A dynamic population monitoring method, system and medium based on multi-source heterogeneous data

Abstract

The application provides a kind of dynamic population monitoring method, system and medium based on multi-source heterogeneous data, first, by receiving the multi-source monitoring data from mobile communication base station, internet of things sensing device, public transport system and image acquisition device;After deduplication, space-time unification and anonymization preprocessing;According to the space-time grid unit, the data is extracted and input into the pre-trained estimation model to generate the estimation evidence matrix;Then, a consistency threshold and a continuity threshold are used to identify short-time data anomalies and sustainable events, dynamically adjust the model weight or generate evidence replacement value, when there is no anomaly, adjacent grid evidence matrix is fused to calculate confidence, and when the threshold is lower, the evidence matrix is corrected through backup data;Finally, the weighted fusion algorithm outputs the dynamic population monitoring result;Thus, the accuracy, reliability and real-time performance of population monitoring are improved, and the data security and system robustness are ensured through multi-level verification and privacy protection mechanism.

Description

Technical Field

[0001] This invention relates to the field of comprehensive population monitoring, and more specifically, to a dynamic population monitoring method, system, and medium based on multi-source heterogeneous data. Background Technology

[0002] In existing technologies, comprehensive population monitoring mainly relies on single or limited data sources, such as operator signaling data or video surveillance system data, to achieve population counting and distribution analysis. However, these methods have significant limitations. First, the coverage of a single data source is limited, making it unable to effectively capture populations without terminal devices, such as children, the elderly, or those in signal blind spots, leading to systematic biases and missed detections in the monitoring results. Second, due to the lack of multi-source data cross-validation mechanisms, existing methods struggle to distinguish between real population changes and data acquisition noise, such as base station signal drift, camera obstruction, or light interference, resulting in low reliability and weak anti-interference capabilities. Third, traditional methods typically employ static thresholds or fixed-weight fusion strategies, which cannot adapt to dynamic scenarios such as sudden population gatherings or large-scale movements, leading to poor real-time performance and insufficient scenario adaptability. Finally, regarding data compliance, existing technologies often lack a robust privacy protection chain, posing a risk of leakage of original personal identifiers during transmission and processing, making it difficult to meet increasingly stringent data security regulations. Furthermore, most systems lack hierarchical anomaly handling and self-correction capabilities. Once the main data source malfunctions, monitoring is interrupted or results are distorted, and rapid recovery and result correction cannot be achieved through backup data sources. These shortcomings collectively limit the large-scale application of existing technologies in high-frequency real-time monitoring scenarios such as smart cities and public safety.

[0003] Therefore, there is an urgent need for a dynamic population monitoring technology based on multi-source heterogeneous data. Summary of the Invention

[0004] In view of the above problems, the purpose of this invention is to provide a dynamic population monitoring method, system, and medium based on multi-source heterogeneous data. By fusing multi-source data such as mobile communication signaling, IoT sensing, public transportation card swiping, and image recognition, and after standardized preprocessing and spatiotemporal grid mapping, a multi-dimensional evidence matrix is ​​generated using a pre-trained estimation model. A dual comparison mechanism of consistency and continuity thresholds accurately identifies short-term anomalies and sustainable events, and dynamically adjusts model weights or generates time-series predicted surrogate values. In the absence of anomalies, confidence levels are calculated by fusing evidence from adjacent grids, and a backup data source is introduced to correct and update low-confidence results. Finally, a weighted fusion algorithm outputs the population monitoring results. This invention significantly improves the accuracy, reliability, and real-time performance of population monitoring through multi-source data complementarity, adaptive handling of anomalies, and a multi-level verification mechanism, while ensuring the privacy and security of data processing.

