An intelligent protection system and method for cultural relics safety based on multi-source data fusion

Abstract

This application relates to a system and method for cultural relic safety risk assessment and emergency decision-making based on multi-source data fusion, belonging to the field of safety alarm technology. The system includes: a multi-source data access module: collecting real-time data on the status of the cultural relic itself, environmental parameters, security equipment operation data, and dynamic information on emergency resources through a heterogeneous data fusion interface; an intelligent decision-making module, further including a risk quantification engine submodule, a risk classification submodule, and an emergency resource map submodule; a contingency plan generation and execution module: converting risk quantification results into executable emergency operation instructions; and a drill and self-learning module: linking updated information from the emergency resource map submodule with data from the contingency plan generation and execution module, setting optimization targets based on the threshold of the risk classification mechanism. This application significantly shortens emergency response time through automated risk quantification and dynamic resource scheduling; it promotes the shift of cultural relic protection from passive response to proactive prevention, providing efficient and reliable technical support for cultural relic safety.

Description

Technical Field

[0001] This application relates to the field of security alarm technology, specifically to an intelligent protection system and method for cultural relics based on multi-source data fusion. Background Technology

[0002] Existing early warning systems in the field of cultural relic protection generally adopt a multi-module collaborative architecture, using environmental sensor networks, disaster monitoring terminals, and GIS platforms to achieve basic data collection and early warning classification. However, they rely on preset thresholds to trigger one-way early warning processes, treating multi-source heterogeneous data such as meteorological, structural, and security data in isolation, resulting in a lack of cross-dimensional risk fusion capabilities. Simultaneously, emergency plan generation is disconnected from real-time resource status, forming static response strategies, and the execution process lacks an effect feedback mechanism, making model iteration reliant on manual intervention for extended periods. Existing early warning systems suffer from three major problems: First, risk assessment is fragmented, with differences in data sources, dimensions, and confidence levels preventing the construction of a unified quantitative model, significantly impacting early warning accuracy and timeliness. Second, emergency resource scheduling exhibits static defects, with plans not linked to dynamic resource maps, limiting executability due to information silos. Third, the system is severely rigid, lacking a self-optimizing closed loop and unable to evolve autonomously based on historical execution data, resulting in long-term insufficient adaptability.

[0003] Patent CN114966897A discloses a cultural relic safety analysis and early warning system. It acquires meteorological data through a weather forecast big data access module, and combines this with a comprehensive meteorological station module (monitoring wind speed, wind direction, temperature, humidity, rainfall, etc.) and a flood monitoring module (water level sensing terminal and rainfall sensor) set up along outdoor cultural relic sites to collect environmental data. When the system receives an extreme weather warning, it automatically activates a Level 1 warning; subsequently, it determines whether to escalate to a Level 2 warning based on actual monitoring data. The system includes a GIS map display module, which can deploy monitoring point density according to terrain elevation differences, and provides drone route planning and emergency repair routes in the contingency plan. Although the above patent proposes a multi-module collaborative framework, it has the following shortcomings: 1) Coarse risk assessment: Insufficient risk assessment, relying only on simple thresholds to trigger warnings, lacking quantitative fusion analysis of multi-source data, lacking risk quantification algorithms, relying on manual experience to classify levels, unable to calculate the overall risk score, and having the potential risks of high false alarm rate and delayed response. 2) Static Emergency Response: Emergency response forces are disconnected, not linked to resource databases, and plans are detached from actual available resources. Plan generation does not consider real-time emergency resource status (such as material reserves and personnel deployment capabilities). This leads to low plan executability and a high potential failure rate during large-scale disasters. 3) Lack of Closed-Loop Optimization: The system is relatively rigid, lacking machine learning modules. Rule updates rely on manual debugging, and there is a lack of a mechanism for retrospective evaluation of plan execution effectiveness and system self-learning capabilities, making dynamic model optimization difficult. This results in poor environmental adaptability of emergency plans, which may become ineffective over time. 4) Data Silo Problem: Data is relatively limited, failing to integrate heterogeneous data such as security equipment status (e.g., camera malfunctions) and expert knowledge bases. Dynamic factors such as equipment status and personnel skills are ignored, affecting the comprehensiveness of risk identification and resulting in low risk coverage.

[0004] To address the aforementioned shortcomings, there is an urgent need to develop a new generation of technical solutions. This involves achieving real-time fusion of multi-source data by breaking through the spatiotemporal alignment of heterogeneous data and risk quantification algorithms, establishing a four-dimensional dynamic model of emergency resources to achieve intelligent resource matching, and relying on reinforcement learning mechanisms to form a closed-loop optimization of the entire process of "monitoring-decision-execution-calibration," thereby systematically improving the accuracy and timeliness of cultural relic protection. Summary of the Invention

[0005] To address the aforementioned shortcomings of existing technologies, this application provides a system and method for cultural relic safety risk assessment and emergency decision-making based on multi-source data fusion. The system aims to significantly shorten emergency response time through automated risk quantification and dynamic resource scheduling. Its self-learning capability can continuously optimize model adaptability and reduce false alarm rates and the risk of contingency plan failure.

[0006] The first aspect of this application provides a cultural relic safety risk assessment and emergency decision-making system based on multi-source data fusion, the system comprising: Multi-source data access module: Through a heterogeneous data fusion interface, it collects real-time data on the status of cultural relics, environmental parameters, security equipment operation data, and emergency resource dynamics; it adopts a spatiotemporal alignment engine to unify the data collection frequency through a sliding time window and uses the entropy weight method to dynamically allocate weights to data sources with different confidence levels to generate a weighted matrix. The intelligent decision-making module further includes a risk quantification engine submodule, a risk classification submodule, and an emergency resource mapping submodule. The risk quantification engine submodule receives a weighted matrix from the multi-source data access module and calculates a comprehensive risk score through a weighted factor fusion model, providing dynamic and accurate decision-making basis for the protection of cultural relics. The risk grading submodule is a core converter that transforms the comprehensive score output by the risk quantification engine submodule into actionable decisions, establishing a scientific decision-making chain from risk perception to emergency response. The emergency resource mapping submodule constructs a machine-readable four-dimensional dynamic model to achieve precise matching of risk response and resource supply; Contingency plan generation and execution module: Transforms risk quantification results into executable emergency operation instructions; Exercise and self-learning module: acquire information collected by the multi-source data access module, update information from the linkage emergency resource map sub-module, and data from the contingency plan generation and execution module, and set optimization targets based on the threshold of the risk grading mechanism.

[0007] Furthermore, the data processing of the spatiotemporal alignment engine further includes: Time dimension alignment uses a sliding window function W(T) current ,ΔT), where T current The current timestamp is ΔT, and the preset window duration is ΔT. The window... All input data within the frame are marked as a uniform time frame and then processed by the resampling function R(D) raw ,f s For the original data stream D raw Perform downsampling or interpolation to output the target frequency f. s The standard data stream uses a delay compensation function D(Td) to add an offset ΔT to the transmission delay Td. offset Spatial dimension alignment is achieved through the coordinate mapping function M. geo (X sensor ,Y sensor Z sensor The sensor's three-dimensional coordinates are converted into the surface-mapped coordinates (X) of the artifact model. m′ ,Y m′ Z m′ ), generating the continuous monitoring field function F(X) m′ ,Y m′ Zm′ ), and based on the topological correlation function T(P o Bind the monitoring point to the cultural relic component identifier P.

