Police cloud intelligent management integrated platform based on multi-modal perception and intelligent decision

Abstract

This invention discloses an integrated intelligent management platform for police cloud based on multimodal perception and intelligent decision-making, belonging to the field of police cloud management technology. Its key technical solutions include an acquisition module that acquires basic regional data and perception data of the target monitoring area; processes the basic regional data to obtain normal security factors, factor decision weights, early warning characteristic factors, and early warning diagnostic factors for the target monitoring area; processes the perception data to obtain security perception factors for the target monitoring area; matches and analyzes the security perception factors with normal security factors to obtain abnormal perception factors and abnormal factor parameters for the target monitoring area; obtains the abnormal impact level of the target monitoring area based on the factor decision weights and abnormal factor parameters; and determines the regional assessment level of the target monitoring area based on the abnormal impact level and the basic regional data. This invention can adapt to scenario requirements and continuously improve the overall efficiency of the security system.

Description

Technical Field

[0001] This invention relates to the field of police cloud management technology, and more specifically, to an integrated platform for police cloud intelligent management based on multimodal perception and intelligent decision-making. Background Technology

[0002] In modern society, the demands for public safety and security management are becoming increasingly complex and diverse. Traditional security management models are gradually revealing significant shortcomings. From a security threat perspective, urban areas experience frequent population movement and increasingly complex business formats. Traditional integrated intelligent management platforms for police clouds offer relatively simplistic identification and response models for security risks, focusing solely on simple threshold alarm mechanisms. Taking densely populated areas as an example, they only trigger alarms when the number of people in the area exceeds a certain fixed value, neglecting the dynamic changes in population flow, the spatial characteristics of population distribution, and the differences in reasonable population flow thresholds at different times and in different activity scenarios. Furthermore, they lack effective data fusion and mining capabilities, failing to extract potential correlations from complex data and struggling to uncover risk trends from historical data. Before an abnormal event occurs, the inability to predict potential risks from multi-source data prevents early warnings and intervention measures. Alarms are only triggered after an anomaly exceeds a simple threshold, leaving security management in a perpetually reactive, reactive stage, thus easily missing the optimal response opportunity. Summary of the Invention

[0003] To address the shortcomings of existing technologies, the purpose of this invention is to provide an integrated intelligent management platform for police cloud based on multimodal perception and intelligent decision-making.

[0004] To achieve the above objectives, the present invention provides the following technical solution: an integrated intelligent management platform for police cloud based on multimodal perception and intelligent decision-making, characterized in that it includes:

[0005] Acquisition module: Acquires basic regional data and sensing data of the target monitoring area;

[0006] The first processing module processes the basic regional data to obtain the normal security factors, factor decision weights, early warning characteristic factors, and early warning diagnostic factors for the target monitoring area.

[0007] The second processing module processes the sensed data to obtain the security sense factors of the target monitoring area;

[0008] Matching module: Matches and analyzes security sensing factors with normal security factors to obtain abnormal sensing factors and abnormal factor parameters of the target monitoring area;

[0009] Assessment module: Based on the factor decision weights and abnormal factor parameters, the abnormal impact level of the target monitoring area is obtained. Based on the abnormal impact level and the regional basic data, the assessment level is determined to obtain the regional assessment level of the target monitoring area.

[0010] The third processing module processes the early warning feature factors to obtain target early warning diagnostic factors; based on the early warning feature factors and target early warning diagnostic factors, it obtains the security early warning status of the target monitoring area;

[0011] Decision module: Based on the regional assessment level and security early warning status, obtain the security alert cloud decision results for the target monitoring area.

[0012] Preferably, the basic regional data is processed to obtain the normal security factors, factor decision weights, early warning characteristic factors, and early warning diagnostic factors for the target monitoring area, specifically including the following steps:

[0013] Obtain normal security data for the target monitoring area based on regional basic data;

[0014] Normal security data is processed to obtain the normal security factor corresponding to the target monitoring area and the normal factor parameter range corresponding to the normal security factor;

[0015] Based on the regional basic data, set the factor decision weights for the normal security factors corresponding to the target monitoring area;

[0016] Based on the regional basic data, obtain the corresponding historical security anomaly data, which includes security anomaly perception data and security diagnosis results corresponding to the security anomaly perception data;

[0017] The security diagnostic results are processed to obtain early warning diagnostic factors;

[0018] The early warning feature factors are obtained by processing the security anomaly perception data.

[0019] Preferably, the security anomaly perception data is processed to obtain early warning feature factors, specifically including the following steps:

[0020] Security anomaly detection data is used to extract security features to obtain the security anomaly features corresponding to the target monitoring area;

[0021] Early warning feature factors are generated based on the characteristics of security anomalies.

[0022] Preferably, the normal security data is processed to obtain the normal security factor corresponding to the target monitoring area and the normal factor parameter range corresponding to the normal security factor, specifically including the following steps:

[0023] Based on normal security data, security features are extracted to obtain the normal security features corresponding to the target monitoring area;

[0024] Based on normal security characteristics, normal security factors and normal factor parameter ranges corresponding to the target monitoring area are generated.

