An alarm method, system and storage medium based on dynamic configuration management caused by business rules

Abstract

The present invention provides an alarm method based on dynamic configuration management caused by business rules. According to multiple national standards, this method establishes a knowledge base (including a thesaurus) for urban governance alarm services and a business model, realizes the urban governance AI alarm decision-making management driven by a business rule engine, and provides application functions such as business model, rule definition, rule strategy, and rule automatic adaptation. It also provides the above-mentioned system and storage medium based on dynamic configuration management caused by business rules, eliminates alarm storms, improves the accuracy and reliability of alarms, and reduces false alarms and missed alarms.

Description

Technical Field

[0001] The present invention belongs to the technical field of intelligent communication technology and equipment, and in particular relates to an alarm method, system and storage medium for dynamic configuration management caused by business rules. Background Art

[0002] In recent years, with the rapid development of science and technology, especially the continuous breakthroughs in artificial intelligence technology, video intelligent analysis technology has become an indispensable and important driver of urban governance modernization. However, with the widespread application of AI video surveillance systems in urban governance, the massive alarm data generated by the system is mixed with a large amount of invalid and repeated "noise", forming an "alarm peak", drowning out the real emergency events that need attention, forcing monitoring personnel to spend a lot of time screening and verifying alarms.

[0003] Because the AI ​​video surveillance system still has major deficiencies in recognition accuracy, which not only affects the overall credibility of the system but may also lead to incorrect handling decisions, the lack of deep integration with specific business scenarios limits the full release of AI-assisted decision-making potential, making it difficult to meet the increasingly complex needs of urban governance, and invisibly adding extra burdens to managers and front-line workers. Now let's give examples to illustrate the existing problems.

[0004] (1) Technology does not match actual business needs, technology is separated from business, and practicality is insufficient. AI technology that is separated from application scenarios may not be able to effectively solve practical problems in urban governance. For example, some intelligent video recognition systems may be very advanced in theory, but in actual operation, they are difficult to play a role due to the lack of deep integration with specific scenarios.

[0005] (2) Low decision-making efficiency and poor algorithm adaptability. AI algorithms that are out of touch with the application scenario may not be able to adapt to the complex and changing environment and needs of urban governance. For example, some algorithms may perform well under certain conditions, but may fail or misjudge under other conditions.

[0006] (3) Alarm recognition is limited to a single technology. For example, an alarm is triggered when a single video is intelligently recognized, or when the IoT sensing water level exceeds a threshold. There is no integration of multiple factors, and no integration of multi-modal fusion technologies such as video, audio, IoT sensing, and VR / AR. For example:

[0007] Example 1: For example, if someone is eating while waiting for a bus with a suitcase on the roadside, he or she will be identified as a traveling merchant.

[0008] Example 2: Someone in a residential community poured a bucket of water at a waterlogging monitoring point (where a water level meter was installed), causing a false flood control and drainage warning. Summary of the invention

[0009] Technical solution: To solve the above technical problems, the present invention provides an alarm method based on dynamic configuration management caused by business rules, and the method includes:

[0010] Establish a knowledge base for urban governance alarm services according to known national standards;

[0011] Construct a business model supported by historical alarm event features, and define rules, rule strategies, and rule self-adaptation technologies; perform feature extraction on actual urban governance business scenarios, and compare the features with the business scenario data in the urban governance alarm service knowledge base. Perform alarm filtering according to the rule definition, rule strategy, and rule self-adaptation method through business rule self-adaptation, and issue alarm self-healing processing, alarm prediction processing, and decision analysis visualization processing.

[0012] As an improvement, the alarm filtering includes the following methods:

[0013] (1) Preprocess the alarm classification, adopt a high-aggregation intelligent algorithm, extract the key feature information of the alarm, perform algorithm clustering and noise reduction on the alarm, associate the key feature information with the clustering result, and compress subsequent similar alarms into the alarms that have not been closed yet;

[0014] (2) Based on time series, compress the same event alarms, or based on classification algorithms, automatically label and classify the alarm information, and perform multi-dimensional alarm compression, merging, and noise reduction.