[0008] Furthermore, the weighting matrix further includes: Construct a monitoring matrix A = [m time points × n data sources] aij ], where elements aij This indicates that the j-th source has a normalized value at the i-th position, obtained through the information entropy function. Quantify the uncertainty of the data source and generate a difference coefficient function. The value that represents the utility of information, where g(j) reflects the volatility of the data; a larger value indicates a more significant amount of information from the source, determined by the dynamic weighting function. Allocate comprehensive weights, For real-time confidence Where λ is the attenuation coefficient, Δt is the transmission delay, and the weight w j Simultaneously considering both data value g(j) and reliability C(j), the influence weight of low-confidence sources is automatically reduced when data conflicts occur. base (j)为 Integrating inherent confidence levels, where λ is the transmission delay attenuation factor; the entropy weighting method addresses data quality heterogeneity through the function chain E→G→Wj, generating a weighted matrix. , This represents the standardized data stream from the j-th data source after spatiotemporal alignment. Time synchronization, spatial coordinate mapping, frequency normalization, and latency correction have been completed; the format is unified and can be directly merged. This represents the dynamic weight value of the j-th data source.

[0009] Furthermore, the risk quantification engine submodule further includes: The weighted matrix Aw is generated by spatiotemporal alignment and entropy weighting, containing a spatiotemporally unified and confidence-weighted risk factor data stream. A comprehensive risk score is calculated using a weighted factor fusion model, as follows:

[0010] function Φ( vi () is a normalization processor, physical quantity vi As input, the output range is [0,1], with dynamic weight W. i Generated using the ReliefF algorithm, its update mechanism is controlled by a self-learning algorithm formula:

[0011] Where R actual To determine the actual damage value based on the degree of damage to cultural relics, the system predicts the value R. predictW i =F(P type ), P type The cultural relic type identifier determines the basic weight allocation strategy for each risk factor. The F-weight decision function has a learning rate α to control the weight adjustment range and a constraint ∑Wi=1.

[0012] Furthermore, the risk grading submodule further includes: Receive the risk score value R generated by the risk quantification engine, and based on the numerical range of R, trigger differentiated emergency response strategies, discretizing the risk score value R into three-level response ranges: Low-risk response: When the R value is lower than the preset threshold R_low, preventive monitoring measures are implemented, and only the basic monitoring equipment attached to the cultural relic is activated; Medium-risk response: R value reaches or exceeds R_low but is lower than R_high, pre-deploy emergency resources; High-risk response: When the R value reaches or exceeds R_high, the cultural relic area will be forcibly closed and a joint cross-departmental response will be implemented.

[0013] Furthermore, the emergency resource mapping submodule further includes: It receives data collected in real time by the multi-source data access module. After the risk classification submodule determines the risk level, it sends a resource matching request instruction to the emergency resource map submodule to drive the subsequent scheduling process. The Emergency Resource Mapping submodule provides a standardized description of emergency response forces and resources. 1) Type attribute: Distinguishes the essential categories of emergency resources, including basic categories such as Expert, Material, and Team; 2) Spatial attribute Geo: Records the GPS coordinates of the resource, used to calculate the spatial distance between the disaster point and the resource, supporting optimal path planning; 3) Status attribute: Dynamically reflects availability, represented by binary logic or inventory level, ensuring that scheduling is based on real-time availability status; 4) Capability attribute: Quantifies the level of expertise in resources, expressed numerically.

[0014] Furthermore, the workflow of the emergency resource mapping submodule further includes: Phase 1: Emergency force data fusion, receiving data from multi-source data access modules, and unifying data format and update frequency through a spatiotemporal alignment engine; Phase Two: Dynamic Modeling. The emergency resource mapping submodule generates resource vectors based on four-dimensional attributes. Algorithm Formula. Type represents the type attribute, L lon , L lat"Status" and "Longitude" represent the longitude attribute, latitude attribute, and status attribute, respectively. score Indicates ability attributes; Phase 3: Response command execution. When the risk grading mechanism outputs an R value of medium to high risk, the intelligent matching process of the graph is triggered to extract the current risk feature vector and apply it using the cosine similarity algorithm. , Let be the matching degree scalar, representing the th The degree of compatibility between emergency resources and current risk events. Its value range is [0, 1]. The closer the value is to 1, the more closely the resource matches the characteristics of the current risk and the more suitable it is to be dispatched. A risk feature vector is a machine-parseable, multi-dimensional vector used to quantify the key attributes of a current risk event. It typically includes the following dimensions: 1. Disaster Type Attribute: Identifies whether the risk is due to natural weather or human-caused events. 2. Cultural Relic Vulnerability Attribute: Assesses the degree of vulnerability based on factors such as the material and structural condition of the cultural relic. 3. Risk Level Attribute: A comprehensive score output by the risk quantification engine. The corresponding severity level. Represents the magnitude of a vector; "Vector dot product operator," Indicates the first The feature vector of an emergency resource is generated based on the standard four-dimensional description of the resource from the "Emergency Resource Map" module, specifically including: 1. Type attribute: Identifies the essential category of the resource, such as Expert, Material, or Team. 2. Geo attribute: Records the GPS coordinates of the resource. 3. Status Attribute: Dynamically reflects resource availability, such as binary availability status (0 / 1) or inventory level. 4. Capability Attribute: Statistically expressed as a numerical value. Quantify the professional level or effectiveness of resources. Calculate the matching degree between each resource vector and risk characteristics, where res_idx is the resource index number.

[0015] Furthermore, the contingency plan generation and execution module further includes: The contingency plan template matching submodule uses the discrete risk level output by the risk grading mechanism as input conditions to automatically trigger the retrieval and calling logic of the contingency plan library. Through the preset level-template mapping rules, the continuous risk score value R is converted into discrete operation instructions: when the system determines that the risk level reaches the preset threshold, the matching engine actively indexes the contingency plan template set of the corresponding level and outputs a structured task list. The emergency resource binding submodule is invoked through... The resource matching algorithm calculates the optimal resource, where V E Quantifying the multidimensional attributes of current risk, V C To construct a machine-analyzable feature space for resource feature vectors, the system consists of three dimensions: disaster type attribute, cultural relic vulnerability attribute, and risk level attribute. Abstract risks are transformed into computable vectors to provide input benchmarks for matching algorithms. The machine instruction execution submodule generates a standardized instruction set and distributes it to the terminal device cluster via an API interface.

[0016] Furthermore, the practice and self-learning module further includes: Driven by a dual-engine approach of digital sandbox simulation and reinforcement learning, an extreme disaster scenario is constructed in a virtual environment. By dynamically executing contingency plan strategies and collecting feedback data, the system autonomously optimizes its decision-making model. The core algorithm of this autonomous optimization model consists of a reward function and a weight iteration formula. By quantitatively evaluating the effectiveness of contingency plan implementation, the continuous optimization process of the contingency plan library is driven, and an intelligent optimization engine is established. In this engine, Reward is the comprehensive reward value output, Ta is the actual handling time, Tp is the estimated time of the contingency plan, ΔR is the actual risk reduction value, α represents the response efficiency weight coefficient, and β represents the risk control weight coefficient.