[0025] Preferably, the security diagnosis results are processed to obtain early warning diagnosis factors, specifically including the following steps:

[0026] The security diagnostic results are used to extract security features to obtain the security diagnostic features corresponding to the target monitoring area.

[0027] Early warning diagnostic factors are generated based on security diagnostic characteristics.

[0028] Preferably, the security sensing factors of the target monitoring area are obtained by processing the sensing data, specifically including the following steps:

[0029] Security features are extracted from the perceived data to obtain the security perception features of the target monitoring area;

[0030] Based on the security perception characteristics, security perception factors of the target monitoring area and the corresponding perception factor parameters are generated.

[0031] Preferably, the abnormal sensing factors and abnormal factor parameters of the target monitoring area are obtained by matching and analyzing the security sensing factors with normal security factors, specifically including the following steps:

[0032] Abnormal sensing factors are obtained by comparing the sensing factor parameters of security sensing factors with the normal factor parameter range of normal security factors.

[0033] Obtain the center parameter of the normal factor parameter interval;

[0034] The abnormal factor parameters of the abnormal sensing factor are obtained based on the interval center parameter and the sensing factor parameter.

[0035] Preferably, the abnormal sensing factor is obtained by comparing the sensing factor parameters of the security sensing factor with the normal factor parameter range of the normal security factor, specifically as follows:

[0036] If the sensing factor parameter of a security sensing factor is not within the normal factor parameter range of a normal security factor, then the security sensing factor is marked as an abnormal sensing factor.

[0037] Preferably, the target early warning diagnostic factors are obtained by processing the early warning feature factors, specifically including the following steps:

[0038] Obtain the target factor association path of the early warning feature factors;

[0039] Based on the correlation path of the target factors, the target early warning diagnostic factors corresponding to the early warning feature factors are obtained.

[0040] Preferably, the security warning status of the target monitoring area is obtained based on the warning feature factors and the target warning diagnostic factors, specifically including the following steps:

[0041] Obtain the security diagnostic features corresponding to the target early warning diagnostic factors;

[0042] The security early warning status of the target monitoring area is obtained based on early warning characteristic factors and security diagnostic characteristics.

[0043] Compared with the prior art, the present invention has the following beneficial effects:

[0044] This invention acquires regional basic data and sensing data through an acquisition module, including regional static features and real-time dynamic information. A first processing module and a second processing module process the two types of data respectively, thereby extracting normal security factors and security sensing factors. A matching module identifies abnormal sensing factors and abnormal factor parameters by comparing the two, enabling comprehensive identification of security anomalies, capturing abnormal personnel gatherings, potential equipment malfunctions, and abnormal fluctuations in environmental parameters. An evaluation module introduces factor decision weights and combines them with abnormal factor parameters to determine the impact level of the anomaly; then, it integrates regional basic data to determine the regional assessment level, assigning different levels to different risk degrees and handling priorities. For high-impact anomalies, priority is given to allocating police resources and activating emergency plans; low-impact anomalies are handled through routine inspections. This hierarchical management avoids resource waste and over-response, significantly improving resource allocation efficiency. The decision module outputs customized police cloud decision results based on the regional assessment level and early warning status. This application can adapt to scenario requirements and continuously improve the overall effectiveness of the security system. Attached Figure Description

[0045] Figure 1 This is a schematic diagram of the modules of the integrated police cloud intelligent management platform based on multimodal perception and intelligent decision-making proposed in this invention;

[0046] Figure 2 This is a schematic diagram illustrating the steps involved in calculating the sensing factor parameters in the integrated police cloud intelligent management platform based on multimodal perception and intelligent decision-making proposed in this invention.

[0047] Figure 3 This is a schematic diagram illustrating the steps involved in calculating abnormal factor parameters in the integrated police cloud intelligent management platform based on multimodal perception and intelligent decision-making proposed in this invention. Detailed Implementation

[0048] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0049] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.

[0050] Secondly, the term "an embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in different places throughout this specification does not necessarily refer to the same embodiment, nor is it a single embodiment or an embodiment selectively excluded from other embodiments.

[0051] It should be noted in advance that all acquisition and processing of images, information and data in this invention are carried out in compliance with relevant data protection laws and policies and with the authorization of the respective device owners.

[0052] Reference Figures 1-3 As shown.

[0053] The embodiments further illustrate the integrated police cloud intelligent management platform based on multimodal perception and intelligent decision-making proposed in this invention.

[0054] The integrated police cloud intelligent management platform based on multimodal perception and intelligent decision-making includes:

[0055] Acquisition module: Acquires basic regional data and sensing data of the target monitoring area;

[0056] The first processing module processes the basic regional data to obtain the normal security factors, factor decision weights, early warning characteristic factors, and early warning diagnostic factors for the target monitoring area.