[0015] As an improvement, the business rule self-adaptation includes an identification and discrimination processing method based on the matching degree decision method and an alarm method based on the set priority level. The identification and discrimination processing method based on the matching degree decision method is used for training the historical alarm event features of the business model when there are multiple similar business scenarios under different business rules; the alarm method based on the set priority level is a method of automatically learning through historical alarm data features, alarm handling situations, and industry business rule knowledge, and automatically generating the priority ranking of business rules according to the nine-square grid method.

[0016] As an improvement, the specific steps of the matching degree decision method are:

[0017] (1) Define a matching degree index to measure the matching degree between the rule and the current event;

[0018] (2) Suppose there are two rules R1 and R2, which have a set of conditions C1 = {c11, c12,...} and C2 = {c21, c22,...}, respectively, and the corresponding matching degree weights W1 = {w11, w12,...} and W2 = {w21, w22,...};

[0019] (3) For each rule Ri, calculate the sum of the weighted matching degrees of all its conditions;

[0020] (4) Considering priority, matching degree, and the freshness of the rule comprehensively, select the rule with the highest evaluation score for execution. The formula for comprehensive consideration is as follows:

[0021] EvaluationScore(Ri) = α × Priority(Ri) + β × MatchScore(Ri) + γ × Freshness(Ri)

[0022] Among them, α, β, and γ are weight coefficients; Priority(Ri) is the priority of rule Ri, MatchScore(Ri) is the matching degree of rule Ri with the current alarm event, Freshness(Ri) is the freshness of rule Ri, and EvaluationScore(Ri) is the evaluation score of rule Ri.

[0023] As an improvement, the self - adaptation of business rules includes a multi - modal fusion decision - making strategy. When dealing with the fusion of multi - source heterogeneous alarm data, through feature extraction and fusion, a single - body model is established to identify alarms. A multi - modal fusion decision - making strategy based on the weighted voting method is adopted to complete the final adaptive alarm processing. Among them, the multi - source heterogeneous alarm data includes video image recognition alarms, audio recognition (noise index, sensitive sounds), Internet of Things perception (the physical quantity value measured by sensors exceeds the alarm threshold), and geospatial coordinates (alarm - sensitive locations).

[0024] As an improvement, the specific steps for alarm processing of fusing multi - source heterogeneous alarm data are as follows:

[0025] (1) Determine the alarm category

[0026] Suppose there are N alarm categories, denoted as C1, C2, …, CN respectively;

[0027] (2) Collect modal alarm results

[0028] Set that for each modality j, there is a corresponding alarm result vector v j = [v j 1, v j 2, …, v j N], where the support degree v j , where n is the number of modalities, and v j represents the support degree of modality j for alarm category Ci, which is a probability value from 0 to 1;

[0029] (3) Determine modal weights

[0030] Assign a weight w j to each modality j, where the weight w jis a non - negative real number, and the sum of the weights of all modalities is usually normalized to 1;

[0031] (4) Calculate the weighted sum

[0032] For each alarm category Ci, calculate its weighted sum Si, and the calculation formula is as follows:

[0033]

[0034] where Si reflects the comprehensive support degree of all modalities for the alarm category Ci; n is the number of modalities;

[0035] (5) Compare the weighted sums

[0036] Among all alarm categories C1, C2,..., CN, compare their weighted sums S1, S2,..., SN;

[0037] (6) Select the maximum weighted sum

[0038] Select the alarm category with the maximum weighted sum as the final decision, that is, the final decision is that this event should be the Ci alarm category that makes Si reach the maximum value.

[0039] As an improvement, the specific steps of alarm self - healing processing are based on the self - healing rules in the urban governance business rule engine. After a non - urgent urban governance alarm occurs, the system will automatically adapt the self - healing disposal rules, trigger the corresponding self - healing process, achieve alarm self - healing, and then verify whether the alarm exists through AI vision recognition technology. If it is determined that there is no problem with the self - healing, the alarm is closed.

