Property security inspection method and system and storage medium

Abstract

The invention discloses a property security inspection method and system and a storage medium, and belongs to the technical field of property management. Through combination of intelligent inspection and manual inspection, the intelligent inspection adopts a multi-source data fusion and dynamic threshold determination mode to realize alarm of different risk levels of intelligent inspection abnormity; manual inspection is used for supplementing intelligent inspection and completing regular inspection tasks such as re-checking of intelligent inspection abnormity alarm and inspection of facilities and equipment without sensors; the dynamic threshold value model has the adaptive capacity of different scenes; through machine learning, the system can be continuously optimized, and multi-source data fusion is combined, so that the false alarm rate and the missing report rate of the system are effectively reduced, the accuracy and the dynamic adaptability of intelligent inspection are improved, and accurate alarm is realized; the cost problem caused by blind installation of the sensor can be avoided, and quantitative decision support is provided for property management; the management cost of the property can be reduced, and the intelligent management level and the working efficiency of the property can be improved.

Description

Technical Field

[0001] This application relates to the technical field of property management, and particularly to a property security inspection method, system, and storage medium. Background Art

[0002] Property security inspection is an important task in property management to ensure the safety of places such as residential communities, office buildings, and shopping malls. Its purpose is to discover potential safety hazards, prevent risks, and safeguard the safety of personnel and property through regular inspections.

[0003] For traditional security inspections in property management areas, security personnel are generally arranged to conduct inspections at fixed times and locations, resulting in the inability to promptly detect abnormal security inspection situations and poor security inspection effects in property management. Currently, some improved solutions, such as the Chinese patent with the publication number CN117634837A, disclose a property management security inspection system and method. By installing sensors and monitoring cameras to collect security inspection information, processing the data to extract feature information, and then comparing the feature information with standard information to determine the security inspection analysis result, it can conduct real-time security inspections in the property management area, reduce the workload of security personnel, promptly detect abnormal security inspection situations, and improve the security inspection effect in property management.

[0004] Although the above solutions solve some problems existing in traditional security inspections, there are still the following technical defects: 1. The existing system relies on a single sensor for anomaly detection and lacks the ability to fuse and analyze multi-modal data, resulting in a high false alarm rate; 2. The traditional threshold setting is fixed and cannot be automatically adjusted according to environmental changes, and the system lacks a self-optimization mechanism and cannot learn from historical false alarm or missed alarm events, leading to repeated errors and poor dynamic adaptability; 3. The results of manual inspections are not linked with the intelligent system, and equipment maintenance relies on experience judgment, lacking data-driven decision support; 4. The cost of installing sensors for all equipment is high, especially for the renovation of old properties, which is unrealistic. Moreover, manual inspection records are mostly paper forms, making it difficult to trace and analyze, and unable to form a closed-loop management, resulting in an imbalance between cost and efficiency. Summary of the Invention

[0005] Therefore, in view of the problems and needs existing in the above-mentioned prior art, this application is proposed. The purpose of this application is to provide a property security inspection method, system, and storage medium, which, compared with the defects of the prior art, can effectively reduce the false alarm rate and missed alarm rate of the system, improve the accuracy and dynamic adaptability of intelligent inspections, and achieve accurate alarms; can avoid the cost problems caused by blindly installing sensors, provide quantitative decision support for the property; can reduce the management cost of the property, improve the intelligent management level and work efficiency of the property.

[0006] The purpose of this application is achieved through the following technical solutions:

[0007] In a first aspect, an embodiment of the present application provides a property security inspection method, including the following steps:

[0008] Real-time collection of multi-source environmental data in the property management area through intelligent inspection devices, where the multi-source environmental data includes temperature, humidity, smoke concentration, combustible gas concentration, equipment operating status, infrared images, and video surveillance data;

[0009] Perform spatio-temporal alignment and multi-modal fusion processing on the multi-source environmental data to generate comprehensive inspection feature information;

[0010] According to a preset dynamic threshold model, perform real-time analysis on the comprehensive inspection feature information to determine whether there is an abnormal situation; if it is determined that there is an abnormality, generate an abnormal alarm message and perform task allocation;

[0011] The inspection personnel or mobile inspection robot go to the designated location according to the abnormal alarm message for verification and processing, and feedback the processing result to the system; the inspection personnel also perform inspections according to the regular inspection tasks generated by the system, including equipment function verification and physical damage inspection;

[0012] The system optimizes the dynamic threshold model and subsequent inspection strategies according to the feedback results of manual inspections or mobile inspection robots, and dynamically adjusts the equipment maintenance priority.

[0013] In the above-mentioned property security inspection method, the dynamic threshold model is generated in the following way: based on historical inspection data, combined with factors such as time, weather, and regional function attributes, dynamically set the initial thresholds of each environmental data; analyze the false alarm or missed alarm data in the review results through machine learning algorithms, and dynamically optimize the threshold rules.

[0014] In the above-mentioned property security inspection method, the multi-modal fusion processing includes: performing time synchronization and spatial alignment on data from different sensors; using a weighted fusion algorithm or a deep learning model to fuse the multi-source environmental data into unified inspection feature information.

[0015] In the above-mentioned property security inspection method, the allocation logic of the manual inspection task includes: priority classification: the intelligent inspection abnormal alarm event review is of high priority, and the regular inspection task is of low priority; path planning: generate the optimal inspection path according to the task location to reduce repeated movement; skill matching: electricians are preferentially assigned equipment detection tasks, and security personnel are preferentially assigned intelligent inspection abnormal alarm event review tasks and physical damage inspection tasks.