[0057] The second processing module processes the sensed data to obtain the security sense factors of the target monitoring area;

[0058] Matching module: Matches and analyzes security sensing factors with normal security factors to obtain abnormal sensing factors and abnormal factor parameters of the target monitoring area;

[0059] Assessment module: Based on the factor decision weights and abnormal factor parameters, the abnormal impact level of the target monitoring area is obtained. Based on the abnormal impact level and the regional basic data, the assessment level is determined to obtain the regional assessment level of the target monitoring area.

[0060] The third processing module processes the early warning feature factors to obtain target early warning diagnostic factors; based on the early warning feature factors and target early warning diagnostic factors, it obtains the security early warning status of the target monitoring area;

[0061] Decision module: Based on the regional assessment level and security early warning status, obtain the security alert cloud decision results for the target monitoring area.

[0062] The acquisition module acquires basic regional data and sensing data of the target monitoring area. The basic regional data includes the geographical environment of the area, the configuration of existing security facilities, and historical security event records. The sensing data is real-time dynamic information collected by sensing devices, such as personnel flow, changes in environmental parameters, and equipment operating status. Sensing devices include cameras and sensors.

[0063] The first processing module processes the regional basic data to obtain normal security factors. These factors are a set of features extracted from the region's long-term stable security status, representing the region's security operation under normal conditions, such as population density range. Simultaneously, based on the region's importance and management priority, different normal security factors are assigned decision-making weights, clarifying the influence weight of each factor in subsequent decision-making processes. Warning feature factors and warning diagnostic factors are mined from the basic data. Warning feature factors focus on potential characteristics that could trigger security anomalies, while warning diagnostic factors are information extracted from the diagnostic results of historical security anomalies, used to subsequently identify whether anomalies have occurred in the current scenario and to determine the nature and category of the anomalies.

[0064] The second processing module extracts security sensing factors from the real-time collected sensing data. These factors reflect the current security status of the target monitored area. The matching module compares the security sensing factors with normal security factors. By analyzing the differences in parameters and characteristics between the sensing factors and normal factors, abnormal sensing factors, i.e., factors that deviate from the normal security pattern, are identified. At the same time, abnormal factor parameters are obtained, and the degree and characteristics of the abnormality are quantified through the abnormal factor parameters to clarify the deviation of the area's security status from the normal pattern.

[0065] The assessment module combines factor decision weights and anomaly factor parameters to determine the anomaly impact level. By incorporating regional basic data and comprehensively considering factors such as the region's size, function, and security performance, the anomaly impact level is adjusted to ultimately determine the regional assessment level of the target monitoring area, thereby comprehensively assessing the overall impact of the anomaly on the region.

[0066] The abnormal impact level of the target monitoring area is obtained by weighting and summing the factor decision weights corresponding to each abnormality perception factor with the abnormal factor parameters.

[0067] First, an initial scope is defined based on the level of impact of the anomaly, and then revised by incorporating basic regional data. For example, if the anomaly impact level is moderate, but the area is densely populated with limited escape routes, the regional assessment level is raised; conversely, if the area has abundant security resources and strong capabilities to respond to anomalies, even if the anomaly impact level is moderate, the regional assessment level is relatively lower. A comprehensive assessment of the actual threat posed by the anomaly in a specific regional scenario is conducted to ultimately determine a regional assessment level that closely reflects the actual situation of the target monitoring area.

[0068] The third processing module processes the early warning feature factors to obtain the target early warning diagnostic factors required for the current analysis. Combining the early warning feature factors and the target early warning diagnostic factors, the security early warning status of the target monitoring area is obtained.

[0069] The security alert cloud decision-making results for the target monitoring area are obtained based on the regional assessment level and security early warning status. Based on a pre-set decision database, security alert cloud decision-making results adapted to the current security needs of the target monitoring area are formulated, including police force deployment strategies, security measure adjustment plans, and emergency response procedures. This provides execution guidance for actual security management actions, enabling the security management of the target monitoring area.

[0070] The basic regional data is processed to obtain the normal security factors, factor decision weights, early warning characteristic factors, and early warning diagnostic factors for the target monitoring area. This process includes the following steps:

[0071] Obtain normal security data for the target monitoring area based on regional basic data;

[0072] Normal security data is processed to obtain the normal security factor corresponding to the target monitoring area and the normal factor parameter range corresponding to the normal security factor;

[0073] Based on the regional basic data, set the factor decision weights for the normal security factors corresponding to the target monitoring area;

[0074] Based on the regional basic data, obtain the corresponding historical security anomaly data, which includes security anomaly perception data and the corresponding security diagnosis results.

[0075] The security diagnostic results are processed to obtain early warning diagnostic factors;

[0076] The early warning feature factors are obtained by processing the security anomaly perception data.

[0077] This application uses regional basic data to filter out normal security data generated under normal and stable security conditions in the target monitoring area, including daily personnel flow patterns, normal operating parameters of security equipment, and environmental safety indicators. Personnel flow patterns include the number of people entering and exiting the area at different times on weekdays; normal operating parameters of security equipment include the fault-free operating time of cameras and the threshold range of fire sensors; and environmental safety indicators include the temperature, humidity, and lighting intensity of the area.