[0017] The second aspect of this application provides a method for cultural relic safety risk assessment and emergency decision-making based on multi-source data fusion, the method comprising: Multi-source data access steps: Real-time collection of the status of cultural relics, environmental parameters, security equipment operation data and emergency resource dynamic information through heterogeneous data fusion interface; Using a spatiotemporal alignment engine, the data collection frequency is unified through a sliding time window, and the weights of data sources with different confidence levels are dynamically assigned using the entropy weight method to generate a weighted matrix; The intelligent decision-making process further includes a risk quantification engine sub-step, a risk classification sub-step, and an emergency resource mapping sub-step: The risk quantification engine sub-step receives a weighted matrix from the multi-source data access module and calculates a comprehensive risk score through a weighted factor fusion model, providing a dynamic and accurate decision-making basis for the protection of cultural relics. The risk grading sub-step transforms the comprehensive score output by the risk quantification engine sub-module into a core converter for actionable decisions, establishing a scientific decision-making chain from risk perception to emergency response. The emergency resource mapping sub-steps construct a machine-readable four-dimensional dynamic model to achieve precise matching of risk response and resource supply; Contingency plan generation and execution steps: Transform the risk quantification results into executable emergency operation instructions; Exercise and self-learning steps: Obtain information collected by the multi-source data access module, update information from the linkage emergency resource map sub-module, and data from the contingency plan generation and execution module, and set optimization targets based on the threshold of the risk grading mechanism.

[0018] This application constructs a closed-loop intelligent technology system for the protection of cultural relics, breaking through the technical bottlenecks of traditional systems in three major areas: multi-source data fusion, dynamic resource matching, and self-learning optimization. The beneficial technical effects that can be achieved are as follows: The design incorporates a spatiotemporal alignment engine and an entropy weighting method for dynamic weighting. By unifying the collection frequency of heterogeneous data through a sliding time window and automatically allocating weight coefficients based on the confidence level of the data source, the spatiotemporal inconsistency and dimensional barriers of multi-source data such as environmental sensors, security equipment, and resource databases can be resolved.

[0019] A weighted factor fusion model and a dual-loop self-learning mechanism are established. The normalization function transforms physical monitoring values ​​into dimensionless risk factors. Combined with the ReliefF algorithm, the weight coefficients are dynamically updated to achieve accurate adaptation of risk scores to cultural relic types, climate cycles, and emergencies. Through the linkage of the weight iteration function and the reinforcement learning reward function, a two-layer optimization loop of "prediction-execution-feedback" is formed to continuously improve the robustness of the model.

[0020] A four-dimensional vector model for emergency resources and a cosine similarity matching algorithm are proposed. The types, spatial coordinates, real-time status and capability attributes of resources such as experts, materials and teams are standardized and encoded into machine-readable feature vectors. Based on the similarity calculation of risk features and resource attributes, the automated and precise dispatch of emergency forces is realized.

[0021] This application significantly improves the level of intelligence in cultural relic protection. Through automated risk quantification and dynamic resource scheduling, it greatly shortens emergency response time. Its self-learning ability can continuously optimize model adaptability, reduce false alarm rate and the risk of plan failure, and promote the transformation of cultural relic protection from passive response to proactive prevention, providing efficient and reliable technical support for cultural relic safety. Attached Figure Description

[0022] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0023] Figure 1 This is a schematic diagram of the intelligent protection system for cultural relics security proposed in this application. Detailed Implementation

[0024] To better illustrate the technical solution of this application, this application will be further described in conjunction with the accompanying drawings and specific examples.

[0025] This application discloses a cultural relic safety risk assessment and emergency decision-making system based on multi-source data fusion. It aims to address the shortcomings of existing technologies, such as coarse risk assessment, static emergency resource matching, and open-loop execution of plans. The system constructs an intelligent protection system integrating quantitative risk assessment, dynamic resource matching, and closed-loop self-optimization. First, a spatiotemporal alignment engine enables real-time fusion of heterogeneous data from multiple sources, including the environment, the cultural relic itself, equipment, and personnel. A normalization function eliminates dimensional differences, and an entropy weighting method is used to dynamically allocate confidence weights. A weighted factor fusion model is established to calculate a comprehensive risk score. This model achieves adaptive adjustments driven by space, time, and events through a dynamic weighting mechanism. Second, a dynamic emergency resource map containing four-dimensional attributes (type, location, status, and capability) is constructed. An intelligent matching algorithm based on the similarity calculation of risk feature vectors and resource feature vectors is designed to achieve accurate association between emergency plans and real-time available resources. Finally, a closed-loop optimization mechanism including an execution monitoring layer and a reinforcement learning module is established. Real-time collection of plan execution indicators and the construction of a reward function drive system iterative updates. Simultaneously, a pre-generated optimization strategy library is generated through digital sand table simulation exercises. Through the synergy of the aforementioned technologies, the system achieves fully automated closed-loop management of the entire process, from risk perception, quantitative assessment, resource matching, contingency plan execution to effect feedback, significantly improving the initiative and adaptability of cultural relic security protection.

[0026] like Figure 1 As shown, the cultural relic safety alarm management system of this application includes the following modules: 1. Multi-source data access module: The multi-source data access module collects real-time data on the status of cultural relics, environmental parameters, security equipment operation data, and emergency resource dynamics through a heterogeneous data fusion interface. A spatiotemporal alignment engine addresses data heterogeneity by using a sliding time window to unify data collection frequency and employing entropy weighting to dynamically assign weights to data sources with different confidence levels.

[0027] The spatiotemporal alignment engine addresses the problem of heterogeneous data structure through a functional processing chain. The data processing implementation of the spatiotemporal alignment engine is as follows: time dimension alignment uses a sliding window function W(T) current ,ΔT) is realized, where T current `t` is the current timestamp, and `ΔT` is the preset window duration. This function will... All input data within the frame are marked as a uniform time frame and then processed by the resampling function R(D) raw ,f s For the original data stream D raw Perform downsampling or interpolation to output the target frequency f. sThe standard data stream is used, and the transmission delay Td is supplemented with an offset ΔT using the delay compensation function D(Td). offset Ensure the causal logic of events; where spatial dimension alignment is achieved through the coordinate mapping function M. geo (X sensor ,Y sensor Z sensor The sensor's three-dimensional coordinates are converted into the surface-mapped coordinates (X) of the artifact model. m′ ,Y m′ Z m′ ), generating the continuous monitoring field function F(X) m′ ,Y m′ Z m′ ), and based on the topological correlation function T(P o Binding monitoring points with cultural relic component identifiers P o (P) o (Unique identifier for artifact components). The spatiotemporal alignment engine uses the function chain W→R→D→M. geo Eliminate data structural heterogeneity (time asynchrony, spatial dispersion, and chaotic format) and output a standardized data stream S. std .

[0028] Definition of core formula symbols: Time alignment W(T) current ,ΔT): Time window synchronization function (T current ΔT: millisecond-level timestamp, ΔT: second-level window duration); Resampling R(D) raw ,f s ): Data frequency normalization function (D raw : Raw data, f s (Hertz-level target frequency); Delay compensation D(T) d Network latency correction function (T) d (millisecond-level latency); Space mapping M geo (X,Y,Z): Spatial coordinate mapping function ((X,Y,Z): Sensor coordinates); The entropy weighting method addresses the heterogeneity of data quality through a dynamic weighting function chain. First, it constructs a monitoring matrix A = [m time points × n data sources]. aij ], where elements aij This indicates that the j-th source has a normalized value at the i-th position. Then, the information entropy function is used... (in Quantify the uncertainty of the data source and generate a difference coefficient function. The information utility value is represented by "g(j)", where "g(j)" reflects data volatility; a larger value indicates more significant information from the source. Ultimately, this is determined by a dynamic weighting function. Allocate comprehensive weights, For real-time confidence , where λ is the attenuation coefficient and Δt is the transmission delay. The weight w j Simultaneously considering both data value g(j) and reliability C(j), the influence weight of low-confidence sources is automatically reduced when data conflicts occur. The inherent confidence level C is also incorporated. base (j) (Including the transmission delay attenuation factor λ). The entropy weighting method addresses data quality heterogeneity (precision fluctuations, reliability differences) through the function chain E→G→Wj, generating a weighted matrix. , This represents the standardized data stream from the j-th data source after spatiotemporal alignment. Time synchronization, spatial coordinate mapping, frequency normalization, and latency correction have been completed; the format is unified and can be directly merged. This represents the dynamic weight value of the j-th data source.