[0016] In the above-mentioned property security inspection method, the verification of equipment functions includes: sending control instructions through the manual inspection terminal APP to verify the response speed and normal functions of the equipment; if any abnormality is found, photos of the faulty equipment, fault descriptions, and equipment locations are uploaded to the system, and the system automatically generates maintenance work orders.

[0017] In a second aspect, an embodiment of the present application provides a property security inspection system, including the following modules:

[0018] An intelligent inspection module, configured to collect multi-source environmental data of the property management area in real time through intelligent inspection equipment, then perform spatio-temporal alignment and multi-modal fusion processing on the collected multi-source environmental data to generate comprehensive inspection feature information, and then analyze the comprehensive inspection feature information based on a dynamic threshold model to determine whether there is an abnormal situation. If it is determined that there is an abnormal situation, an abnormal alarm message is generated.

[0019] A human-machine collaborative management module, configured to automatically assign abnormal alarm events discovered by intelligent inspection to the manual inspection terminal APP or mobile inspection robot, and generate a route to the target location by the manual inspection terminal APP; it is also configured to generate regular manual inspection tasks, including verification of equipment functions and inspection of physical damages.

[0020] A feedback optimization module, which optimizes the dynamic threshold model and subsequent inspection strategies according to the abnormal alarm review results and equipment status data submitted by manual inspection or mobile inspection robots, and dynamically adjusts the equipment maintenance priority.

[0021] In the above-mentioned property security inspection system, the human-machine collaborative management module includes:

[0022] A task assignment unit, configured to assign inspection tasks according to the task assignment logic, and the task assignment logic includes:

[0023] Priority classification: The review of intelligent inspection abnormal alarm events is of high priority, and regular inspection tasks are of low priority;

[0024] Path planning: Generate the optimal inspection path according to the task location to reduce repeated movement;

[0025] Skill matching: Electricians are preferentially assigned equipment inspection tasks, and security personnel are preferentially assigned tasks for reviewing intelligent inspection abnormal alarm events and inspecting physical damages;

[0026] A review verification unit, configured to feed back the review results of abnormal alarm events by manual or mobile inspection robots to the system;

[0027] A cost optimization unit, configured to calculate the cost-benefit ratio of installing intelligent inspection equipment for equipment and facilities that require manual inspection, and generate a recommended plan.

[0028] In the above-mentioned property security inspection system, the system further includes:

[0029] A knowledge base module, which is used to store equipment detection standards and push operation guides to the manual inspection terminal in real time;

[0030] A visualization display module, which is used for visualizing abnormal alarm information, abnormal alarm accuracy rate, equipment failure rate, inspection data, inspection tasks and task progress, and maintenance work orders, and three-dimensional marking of the sensor coverage area and blind spots for managers to monitor and analyze in real time;

[0031] A maintenance prediction module, which is used to analyze historical maintenance data, combine with the life parameters of the equipment, and calculate the remaining service life; when the remaining life of the equipment is lower than the threshold, a purchase requisition is automatically generated.

[0032] In the above-mentioned property security inspection system, the manual inspection terminal has the following functions: AR-assisted detection, which is used to automatically compare the current status based on standard parameters when scanning equipment and find abnormal situations; voice reporting, which describes the fault phenomenon through natural language, and the system automatically converts it into a maintenance work order.

[0033] In a third aspect, an embodiment of the present application provides a computer-readable storage medium, on which computer program instructions are stored. When the computer program instructions are run by a processor, the processor is caused to execute the property security inspection method as described above.

[0034] Compared with the prior art, the beneficial effects of the present invention are at least as follows:

[0035] The property security inspection method, system and storage medium of the present application combine intelligent inspection with manual inspection. Among them, intelligent inspection uses multi-source data fusion and dynamic threshold determination methods to achieve different risk level alarms for intelligent inspection anomalies; manual inspection is a supplement to intelligent inspection, completing the review of intelligent inspection anomalies and regular inspection tasks such as inspecting facilities and equipment without installed sensors, solving the problem of high cost for installing sensors to cover all equipment; the dynamic threshold model dynamically sets the initial threshold of each environmental data based on historical inspection data, combined with factors such as time, weather, and regional function attributes, and has the adaptive ability for different scenarios; by dynamically optimizing the threshold rules through machine learning, the system can continuously optimize, combined with multi-source data fusion, effectively reducing the false alarm rate and missed alarm rate of the system, improving the accuracy and dynamic adaptability of intelligent inspection, and achieving accurate alarm.

[0036] In the property security inspection system of the present application, the cost optimization unit avoids the cost problems caused by blindly installing sensors, and provides quantitative decision-making support for the property; the knowledge base module converts manual experience into system rules, and with the AR-assisted detection function, non-professional technicians can also conduct inspections, improving the inspection efficiency and detection accuracy; the visual display module realizes the visualization of data and task progress, the progress can be traced, and the visual monitoring blind area guides manual key inspections, improving the intelligent management level and work efficiency of the property; the maintenance prediction module facilitates property management personnel to maintain equipment before security incidents occur, reducing the management cost of the property and improving work efficiency; when abnormalities are found during manual regular inspections, photos of faulty equipment and fault descriptions (which can be reported by voice to reduce the work intensity of inspectors) are uploaded, automatically associating the equipment location and equipment model to the system. The system automatically generates maintenance level marks and repair work orders and triggers subsequent repair processes, and can dynamically adjust the equipment maintenance priority, improving the intelligent level of property management while reducing the management cost of the property. The system dynamically adjusts the inspection cycle of corresponding equipment according to the failure rate and usage frequency data of facilities and equipment, optimizes the inspection strategy, and ensures the efficient operation of property security inspections. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] The above and other objects, features, and advantages of the present application will become more obvious by describing the embodiments of the present application in more detail in conjunction with the accompanying drawings. The accompanying drawings are used to provide a further understanding of the embodiments of the present application, and constitute a part of the specification. Together with the embodiments of the present application, they are used to explain the present application and do not constitute a limitation to the present application.