[0078] By employing data mining and feature engineering techniques, normal security factors reflecting the essential patterns of regional security are extracted from routine security data. These factors include, for example, the personnel density in core access areas during the morning rush hour and the average power consumption of security equipment at night. Simultaneously, by calculating the mean and standard deviation of each factor, the parameter ranges for normal factors are defined, clarifying the quantitative boundaries of normal conditions. For instance, the personnel density in core access areas during the morning rush hour is 30-80 people per access area. This clearly defines the numerical range for safe and stable operation, providing a reference standard for subsequent identification of abnormal situations. If real-time data exceeds this range, a preliminary assessment indicates the existence of security risks.

[0079] Factor decision weights aim to differentiate the varying impacts of different normal security factors on regional security decisions. Each normal security factor is assigned a corresponding weight based on regional basic data, including regional attributes and management needs. These attributes include whether the region is a densely populated commercial complex or a warehouse park storing goods. Management needs include prioritizing personnel safety, material safety, and facility safety. In densely populated commercial plazas, the decision weight for the crowd congestion factor is higher than that for the equipment temperature factor, because abnormal crowd gatherings have a more significant direct impact on public safety. Conversely, in warehouses storing flammable materials, the fire safety facility operation factor has a higher weight, as ensuring the safety of the material storage environment is a core management priority.

[0080] The decision-making weights are optimized and adjusted based on dynamic changes in regional basic data, such as upgrades to security facilities and shifts in management priorities. If a smart facial recognition access control system is added to the target monitoring area, the importance of the personnel identity verification factor increases, meaning the decision-making weight is adjusted accordingly. This ensures that the factor weights continuously adapt to actual security management needs, making the influence of each factor more relevant to the scenario in subsequent decision-making processes, thereby improving the scientific nature of the decision-making.

[0081] Security anomaly perception data and corresponding security diagnostic results were extracted from historical security anomaly perception data generated during abnormal security events in the region. Security anomaly perception data consists of raw information collected by sensing devices during the anomaly occurrence process, such as keyframes from video footage of sudden crowd gatherings, abnormal readings uploaded by sensors during equipment malfunctions (e.g., sudden current surges and temperature exceeding limits), and access control alarm records triggered by illegal intrusions. This type of data carries the scene characteristics of the anomaly occurrence. Security diagnostic results include post-event analysis reports and handling records, clearly defining the cause of the anomaly, its development process, handling plan, and diagnostic results. The cause of the anomaly includes whether the crowd gathering was due to a promotional activity without prior warning; the development process includes the evolution from localized gathering to passageway congestion; and the handling plan and its effects include police deployment methods and the time taken for congestion evacuation.

[0082] The security diagnostic results are processed to obtain early warning diagnostic factors; the security anomaly perception data is processed to obtain early warning feature factors. Semantic parsing and feature extraction are performed on the security diagnostic results to obtain early warning diagnostic factors, such as the risk of crowd gatherings caused by promotional activities and equipment aging leading to malfunctions, which serve as the basis for rapid matching and judgment in subsequent anomaly diagnosis. Anomaly detection and feature engineering techniques are used to extract key scene features that trigger anomalies from the security anomaly perception data, thereby generating early warning feature factors, such as a 200% increase in personnel density in a passageway within 10 minutes and a sudden 15°C increase in equipment temperature within 5 minutes. If real-time data matches these features during subsequent real-time monitoring, potential anomalies can be identified in advance, thus triggering the early warning process.

[0083] The parameter range of normal security factors is compared with the abnormal perception features in historical abnormal data. If the data collected in real time exceeds the normal parameter range, the corresponding early warning feature factors and early warning diagnostic factors are matched after quickly associating with the historical abnormal case library to determine the type of current abnormality and its consequences. Abnormality types include crowd gathering and equipment failure, and the consequences include local congestion and large-scale power outage.

[0084] Factor decision weights play a regulatory role; when factors with high decision weights exhibit abnormalities, the overall risk assessment level is raised, triggering more urgent response procedures. For example, if a factor with a high decision weight for crowd gathering and congestion exceeds its normal range, even if other factors are normal, it is considered high-risk, thus prioritizing the allocation of police resources and the activation of emergency plans. As regional basic data is continuously updated, the parameter ranges, decision weights, and early warning diagnostic factors of normal security factors will all be iteratively optimized, promoting the efficient operation of the integrated police-cloud intelligent management platform.

[0085] The process of processing security anomaly detection data to obtain early warning feature factors includes the following steps:

[0086] Security anomaly detection data is used to extract security features to obtain the security anomaly features corresponding to the target monitoring area;

[0087] Early warning feature factors are generated based on the characteristics of security anomalies.