[0005] The first aspect of this invention provides a dynamic population monitoring method based on multi-source heterogeneous data, the method comprising: Receive monitoring data from multiple independent sources; Based on the preset data preprocessing logic, multi-source heterogeneous data is obtained from multi-source monitoring data; Based on the preset spatiotemporal grid partitioning rules, the multi-source heterogeneous data is extracted, and the estimation evidence matrix of the same spatiotemporal grid unit is obtained according to the pre-trained estimation model. The comparison is based on a preset consistency threshold. When a short-term data anomaly is detected in the estimated evidence matrix, the comparison is based on a preset continuity threshold to obtain the anomaly event information. The weight coefficients of the estimation model are dynamically adjusted based on the abnormal event information. If no abnormal event information is identified, the estimated evidence matrix of adjacent spatiotemporal grids is fused to obtain the confidence information of multi-source heterogeneous data; If the confidence level is lower than the preset confidence level threshold, the estimated evidence matrix is ​​updated based on the backup data; Based on the updated estimated evidence matrix and a weighted fusion algorithm, the dynamic population monitoring results of the spatiotemporal grid unit are output.

[0006] In this solution, based on preset data preprocessing logic, multi-source heterogeneous data is obtained from multi-source monitoring data, specifically including: The multi-source monitoring data is subjected to invalid data removal and duplicate data deduplication operations. Based on timestamp alignment rules, time alignment is performed on the deduplicated multi-source monitoring data; Based on the location-unified mapping rules, spatial unification is performed on the time-aligned multi-source monitoring data; Irreversible cryptographic hashing is applied to data involving identity information to achieve data anonymization. The processed data is standardized and stored according to its spatiotemporal structure to form the multi-source heterogeneous data.

[0007] In this scheme, the extraction of multi-source heterogeneous data based on preset spatiotemporal grid partitioning rules, and the obtaining of the estimation evidence matrix for the same spatiotemporal grid unit according to a pre-trained estimation model, specifically includes: Based on preset latitude and longitude coordinates, the monitoring area is divided into geographic subgrids of fixed size; Based on a preset time slice, combined with the geographic sub-grid, a spatiotemporal grid unit is formed; For each spatiotemporal grid cell, each type of data is input into the corresponding pre-trained estimation model to obtain the corresponding estimated value representing the population size. All population estimates are combined into a multidimensional evidence vector to obtain the estimation evidence matrix of the spatiotemporal grid cell.

[0008] In this scheme, the comparison is based on a preset consistency threshold. When a short-term data anomaly is detected in the estimated evidence matrix, a comparison is then performed based on a preset continuity threshold to obtain anomaly event information, specifically including: Calculate the pairwise dissimilarity between each evidence vector in the estimated evidence matrix of the current spatiotemporal grid cell; Determine whether the degree of difference is lower than a preset consistency threshold; If so, then it is determined that there is no short-term data anomaly; If not, mark the current evidence vector as short-term anomalous evidence; Extract the historical evidence vector of a preset number of time slices prior to the time slice in which the short-term anomaly evidence is located; Based on short-term anomaly evidence and historical evidence vectors, the changing trend and fluctuation amplitude are obtained; Based on the comparison between the changing trend and the jitter amplitude and a preset continuity threshold, it is determined whether it conforms to the preset continuity rule. If the conditions are met, the change is determined to be caused by a real event, and information on a persistent abnormal event is generated. If the condition does not meet the requirements, the change is determined to be transient data interference, and transient abnormal event information is generated.

[0009] In this solution, dynamically adjusting the weight coefficients of the estimation model based on abnormal event information specifically includes: In response to the information on the persistent anomalous events, the weight coefficient of the estimation model corresponding to the short-term anomalous evidence is increased; In response to the instantaneous anomalous event information, the weight coefficient of the estimation model corresponding to the short-term anomalous evidence is reduced, or an alternative value for the short-term anomalous evidence is generated based on the historical evidence vector using a pre-trained time series prediction model.

[0010] In this scheme, if no abnormal event information is identified, the estimated evidence matrix of adjacent spatiotemporal grids is fused to obtain the confidence information of multi-source heterogeneous data, specifically including: In response to the absence of short-term data anomalies, the estimated evidence matrix of adjacent spatiotemporal grid cells in the same time slice is obtained; Calculate the spatial distribution deviation between the current grid estimation evidence matrix and the estimation evidence matrices of adjacent grids; Based on the deviation, calculate the confidence information of the multi-source heterogeneous data in the current grid.