[0038] Figure 1 is a schematic diagram of a property security inspection method provided by an embodiment of the present application;

[0039] Figure 2 is a schematic structural diagram of a property security inspection system provided by an embodiment of the present application. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0040] Next, exemplary embodiments according to the present application will be described in detail with reference to the accompanying drawings. It should be understood that the specific embodiments described herein are only a part of the embodiments of the present application, and the present application is not limited by the exemplary embodiments described herein.

[0041] APPLICATION OVERVIEW

[0042] The inventor has found that for the security patrol in the property management area, security personnel are generally used to conduct patrols at fixed times and locations, resulting in the inability to detect abnormal situations in security patrols in a timely manner, and the poor effect of property management security patrols. At present, some improvement schemes, such as the Chinese patent with the publication number CN117634837A, disclose a property management security patrol system and method. By installing sensors and monitoring cameras to collect security patrol information, processing the data to extract feature information, and then comparing the feature information with the standard information to determine the security patrol analysis result, it can conduct security patrols in the property management area in real time, reduce the workload of security personnel, detect abnormal situations in security patrols in a timely manner, and improve the effect of property management security patrols. Although this scheme solves some problems existing in traditional security patrols, there are still the following technical defects: 1. The existing system relies on a single sensor for abnormal detection, lacking the ability of multi-modal data fusion analysis, resulting in a high false alarm rate; 2. The traditional threshold setting is fixed and cannot be automatically adjusted according to environmental changes, and the system lacks a self-optimization mechanism and cannot learn from historical false alarm or missed alarm events, resulting in repeated errors and poor dynamic adaptability; 3. The results of manual inspections are not linked with the intelligent system, and equipment maintenance relies on empirical judgments, lacking data-driven decision support; 4. The cost of installing sensors for all equipment is high, especially for the renovation of old properties, which is unrealistic. Moreover, the manual patrol records are mostly paper forms, which are difficult to trace and analyze, and cannot form a closed-loop management, resulting in an imbalance between cost and efficiency.

[0043] Therefore, in view of the above defects in the prior art, the inventor has proposed a property security patrol method, system and storage medium. The basic idea is to combine intelligent patrols with manual patrols. Among them, intelligent patrols adopt the method of multi-source data fusion and dynamic threshold determination to improve the accuracy of intelligent patrols. Manual patrols are a supplement to intelligent patrols, used for the review of abnormal alarms in intelligent patrols and the completion of regular inspection tasks such as the inspection of facilities and equipment without installed sensors. The feedback results of manual patrols are used to optimize the dynamic threshold model and subsequent patrol strategies.

[0044] After introducing the basic idea of the present application, the various non-limiting embodiments of the present application will be specifically introduced below with reference to the accompanying drawings.

[0045] Embodiment 1

[0046] As Figure 1 shown, the embodiment of the present application provides a property security patrol method, including steps S1 - S5.

[0047] In step S1, multi-source environmental data of the property management area is collected in real time through intelligent patrol equipment. The multi-source environmental data includes temperature, humidity, smoke concentration, combustible gas concentration, equipment operation status, infrared images and video surveillance data.

[0048] The environmental data of the property management area is the basis for subsequent intelligent patrol inspection analysis. Temperature, humidity, smoke concentration, combustible gas concentration, equipment operation status, infrared images, and video surveillance data can be collected through intelligent patrol inspection devices such as fixed monitoring devices, such as temperature sensors, smoke sensors, humidity sensors, combustible gas detectors, infrared cameras, and high-definition surveillance cameras. It can also be obtained through mobile patrol robots equipped with various sensors. Fixed monitoring devices are reasonably installed at key positions in the property management area. At the same location, data can be collected by mixing multiple monitoring devices or a single type of monitoring device based on the characteristics of the building area. For example, combustible gas detectors and smoke sensors are deployed in areas where combustible gas is likely to accumulate, such as kitchens and boiler rooms; smoke sensors are installed every three floors in the stairwells of high-rise residential buildings; distributed optical fiber temperature sensors and electrical fire detectors are densely laid in power distribution rooms and electrical shafts; high-definition surveillance cameras are installed in high-risk areas. The data collected by each device is transmitted to the data processing unit for subsequent processing based on a preset frequency.

[0049] In step S2, spatio-temporal alignment and multi-modal fusion processing are performed on the multi-source environmental data to generate comprehensive patrol inspection feature information.

[0050] In the data processing unit of the intelligent patrol inspection module, first, time synchronization and spatial alignment processing are performed on the data collected from different sensors or surveillance cameras, and a filtering algorithm is used to process the data of a single sensor to remove noise interference, ensuring the relative smoothness and reliability of the input data and laying a foundation for subsequent accurate analysis. Then, fusion processing is performed on the data of multiple sensors within the same spatial range to generate comprehensive patrol inspection feature information. For scenarios where only a single monitoring device is installed within the same spatial range, a filtering algorithm is used to process the sensor data to remove noise interference or extract video surveillance image features for subsequent analysis.

[0051] In step S3, according to the preset dynamic threshold model, real-time analysis is performed on the comprehensive patrol inspection feature information to determine whether there is an abnormal situation. If it is determined that there is an abnormality, an abnormal alarm information is generated and task allocation is performed.