[0088] Security anomaly detection data includes abnormal images captured by surveillance cameras, abnormal readings returned by various sensors, and abnormal access recorded by access control systems. Abnormal images captured by cameras include sudden gatherings of people and abnormal movement of objects; abnormal readings returned by sensors include excessive smoke concentration and sudden rises in equipment temperature; and abnormal access recorded by access control systems include frequent attempts by unauthorized personnel to enter.

[0089] For video data, this system identifies changes in key elements, such as detecting rapid changes in crowd density within a short period and marking them as abnormal crowd gathering points. It also determines whether object trajectories deviate from normal paths and extracts abnormal movement trajectory features. By comparing sensor data with historical normal data, it identifies significant deviations in numerical fluctuations, such as a sudden jump in smoke concentration from the normal 0-5 ppm to 20 ppm, extracting abnormal peak smoke concentration features. Through this processing, it accurately extracts security anomaly features from security anomaly perception data, thus outlining the key manifestations when security anomalies occur in the target monitored area.

[0090] After extracting security anomaly features, they need to be transformed into early warning feature factors that can be used for intelligent decision-making. These features are classified and quantified based on their type, severity, and scope of impact. For anomaly features related to crowd gathering, the crowd density and the proportion of gathering areas are calculated using regional basic data, thus transforming them into crowd gathering early warning factors to clarify their impact on regional security risks. Similarly, for features such as sudden temperature rises in equipment, associated equipment types, and their environmental conditions, equipment overheating early warning factors are generated.

[0091] The normal security data is processed to obtain the normal security factor corresponding to the target monitoring area and the normal factor parameter range corresponding to the normal factor. The specific steps include:

[0092] Based on normal security data, security features are extracted to obtain the normal security features corresponding to the target monitoring area;

[0093] Based on normal security characteristics, normal security factors and normal factor parameter ranges corresponding to the target monitoring area are generated.

[0094] Key information representing the stable operation of regional security is extracted from normal security data. For example, by analyzing pedestrian flow data to determine the distribution patterns of people in different time periods and areas, features such as the average travel time per person in core channels during weekday morning rush hours and the average pedestrian density in areas during off-peak hours are extracted. For equipment operation data, the range of voltage, current, and temperature fluctuations during normal equipment operation is analyzed, and features such as the average daily online time of security cameras and the normal response delay range of fire sensors are summarized. These extracted contents collectively constitute the normal security features corresponding to the target monitoring area. Normal security factors and corresponding normal factor parameter ranges are generated from these extracted normal security features. Normal security factors are core indicators that transform normal security features into features usable for security analysis and decision-making. For example, pedestrian flow-related features such as the average travel time per person in core channels during weekday morning rush hours and the average pedestrian density in areas during off-peak hours are integrated into a regional pedestrian flow dynamic distribution factor; equipment operation features such as the average daily online time of security cameras and the normal response delay range of fire sensors are summarized into a security equipment steady-state operation factor. The normal factor parameter range corresponding to each normal security factor is then determined. By calculating the mean, standard deviation, and extreme value distribution of characteristic data, the numerical range of factors under safe and stable operation is determined. For example, the parameter range corresponding to the regional dynamic distribution factor of people flow is defined as 15-30 seconds per person in the core channel during the morning peak on weekdays, and 5-15 people per 100 square meters per person during off-peak hours. This range clearly defines the fluctuation range of factors under normal security conditions. If the subsequent real-time monitoring data exceeds this range, it is determined that the regional security status is abnormal.

[0095] The security diagnostic results are processed to obtain early warning diagnostic factors, specifically including the following steps:

[0096] The security diagnostic results are used to extract security features to obtain the security diagnostic features corresponding to the target monitoring area.

[0097] Early warning diagnostic factors are generated based on security diagnostic characteristics.

[0098] Security diagnostic results include the causes of anomalies, their development process, and the conclusions of their handling. For example, from a diagnostic report on fire warning delays caused by equipment aging, the core elements linking equipment aging, warning delays, and fire risk can be extracted and transformed into structured security diagnostic features, such as functional anomalies caused by equipment maintenance deficiencies and security response chain failures.

[0099] Early warning diagnostic factors are generated based on extracted security diagnostic features. These features are then transformed into quantitative indicators for security decision-making. Features are categorized and assigned values ​​based on their risk level, impact range, and correlation degree. For features indicating functional abnormalities caused by equipment maintenance deficiencies, regional basic data is combined to transform them into equipment maintenance health factors. Historical data is used to train a model to determine the corresponding risk threshold for these factors; for example, an early warning is triggered when the maintenance deficiency rate exceeds 30%. Simultaneously, the effectiveness of handling past abnormal events is correlated to assign early warning weights to these factors, enabling them to not only identify abnormal patterns but also predict risk consequences. For instance, fire risk correlation features are transformed into fire early warning correlation factors. If real-time data matches this factor, historical fire response plans are quickly linked.

[0100] The security perception factors of the target monitoring area are obtained by processing the perceived data, specifically including the following steps:

[0101] Security features are extracted from the perceived data to obtain the security perception features of the target monitoring area;

[0102] Based on the security perception characteristics, security perception factors of the target monitoring area and the corresponding perception factor parameters are generated.