[0029] Information entropy E(j): Information entropy function (κ=1 / lnm: normalization coefficient, pij (Numerical distribution probability); Difference coefficient G(j): Difference coefficient function (1 E(j): Data value weighting index); Dynamic weights C(j): Real-time confidence function (C base (j) : Inherent confidence level, λ: second-level decay factor, Δt: second-level delay duration); Real-time confidence level Wj: Dynamic weighting function (outputs the final weights of fused data quality and value); The cascading of the two mechanisms forms a complete processing pipeline, providing the risk quantification engine with spatiotemporally unified and quality-controllable input data.

[0030] 2. Intelligent Decision-Making Module: The intelligent decision-making module includes a risk quantification engine, a risk classification mechanism, and an emergency resource map; wherein: The risk quantification engine specifically receives a weighted matrix Aw from the preprocessed output of the multi-source data access module. Aw is generated by spatiotemporal alignment and entropy weighting, and contains a spatiotemporally unified and confidence-weighted risk factor data stream. A comprehensive risk score is calculated using a weighted factor fusion model, as follows:

[0031] As the core engine for unified risk assessment of the system, it transforms monitoring data with different physical dimensions into comparable risk values. The function Φ( vi (This is a normalization processor that receives physical quantities) viInputs include temperature deviation ΔT or crack width S; the output is a dimensionless risk value ranging from [0,1]. It employs a piecewise linear mapping, where the monitored value... vi Below the cultural relic safety threshold V safe When the output is 0 (for example, the safety threshold for cracks in wooden structures is 1.5mm, and the initial value is set according to the requirements of national standards such as the "Technical Specification for Maintenance and Reinforcement of Ancient Wooden Structures" (GB 50165) and "Classification and Illustration of Diseases of Stone Cultural Relics" (WW / T 0063), the output is 0. When the output is between the safety threshold V and the threshold V, the output is 0. safe With the critical threshold of damage V critical The values ​​are mapped proportionally between the specified values ​​(e.g., the critical value for a crack is 3.0 mm), and when the value exceeds the critical value, the output is 1, thereby quantifying the degree of physical damage to the cultural relic.

[0032] Dynamic weight W i Generated using the ReliefF algorithm, it reflects the importance of different risk factors. Its update mechanism is controlled by a self-learning algorithm formula:

[0033] This algorithm formula forms the mathematical basis for closed-loop optimization, by comparing it with the actual disaster loss value R. actual (Based on the degree of damage to cultural relics) and the system's predicted value R predict The deviation is used to dynamically adjust the weights. Where Wi = F(P) type ), P type The cultural relic type identifier determines the basic weight allocation strategy for each risk factor. The F-weight decision function, when P... type When P = 1 (for grotto-type cultural relics), the base value of the output crack weight Wcrack is increased from the preset value of 0.25 to 0.32; when P type When =2 (wooden architectural relics), output temperature and humidity weight W. env The base value is increased from the preset value of 0.25 to 0.35. The parameter α is the learning rate (default value 0.05), used to control the magnitude of weight adjustment. When the actual disaster loss value is higher than the predicted value, the relevant weights are increased to correct the underestimation of risk; conversely, the weights are decreased to avoid overfitting. After each iteration, the constraint ∑Wi=1 is applied to maintain the balance of the weight system. The weight allocation has a triple adaptive mechanism: 1) Spatial adaptability: Based on the cultural relic type identifier P type Differentiated allocation of weights, for example, for grotto artifacts (P type =1) Automatically increase fracture weight W crack ; 2) Temporal evolution: Combined with the climate cycle function C(t), dynamic adjustment is made, such as automatically increasing the weight of wind and rain factors during the typhoon season (t∈July-September); 3) Event-driven property: Weights are adjusted in real time in response to unexpected events. For example, when a security device failure occurs, the human-induced risk coefficient is increased via the incremental parameter β (default value 0.1).

[0034] This engine realizes quantitative mapping of physical damage through a normalization function, converting physical changes into computable risk values; it builds continuous evolution capabilities through a self-learning closed-loop formula, and optimizes model weights using historical disaster data. This forms a complete optimization chain of "data input → risk calculation → disaster loss feedback → weight adjustment", breaking through the static limitations of traditional threshold methods (such as fixed temperature alarm values) and providing dynamic and accurate decision-making basis for cultural relic safety protection.

[0035] The risk grading mechanism is specifically as follows: The risk grading mechanism in this application is a core converter that converts the comprehensive score output by the risk quantification engine into executable decisions, and its function is to establish a scientific decision-making chain from risk perception to emergency response. This mechanism directly receives the risk score value R generated by the risk quantification engine, wherein the calculation basis of the R value completely depends on the preprocessed data provided by the multi-source data access module. The risk grading mechanism automatically triggers differentiated emergency response strategies based on the numerical interval of the R value, realizing the accurate conversion of risk assessment results into operation instructions. The risk grading mechanism discretizes the continuous risk score R into three levels of response intervals: 1) Low-risk response (R<R_low): When the R value is lower than the preset threshold R_low (R_low represents the upper limit of risk that cultural relics can bear, its value is calibrated and generated by the cultural relic material aging model, and the initial typical value is set to 4), preventive monitoring measures are implemented, and only the basic monitoring devices attached to the cultural relics are activated, such as environment regulating devices, fixed security devices, standard manual patrols, etc.

[0036] 2) Medium-risk response (R_low ≤ R<R_high): The R value reaches or exceeds R_low but is lower than R_high (R_high represents the critical point from quantitative change to qualitative change, and exceeding this value will cause irreversible damage. The initial typical value of R_high is set to 7), emergency resources are pre-deployed, the system activates the pre-plan preloading process, dispatches emergency forces to stand by and check the material inventory.

[0037] 3) High-risk response (R ≥ R_high): When the R value reaches or exceeds R_high, the cultural relic area is forcibly closed, cross-departmental joint disposal is implemented, and external institutions such as fire departments and medical institutions can be linked according to the generated emergency plan, and emergency forces are mobilized to perform operations such as cultural relic reinforcement and evacuation. R_high is associated with the cultural relic structure failure threshold, and is dynamically corrected through a mechanical model.

[0038] The risk grading mechanism relies on an R-value derived from the risk quantification engine. The input data for this engine is entirely provided by a multi-source data access module, with the data transmission path being: raw data → spatiotemporal alignment engine → φ(v_i) normalization → R-value calculation → grading judgment. The calibration data for the grading thresholds R_low / R_high originates from the historical analysis library of the quantification engine. The risk grading mechanism outputs grading results to the emergency execution layer and directly converts them into machine-readable operation codes, driving the contingency plan generation module to initiate the resource matching algorithm. The risk grading mechanism discretizes continuous risk variables into operational instructions through preset thresholds. Its innovation lies in establishing a dynamically adjustable threshold calibration model, ensuring the system maintains responsiveness while avoiding frequent false triggers. This mechanism serves as the connecting module between the risk quantification engine and the emergency execution layer, ensuring a closed-loop machine-driven decision-making process for cultural relic protection, from risk perception to response actions.