[0052] For a scenario where only a single monitoring device is installed within the same spatial range, if it is a sensor device, the filtered data is compared and analyzed with a preset safety threshold. If it exceeds the preset safety threshold range, it is determined that an anomaly has occurred, and anomaly alarm information of different risk levels is generated according to different multiples of exceeding the preset safety threshold. For example, if the filtered data is greater than or equal to 2 times the preset safety threshold, high-risk anomaly alarm information is generated; within the range of 1 to 2 times the preset safety threshold, low-risk anomaly alarm information is generated. If it is a video surveillance or infrared image, the extracted image features are compared and analyzed with the preset dangerous behavior or event features to detect anomalies, and anomaly alarm information of different levels is generated according to the preset risk levels.

[0053] For a scenario where there are multiple monitoring devices within the same spatial range, the comprehensive inspection feature information generated after multi-modal fusion processing in step S2 is compared and analyzed with a preset dynamic threshold model. If it exceeds the safety threshold, it is determined that a high-confidence anomaly event has occurred, and high-risk level anomaly alarm information is generated. If the comprehensive inspection feature information does not exceed the safety threshold but the data of a certain sensor is abnormal, low-risk level anomaly alarm information is generated.

[0054] The system performs priority management of task allocation according to the risk levels of the anomaly alarm information. For high-risk level anomaly alarm information, personnel are preferentially allocated for inspection and verification, and the allocated personnel comply with the principle of proximity. For low-risk level anomaly alarm information, in the scenario where a mobile inspection robot exists, the robot is preferentially allocated for verification to reduce the labor intensity of personnel. If there is no mobile inspection robot, nearby personnel are allocated for inspection and verification.

[0055] In step S4, the inspection personnel or the mobile inspection robot go to the designated location according to the anomaly alarm information for verification and processing, and feedback the processing results to the system. The inspection personnel also perform inspections according to the regular inspection tasks generated by the system, including equipment function verification and physical damage inspection.

[0056] The mobile inspection robot constructs an environmental map through SLAM technology and generates an optimal path according to the anomaly alarm information to go to the designated location for verification, and uploads the on-site video captured by the equipped high-definition camera to the system for event verification. The inspection personnel go to the designated location for anomaly alarm verification based on the route guidance generated by the anomaly alarm information through the inspection terminal, and feedback false alarms or normal alarm events to the system for recording and storage.

[0057] The patrol personnel also execute the regular patrol tasks generated by the system, including equipment function verification and physical damage inspection. It is understandable that the cost of installing sensors for all equipment is high, especially for the renovation of old properties, which is not practical. For example, for non-emergency patrol tasks such as the inspection of lighting equipment, fire protection facilities, and public facilities, the cost of installing sensors for detection is relatively high, and subjective judgment by humans is required in some scenarios. Therefore, the method of regular personnel patrol can be adopted to solve this problem. The system generates non-emergency patrol tasks according to the preset cycle. The patrol personnel receive the task list through the terminal device, execute the tasks in the order of priority, record and feedback the inspection results. The system automatically marks the inspection results as "normal", "to be maintained", and "urgent repair" at three levels according to the feedback information. The equipment marked as "to be maintained" automatically generates a maintenance work order, and the equipment marked as "urgent repair" immediately triggers an alarm and notifies the management personnel.

[0058] Preferably, the allocation logic of the manual patrol tasks by the system includes:

[0059] Priority classification: The review of intelligent patrol abnormal alarm events is of high priority, and the regular patrol tasks are of low priority. Moreover, the review of intelligent patrol abnormal alarm events is preferentially assigned to the nearest personnel;

[0060] Path planning: Generate the optimal patrol path according to the task location, and generate associated inspection tasks for other equipment near the equipment to be inspected, reducing the repeated movement of personnel;

[0061] Skill matching: Electricians are preferentially assigned equipment detection tasks, and security personnel are preferentially assigned the tasks of reviewing intelligent patrol abnormal alarm events and physical damage inspection tasks.

[0062] Preferably, the equipment function verification in the manual patrol includes:

[0063] Send a control instruction through the manual patrol terminal APP to verify whether the response speed and function of the equipment are normal; if an abnormality is found, upload photos of the faulty equipment and the fault description, automatically associate the equipment location and equipment model to the system. The system automatically generates a maintenance work order for the equipment marked as "to be maintained" and triggers the subsequent maintenance process. For the equipment marked as "urgent repair", an alarm is immediately triggered and the management personnel are notified.

[0064] In step S5, the system optimizes the dynamic threshold model and subsequent patrol strategies according to the feedback results of the manual patrol or mobile patrol robot, and dynamically adjusts the equipment maintenance priority.

[0065] An artificial or mobile inspection robot conducts intelligent inspection anomaly alarm review and feeds back the review results to the system. The system, based on the feedback of false alarms or normal alarm results and intelligent inspection missed alarm events added through the management background, uses machine learning algorithms to dynamically optimize and adjust the threshold ranges of various environmental data. For example, if the abnormal alarms of a smoke sensor in a certain area are finally confirmed as false alarms for 2 consecutive times, the alarm threshold of this smoke sensor will be increased by 10% - 15%. If the alarm in this area is triggered by multi-source data fusion processing at the same time, the weight of the smoke sensor during weighted fusion processing will be reduced; if a smoke sensor in a certain area misses an alarm during a fire incident, the alarm threshold of this smoke sensor will be decreased by 10% - 15%, and at the same time, the weight of the smoke sensor during weighted fusion processing will be increased.