[0103] Security perception features of the target monitoring area are obtained by extracting security features from the perceived data. For video data, the system identifies personnel behavior and object states in the footage. Personnel behavior includes gathering and running, while object states include stacking and equipment movement, extracting features of abnormal personnel gathering and illegal object movement. For sensor data, the system judges numerical fluctuations, such as sudden temperature increases and excessive smoke concentration, extracting features of abnormal environmental parameters and equipment malfunction warnings. Access control data is analyzed to determine passage frequency and personnel permission matching, yielding features of abnormal personnel passage and permission overreach warnings. These extracted elements collectively constitute the security perception features of the target monitoring area.

[0104] Based on these security perception features, security perception factors and corresponding perception factor parameters are generated. For example, abnormal personnel gathering characteristics and abnormal personnel passage characteristics are integrated into regional personnel dynamic security factors; abnormal environmental parameter characteristics and equipment failure early warning characteristics are summarized into regional facility environment security factors. During factor generation, the perception factor parameters for each security perception factor are determined by quantifying the perception features. The numerical performance of the factor in its current state is clarified by statistically analyzing the mean, fluctuation range, and duration of the feature data. For example, the parameters corresponding to the regional personnel dynamic security factor may include the number of abnormally gathered individuals and the duration of the gathering; the parameters of the regional facility environment security factor include the environmental parameter exceedance value and the duration of equipment failure. These parameters quantify the state of the perception factors and will subsequently be compared with normal security factors, early warning feature factors, etc., providing data support for identifying anomalies and assessing risks.

[0105] The abnormal sensing factors and abnormal factor parameters of the target monitoring area are obtained by matching and analyzing the security sensing factors with normal security factors. The specific steps include:

[0106] Abnormal sensing factors are obtained by comparing the sensing factor parameters of security sensing factors with the normal factor parameter range of normal security factors.

[0107] Obtain the center parameter of the normal factor parameter interval;

[0108] The abnormal factor parameters of the abnormal sensing factor are obtained based on the interval center parameter and the sensing factor parameter.

[0109] Abnormal sensing factors are obtained by comparing the sensing factor parameters of security sensing factors with the normal factor parameter range of normal security factors. Specifically:

[0110] If the sensing factor parameter of a security sensing factor is not within the normal factor parameter range of a normal security factor, then the security sensing factor is marked as an abnormal sensing factor.

[0111] Security sensing factors carry information about the current security status of the target monitoring area, and their corresponding sensing factor parameters are a quantitative representation of the real-time status. Normal security factors, on the other hand, have normal parameter ranges defined as safe operating boundaries based on historical normal data. By comparing these two sets of parameters one by one, when a sensing factor parameter exceeds the normal parameter range, the corresponding security sensing factor is determined to be an abnormal sensing factor. For example, if the normal factor parameter range specifies a normal range for population density in an area as 10-50 people / 100 square meters, and the real-time sensed population density parameter is 60 people / 100 square meters, then the population density sensing factor is marked as an abnormal sensing factor, thus quickly identifying deviations from the normal security status of the area.

[0112] The interval center parameter is a core feature extracted from the normal factor parameter interval and is obtained by calculating the mean of the interval. For example, if the normal factor parameter interval is 10-50 people / 100 square meters, its interval center parameter is 30 people / 100 square meters. This parameter is used to measure the degree of abnormal deviation.

[0113] Anomaly factor parameters are calculated based on interval center parameters and perception factor parameters. The specific numerical value of the abnormal deviation is obtained by calculating the difference between the perception factor parameter and the interval center parameter. For example, if the perception factor parameter is 60 people / 100 square meters and the interval center parameter is 30 people / 100 square meters, the difference is 30 people / 100 square meters. This value is the anomaly factor parameter of the abnormal perception factor, clearly showing the deviation of the current state from the normal core state.

[0114] The target early warning diagnostic factors are obtained by processing the early warning feature factors, specifically including the following steps:

[0115] Obtain the target factor association path of the early warning feature factors;

[0116] Based on the correlation path of the target factors, the target early warning diagnostic factors corresponding to the early warning feature factors are obtained.

[0117] Early warning feature factors are key characteristic identifiers extracted from historical security anomaly perception data to identify potential risks. Target factor association paths are pre-constructed logical links that describe the relationships between early warning feature factors and early warning diagnostic factors. Actively acquiring the target factor association paths corresponding to early warning feature factors involves mining these paths using machine learning techniques based on a large amount of historical security data. This records which early warning diagnostic factors are typically associated with when early warning feature factors appear historically, as well as the triggering conditions and logical order of these associations. For example, when there is an early warning feature factor indicating a rapid increase in personnel density in a certain passageway within a short period, the target factor association path reveals that this feature factor will sequentially associate with intermediate nodes related to the risk of personnel gathering and the potential for passageway congestion, ultimately pointing to a stampede risk early warning diagnostic factor.