[0039] The emergency resource map specifically includes: It serves as a hub for standardized resource management and intelligent scheduling within the system. This module primarily addresses the technical problems of fragmented resource information and delayed status updates in traditional emergency management. By constructing a machine-readable four-dimensional dynamic model, it achieves precise matching between risk response and resource supply. The map establishes a data supply relationship with the multi-source data access layer and an instruction triggering relationship with the risk grading mechanism. The multi-source data access layer collects and transmits raw resource data such as personnel location, material inventory, and equipment status to the map database in real time. When the risk grading mechanism determines the risk level, it automatically sends a resource matching request instruction to the map, driving subsequent scheduling processes.

[0040] The emergency resource map provides a standardized description of emergency response forces and resources. Its basic framework is as follows: 1) Type attribute: Distinguish the essential categories of emergency resources, including basic categories such as Expert, Material, and Team, to ensure that the system can identify the essential functions of the resources.

[0041] 2) Spatial attribute (Geo): Records the GPS coordinates of the resource, used to calculate the spatial distance D between the disaster point and the resource, supporting optimal path planning.

[0042] 3) Status attribute: Dynamically reflects availability, using binary logic (0 = unavailable, 1 = available) or inventory Q_{stock} to ensure scheduling is based on real-time availability status.

[0043] 4) Capability attribute: Quantifies the professional level of resources, represented by a numerical value C_{score} \in [0,10] (e.g., ancient building restoration expert C_{score}=8.5, water pump drainage efficiency C_{score}=200 / 25=8.0), solving the problem that traditional descriptions cannot quantify and match.

[0044] The emergency mapping workflow consists of three stages: Phase 1: Emergency response force data fusion. The multi-source data access layer transmits scattered resource information to the map, and the spatiotemporal alignment engine unifies the data format and update frequency.

[0045] Phase Two: Dynamic Modeling. Emergency resource maps generate resource vectors based on four-dimensional attributes. Algorithm Formula. Taking flood control sandbags as an example, the resource is vectorized into a machine-readable `Material`, (116.39^E, 39.91^N), Q_{stock}=1200, C_{score}=7.2. Its complete description includes: the resource type is marked as "material," the spatial location is recorded as the coordinates of 116.39 degrees east longitude and 39.91 degrees north latitude, the status attribute is represented as the current inventory of 1200 units, and the capability attribute is quantified as 7.2 points based on flood control effectiveness. Logical coupling is formed between these attributes; the `Type` attribute defines the resource's functional scope, (L... lon ,L lat Resource schedulability is jointly determined by spatial and status attributes (such as coordinates determining transportation time and inventory quantity determining the available scale), C score Capability attributes ensure the professional suitability of resources for disaster scenarios.

[0046] Phase Three: Response Command Execution. When the risk grading mechanism outputs an R value of medium to high risk, the intelligent matching process of the map is triggered. The system extracts the current risk feature vector \overrightarrow{RiskVector} (containing dimensions such as disaster type and cultural relic vulnerability), and uses the cosine similarity algorithm... Calculate the matching degree between each resource vector and the risk feature, where res_idx is the resource index number. When M_{res_idx} > α (preset threshold), the matching degree exceeds the preset threshold, and the optimal resource set is automatically bound. Let be the matching degree scalar, representing the th The degree of compatibility between emergency resources and current risk events. Its value range is [0, 1]. The closer the value is to 1, the more closely the resource matches the characteristics of the current risk and the more suitable it is to be dispatched. A risk feature vector is a machine-parseable, multi-dimensional vector used to quantify the key attributes of a current risk event. It typically includes the following dimensions: 1. Disaster Type Attribute: Identifies whether the risk is due to natural weather or human-caused events. 2. Cultural Relic Vulnerability Attribute: Assesses the degree of vulnerability based on factors such as the material and structural condition of the cultural relic. 3. Risk Level Attribute: A comprehensive score output by the risk quantification engine. The corresponding severity level. Represents the magnitude of a vector; "Vector dot product operator" ， Indicates the first The feature vector of an emergency resource is generated based on the standard four-dimensional description of the resource from the "Emergency Resource Map" module, specifically including: 1. Type attribute: Identifies the essential category of the resource, such as Expert, Material, or Team. 2. Geo attribute: Records the GPS coordinates of the resource. 3. Status Attribute: Dynamically reflects resource availability, such as binary availability status (0 / 1) or inventory level. 4. Capability Attribute: Statistically expressed as a numerical value. Quantify the professional level or effectiveness of resources.

[0047] This mapping system, through the dynamic coupling of four-dimensional attributes, upgrades resource scheduling from experience-based decision-making to a data-driven paradigm. Spatial coordinates support path optimization to reduce response time, status monitoring avoids invalid scheduling instructions, and capability quantification ensures the professionalism of response measures. Ultimately, it forms an automated closed loop with the risk classification mechanism, encompassing "risk assessment → resource retrieval → instruction generation."

[0048] The contingency plan generation and execution module specifically includes: This module transforms risk quantification results into executable emergency operation instructions, creating a closed-loop machine process from risk perception to resource allocation. This module operates through a dual-coupling mechanism. On the input side, it dynamically links with the risk grading mechanism, triggering the corresponding level of the contingency plan template library based on the risk score value R. It receives the score value R (R∈[0,10]) output by the risk quantification engine and activates the corresponding contingency plan generation process based on the R value, which directly determines the response level and resource allocation scale of the contingency plan. On the resource allocation side, it interacts in real-time with the emergency resource map, reading the four-dimensional resource vector R=〈type,geo,status,capability〉 in the map, where type is the type attribute, geo is the spatial attribute, status is the state attribute, and capability is the capability attribute. These vectors serve as the basis for the matching algorithm, calling the resource matching algorithm to bind the optimal emergency forces. On the execution side, it connects to terminal devices in the multi-source data access layer through a standardized interface, directly calling the terminal devices in the access layer through the API instruction set, including environmental sensor data verification, environmental control system startup, and emergency communication protocols, ensuring that contingency plan execution is synchronized with the real-time data stream.

[0049] The contingency plan generation and execution module consists of a contingency plan matching engine and an instruction execution engine. The implementation process is as follows: Step 1: Contingency Plan Template Matching. Based on the discrete risk levels (low, medium, and high) output by the risk grading mechanism, this step automatically triggers the retrieval and retrieval logic of the contingency plan library. This function converts continuous risk scores R into discrete operation instructions through preset level-template mapping rules. When the system determines that a risk level reaches a preset threshold, the matching engine actively indexes the corresponding level's contingency plan template set and outputs a structured task list. The input linkage mechanism receives the score value R calculated by the risk quantification engine and generates discrete risk level identifiers according to the grading rules (R<4 → low risk, 4≤R<7 → medium risk, R≥7 → high risk). The template mapping logic establishes a binding relationship between the level identifier and the contingency plan template. High-risk levels automatically activate high-intensity protection task sets, medium-risk levels call pre-deployed task sets, and low-risk levels only load the basic monitoring task set. Structured Output: This generates an operable instruction set containing task sequences, execution standards, and acceptance indicators. For example, a high-risk contingency plan might output multi-threaded task items such as cultural relic reinforcement, drainage system activation, and personnel evacuation coordination.