[0066] The system, based on the inspection results of non-emergency tasks fed back by artificial inspections or mobile robots, statistically analyzes the failure rate and usage frequency data of facilities and equipment, and dynamically adjusts the inspection cycles of corresponding equipment; based on the feedback information marked as "to be maintained" and "urgent repair" in the inspection results, it dynamically adjusts the equipment maintenance priorities.

[0067] The property security inspection method of this embodiment collects multi-source environmental data through fixed monitoring devices and mobile inspection robots, providing solid data support for subsequent intelligent inspection analysis; for the scenario where there is a single monitoring device in the same monitoring area, the sensor data is processed through a filtering algorithm to remove noise interference or extract video image features, and then different risk-level alarm information is generated by comparing with preset thresholds or preset dangerous behaviors. Based on the review feedback of artificial or inspection robots, the preset thresholds are adjusted, reducing the false alarm rate and improving the accuracy and dynamic adaptability of intelligent inspections.

[0068] For the scenario where there are multiple monitoring devices in the same monitoring area, the multi-source environmental data is subjected to spatio-temporal alignment and multi-modal fusion processing to generate comprehensive inspection feature information. Then, according to the preset dynamic threshold model, the comprehensive inspection feature information is analyzed in real time to determine whether there are abnormal situations and generate abnormal alarm information of different risk levels. Based on the task assignment rules, artificial or mobile robots are assigned for review feedback. Then, based on the false alarm information in the feedback and the missed alarm information added in the background, machine learning algorithms are used to dynamically optimize and adjust the threshold ranges of various environmental data and the corresponding weights during weighted fusion processing, avoiding misjudgment caused by relying on single-sensor data, reducing the false alarm rate and missed alarm rate of the system, and improving the accuracy and dynamic adaptability of intelligent inspections.

[0069] The manual inspection and execution system also performs regular inspection tasks for non-emergency events such as inspections of lighting equipment, fire protection facilities, and public facilities assigned based on task allocation rules, solving the problem of the high cost of installing sensors for all equipment coverage. The manual inspection terminal can send control instructions to verify the response speed and normal function of the equipment, and has a knowledge base function that stores equipment detection standards and pushes operation guidelines to the manual inspection terminal in real time. It also has an AR-assisted detection function, improving the detection accuracy and efficiency of non-technical personnel. If an abnormality is found, it uploads photos of the faulty equipment and a fault description, automatically associates the equipment location and equipment model to the system. The system automatically generates a maintenance level mark, automatically generates a repair work order for the equipment marked as "to be maintained" and triggers the subsequent repair process. For the equipment marked as "urgent repair", it immediately triggers an alarm and notifies the management personnel, dynamically adjusts the equipment maintenance priority, improves the intelligent level of property management, and at the same time reduces the property management cost. The system dynamically adjusts the inspection cycle of the corresponding equipment according to the failure rate and usage frequency data of the facilities and equipment, optimizes the inspection strategy, and ensures the efficient operation of the property security inspection.

[0070] Embodiment 2

[0071] The dynamic threshold model is generated in the following way:

[0072] Based on historical inspection data, combined with factors such as time, weather, and regional function attributes, the initial thresholds of various environmental data are dynamically set;

[0073] Analyze the false alarm or missed alarm data in the review results through machine learning algorithms, and dynamically optimize the threshold rules.

[0074] Specifically, for example, in the property security fire alarm monitoring scenario of a commercial complex, it is necessary to set dynamic thresholds for environmental data (temperature, smoke concentration, combustible gas concentration) in different areas (such as kitchens, warehouses, and office areas) to achieve accurate fire alarm monitoring. Traditional fixed thresholds result in frequent false alarms in the kitchen area due to daily cooking, and there is a risk of missed alarms in the warehouse due to differences in the ignition points of stored items. This embodiment solves the above problems through a dynamic threshold model. Among them, it is dynamically set based on regional function attributes. For example, the combustible gas concentration threshold in the kitchen area is higher than that in the ordinary office area, and the smoke concentration threshold in the warehouse area is dynamically adjusted according to the ignition points of the stored items. Weather and time factors also affect the detection accuracy of sensors. For example, in high-temperature and dry weather, the alarm threshold of the smoke sensor should be correspondingly reduced. During the peak cooking period at noon on weekdays, the smoke concentration alarm threshold in the kitchen area should be correspondingly increased to avoid frequent false alarms, and it needs to be set according to the historical inspection data (environmental data, false alarm or missed alarm records) of this area.

[0075] Using historical patrol inspection data as the data source, through the random forest algorithm of machine learning, input environmental data, external factors (weather type), regional type, and time factors (time periods, such as weekdays or holidays, day or night). After preprocessing the data by normalizing (scaling numerical features such as temperature and smoke concentration to the [0,1] interval) and time encoding (converting time periods into binary labels), 80% of the historical data is selected as the training set and 20% as the test set. Taking whether it is a real fire alarm as the target variable and a prediction accuracy rate of 95% as the training goal, the model is trained.

[0076] Among them, the dynamic threshold optimization rules include:

[0077] False alarm: If a certain area has false alarms continuously for 2 times, then the smoke threshold is increased by 10% - 15%, and the temperature threshold is increased by 5% - 10%.

[0078] Missed alarm: If a missed alarm occurs in a certain area, adjust the weights and thresholds during weighted fusion processing according to the feature importance. For example, if the smoke sensor has a missed alarm, then increase the weight during weighted fusion processing of the smoke feature, and lower the smoke threshold by 10% - 15%.

[0079] The trained dynamic threshold model is applied in daily intelligent patrol inspections, and then the results of manual or mobile patrol robots' rechecks after each abnormal alarm are entered into the database as the training data for the next round; the system triggers model incremental training regularly, such as once a week, to continuously update the dynamic threshold model.