[0118] Specifically, the target factor association path is constructed through the following steps:

[0119] Based on historical data, first-series data on early warning feature factors and target early warning diagnostic factors are obtained. The first-series data is split into multiple second-series data. The occurrence time of each element in the second-series data is obtained. The second-series data are clustered based on the occurrence time as a feature to obtain multiple clustering results. The early warning diagnostic factors of the early warning feature factors and the association path of the target factors are determined according to the number of second-series data included in the clustering results.

[0120] Historical data includes temperature data collected by various sensors before an accident occurs, as well as changes in crowd density obtained through video footage. This data is encoded and a first sequence of data is constructed based on the time of occurrence. For example, the first sequence of data is ABCDE, where E represents the accident that occurred, i.e., the target early warning diagnostic factor. For example, if someone faints, ABCD represents data that occurred before the accident, i.e., early warning characteristic factors. For example, A represents a 200% increase in crowd density, and B represents a 2°C increase in indoor temperature.

[0121] The first sequence data is then split to obtain the second sequence data. For example, ABCDE is split into AE, ABE, ABCE, and ABCDE. The purpose is to analyze the correlation between each data point and the final accident. Then, the occurrence time of each element in the second sequence data is obtained, such as A occurring at 15:00:02 and B occurring at 15:03:02. The time differences between elements are then obtained, such as the time difference between A and B, and the time difference between A and C. It should be noted that historical data contains a large amount of accident data; therefore, the first sequence data will contain multiple sequences. For example, there may also be ABCDF, where F represents another type of accident. That is, the second sequence data formed after splitting will also include multiple sequences of different accidents.

[0122] Then, clustering is performed on each second time series data. For example, clustering is performed on the second time series AE. The time difference of the elements is used as the clustering feature. The clustering algorithm can be K-means or DBSCAN.

[0123] After clustering is completed, multiple clusters will be obtained. The number of second sequence data included in each cluster will be obtained. The cluster with the most second sequence data will be selected. The target factor association path will be determined based on the second sequence data it includes. For example, if the analysis finds that a certain cluster includes the most second sequence data and its second sequence data is ABE, then when warning feature factors A and B appear, the target warning diagnostic factor E will be associated.

[0124] Following the established logical path along the obtained target factor association path, the warning feature factors are matched and associated with their corresponding warning diagnostic factors. Because the path clearly defines how to connect the warning feature factor to the specific warning diagnostic factor, it is possible to find the target warning diagnostic factor corresponding to the current warning feature factor. When a warning feature factor is detected in real time, the corresponding target warning diagnostic factor can be quickly located through the association path, clarifying the specific security risk type and diagnostic conclusion that the warning feature factor may trigger. This provides direct and crucial diagnostic basis for subsequent security warning status analysis and decision-making.

[0125] The security warning status of the target monitoring area is obtained based on the warning feature factors and target warning diagnostic factors, specifically including the following steps:

[0126] Obtain the security diagnostic features corresponding to the target early warning diagnostic factors;

[0127] The security early warning status of the target monitoring area is obtained based on early warning characteristic factors and security diagnostic characteristics.

[0128] Target early warning diagnostic factors are core factors extracted from historical security anomaly data to predict potential risks. Their corresponding security diagnostic features originate from the analysis of past diagnostic results of similar security anomalies, including key information such as the patterns of anomaly occurrence, triggering conditions, and the scope of impact. Through data retrieval and matching mechanisms, security diagnostic features associated with the target early warning diagnostic factors are extracted from the historical security diagnostic database. For example, when a target early warning diagnostic factor points to the risk of stampedes caused by crowd gathering, the corresponding security diagnostic features might include abnormal characteristics such as the population density threshold and crowd flow rate in a specific area.

[0129] The security alert status of the target monitoring area is determined based on early warning feature factors and security diagnostic features. Early warning feature factors are characteristics extracted from real-time perceived data of the current monitoring area that may indicate risk. These features are then compared and fused with security diagnostic features to determine the degree of matching and the strength of the correlation. If the early warning feature factors and security diagnostic features are highly consistent—for example, if the early warning feature factors of population density and crowd flow rate in a certain area match the security diagnostic features corresponding to the risk of stampedes caused by crowd gathering—then it is determined that the area poses a corresponding security risk, and the early warning level and risk type are clearly defined.

[0130] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.

[0131] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.