[0040] As an improvement, the specific steps of alarm prediction processing are as follows:

[0041] (1) After real - time alarm events are accessed, filter the alarm events to generate a real - time alarm knowledge graph;

[0042] (2) Construct an event graph model for urban governance, and use a graph database to store and manage the graph model; define each alarm event as a node in the graph model, including the detailed information of the alarm;

[0043] (3) In the event graph model, obtain the relevant historical alarm knowledge graph in (1), and use graph algorithms to calculate the similarity; the historical alarm knowledge graph with a similarity exceeding the set threshold is used as the candidate root cause graph;

[0044] (4) In the candidate root cause graph, through the association analysis algorithm, calculate the support degree and confidence degree of the alarm event indicators; sort the root causes according to the severity and influence range of the alarm root cause;

[0045] (5) Make an auxiliary decision - making analysis and an event handling plan.

[0046] Meanwhile, the present invention also provides an alarm system based on dynamic configuration management caused by business rules, including an alarm aggregation module, an alarm filtering module, and an alarm application scenario module;

[0047] The alarm aggregation module is used to collect alarm information through AI visual analysis alarms and multiple types of alarm methods;

[0048] The alarm filtering module is used to send decision-making alarm information according to any of the above alarm methods after receiving the information of the alarm aggregation module;

[0049] The alarm application scenario module is used to perform self-healing processing, alarm prediction processing, conversion into work order events, or analysis and visualization processing on the information sent by the alarm filtering module.

[0050] Meanwhile, the present invention also provides a storage medium, and the computer program stored in this storage medium can be executed by one or more processors and can be used to implement any of the above alarm methods.

[0051] Beneficial effects: The alarm method based on dynamic configuration management caused by business rules proposed by the present invention establishes a knowledge base (including a thesaurus) and a business model for urban governance alarm services according to multiple national standards, realizes urban governance AI alarm decision-making management driven by a business rule engine, and provides application functions such as business models, rule definitions, rule strategies, and rule automatic adaptation; eliminates alarm storms, improves the accuracy and reliability of alarms, and reduces false alarms and missed alarms.

[0052] Compared with conventional technologies, the alarm method proposed by the present invention has the following advantages:

[0053] (1) Through the aggregation of multi-source heterogeneous alarm data and standard alarm data management, it solves the problems of each scenario in urban governance being managed separately, different manufacturers' platforms being different, and alarm information being scattered and messy.

[0054] (2) Through the business rule engine and the use of alarm self-healing applications, the need for manual intervention is greatly reduced, the risk of human errors is lowered, and the alarm handling process becomes more efficient and accurate. While reducing the burden, it also brings significant efficiency improvement to users.

[0055] (3) Through the root cause analysis alarm prediction application of the graph algorithm, the source of the problem can be quickly located, reducing the situation of false alarms and missed alarms; through the learning and analysis of historical alarm data, the graph algorithm can discover the association rules and potential problems between alarms. These rules and patterns can help predict possible urban governance problems in the future, so as to take preventive measures in advance.

[0056] (4) Through the multi-modal fusion decision-making alarm application, the accuracy and reliability of the alarm can be improved, and false alarms and missed alarms can be reduced.

[0057] (5) The accurate positioning of the alarm location is realized through the application of the geospatial technology positioning function; the intelligent recognition technology of the AI vision technology can apply multiple algorithms for alarm at the same time, and obtain the actual alarm picture at the same time; the real-time and high-precision technical characteristics of the Internet of Things perception technology can detect alarms more real-time and with higher precision.

[0058] In the present invention, these multiple advanced technologies are integrated to accurately screen out effective alarms from a large amount of multi-source heterogeneous alarm data, reduce the occurrence of false alarms and missed alarms, improve the accuracy and effectiveness of the alarms, and bring a significant burden reduction and efficiency increase experience to urban governance staff. BRIEF DESCRIPTION OF THE DRAWINGS

[0059] Figure 1 It is a schematic diagram of the data flow of the system of the present invention.

[0060] Figure 2 It is a schematic diagram of the logical flow of the alarm method of the present invention.

[0061] Figure 3 It is a schematic flow chart of the self-adaptive technology method of the alarm method of the present invention.