[0050] Step 2: Emergency resource binding, via invocation The resource matching algorithm calculates the optimal resource, where V EThis study quantifies the multidimensional attributes of current risks and constructs a machine-analyzable feature space, consisting of three dimensions: disaster type attributes (scaled to natural climate or man-made event categories), cultural relic vulnerability attributes (scaled to the material and structural sensitivity of the artifact), and risk level attributes (scaled to the severity of the comprehensive score R). Abstract risks are transformed into computable vectors, providing an input benchmark for the matching algorithm. Where V... C This is a resource feature vector representing a four-dimensional capability model of emergency resources, enabling a mathematical mapping of resource attributes. By dynamically reading standardized fields from the emergency resource map, the type attribute (type) categorizes resource functions, the spatial attribute (geo) provides resource location coordinates, the status attribute (status) provides real-time availability, and the capability attribute (capability) quantifies skill levels. A machine-readable description of resource capabilities is established to support adaptability calculations. M is a matching degree scalar (M∈[0,1]), which measures the adaptability of resources and risks using the cosine similarity principle. An M value approaching 1 indicates a high degree of compatibility between resource and risk characteristics, while an M value approaching 0 indicates no significant correlation. When M>θ (θ is a preset matching threshold), the resource is automatically bound. The threshold θ serves as a confidence boundary for resource binding, preventing mis-scheduled low-adaptability resources. The value of θ is generated through training on historical exercise data; when M<θ, a resource pool expansion warning is triggered.

[0051] Step 3: Machine Instruction Execution. This module generates a standardized instruction set and distributes it to the terminal device cluster via API. Specific operations include: transmitting task spatial coordinates and standardized operation protocols to the bound emergency teams; activating IoT devices and systems to perform artifact status monitoring and task execution according to predefined protocols; linking with security systems to perform area access control or evacuation route management; and automatically configuring equipment operating parameters based on material capability attributes. Simultaneously, through a real-time two-way communication channel between resource status changes and map data, the four-dimensional vector of the emergency resource map is dynamically updated, synchronizing resource status and material capabilities in real time. When material inventory falls below a preset threshold, an automatic replenishment prompt is triggered.

[0052] This module optimizes the delays inherent in traditional systems that involve "manual risk analysis → experience-based resource matching → multi-level approval and execution" through an algorithmic decision-making mechanism. Its innovation lies in establishing a triple machine closed loop: an automatic mapping loop between risk classification and contingency plan templates, a mathematical matching loop between resource characteristics and risk characteristics, and an execution feedback loop between API commands and terminal devices, enabling automated handling from risk perception to emergency response.

[0053] 4. Practice and Self-Learning Module: The drill and self-learning module serves the function of continuous system evolution in the patent, and breaks through the static defect of traditional preplans by establishing a closed-loop mechanism of "execution-feedback-optimization". This module forms deep collaboration with other components of the system: it receives practical operation data from the preplan generation and execution module (such as emergency response time consumption and resource consumption), links the emergency resource graph to update the capability status (such as the location and quantity of emergency supplies and personnel capabilities), obtains the historical disaster case training simulation environment from the multi-source data access module, and optimizes objectives based on the threshold setting of the risk grading mechanism. Its working principle is driven by dual engines of digital sandbox simulation and reinforcement learning, constructing extreme disaster scenarios in a virtual environment, and driving the system to autonomously optimize the decision model by dynamically executing preplan library strategies and collecting feedback data. The core algorithm of the module consists of a reward function and a weight iteration formula , which quantifies and evaluates the execution effect of preplans, drives the continuous optimization process of the preplan library, and establishes an intelligent optimization engine.

[0054] where Ta (actual response time) is derived from the data returned by the IoT devices of the preplan execution module, reflecting the practical operation efficiency of the preplan, and the ratio of Ta to Tp (preplan estimated time) reflects the execution deviation in the time dimension. The purpose of setting Tp is to establish an efficiency baseline, 1 The TpTa term quantifies the optimization space of response speed, and positive incentive is generated when Ta<Tp.

[0055] where ΔR (actual risk reduction value) is obtained by monitoring the difference between pre-disaster and post-disaster risk scores through the multi-source data access module, and is calculated from the difference between the risk score before and after execution R pre R post , which directly measures the effect of preplan implementation.

[0056] where α represents the response efficiency weight coefficient, with a definition range of α∈[0,1], used to adjust the contribution weight of timeliness in the reward function. β represents the risk control weight coefficient, with a definition range of β∈[0,1], used to adjust the contribution weight of the risk reduction value in the reward function. The weight coefficients α and β respectively adjust the optimization weights of timeliness and risk control, and the proportion of β is increased when cultural relics are in a rapid deterioration state to strengthen risk suppression.

[0057] where α and β (balance coefficients) are artificially preset weight parameters, α regulates the contribution of the efficiency term, β regulates the contribution of the effect term, and the two satisfy α + β = 1 to ensure the normalization of the reward value. Its purpose of setting is to adapt to different cultural relic value levels, and the β weight can be increased for high-level cultural relics to strengthen risk control.

[0058] The Reward is the overall reward value output. The output value of the reward function is set to a continuous value between -1 and 1, which serves as the core driving signal for policy optimization in the contingency plan database: when the output value Reward > 0, the system automatically increases the weight of the currently executed policy; when the output value Reward < 0, the system triggers the corresponding policy elimination mechanism; when the output value Reward = 0, the system maintains the existing policy unchanged. This design establishes an objective standard for the dynamic screening of the contingency plan database by quantitatively evaluating policy effectiveness and risk control results.

[0059] The reward function and the weight iteration formula constitute a dual-closed-loop optimization mechanism, which achieves system co-evolution through data coupling. The weight iteration formula is as follows: ,in wi The formula represents the weight coefficient of the i-th type of risk factor; k represents the iteration number; η is a preset learning rate parameter, the value of which is preset according to the type of cultural relic and the convergence characteristics of historical data, used to control the magnitude of weight adjustment; Ra represents the actual disaster damage quantification value obtained from the analysis of the multi-source data module; Rp represents the predicted risk value output by the risk quantification engine. The formula dynamically adjusts the risk factor weights by receiving the deviation (Ra-Rp) between the actual disaster damage value Ra and the predicted value Rp. wi Its working logic is as follows: when Ra > Rp, it indicates that the system underestimates the actual risk, and the formula positively increases the weight values ​​of relevant factors through the η parameter. When Ra < Rp, it indicates that the system has an overfitting risk, and the formula negatively decreases the corresponding weight values. After each iteration, weight normalization is performed to ensure that the sum of all weight coefficients is constant at 1.

[0060] Through a dual-loop mechanism of reward function algorithm and weight iteration algorithm, the system has the ability to evolve globally. The contingency plan library is optimized to improve short-term response efficiency, and weight iteration enhances the accuracy of long-term risk assessment. Both are continuously verified through the reenactment of historical scenarios in the exercise module, ultimately achieving a global optimization and upgrade from passive response to proactive defense.