[0011] A second aspect of the present invention provides a dynamic population monitoring system based on multi-source heterogeneous data, including a dynamic population monitoring method program based on multi-source heterogeneous data, wherein the dynamic population monitoring method program based on multi-source heterogeneous data, when executed by the processor, implements the following steps: Receive monitoring data from multiple independent sources; Based on the preset data preprocessing logic, multi-source heterogeneous data is obtained from multi-source monitoring data; Based on the preset spatiotemporal grid partitioning rules, the multi-source heterogeneous data is extracted, and the estimation evidence matrix of the same spatiotemporal grid unit is obtained according to the pre-trained estimation model. The comparison is based on a preset consistency threshold. When a short-term data anomaly is detected in the estimated evidence matrix, the comparison is based on a preset continuity threshold to obtain the anomaly event information. The weight coefficients of the estimation model are dynamically adjusted based on the abnormal event information. If no abnormal event information is identified, the estimated evidence matrix of adjacent spatiotemporal grids is fused to obtain the confidence information of multi-source heterogeneous data; If the confidence level is lower than the preset confidence level threshold, the estimated evidence matrix is ​​updated based on the backup data; Based on the updated estimated evidence matrix and a weighted fusion algorithm, the dynamic population monitoring results of the spatiotemporal grid unit are output.

[0012] In this solution, based on preset data preprocessing logic, multi-source heterogeneous data is obtained from multi-source monitoring data, specifically including: The multi-source monitoring data is subjected to invalid data removal and duplicate data deduplication operations. Based on timestamp alignment rules, time alignment is performed on the deduplicated multi-source monitoring data; Based on the location-unified mapping rules, spatial unification is performed on the time-aligned multi-source monitoring data; Irreversible cryptographic hashing is applied to data involving identity information to achieve data anonymization. The processed data is standardized and stored according to its spatiotemporal structure to form the multi-source heterogeneous data.

[0013] In this scheme, the extraction of multi-source heterogeneous data based on preset spatiotemporal grid partitioning rules, and the obtaining of the estimation evidence matrix for the same spatiotemporal grid unit according to a pre-trained estimation model, specifically includes: Based on preset latitude and longitude coordinates, the monitoring area is divided into geographic subgrids of fixed size; Based on a preset time slice, combined with the geographic sub-grid, a spatiotemporal grid unit is formed; For each spatiotemporal grid cell, each type of data is input into the corresponding pre-trained estimation model to obtain the corresponding estimated value representing the population size. All population estimates are combined into a multidimensional evidence vector to obtain the estimation evidence matrix of the spatiotemporal grid cell.

[0014] A third aspect of the present invention provides a computer-readable storage medium comprising a dynamic population monitoring method program based on multi-source heterogeneous data, wherein when the dynamic population monitoring method program based on multi-source heterogeneous data is executed by a processor, it implements the steps of the dynamic population monitoring method based on multi-source heterogeneous data as described in any of the preceding claims.

[0015] This invention provides a dynamic population monitoring method, system, and medium based on multi-source heterogeneous data. First, it receives multi-source monitoring data from mobile communication base stations, IoT sensing devices, public transportation systems, and image acquisition devices. This data undergoes deduplication, spatiotemporal unification, and anonymization preprocessing. Then, data is extracted according to spatiotemporal grid units and input into a pre-trained estimation model to generate an estimation evidence matrix. Next, a dual comparison mechanism using consistency and continuity thresholds is employed to identify short-term data anomalies and sustainable events, dynamically adjusting model weights or generating alternative evidence values. When no anomalies are found, adjacent grid evidence matrices are fused to calculate confidence levels; when the confidence level is below the threshold, backup data is used to correct the evidence matrix. Finally, a weighted fusion algorithm outputs the dynamic population monitoring results. This improves the accuracy, reliability, and real-time performance of population monitoring, while multi-level verification and privacy protection mechanisms ensure data security and system robustness. Attached Figure Description

[0016] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the embodiments will be briefly described below. It should be understood that the following drawings only show some embodiments of the present invention and should not be regarded as a limitation on the scope.