[0062] Figure 4 It is a schematic flow chart of the management method based on the rules of multi-modal fusion decision-making in the alarm method of the present invention.

[0063] Figure 5 It is a flow chart of the alarm self-healing process in the alarm method of the present invention.

[0064] Figure 6 It is a flow chart of the alarm prediction in the alarm method of the present invention.

[0065] Figure 7 It is a schematic diagram of the multi-modal fusion decision-making scheme for fire recognition and fireworks display events in the embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0066] The technical solutions in the embodiments of the present invention will be clearly and completely described below, so that those skilled in the art can better understand the advantages and features of the present invention, and thus make a clearer definition of the protection scope of the present invention. The embodiments described in the present invention are only a part of the embodiments of the present invention, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0067] The alarm method for dynamic configuration management caused by business rules in the present invention implements two sets of alarm engines, namely a business rule engine and an AI algorithm engine; the alarm application is an application function of the alarm system, mainly including alarm filtering, intelligent noise reduction for priority alarms, multi-modal fusion decision-making alarms, alarm self-healing, i.e., automatic verification based on AI vision recognition technology, alarm prediction, i.e., root cause prediction analysis, and decision analysis visualization, etc.

[0068] See Figure 1 As shown, it is a data flow diagram of the alarm system for dynamic configuration management caused by business rules in the present invention. Specifically, the system includes an alarm aggregation module, an alarm filtering module, and an alarm application scenario module;

[0069] The alarm aggregation module is used to collect alarm information through AI vision analysis alarms and multiple types of alarm methods;

[0070] The alarm filtering module is used to send decision-making alarm information according to any of the above alarm methods after receiving the information from the alarm aggregation module;

[0071] The alarm application scenario module is used to perform in-system alarm self-healing processing, alarm prediction processing, conversion into work order events, or analysis visualization processing on the information sent by the alarm filtering module.

[0072] Further, see Figure 2 As shown, it is a schematic flowchart of the alarm method for dynamic configuration management caused by business rules in the present invention. The method includes:

[0073] Establish a knowledge base for urban governance alarm services according to known national standards; for example, "GB / T 30428.2-2013 Digital Urban Governance Information System - Part 2: Management Components and Events".

[0074] Construct a business model supported by historical alarm event features, and define rules, rule strategies, and rule self-adaptation technologies; perform feature extraction for actual urban governance business scenarios, and compare the features with the business scenario data in the urban governance alarm service knowledge base. Perform alarm filtering according to the rule definition, rule strategy, and rule self-adaptation method through business rule self-adaptation, and issue alarm self-healing processing, alarm prediction processing, and decision analysis visualization processing. The self-adaptation method is shown in Figure 3 As shown.

[0075] As a business rule self-adaptation technology, the method of setting priority levels for alarms is used when there are at least two business rules in the same business application scenario. In the case of alarm according to the set priority levels, specifically:

[0076] (1) Priority setting method: Automatically learn through historical alarm data characteristics, alarm handling situations, and industry business rule knowledge, and automatically generate the priority ranking of business rules according to the nine-square grid method:

[0077] ① Collect all alarm data generated by the system in the past period of time, and extract key characteristics such as alarm frequency, duration, and influence scope. These data are important bases for analyzing business rules.

[0078] ② Record information such as the handling process, handling results, and time consumption after each alarm occurs. Analyze the correlation relationships between different alarms, identify possible causal relationships or co-occurrence relationships, understand the triggering mechanism and influence path of alarms, and evaluate the emergency level and handling efficiency of different alarms.

[0079] ③ Obtain business rule data for the urban governance industry, including industry standards, best practices, policies and regulations, etc., and extract key characteristics such as alarm frequency, alarm emergency level, and influence scope. As a supplementary basis for the priority ranking of business rules.

[0080] ④ Use a classification algorithm to input the extracted feature data into the model for training, and then use the nine-square grid method to divide the rules into nine regions according to two key dimensions (such as emergency level and influence scope), and each region corresponds to a priority level.

[0081] ⑤ Define the priority level of each business rule according to the results output by the model and the division results of the nine-square grid method.