[0061] The beneficial effects of the technical solution presented in this application are as follows: It constructs a fully intelligent cultural relic protection system. Through a multi-source data spatiotemporal alignment engine, it achieves full-domain information fusion, eliminating the cost of manual data integration and providing an efficient data foundation for accurate decision-making. By dynamically generating risk scores through a weighted factor fusion model combined with an adaptive weighting algorithm, it significantly reduces the waste of emergency resources caused by experience-based misjudgments. The standardized definition of resource attributes in a four-dimensional emergency resource map, along with the optimization of scheduling paths through a machine-readable framework, reduces the time spent on human communication and coordination, improving overall efficiency. The two-level algorithm linkage between plan generation and execution enables automated decision-making, compressing the time required for traditional manual assessment and avoiding damage to cultural relics caused by delayed disaster response. The reinforcement learning dual-loop mechanism continuously optimizes the deployment of monitoring equipment and material reserve strategies, simultaneously reducing redundant equipment investment and optimizing emergency material inventory. This system fundamentally changes the traditional high-cost, manually-led operation and maintenance model of protection. Through machine autonomous decision-making, accurate risk prediction reduces ineffective scheduling losses, intelligent resource matching improves the utilization rate of human and material resources, and an automated response mechanism avoids damage to cultural relics, providing sustainable intelligent support for the preventive protection of cultural relics.

[0062] This application proposes a multi-source data fusion method based on spatiotemporal alignment and confidence weighting: It aligns the collection frequencies of heterogeneous data sources such as environmental sensors, security equipment, and resource databases through a sliding time window, and dynamically assigns weight coefficients to data with different confidence levels using the entropy weighting method; further, a normalization function is designed to eliminate dimensional differences, and a dynamic weighting algorithm is combined to achieve a quantitative scoring of cultural relic risks. This mechanism solves the problem of one-sided risk assessment caused by data fragmentation in traditional systems, significantly improving the comprehensiveness and accuracy of risk perception.

[0063] Furthermore, this application proposes a risk assessment model with spatiotemporal evolution characteristics: it employs a feature selection algorithm to dynamically adjust the weights of risk factors, achieving differentiated configuration of the spatial dimension for different types of cultural relics, automatic adjustment of the temporal dimension according to natural climate cycles, and adaptive capability in response to emergencies in the event dimension; it establishes a three-level response mechanism based on quantitative scoring, realizing automated decision-making transitions from routine monitoring to emergency response. This system overcomes the limitations of traditional static models, enabling the system to possess environmental adaptability and real-time decision-making.

[0064] Furthermore, this application proposes a dynamic modeling method for emergency resources based on a four-dimensional vector model: a standardized resource description framework including type, spatial, state, and capability attributes is constructed, and resource state changes are synchronized in real time via an IoT interface; a cosine similarity matching algorithm based on feature vectors is designed to achieve accurate mapping between disaster characteristics and resource attributes. This method solves the problem of the disconnect between contingency plans and resources, ensuring the timeliness and adaptability of emergency force deployment.

[0065] Furthermore, this application proposes a dual-loop optimization architecture that integrates reinforcement learning and weight iteration: the inner loop evaluates the effectiveness of contingency plan execution through a reward function, optimizing resource allocation and response strategies; the outer loop corrects the parameters of the risk assessment model through a weight iteration function, enabling continuous system evolution; the two loops share actual disaster loss data to form collaborative optimization. This mechanism establishes a full-process self-learning system of "monitoring-decision-execution-feedback-evolution," breaking through the technical bottleneck of traditional systems' lack of backtracking optimization.

[0066] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their likenesses.

[0067] Furthermore, this application describes a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of a method for managing cultural relic safety alarms.

[0068] Furthermore, this application describes a computer device, comprising: a memory for storing a computer program; and a processor for executing the computer program to implement the steps of a cultural relic safety alarm management method.

[0069] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0070] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0071] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0072] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0073] The above is an example of the working process and results of one specific embodiment of this application, for reference only. This application is not limited to this embodiment, and other variations or equivalent substitutions are possible. All such variations or substitutions, without departing from the technical solution and inventive concept of this application, fall within the protection scope of this application.

[0074] It should be noted that the above detailed descriptions are exemplary and intended to provide further explanation of this application. Unless otherwise specified, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application pertains.

[0075] In the detailed description above, reference has been made to the accompanying drawings, which form part of this document. In the drawings, similar symbols typically identify similar parts unless the context otherwise indicates otherwise. The illustrated embodiments described in the detailed specification, drawings, and claims are not intended to be limiting. Other embodiments may be used and other changes may be made without departing from the spirit or scope of the subject matter presented herein.

Claims

1. A cultural relic safety risk assessment and emergency decision-making system based on multi-source data fusion, characterized in that, The system includes: The multi-source data access module collects real-time data on the status of cultural relics, environmental parameters, security equipment operation data, and emergency resource dynamics through a heterogeneous data fusion interface. It adopts a spatiotemporal alignment engine to unify the data collection frequency through a sliding time window and dynamically assigns weights to data sources with different confidence levels using the entropy weight method to generate a weighted matrix. The intelligent decision-making module further includes a risk quantification engine submodule, a risk classification submodule, and an emergency resource mapping submodule. The risk quantification engine submodule receives a weighted matrix from the multi-source data access module and calculates a comprehensive risk score through a weighted factor fusion model, providing dynamic and accurate decision-making basis for the protection of cultural relics. The risk grading submodule is a core converter that transforms the comprehensive score output by the risk quantification engine submodule into actionable decisions, establishing a scientific decision-making chain from risk perception to emergency response. The emergency resource mapping submodule constructs a machine-readable four-dimensional dynamic model to achieve precise matching of risk response and resource supply; Contingency plan generation and execution module: Transforms risk quantification results into executable emergency operation instructions; Exercise and self-learning module: acquire information collected by the multi-source data access module, update information from the linkage emergency resource map sub-module, and data from the contingency plan generation and execution module, and set optimization targets based on the threshold of the risk grading mechanism.

2. The cultural relic safety risk assessment and emergency decision-making system based on multi-source data fusion as described in claim 1, characterized in that, The data processing procedure of the spatiotemporal alignment engine further includes: Time dimension alignment uses a sliding window function W(T) current ,ΔT), where T current The current timestamp is ΔT, and the preset window duration is ΔT. The window... All input data within the frame are marked as a uniform time frame and then processed by the resampling function R(D) raw ,f s For the original data stream D raw Perform downsampling or interpolation to output the target frequency f. s The standard data stream uses a delay compensation function D(Td) to add an offset ΔT to the transmission delay Td. offset Spatial dimension alignment is achieved through the coordinate mapping function M. geo (X sensor ,Y sensor Z sensor The sensor's three-dimensional coordinates are converted into the surface-mapped coordinates (X) of the artifact model. m′ ,Y m′ Z m′ ), generating the continuous monitoring field function F(X) m′ ,Y m′ Z m′ ), and based on the topological correlation function T(P o Bind the monitoring point to the cultural relic component identifier P.

3. The cultural relic safety risk assessment and emergency decision-making system based on multi-source data fusion as described in claim 2, characterized in that, The weighting matrix further includes: Construct a monitoring matrix A = [m time points × n data sources] aij ], where elements aij This indicates that the j-th source has a normalized value at the i-th position, obtained through the information entropy function. Quantify the uncertainty of the data source and generate a difference coefficient function. The value that represents the utility of information, where g(j) reflects the volatility of the data; a larger value indicates a more significant amount of information from the source, determined by the dynamic weighting function. Allocate comprehensive weights, For real-time confidence Where λ is the attenuation coefficient, Δt is the transmission delay, and the weight w j Simultaneously considering both data value g(j) and reliability C(j), the influence weight of low-confidence sources is automatically reduced when data conflicts occur. base (j) To integrate inherent confidence levels, λ is the transmission delay attenuation factor; the entropy weighting method addresses data quality heterogeneity through the function chain E→G→Wj, generating a weighting matrix. ,in This represents the normalized data stream from the j-th data source after spatiotemporal alignment. This represents the dynamic weight value of the j-th data source.