[0017] Figure 1 A flowchart of a dynamic population monitoring method based on multi-source heterogeneous data according to the present invention is shown; Figure 2 This invention provides a flowchart of a preprocessing procedure for multi-source monitoring data according to an embodiment of the invention. Figure 3 This invention provides a flowchart of an estimated evidence matrix for generating spatiotemporal grid cells, according to an embodiment of the present invention. Figure 4 A block diagram of a dynamic population monitoring system based on multi-source heterogeneous data according to the present invention is shown. Detailed Implementation

[0018] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0019] Unless otherwise defined, all terms (including technical and scientific terms) used in embodiments of this invention shall have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains. It should also be understood that terms such as those defined in a common dictionary shall be interpreted as having a meaning consistent with their meaning in the context of the relevant art, and not as being interpreted in an idealized or highly formalized sense, unless expressly defined in this embodiment of the invention.

[0020] The terms "first," "second," and similar terms used in the embodiments of this invention do not indicate any order, quantity, or importance, but are merely used to distinguish different components. Terms such as "an," "a," or "the" do not indicate a quantity limitation, but rather indicate the presence of at least one. Similarly, terms such as "including" or "comprising" mean that the element or object preceding the word encompasses the elements or objects listed after the word and their equivalents, without excluding other elements or objects. Terms such as "connected" or "linked" are not limited to physical or mechanical connections, but can include electrical connections, whether direct or indirect. The steps preceding or following the steps in the method of the embodiments of this invention are not necessarily performed precisely in sequence. Instead, various steps can be processed in reverse order or simultaneously. Furthermore, other operations can be added to these processes, or one or more steps can be removed from these processes.

[0021] In addition, the functional modules in the various embodiments of the present invention can be integrated together to form an independent part, or each module can exist independently, or two or more modules can be integrated to form an independent part.

[0022] Figure 1 A flowchart of a dynamic population monitoring method based on multi-source heterogeneous data according to the present invention is shown.

[0023] like Figure 1 As shown, the first aspect of this invention discloses a dynamic population monitoring method based on multi-source heterogeneous data, the method comprising: S102, receives monitoring data from multiple independent sources; S104, based on the preset data preprocessing logic, obtain multi-source heterogeneous data from multi-source monitoring data; S106, Based on the preset spatiotemporal grid partitioning rules, extract the multi-source heterogeneous data, and according to the pre-trained estimation model, obtain the estimation evidence matrix of the same spatiotemporal grid unit; S108, compare based on a preset consistency threshold. When a short-term data anomaly is detected in the estimated evidence matrix, compare based on a preset continuity threshold to obtain anomaly event information. S110, dynamically adjust the weight coefficients of the estimation model based on the abnormal event information; S112, If no abnormal event information is identified, the estimated evidence matrix of adjacent spatiotemporal grids is fused to obtain the confidence information of multi-source heterogeneous data; S114, If the confidence level information is lower than the preset confidence level threshold, then update the estimated evidence matrix based on the backup data; S116, Based on the updated estimated evidence matrix and a weighted fusion algorithm, output the dynamic population monitoring results of the spatiotemporal grid unit.

[0024] It should be noted that this embodiment provides a comprehensive dynamic population monitoring method, the core of which lies in improving the accuracy and reliability of monitoring results through multi-source data fusion and a multi-level verification mechanism. In this embodiment, raw monitoring data is first received from multiple independent data sources, which typically include mobile communication base stations, IoT sensing devices, public transportation systems, and image acquisition devices, to ensure data diversity and complementarity. Subsequently, the raw data is preprocessed, including invalid data removal, duplicate data deduplication, timestamp alignment and conversion, geographic location mapping, and identity information anonymization, forming a standardized multi-source heterogeneous data set. Based on this, the monitoring area is divided into several grid units according to preset spatiotemporal grid rules, and corresponding multi-source data is extracted for each grid unit. Multiple evidence vectors reflecting population size are generated through a pre-trained estimation model, and then combined into an estimation evidence matrix. Then, a consistency threshold comparison is used to determine whether there are short-term data anomalies within the matrix. If anomalies are found, further analysis is conducted based on a continuity threshold to determine whether it is a sustainable event, and the weight coefficients of the estimation model are dynamically adjusted accordingly. If no anomalies are found, the evidence matrices of adjacent spatiotemporal grids are fused to assess the confidence level of the current data. If the confidence level is lower than a preset threshold, backup data is used to correct and update the evidence matrix. Finally, the population monitoring results are dynamically calculated based on the estimated evidence matrix of the current spatiotemporal grid unit using a weighted fusion algorithm. This embodiment ensures the robustness and adaptability of the monitoring process by providing a complete closed loop from data input, anomaly detection, adaptive adjustment to result output.

