as the change identification target; According to the number of switching processes in the recent preset time period and the change identification target of different switching processes, and combined with the control strategy of the repair site, determine the early warning processing method of the behavior record terminal of the repair site.

9. The image recognition-based violation pre-warning processing method of claim 8, wherein, According to the number of switching processes in the recent preset time period and the change identification target of different switching processes, and combined with the control strategy of the repair site, determine the early warning processing method of the behavior record terminal of the repair site, specifically including: Obtain the number of switching processes in the recent preset time period, and when the number of switching processes in the recent preset time period is greater than the preset switching process quantity threshold, determine the early warning processing method of the behavior record terminal of the repair site as identifying and warning all behavior record terminals of violation behavior.

10. A computer system comprising: The memory and processor connected by communication, and the computer program stored on the memory and capable of running on the processor, characterized in that the processor executes the computer program to perform the image recognition-based violation behavior early warning processing method of any one of claims 1-9.

Images