[0082] (2) Select the rule with the highest priority, execute the one with higher priority, and merge the ones with the same priority into the first one;

[0083] (3) Select the rules for comparing logical priorities, execute the one with higher priority, and merge the ones with the same priority under the first one.

[0084] Preferably, a method for setting the priority level for alarms, setting rules: Suppose there are two or more alarm judgment rules R1, R2, Rn in the same business scenario, and they are all triggered to respond to an event. Each rule has a priority attribute, denoted as P1, P2, Pn respectively.

[0085] Strategy: Select the rule with the highest priority to execute. If the priorities are the same, merge them into the first one to achieve alarm filtering and deduplication.

[0086] Comparison logic: If P1 > P2 > Pn. Keep P1 as the alarm result and eliminate other alarms; If P1 = P2 = Pn, then keep the P1 alarm and merge other alarms under P1.

[0087] Example 1: To implement, for example, intelligent video recognition of warnings about chefs not wearing uniforms properly and chefs not wearing masks, based on the "transparent kitchen" business rules, the two warnings can be combined into one warning.

[0088] Table 1 Example 1 of the method for issuing warnings according to the priority levels set in the present invention

[0089]

[0090]

[0091] As a business rule self - adaptation technology, it also includes a rule self - adaptation warning management method for identifying different rules of multiple similar business scenarios based on the matching degree. Specifically: when there are multiple similar business scenarios under different business rules, train the characteristics of historical warning events for business model learning, and perform identification and discrimination processing according to the matching degree decision method.

[0092] In urban governance warning events, there are many warning events with similar business characteristics but completely different regulatory disposals. Often, AI visual analysis warnings are prone to confusion. It is necessary to update the business model in a timely manner according to regulations and policies by learning the characteristics of historical warning events, and also self - adapt relevant rules. It may be necessary to accurately identify and distinguish based on multiple conditions, and the matching degree of each condition may be different.

[0093] Preferably, the specific steps of the matching degree decision method are as follows:

[0094] (1) Define a matching degree index to measure the matching degree between the rule and the current event;

[0095] (2) Suppose there are two rules R1 and R2, which have a set of conditions C1 = {c11, c12,...} and C2 = {c21, c22,...} respectively, and the corresponding matching degree weights W1 = {w11, w12,...} and W2 = {w21, w22,...};

[0096] (3) For each rule Ri, calculate the sum of the weighted matching degrees of all its conditions;

[0097] (4) Considering the priority, matching degree, and freshness of the rule comprehensively, select the rule with the highest evaluation score to execute. The comprehensive consideration formula is as follows:

[0098] EvaluationScore(Ri) = α × Priority(Ri)+β × MatchScore(Ri)+γ × Freshness(Ri)

[0099] Among them, α, β, and γ are weight coefficients, Priority(Ri) is the priority of rule Ri, MatchScore(Ri) is the matching degree of rule Ri with the current alarm event, Freshness(Ri) is the freshness of rule Ri, and EvaluationScore(Ri) is the evaluation score of rule Ri.

[0100] Furthermore, the evaluation strategy is to select the rule with the highest evaluation score for execution; Example 2 of this method is as follows:

[0101] For example, in street order management, there are three types of event alarms: out-of-store business operations, occupancy of roads for business, and unlicensed mobile vendors. These forms of business activities are very similar in rules, and conflicts are likely to occur in the automatic adaptation of rules. It is difficult to accurately determine which type of alarm it belongs to solely based on AI video recognition, and it is often misjudged. Moreover, the law enforcement management for the three is significantly different.

[0102] In the above Example 2, the similar features of the three are: unauthorized or out-of-scope business operations; visual activities: shelves, goods, customers; dynamic changes; occupation of public resources; affecting the city appearance; the differences between the three are shown in the following table.

[0103] Table 2 Differences among the three in Example 2 of the present invention

[0104]

[0105] As a business rule self-adaptation technology, it also includes a rule self-adaptation alarm management method based on multi-modal fusion decision-making. See Figure 4 As shown, when it is used for containing alarm data that fuses multi-source heterogeneous data, through feature extraction and fusion, a single model is established to identify alarms, and a multi-modal fusion decision-making strategy based on a weighted voting method is adopted to complete the final adaptation alarm processing. Among them, the multi-source heterogeneous alarm data includes video, audio, Internet of Things perception, and geospatial information.