[0025] Figure 2 A flowchart illustrating a preprocessing procedure for multi-source monitoring data provided by an embodiment of the present invention is shown.

[0026] According to embodiments of the present invention, such as Figure 2 As shown, based on the preset data preprocessing logic, multi-source heterogeneous data is obtained from multi-source monitoring data, specifically including: S202, perform invalid data removal and duplicate data deduplication operations on the multi-source monitoring data; S204, based on timestamp alignment rules, performs time alignment on deduplicated multi-source monitoring data; S206, based on the location-unified mapping rule, performs spatial unification on time-aligned multi-source monitoring data; S208, irreversible cryptographic hashing is performed on data involving identity information to achieve data anonymization; S210, the processed data is standardized and stored according to the spatiotemporal structure to form the multi-source heterogeneous data.

[0027] It should be noted that this embodiment provides a data preprocessing operation logic to ensure the availability and consistency of multi-source data through standardized operations. In this embodiment, firstly, monitoring data obtained from different sources is cleaned, removing invalid data with incorrect formats, missing data, or obviously illogical data, and duplicate data records are deduplicated to ensure the uniqueness and accuracy of the data source. Secondly, the time information of all data is standardized and converted based on a unified timestamp rule, for example, aligning timestamps from different time zones or formats to UTC time, to facilitate subsequent time series analysis. Spatially, location information in various types of data, such as base station cell IDs, device GPS coordinates, and transit station numbers, is converted to a unified geographic coordinate system through geographic location mapping rules, enabling all data to be accurately mapped to a preset spatiotemporal grid. Then, to further protect personal privacy, identifiers in the data that may involve personal identification, such as mobile phone IMSI, device MAC address, and transit card number, are subjected to irreversible encrypted hashing, making the original information unrecoverable while ensuring the consistency of the same identifier across different data sources. Finally, the processed data is classified and stored according to its spatiotemporal attributes, forming a structured multi-source heterogeneous data set, which provides high-quality input for the subsequent generation of the evidence matrix.

[0028] Figure 3 A flowchart illustrating an estimation evidence matrix for generating spatiotemporal grid cells, provided by an embodiment of the present invention, is shown.

[0029] According to embodiments of the present invention, such as Figure 3 As shown, the extraction of multi-source heterogeneous data based on preset spatiotemporal grid partitioning rules, and the obtaining of the estimation evidence matrix for the same spatiotemporal grid unit according to a pre-trained estimation model, specifically includes: S302, based on preset latitude and longitude coordinates, divides the monitoring area into geographic subgrids of fixed size; S304, Based on a preset time slice, combined with the geographic sub-grid, a spatiotemporal grid unit is formed; S306, for each spatiotemporal grid cell, each type of data is input into the corresponding pre-trained estimation model to obtain the corresponding estimated value representing the population size; S308, all population size estimates are combined into a multidimensional evidence vector to obtain the estimation evidence matrix of the spatiotemporal grid cell.

[0030] It should be noted that this embodiment provides a process for spatiotemporal grid division and evidence matrix generation to transform abstract data into a concrete spatial population distribution representation. In this embodiment, firstly, the entire monitoring area is divided into multiple fixed-size geographic sub-grids based on a preset latitude and longitude coordinate range, each sub-grid representing an independent spatial unit. Simultaneously, the continuous time axis is divided into time slices of equal length, thereby expanding the geographic sub-grids into spatiotemporal three-dimensional grid units, which serve as the basic analytical unit for population monitoring. Secondly, for each spatiotemporal grid unit, data falling within the unit's range is extracted from preprocessed multi-source heterogeneous data, and each type of data is input into its corresponding pre-trained estimation model. The estimation models for each type of data are trained based on historical data and can output an estimated population size represented by that type of data based on the input features. Then, multiple estimates from different data sources are combined into a multi-dimensional evidence vector, where each dimension represents evidence provided by an independent data source. Finally, the evidence vectors from all time slices within the same spatiotemporal unit are organized sequentially to form an estimation evidence matrix. Among them, the estimated evidence matrix serves as the basis for subsequent consistency comparison and data fusion, reflecting the integration and expression of multi-source data in the spatiotemporal dimension.