[0132] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. An integrated intelligent management platform for police cloud based on multimodal perception and intelligent decision-making, characterized in that: include: Acquisition Module: Acquires basic regional data and sensing data of the target monitoring area. Basic regional data includes the geographical environment of the area, the configuration of existing security facilities, and historical security event records. Sensing data consists of real-time dynamic information collected by sensing devices, including personnel flow, changes in environmental parameters, and equipment operating status. The first processing module processes the regional basic data to obtain the normal security factors, factor decision weights, early warning characteristic factors, and early warning diagnostic factors for the target monitoring area, including: Obtain normal security data for the target monitoring area based on regional basic data; Normal security data is processed to obtain the normal security factor corresponding to the target monitoring area and the normal factor parameter range corresponding to the normal security factor; Based on the regional basic data, set the factor decision weights for the normal security factors corresponding to the target monitoring area; Based on the regional basic data, obtain the corresponding historical security anomaly data, which includes security anomaly perception data and security diagnosis results corresponding to the security anomaly perception data; The security diagnostic results are processed to obtain early warning diagnostic factors; The process of processing security anomaly perception data to obtain early warning feature factors includes: extracting security features from the security anomaly perception data to obtain security anomaly features corresponding to the target monitoring area; and generating early warning feature factors based on the security anomaly features. Early warning feature factors characterize the potential features that could trigger security anomalies, while early warning diagnostic factors are information extracted from the diagnostic results of historical security anomalies. These factors are used to identify whether anomalies have occurred in the current scenario and to determine the nature and type of the anomalies. The second processing module processes the sensed data to obtain the security sense factor of the target monitoring area, which reflects the current security status of the target monitoring area. Matching module: Matches and analyzes security sensing factors with normal security factors to obtain abnormal sensing factors and abnormal factor parameters of the target monitoring area; Anomaly detection factors are those that deviate from normal security patterns; Assessment module: Based on the factor decision weights and abnormal factor parameters, the abnormal impact level of the target monitoring area is obtained. Based on the abnormal impact level and the regional basic data, the assessment level is determined to obtain the regional assessment level of the target monitoring area. The third processing module processes the early warning feature factors to obtain target early warning diagnostic factors, including: obtaining the target factor association path of the early warning feature factors; and obtaining the target early warning diagnostic factors corresponding to the early warning feature factors based on the target factor association path. The security early warning status of the target monitoring area is obtained based on early warning characteristic factors and target early warning diagnostic factors; Early warning feature factors are key feature identifiers extracted from historical security anomaly perception data to identify potential risks. Target factor association paths are logical links constructed in advance that describe the relationship between early warning feature factors and early warning diagnostic factors. Actively obtaining the target factor association paths corresponding to early warning feature factors is based on historical security data and is obtained through machine learning. It records how early warning feature factors are associated with corresponding early warning diagnostic factors when they appear in history, as well as the triggering conditions and logical order of the association. Decision module: Based on the regional assessment level and security early warning status, obtain the security alert cloud decision results for the target monitoring area.

2. The integrated police cloud intelligent management platform based on multimodal perception and intelligent decision-making as described in claim 1, characterized in that, The normal security data is processed to obtain the normal security factor corresponding to the target monitoring area and the normal factor parameter range corresponding to the normal factor. The specific steps include: Based on normal security data, security features are extracted to obtain the normal security features corresponding to the target monitoring area; Based on normal security characteristics, normal security factors and normal factor parameter ranges corresponding to the target monitoring area are generated.

3. The integrated police cloud intelligent management platform based on multimodal perception and intelligent decision-making as described in claim 2, characterized in that, The security diagnostic results are processed to obtain early warning diagnostic factors, specifically including the following steps: The security diagnostic results are used to extract security features to obtain the security diagnostic features corresponding to the target monitoring area. Early warning diagnostic factors are generated based on security diagnostic characteristics.

4. The integrated police cloud intelligent management platform based on multimodal perception and intelligent decision-making as described in claim 1, characterized in that, The security perception factors of the target monitoring area are obtained by processing the perceived data, specifically including the following steps: Security features are extracted from the perceived data to obtain the security perception features of the target monitoring area; Based on the security perception characteristics, security perception factors of the target monitoring area and the corresponding perception factor parameters are generated.

5. The integrated police cloud intelligent management platform based on multimodal perception and intelligent decision-making according to claim 2, characterized in that, The abnormal sensing factors and abnormal factor parameters of the target monitoring area are obtained by matching and analyzing the security sensing factors with normal security factors. The specific steps include: Abnormal sensing factors are obtained by comparing the sensing factor parameters of security sensing factors with the normal factor parameter range of normal security factors. Obtain the center parameter of the normal factor parameter interval; The abnormal factor parameters of the abnormal sensing factor are obtained based on the interval center parameter and the sensing factor parameter.

6. The integrated police cloud intelligent management platform based on multimodal perception and intelligent decision-making according to claim 5, characterized in that, Abnormal sensing factors are obtained by comparing the sensing factor parameters of security sensing factors with the normal factor parameter range of normal security factors. Specifically: If the sensing factor parameter of a security sensing factor is not within the normal factor parameter range of a normal security factor, then the security sensing factor is marked as an abnormal sensing factor.

7. The integrated police cloud intelligent management platform based on multimodal perception and intelligent decision-making as described in claim 1, characterized in that, The security warning status of the target monitoring area is obtained based on the warning feature factors and target warning diagnostic factors, specifically including the following steps: Obtain the security diagnostic features corresponding to the target early warning diagnostic factors; The security early warning status of the target monitoring area is obtained based on early warning characteristic factors and security diagnostic characteristics.

Images