[0106] The specific steps for alarm processing of fusing multi-source heterogeneous alarm data are as follows:

[0107] (1) Determine the alarm category

[0108] Suppose there are N alarm categories, which are represented by C1, C2,..., CN respectively;

[0109] (2) Collect modal alarm results

[0110] Set that for each modal j, there is a corresponding alarm result vector v j =[v j 1, v j 2,..., v j N], the support degree v j i, where n is the number of modalities, vj i represents the degree of support of modality j for alarm category Ci, which is a probability value from 0 to 1;

[0111] (3) Determine modality weights

[0112] Assign a weight w to each modality j j , where the weight w j is a non - negative real number, and the sum of the weights of all modalities is usually normalized to 1;

[0113] (4) Calculate the weighted sum

[0114] For each alarm category Ci, calculate its weighted sum Si, and the calculation formula is as follows:

[0115]

[0116] where Si reflects the comprehensive support degree of all modalities for alarm category Ci; n is the number of modalities;

[0117] (5) Compare the weighted sums

[0118] Among all alarm categories C1, C2, …, CN, compare their weighted sums S1, S2, …, SN;

[0119] (6) Select the maximum weighted sum

[0120] Select the alarm category with the maximum weighted sum as the final decision, that is, the final decision is that this event should be the Ci alarm category that makes Si reach the maximum value.

[0121] Preferably, for the rule - self - adapting alarm management method based on multi - modal fusion decision, a specific case:

[0122] For example, in the fusion process of video and audio signals in the [Fighting] scenario of urban governance, first perform feature extraction, feature fusion, and single - model training based on video signals and audio signals respectively, generate model - predicted alarms respectively, and finally use the weighted voting method to predict the results of each model to obtain a better final recognition result, and finally recognize whether it is roughhousing or fighting.

[0123] See Figure 5 As shown, based on the self - healing rules in the urban governance business rule engine, after an alarm in non - emergency urban governance occurs, the system will automatically adapt the self - healing disposal rules, trigger the corresponding self - healing process, achieve alarm self - healing, and then verify whether the alarm exists through AI vision recognition technology. If it is determined that there is no problem with self - healing, the alarm is closed.

[0124] For example, if the self-healing time for illegal parking of a motor vehicle is set to 15 minutes, the system will issue a voice message reminder and send a mobile phone text message to notify the vehicle owner to move the vehicle. After 15 minutes of warning, the system will automatically trigger the front-end to collect image data and use AI vision recognition technology to verify whether the warning exists. If it is found that the illegally parked vehicle has left, the system will automatically close the warning.

[0125] See Figure 6 As shown, it is a root cause analysis warning prediction based on graph algorithms. The specific steps for warning prediction processing are as follows:

[0126] (1) After real-time warning events are accessed, filter the warning events to generate a real-time warning knowledge graph;

[0127] (2) Build an event graph model for urban governance and use a graph database to store and manage the graph model; define each warning event as a node in the graph model, including detailed information about the warning, such as time, type, address, text description, etc.;

[0128] (3) Obtain the relevant historical warning knowledge graph in (1) in the event graph model and calculate the similarity using graph algorithms; the historical warning knowledge graph with a similarity exceeding the set threshold is used as a candidate root cause graph;

[0129] (4) In the candidate root cause graph, calculate the support and confidence of the warning event indicators through association analysis algorithms; sort the root causes according to the severity and impact range of the warning root cause;

[0130] (5) Make auxiliary decision-making analysis and event handling plans.

[0131] In the present invention, decision-making analysis visualization is the visualization of real-time warning filtering, warning self-healing, and warning prediction data, which helps to achieve more refined urban governance operation and management. Through rich ready-to-use multi-dimensional reports, unified analysis of all warning sources and warning management is realized, providing analysis of warnings and work efficiency improvement for business personnel and leaders.