[0031] According to an embodiment of the present invention, the comparison based on a preset consistency threshold, when a short-term data anomaly is identified in the estimated evidence matrix, is then compared based on a preset continuity threshold to obtain anomaly event information, specifically including: Calculate the pairwise dissimilarity between each evidence vector in the estimated evidence matrix of the current spatiotemporal grid cell; Determine whether the degree of difference is lower than a preset consistency threshold; If so, then it is determined that there is no short-term data anomaly; If not, mark the current evidence vector as short-term anomalous evidence; Extract the historical evidence vector of a preset number of time slices prior to the time slice in which the short-term anomaly evidence is located; Based on short-term anomaly evidence and historical evidence vectors, the changing trend and fluctuation amplitude are obtained; Based on the comparison between the changing trend and the jitter amplitude and a preset continuity threshold, it is determined whether it conforms to the preset continuity rule. If the conditions are met, the change is determined to be caused by a real event, and information on a persistent abnormal event is generated. If the condition does not meet the requirements, the change is determined to be transient data interference, and transient abnormal event information is generated.

[0032] It should be noted that this embodiment provides a logic for consistency comparison and anomaly event determination, which distinguishes real events from data noise through multi-level verification. In this embodiment, firstly, the difference degree between each pair of evidence vectors in the estimation evidence matrix of the current spatiotemporal grid cell is calculated. This difference degree reflects the consistency of different data sources in population estimation. Subsequently, the difference degree is compared with a preset consistency threshold: if all difference degrees are below the threshold, the data is determined to be normal, and the process enters the adjacent grid fusion stage; if there is a difference degree exceeding the threshold, the corresponding evidence vector is marked as short-term anomaly evidence. To analyze the cause of short-term anomalies, historical evidence vectors from several time slices before the time slice where the short-term anomaly evidence is located are extracted. By analyzing the changing trend and jitter amplitude of the anomaly evidence and the historical sequence, it is determined whether it conforms to the preset continuity rules. Based on the continuity threshold comparison, if the changing trend is stable and the amplitude is within a reasonable range, it is determined that the anomaly is caused by a sustainable real event, and sustainable anomaly event information is generated, such as data changes caused by crowd gathering; if the change is drastic and does not conform to historical patterns, it is determined to be instantaneous data interference, and instantaneous anomaly event information is generated, such as data anomalies caused by equipment failure. This embodiment achieves precise identification of anomaly types by combining temporal and spatial logic, providing a basis for decision-making in subsequent weight adjustment and data correction.

[0033] According to an embodiment of the present invention, dynamically adjusting the weight coefficients of the estimation model based on abnormal event information specifically includes: In response to the information on the persistent anomalous events, the weight coefficient of the estimation model corresponding to the short-term anomalous evidence is increased; In response to the instantaneous anomalous event information, the weight coefficient of the estimation model corresponding to the short-term anomalous evidence is reduced, or an alternative value for the short-term anomalous evidence is generated based on the historical evidence vector using a pre-trained time series prediction model.

[0034] It should be noted that this embodiment provides a specific process for dynamically adjusting the weights of the estimation model based on the type of anomalous event. An adaptive strategy optimizes the contribution of multi-source data in the fusion process to improve the accuracy of the final result. In this embodiment, when sustainable anomalous event information is generated, it indicates that the anomalous evidence provided by a certain data source is caused by a real event rather than noise interference. Therefore, the system will increase the weight coefficient of the estimation model corresponding to this anomalous evidence, giving it a higher proportion in the subsequent weighted fusion algorithm, thus more accurately reflecting the actual population change trend. When it is determined to be transient anomalous event information, it means that the data source may have experienced a temporary failure or been subject to random interference. In this case, the system will adopt two strategies: first, directly reduce the weight coefficient of the corresponding estimation model to reduce the impact of unreliable evidence on the fusion result; second, call the pre-trained time series prediction model to generate a reasonable replacement value based on the historical evidence vector sequence of the data source to replace the current anomalous evidence value, thereby maintaining the integrity and consistency of the evidence matrix. The weight adjustment mechanism provided in this embodiment is not a one-time operation, but is embedded in the continuous monitoring process. It iteratively optimizes the benchmark weights of various estimation models in combination with historical accuracy records, thereby forming an intelligent fusion system with self-learning and improvement capabilities, ensuring that the population monitoring model can adapt to changes and evolution of the data environment in the long term.