4. The cultural relic safety risk assessment and emergency decision-making system based on multi-source data fusion as described in claim 1, characterized in that, The risk quantification engine submodule further includes: The weighted matrix Aw is generated by spatiotemporal alignment and entropy weighting, containing a spatiotemporally unified and confidence-weighted risk factor data stream. A comprehensive risk score is calculated using a weighted factor fusion model, as follows: function Φ( vi () is a normalization processor, physical quantity vi As input, the output range is [0,1], with dynamic weight W. i Generated using the ReliefF algorithm, its update mechanism is controlled by a self-learning algorithm formula: Where R actual To determine the actual damage value based on the degree of damage to cultural relics, R predict W is the system's predicted value. i =F(P type ), P type The cultural relic type identifier determines the basic weight allocation strategy for each risk factor. The F-weight decision function has a learning rate α to control the weight adjustment range and a constraint ∑Wi=1.

5. The cultural relic safety risk assessment and emergency decision-making system based on multi-source data fusion as described in claim 1, characterized in that, The risk classification submodule further includes: Receive the risk score value R generated by the risk quantification engine, and based on the numerical range of R, trigger differentiated emergency response strategies, discretizing the risk score value R into three-level response ranges: Low-risk response: When the R value is lower than the preset threshold R_low, preventive monitoring measures are implemented, and only the basic monitoring equipment attached to the cultural relic is activated; Medium-risk response: R value reaches or exceeds R_low but is lower than R_high, pre-deploy emergency resources; High-risk response: When the R value reaches or exceeds R_high, the cultural relic area will be forcibly closed and a joint cross-departmental response will be implemented.

6. The cultural relic safety risk assessment and emergency decision-making system based on multi-source data fusion as described in claim 1, characterized in that, The emergency resource mapping submodule further includes: It receives data collected in real time by the multi-source data access module. After the risk classification submodule determines the risk level, it sends a resource matching request instruction to the emergency resource map submodule to drive the subsequent scheduling process. The Emergency Resource Mapping submodule provides a standardized description of emergency response forces and resources. 1) Type attribute: Distinguishes the essential categories of emergency resources, including basic categories such as Expert, Material, and Team; 2) Spatial attribute Geo: Records the GPS coordinates of the resource, used to calculate the spatial distance between the disaster point and the resource, supporting optimal path planning; 3) Status attribute: Dynamically reflects availability, represented by binary logic or inventory level, ensuring that scheduling is based on real-time availability status; 4) Capability attribute: Quantifies the level of expertise in resources, expressed numerically.

7. The cultural relic safety risk assessment and emergency decision-making system based on multi-source data fusion as described in claim 6, characterized in that, The workflow of the emergency resource mapping submodule further includes: Phase 1: Emergency force data fusion, receiving data from multi-source data access modules, and unifying data format and update frequency through a spatiotemporal alignment engine; Phase Two: Dynamic Modeling. The emergency resource mapping submodule generates resource vectors based on four-dimensional attributes. Algorithm Formula. Type represents the type attribute, L lon , L lat "Status" and "Longitude" represent the longitude attribute, latitude attribute, and status attribute, respectively. score Indicates ability attributes; Phase 3: Response command execution. When the risk grading mechanism outputs an R value of medium to high risk, the intelligent matching process of the map is triggered. The current risk feature vector is extracted, and the matching degree between each resource vector and the risk feature is calculated using the cosine similarity algorithm. , Let be the matching degree scalar, representing the th The degree of compatibility between emergency resources and current risk events; This represents a risk feature vector, used to quantitatively describe the key attributes of the current risk event; Represents the magnitude of a vector; " represents the vector dot product operator. Indicates the first The feature vector of the emergency resource; where res_idx is the resource index number.

8. The cultural relic safety risk assessment and emergency decision-making system based on multi-source data fusion as described in claim 7, characterized in that, The contingency plan generation and execution module further includes: The contingency plan template matching submodule uses the discrete risk level output by the risk grading mechanism as input conditions to automatically trigger the retrieval and calling logic of the contingency plan library. Through the preset level-template mapping rules, the continuous risk score value R is converted into discrete operation instructions: when the system determines that the risk level reaches the preset threshold, the matching engine actively indexes the contingency plan template set of the corresponding level and outputs a structured task list. The emergency resource binding submodule is invoked through... The resource matching algorithm calculates the optimal resource, where V E Quantifying the multidimensional attributes of current risk, V C To construct a machine-analyzable feature space for resource feature vectors, the system consists of three dimensions: disaster type attribute, cultural relic vulnerability attribute, and risk level attribute. Abstract risks are transformed into computable vectors to provide input benchmarks for matching algorithms. The machine instruction execution submodule generates a standardized instruction set and distributes it to the terminal device cluster via an API interface.

9. The cultural relic safety risk assessment and emergency decision-making system based on multi-source data fusion as described in claim 8, characterized in that, The exercise and self-learning module further includes: Driven by a dual-engine approach of digital sandbox simulation and reinforcement learning, an extreme disaster scenario is constructed in a virtual environment. By dynamically executing contingency plan strategies and collecting feedback data, the system autonomously optimizes its decision-making model. The core algorithm of this autonomous optimization model consists of a reward function and a weight iteration formula. By quantitatively evaluating the effectiveness of contingency plan implementation, the continuous optimization process of the contingency plan library is driven, and an intelligent optimization engine is established. In this engine, Reward is the comprehensive reward value output, Ta is the actual handling time, Tp is the estimated time of the contingency plan, ΔR is the actual risk reduction value, α represents the response efficiency weight coefficient, and β represents the risk control weight coefficient.

10. A method for cultural relic safety risk assessment and emergency decision-making based on multi-source data fusion, characterized in that, The method includes: Multi-source data access steps: Real-time collection of the status of cultural relics, environmental parameters, security equipment operation data and emergency resource dynamic information through heterogeneous data fusion interface; Using a spatiotemporal alignment engine, the data collection frequency is unified through a sliding time window, and the weights of data sources with different confidence levels are dynamically assigned using the entropy weight method to generate a weighted matrix; The intelligent decision-making process further includes a risk quantification engine sub-step, a risk classification sub-step, and an emergency resource mapping sub-step. The risk quantification engine sub-step receives a weighted matrix from the multi-source data access module and calculates a comprehensive risk score through a weighted factor fusion model, providing dynamic and accurate decision-making basis for cultural relic safety protection. The risk grading sub-step transforms the comprehensive score output by the risk quantification engine sub-module into a core converter for actionable decisions, establishing a scientific decision-making chain from risk perception to emergency response. The emergency resource mapping sub-steps construct a machine-readable four-dimensional dynamic model to achieve precise matching of risk response and resource supply; Contingency plan generation and execution steps: Transform the risk quantification results into executable emergency operation instructions; Exercise and self-learning steps: Obtain information collected by the multi-source data access module, update information from the linkage emergency resource map sub-module, and data from the contingency plan generation and execution module, and set optimization targets based on the threshold of the risk grading mechanism.

Images