[0080] The dynamic threshold model of this embodiment is based on historical patrol inspection data, combines time, weather, and regional functional attribute factors, dynamically sets the initial thresholds of various environmental data, and has the adaptive ability for different scenarios; by analyzing false alarm or missed alarm data in the recheck results through machine learning algorithms and dynamically optimizing the threshold rules, the system can continuously optimize, reduce the false alarm rate and missed alarm rate, improve the accuracy of abnormal alarms, and achieve accurate alarms.

[0081] Embodiment 3

[0082] The multi-modal fusion processing includes:

[0083] Perform time synchronization and spatial alignment on data from different sensors;

[0084] Adopt a weighted fusion algorithm or a deep learning model to fuse multi-source environmental data into unified patrol inspection feature information.

[0085] Specifically, spatio-temporal alignment is a core technology for multi-sensor data fusion, aiming to solve the data association problems caused by inconsistent timestamps and spatial position deviations among different sensors. Time synchronization needs to ensure that multi-source data is aligned within a millisecond-level error range to support real-time analysis. The implementation method is to set up a buffer in the data processing unit to temporarily store multi-source data according to a time window (such as 100 ms). If the data of a certain sensor is missing within the window, linear interpolation is used to complete it (for example, when temperature data is missing, the current value is calculated based on the data at the previous and subsequent moments), and the system robustness is significantly enhanced; downsampling or upsampling is performed on high-frequency data (such as video stream at 30 fps) and low-frequency data (such as gas sensor at 1 Hz) to unify them to the same time granularity.

[0086] The implementation method of spatial alignment is to record the physical position (such as XYZ coordinates) of each fixed sensor during deployment and associate it with the system map. The mobile inspection robot constructs the environmental map in real time through lidar and visual SLAM (simultaneous localization and mapping), locates its own coordinates, and then converts the sensor data from the local coordinate system to the global coordinate system through coordinate system transformation. For example, when the mobile inspection robot detects a combustible gas leak on the garage floor B1 (local coordinates: X = 10.2 m, Y = 5.3 m), it needs to fuse the data with that of a fixed sensor (coordinates: X = 10.5 m, Y = 5.1 m). The spatial alignment process includes:

[0087] (1) The garage map constructed by the mobile inspection robot through SLAM is matched with the system global map, and its own coordinates are corrected to X = 10.4 m, Y = 5.2 m;

[0088] (2) The system maps the combustible gas concentration data detected by the robot to the global coordinate system and superimposes and analyzes it with the fixed sensor data;

[0089] (3) If the data of the two overlaps in space (error ≤ 0.5 m), it is determined to be the same spatial area.

[0090] The following describes the process of multi-source data fusion processing in combination with a specific scenario. For example, when a cooking oil fire breaks out in the dining area of a large shopping mall, the system needs to fuse the real-time data of temperature sensors, smoke sensors, combustible gas sensors, and surveillance cameras to quickly determine a fire alarm and trigger an emergency response. The original data of the temperature sensor is 85 °C, and the threshold is 60 °C. The original data of the smoke sensor is 15 mg / m 3 , and the threshold is 20 mg / m 3 , the original data of the combustible gas sensor is 200 ppm, and the threshold is 500 ppm. First, the data collected by the temperature sensor, smoke sensor, and combustible gas sensor are normalized, and the normalized values are 0.85, 0.75, and 0.40 respectively. Then, according to the preset weights of each sensor data in this scenario (assuming the temperature weight wT = 0.4, smoke weight wS = 0.4, combustible gas weight wG = 0.2) to calculate the comprehensive risk value, and its calculation formula is:

[0091] R = w T ×T norm + w S ×S norm + w G ×G norm ;

[0092] Among them, T norm is the temperature normalization value, S norm is the smoke concentration normalization value, G norm is the combustible gas concentration normalization value. Then, according to the preset comprehensive risk value threshold R th compare the comprehensive risk value R. If R ≥ R th , it is determined as a high-confidence fire alarm, and a high-risk level abnormal alarm message is generated; otherwise, if the data of a certain sensor exceeds the threshold, a low-risk level abnormal alarm message is generated.

[0093] Identify the flame characteristics in the monitoring video through the deep learning model, and output the flame probability. If the result is greater than the preset probability threshold, it is determined as a real flame. Combine the comprehensive risk value to determine the risk level. If the comprehensive risk value is greater than the preset threshold, it is determined as a real fire alarm event, generate a high-risk level abnormal alarm message, and immediately trigger the review mechanism; if only the flame probability is greater than the preset probability threshold, generate a low-risk level abnormal alarm message.

[0094] The multi-modal fusion processing in this embodiment performs time synchronization and spatial alignment on the data from different sensors, adopts a weighted fusion algorithm or a deep learning model, and fuses the multi-source environmental data into unified inspection feature information, which can effectively balance efficiency and accuracy, effectively solve the problem of high false alarm rate caused by relying on a single sensor through data fusion, improve the accuracy of fire alarm determination, and achieve accurate hierarchical alarm.