[0035] According to an embodiment of the present invention, if no abnormal event information is identified, the step of fusing the estimated evidence matrices of adjacent spatiotemporal grids to obtain confidence information of multi-source heterogeneous data specifically includes: In response to the absence of short-term data anomalies, the estimated evidence matrix of adjacent spatiotemporal grid cells in the same time slice is obtained; Calculate the spatial distribution deviation between the current grid estimation evidence matrix and the estimation evidence matrices of adjacent grids; Based on the deviation, calculate the confidence information of the multi-source heterogeneous data in the current grid.

[0036] It should be noted that this embodiment provides a process for assessing data confidence through spatial consistency verification under anomaly-free conditions, using the continuity patterns of geospatial data to provide auxiliary criteria for data reliability. In this embodiment, when the estimated evidence matrix within a specific spatiotemporal grid cell passes the consistency threshold comparison and no short-term data anomalies are found, the estimated evidence matrices of multiple adjacent spatiotemporal grid cells at the same time slice are further obtained. These adjacent grids typically have similar population distribution characteristics or change patterns in space. By calculating the deviation of the current grid evidence matrix from the evidence matrices of each adjacent grid in spatial distribution, the consistency between the current data and the surrounding environment is quantified. If the deviation is low, it indicates that the population estimate of the current grid is consistent with the surrounding area, and the data has high confidence. If the deviation is high, it suggests that the current data may have potential biases or unidentified anomalies. Finally, based on the deviation calculation results, a preset confidence assessment model is used to generate the overall confidence information of the multi-source heterogeneous data of the current grid. The confidence level information will serve as an important input for subsequent decision nodes, determining whether further verification or correction of the current evidence matrix is ​​needed, thereby enhancing the system's ability to perceive potential risks without triggering anomaly handling procedures.

[0037] It is worth mentioning that if the confidence level information is lower than a preset confidence threshold, the estimated evidence matrix is ​​updated based on the backup data, specifically including: If the confidence level information is lower than the preset confidence level threshold, the current fusion result is determined to be unreliable, and a backup data source is activated. Based on historical or backup data, backup data is extracted. Based on the preset correction rules, the corresponding evidence vector of the current spatiotemporal grid cell is corrected or replaced according to the backup data; The fusion result of the updated evidence matrix is ​​recalculated and its confidence level is reassessed until the requirements are met or the manual verification mechanism is triggered.

[0038] It should be noted that this embodiment provides a control logic for enabling a backup data source to correct the evidence matrix when the confidence level is insufficient, thereby improving the reliability of the fusion result by introducing a backup data source. In this embodiment, when the evaluation finds that the confidence level of the data in the current spatiotemporal grid cell is lower than a preset confidence level threshold, it is determined that the credibility of the fusion result obtained based on the existing evidence matrix is ​​insufficient, and a backup data source intervention mechanism needs to be activated. First, backup data is extracted from a historical database or a dedicated backup platform. The backup data usually comes from a third-party authoritative platform or a highly reliable data source that has been independently verified, and has high accuracy and stability. Subsequently, according to preset correction rules, the backup data is compared and analyzed with the corresponding evidence vector in the current evidence matrix: if there is a significant difference between the backup data and a certain evidence vector, the vector is corrected or directly replaced; if the backup data provides a completely new evidence dimension, it can be added to the matrix as a new evidence vector. After the correction is completed, the weighted fusion result of the updated evidence matrix is ​​recalculated and its confidence level is evaluated again. If the confidence level still does not meet the requirements, the above correction process is repeated or the manual verification mechanism is finally triggered, thus forming a closed-loop correction process with multi-level protection to ensure the robustness and accuracy of the output results.