[0132] Based on geospatial information, with the help of VR and AR technologies, the monitoring screen can be virtualized and augmented reality processed to provide more intuitive and comprehensive monitoring and warning information. Combined with Internet of Things perception warnings, multi-source warning data feature fusion is achieved, and weighted voting analysis is used for verification and decision-making to improve the accuracy and reliability of warnings and reduce false alarms and missed alarms. For example, just because a water level gauge senses that the water level is too high does not mean that a city waterlogging warning should be issued. Through video fusion technology, it can be identified whether the situation of too high water level at this location is caused by city waterlogging or an accidental water accumulation problem. For example, Figure 7 the multi-modal fusion decision-making scheme for fire recognition and fireworks display events in

[0133] See Figure 7As shown, for the present invention, firstly, data of image recognition alarms and smoke sensors are collected and transmitted to the fire and smoke recognition module as data for fire recognition; data of audio recognition alarms and geospatial information are collected and transmitted to the noise alarm, such as the key words being "so beautiful", fixed smoke release points, such as by the river, and the video fusion information module of VR and AR as data for fireworks shows.

[0134] Then, the data for fire recognition and the data for fireworks shows are combined into data for determining whether alarm filtering is required; finally, a determination is made on whether alarm processing is required.

[0135] The embodiments described above only represent several implementation manners of the present invention. The description is relatively specific and detailed, but it should not be construed as a limitation on the scope of the invention patent. It should be noted that for those of ordinary skill in the art, without departing from the concept of the present invention, several modifications and improvements can still be made, and these all fall within the protection scope of the present invention. Therefore, the protection scope of the present invention patent shall be subject to the appended claims.

Claims

1. An alarm method based on dynamic configuration management caused by business rules, characterized in that: The method includes establishing a knowledge base for urban governance warning services according to known national standards; constructing a business model supported by historical warning event features, and defining rules, rule strategies, and rule self-adaptation technologies; extracting features of the actual urban governance business scenario, comparing the features with the business scenario data in the urban governance warning service knowledge base, and performing warning filtering according to the rule definition, rule strategy, and rule self-adaptation method through business rule self-adaptation, and issuing warning self-healing processing, warning prediction processing, and decision analysis visualization processing; Business rule self-adaptation includes warning processing for integrating multi-source heterogeneous warning data. The specific steps for warning processing of integrating multi-source heterogeneous warning data are as follows: (1) Determine the warning category Suppose there are N warning categories, which are represented by C1, C2, …, CN respectively; (2) Collect modal warning results For each modality j, there is a corresponding alarm result vector v j = [v j 1, v j 2, …, v j i, …, v j N], where v j i represents the degree of support of modality j for the alarm category Ci, which is a probability value from 0 to 1; (3) Determine modal weights Assign each modality j a weight w j , where the weight w j is a non - negative real number, and the sum of the weights of all modalities is normalized to 1; (4) Calculate the weighted sum For each warning category Ci, calculate its weighted sum Si, and the calculation formula is as follows: where Si reflects the comprehensive support degree of all modalities for the warning category Ci; n is the number of modalities; (5) Compare the weighted sums Among all warning categories C1, C2, …, CN, compare their weighted sums S1, S2, …, SN; (6) Select the maximum weighted sum Select the warning category with the maximum weighted sum as the final decision, that is, the final decision is that this event should be the Ci warning category that makes Si reach the maximum value.

2. The warning method based on dynamic configuration management caused by business rules according to claim 1, characterized in that: Warning filtering includes the following methods: (1) Preprocess the warning classification, adopt a high-aggregation intelligent algorithm, extract the key feature information of the warning, perform algorithm clustering and noise reduction on the warning, associate the key feature information with the clustering result, and compress subsequent similar warnings into the still-unclosed warnings; (2) Based on time series, compress the same event warnings, or based on classification algorithms, automatically label and classify the warning information, and perform multi-dimensional warning compression, merging, and noise reduction.