[0095] Embodiment 4

[0096] As Figure 2 shown, the embodiment of the present application provides a property security inspection system, including the following modules:

[0097] Intelligent inspection module, which is used to collect multi-source environmental data (temperature and humidity, combustible gas concentration, smoke concentration, infrared images, surveillance videos, etc.) of the property management area in real time through intelligent inspection devices (temperature and humidity sensors, smoke sensors, combustible gas sensors, infrared cameras, high-definition surveillance cameras, mobile inspection robots, etc.) in the data acquisition unit, then perform spatio-temporal alignment and multi-modal fusion processing on the collected multi-source environmental data in the data processing unit to generate comprehensive inspection feature information, and then analyze the comprehensive inspection feature information by the data analysis and comparison unit based on the dynamic threshold model to determine whether there is an abnormal situation. If it is determined that there is an abnormality, an abnormal alarm message will be generated;

[0098] Human-machine collaborative management module, which is used to automatically allocate abnormal alarm events found by intelligent inspection to the manual inspection terminal APP or mobile inspection robot, and generate a route to the target location by the manual inspection terminal APP; it is also used to generate regular manual inspection tasks, including equipment function verification and physical damage inspection;

[0099] Feedback optimization module, which optimizes the dynamic threshold model and subsequent inspection strategies according to the abnormal alarm review results and equipment status data submitted by manual inspection or mobile inspection robots, and dynamically adjusts the equipment maintenance priority;

[0100] Among them, the human-machine collaborative management module includes:

[0101] Task allocation unit, which is used to allocate inspection tasks according to the task allocation logic. For abnormal alarms with a low risk level in intelligent inspection, mobile inspection robots are preferentially arranged for review, and the mobile inspection robot also collects environmental data according to the daily inspection plan to provide data support for the multi-source data fusion step and cover the inspection of blind spots. Among them, the manual inspection task allocation logic includes:

[0102] Priority classification: The review of abnormal alarm events in intelligent inspection is of high priority, and regular inspection tasks are of low priority;

[0103] Path planning: Generate the optimal inspection path according to the task location to reduce repeated movement;

[0104] Skill matching: Electricians are preferentially allocated equipment detection tasks, and security personnel are preferentially allocated tasks for reviewing abnormal alarm events in intelligent inspection and physical damage inspection;

[0105] Review and verification unit, which is used for manual or mobile inspection robots to feedback the review results of abnormal alarm events to the system. The review by the mobile inspection robot is to collect environmental data or on-site videos and upload them for the system or administrator to judge the authenticity of the abnormal alarm, and record the final determination result of the review;

[0106] A cost optimization unit is used to calculate the cost-benefit ratio of installing intelligent inspection devices for equipment and facilities that require manual inspection, and generate a recommended solution. For example, if an important lighting device frequently malfunctions and requires frequent manual inspections, the cost of installing sensors can be calculated at this time, and then compared with the manual inspection cost. If the cost of installing sensors is less than the manual cost, it is recommended to install sensors to avoid cost problems caused by blindly installing sensors and provide quantitative decision-making support;

[0107] Preferably, the system further includes:

[0108] A knowledge base module is used to store equipment detection standards and push operation guides to the manual inspection terminal in real time, so that non-professional technical personnel can also perform inspections, improving inspection efficiency and detection accuracy;

[0109] A visualization display module is used for visualizing abnormal alarm information, abnormal alarm accuracy rate, equipment failure rate, inspection data, inspection tasks and task progress, and maintenance work orders, and three-dimensional marking of the sensor coverage area and blind spots, for managers to monitor and analyze in real time. The data and task progress are visualized, the progress can be traced, and the intelligent management level and work efficiency of the property are improved;

[0110] A maintenance prediction module is used to analyze historical maintenance data, combine with the life parameters of the equipment, and calculate the remaining service life; when the remaining life of the equipment is lower than the threshold, a purchase requisition form is automatically generated, facilitating property managers to maintain the equipment before a safety incident occurs, reducing the management cost of the property and improving the work efficiency of the property.

[0111] Preferably, the manual inspection terminal has the following functions:

[0112] AR-assisted detection is used to automatically compare the current status based on standard parameters when scanning the equipment, and detect abnormal situations. For example, when an artificial inspection terminal integrated with the function of identifying pressure sensors checks a fire extinguisher and scans the fire extinguisher label, the AR interface overlays the standard pressure range, and the current pressure value of the fire extinguisher is detected in real time through the pressure sensor. If the current pressure value is less than the standard value, a maintenance work order is automatically generated and marked as "urgent repair";

[0113] Voice reporting, describing the fault phenomenon through natural language, and the system automatically converts it into a maintenance work order, reducing the manual input work of the inspection personnel and improving the inspection efficiency.

[0114] Those skilled in the art can understand that the specific functions and operations of each module and each unit in the property security inspection system of this embodiment are introduced in detail in a property security inspection method described above. Therefore, the repeated description will be omitted here.

[0115] The property security inspection system of this embodiment has a cost optimization unit that avoids the cost problems caused by blindly installing additional sensors and provides quantitative decision-making support for the property; a knowledge base module that converts manual experience into system rules, enabling non-professional technical personnel to conduct inspections, improving the inspection efficiency and detection accuracy; a visual display module that visualizes data and task progress, with the progress being traceable, and visualizing monitoring blind spots to guide manual key inspections, improving the intelligent management level and work efficiency of the property; a maintenance prediction module that facilitates property management personnel to maintain equipment before security incidents occur, reducing the management cost of the property and improving the work efficiency of the property; AR-assisted detection and voice reporting, reducing the work intensity of inspection personnel and improving the inspection efficiency.

[0116] Embodiment 5

[0117] In addition to the above methods and systems, an embodiment of the present application may also be a computer-readable storage medium storing computer program instructions, which, when run by a processor, cause the processor to execute the steps in the property security inspection method according to various embodiments of the present application described above in this specification.

[0118] The computer-readable storage medium may adopt any combination of one or more readable media. The readable media may be a readable signal medium or a readable storage medium. The readable storage medium may, for example, include but is not limited to an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination of the above. More specific examples (non-exhaustive list) of the readable storage medium include: an electrical connection having one or more wires, a portable disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above.

[0119] The basic principles of the present application have been described above in conjunction with specific embodiments. It should be understood that the above-disclosed specific details are only for the purpose of illustration and easy understanding, rather than limitation, and are not used to limit the protection scope of the present invention. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principles of the present invention shall be included within the protection scope of the present invention.

Claims

1. A property security inspection method, characterized in that: The following steps are involved: Collect multi-source environmental data of the property management area in real time through intelligent inspection equipment, including temperature, humidity, smoke concentration, combustible gas concentration, equipment operation status, infrared images and video monitoring data; Perform spatiotemporal alignment and multimodal fusion processing on multi-source environmental data to generate comprehensive inspection feature information; According to the preset dynamic threshold model, the comprehensive inspection feature information is analyzed in real time to determine whether there are any abnormal conditions; If an abnormality is determined, an abnormality alarm message is generated and a task is assigned; Inspection personnel or mobile inspection robots go to designated locations to verify and handle abnormal alarm information, and feed back the processing results to the system; Inspectors also conduct inspections based on regular inspection tasks generated by the system, including equipment function verification and physical damage inspection; The system optimizes the dynamic threshold model and subsequent inspection strategy based on the feedback results of manual inspections or mobile inspection robots, and dynamically adjusts the equipment maintenance priority.

2. The property security inspection method according to claim 1, characterized in that: The dynamic threshold model is generated in the following way: Based on historical inspection data, combined with time, weather, and regional functional attribute factors, the initial threshold of each environmental data is dynamically set; The false positive or negative data in the review results are analyzed through machine learning algorithms, and the threshold rules are dynamically optimized.

3. The property security inspection method according to claim 1, characterized in that: The multimodal fusion processing includes: Temporal synchronization and spatial alignment of data from different sensors; Use weighted fusion algorithms or deep learning models to fuse multi-source environmental data into unified inspection feature information.

4. The property security inspection method according to claim 1, characterized in that: The allocation logic of the manual inspection task includes: Priority classification: The review of abnormal alarm events in intelligent inspections is of high priority, and regular inspection tasks are of low priority; Path planning: Generate the optimal inspection path based on the task location to reduce repeated movements; Skill matching: Electricians are given priority in equipment inspection tasks, and security personnel are given priority in intelligent patrol abnormal alarm event review tasks and physical damage inspection tasks.

5. The property security inspection method according to claim 1, characterized in that: The device function verification includes: Control commands are sent through the manual inspection terminal APP to verify whether the response speed and function of the equipment are normal; if any abnormality is found, upload photos of the faulty equipment, fault description and equipment location to the system, and the system automatically generates a maintenance work order.

6. A property security inspection system, characterized in that: The system comprises: The intelligent inspection module is used to collect multi-source environmental data of the property management area in real time through intelligent inspection equipment, and then perform spatiotemporal alignment and multimodal fusion processing on the collected multi-source environmental data to generate comprehensive inspection feature information, and then analyze the comprehensive inspection feature information based on the dynamic threshold model to determine whether there is an abnormal situation. If it is determined that there is an abnormality, an abnormal alarm information is generated; The human-machine collaborative management module is used to automatically distribute abnormal alarm events found by intelligent inspection to the manual inspection terminal APP or mobile inspection robot, and the manual inspection terminal APP generates a route to the target location; it is also used to generate regular manual inspection tasks, including equipment function verification and physical damage inspection; The feedback optimization module optimizes the dynamic threshold model and subsequent inspection strategy, and dynamically adjusts the equipment maintenance priority based on the abnormal alarm review results and equipment status data submitted by manual inspections or mobile inspection robots.

7. The property security inspection system according to claim 6, characterized in that: The human-machine collaborative management module includes: The task allocation unit is used to allocate inspection tasks according to the task allocation logic, wherein the task allocation logic includes: Priority classification: The review of abnormal alarm events in intelligent inspections is of high priority, and regular inspection tasks are of low priority; Path planning: Generate the optimal inspection path based on the task location to reduce repeated movements; Skill matching: Electricians are given priority in equipment inspection tasks, while security personnel are given priority in intelligent patrol abnormal alarm event review tasks and physical damage inspection tasks; The review and verification unit is used for manual or mobile inspection robots to feed back the review results of abnormal alarm events to the system; The cost optimization unit is used to calculate the cost-benefit ratio of installing intelligent inspection equipment for equipment and facilities that require manual inspection, and generate recommended solutions.

8. The property security inspection system according to claim 6, characterized in that: The system further comprises: The knowledge base module is used to store equipment testing standards and push operation instructions to manual inspection terminals in real time; Visual display module, used for visual display of abnormal alarm information, abnormal alarm accuracy, equipment failure rate, inspection data, inspection tasks and task progress, maintenance work orders, and 3D annotation of sensor coverage areas and blind spots, for real-time monitoring and analysis by management personnel; The maintenance prediction module is used to analyze historical maintenance data and calculate the remaining service life in combination with the equipment's life parameters; when the equipment's remaining service life is lower than the threshold, a purchase requisition is automatically generated.

9. The property security inspection system according to claim 6, characterized in that: The manual inspection terminal has the following functions: AR-assisted detection is used to automatically compare the current status of the device based on standard parameters when scanning and detect abnormal conditions; Voice reporting: describe the fault phenomenon through natural language, and the system automatically converts it into a maintenance work order.

10. A computer-readable storage medium having computer program instructions stored thereon, wherein the computer program instructions, when executed by a processor, enable the processor to execute the property security inspection method according to any one of claims 1 to 5.

Images