3. The alarm method for dynamic configuration management caused by business rules according to claim 1, wherein: Business rule self-adaptation includes an identification and discrimination processing method based on a matching degree decision method and a method for warning according to a set priority level. The identification and discrimination processing method based on the matching degree decision method is used when there are multiple similar business scenarios under different business rules and for training the historical warning event features of the business model; the method for warning according to the set priority level is a method that automatically learns through historical warning data features, warning handling situations, and industry business rule knowledge, and automatically generates the priority ranking of business rules according to the nine-square grid method.

4. The alert method for dynamic configuration management caused by business rules according to claim 3, characterized in that: The specific steps of the matching degree decision method are as follows: (1) Define a matching degree index to measure the matching degree between the rule and the current event; (2) Suppose there are two rules R1 and R2, which have a set of conditions C1 = {c11, c12, …} and C2 = {c21, c22, …} respectively, and the corresponding matching degree weights W1 = {w11, w12, …} and W2 = {w21, w22, …}; (3) For each rule Ri, calculate the sum of the weighted matching degrees of all its conditions; (4) Considering priority, matching degree, timestamp, and the freshness of the rule comprehensively, select the rule with the highest evaluation score for execution. The formula for comprehensive consideration is as follows: EvaluationScore(Ri) = α × Priority(Ri) + β × MatchScore(Ri) + γ × Freshness(Ri) where, α, β, γ are weight coefficients; Priority(Ri) is the priority of rule Ri, MatchScore(Ri) is the matching degree between rule Ri and the current alarm event, and Freshness(Ri) is the freshness of rule Ri. EvaluationScore(Ri) is the evaluation score of rule Ri.

5. The alarm method for dynamic configuration management caused by business rules according to claim 1, characterized in that: The self - adaptation of business rules includes a multi - modal fusion decision - making strategy, which is used when fusing multi - source heterogeneous alarm data. By performing feature extraction and fusion, a single - entity model is established to identify alarms, and a multi - modal fusion decision - making strategy based on the weighted voting method is adopted to complete the final adaptive alarm processing. Among them, the multi - source heterogeneous alarm data includes video, audio, Internet of Things perception, and geospatial information.

6. The alarm method for dynamic configuration management caused by business rules according to claim 1, wherein: The specific steps of alarm self - healing processing are based on the self - healing rules in the urban governance business rule engine. After a non - emergency urban governance alarm occurs, the system will automatically adapt the self - healing disposal rules, trigger the corresponding self - healing process, achieve alarm self - healing, and then verify whether the alarm exists through AI vision recognition technology. If it is determined that there is no problem with self - healing, the alarm is closed.

7. The alarm method for dynamic configuration management caused by business rules according to claim 1, characterized in that: The specific steps of alarm prediction processing are as follows: (1) After real - time alarm events are accessed, filter the alarm events to generate a real - time alarm knowledge graph. (2) Construct an event graph model for urban governance and use a graph database to store and manage the graph model. Define each alarm event as a node in the graph model, including the detailed information of the alarm. (3) In the event graph model, obtain the relevant historical alarm knowledge graph in (1) and calculate the similarity using graph algorithms; the historical alarm knowledge graph with a similarity exceeding the set threshold is used as the candidate root - cause graph. (4) In the candidate root - cause graph, calculate the support degree and confidence degree of the alarm event indicators through the association analysis algorithm; sort the root causes according to the severity and influence range of the alarm root cause. (5) Make an auxiliary decision - making analysis and an event handling plan.

8. An alarm system based on dynamic configuration management caused by business rules, characterized in that: It includes an alarm aggregation module, an alarm filtering module, and an alarm application scenario module. The alarm aggregation module is used to collect alarm information through AI vision analysis of alarms and multiple types of alarm methods. The alarm filtering module is used to send decision - making alarm information according to any one of the alarm methods described in claims 1 - 7 after receiving the information from the alarm aggregation module. The alarm application scenario module is used to perform in - system alarm self - healing processing, alarm prediction processing, convert to work order events, or analyze visualization processing on the information sent by the alarm filtering module.

9. A storage medium, characterized in that: The computer program stored in this storage medium can be executed by one or more processors and can be used to implement the alarm method described in any one of claims 1 to 7.

Images