[0039] It is worth mentioning that the receipt of monitoring data from multiple independent sources specifically includes: Receive signaling data from a mobile communication base station, wherein the signaling data includes at least an anonymous user identifier, a timestamp, and a base station identifier; Receive detection data uploaded by IoT sensing devices deployed within the monitoring area, wherein the detection data includes at least device identifier, signal strength, and acquisition time point; Receive card-swipe passage record data from the public transportation system, wherein the record data includes at least the anonymous card number, entry / exit identifier, and timestamp; Receive and analyze the population statistics output by image acquisition devices deployed in key areas.

[0040] It should be noted that this embodiment provides the types and content of the multi-source monitoring data received, which is used to clearly define the scope and format of data input to ensure the comprehensiveness and standardization of the monitoring basis. First, the first type of data received comes from signaling data from mobile communication base stations. This signaling data includes at least anonymized user identifiers, precise timestamps, and base station cell identifiers, reflecting the real-time location and movement trajectory of mobile phone users. Second, the second type of data received comes from IoT sensing devices deployed in the monitoring area, including but not limited to detection data uploaded by devices such as Wi-Fi probes and Bluetooth sniffers. This detection data includes at least anonymized device identifiers, signal strength indicators, and data collection time points, used to capture the distribution of people carrying smart terminal devices. Third, the third type of data received comes from card-swiping records of public transportation systems, including anonymous card numbers, entry and exit identifiers, and corresponding timestamps for subways, buses, and other transportation, providing key information about people traveling via public transportation. Finally, the fourth type of data received is output by image acquisition devices deployed in key areas after intelligent recognition and analysis. This mainly consists of statistical data on the number of people, without any personal identification information, supplementing crowd density and distribution information through visual perception. These multi-source data undergo format verification and security audits during the input stage to ensure data compliance and usability, laying a solid foundation for subsequent processing.

[0041] It is worth mentioning that it also includes: Real-time monitoring of data input interfaces ensures that all input monitoring data has undergone compliant anonymization processing. Sampling checks were conducted on intermediate data states at each stage of data preprocessing, evidence matrix generation, and fusion analysis to determine the risk of leakage of original personal information. Record all data access, model calls, and weight adjustment operations to form a complete operation log for auditing and tracing.

[0042] It should be noted that this embodiment provides a mechanism for privacy and security auditing throughout the entire process. By embedding compliance control mechanisms at each stage of data processing, it ensures personal information security and system compliance. In this embodiment, firstly, the data input interface is monitored in real time to ensure that all incoming monitoring data has undergone compliant anonymization processing before entering the processing pipeline. Any data stream containing raw personal information will be intercepted and trigger an alarm. Then, during data processing, the intermediate states of data at each stage—data preprocessing, evidence matrix generation, and fusion analysis—are periodically sampled and checked. A rule engine and risk model are used to determine if there is a risk of raw personal information leakage. If a potential risk is found, the process is immediately interrupted and a corrective procedure is initiated. Simultaneously, the system records all data access operations, model call processes, and weight adjustment actions, forming a structured and complete operation log. These logs contain key information such as timestamps, operation types, involved data identifiers, and execution results, providing a complete basis for post-event auditing and source tracing analysis. This constructs a compliant and reliable security governance mechanism.

[0043] Figure 4 A block diagram of a dynamic population monitoring system based on multi-source heterogeneous data according to the present invention is shown.

[0044] like Figure 4 As shown, the second aspect of the present invention discloses a dynamic population monitoring system 4 based on multi-source heterogeneous data, including a memory 41 and a processor 42. The memory includes a dynamic population monitoring method program based on multi-source heterogeneous data. When the processor executes the dynamic population monitoring method program based on multi-source heterogeneous data, it performs the following steps: Receive monitoring data from multiple independent sources; Based on the preset data preprocessing logic, multi-source heterogeneous data is obtained from multi-source monitoring data; Based on the preset spatiotemporal grid partitioning rules, the multi-source heterogeneous data is extracted, and the estimation evidence matrix of the same spatiotemporal grid unit is obtained according to the pre-trained estimation model. The comparison is based on a preset consistency threshold. When a short-term data anomaly is detected in the estimated evidence matrix, the comparison is based on a preset continuity threshold to obtain the anomaly event information. The weight coefficients of the estimation model are dynamically adjusted based on the abnormal event information. If no abnormal event information is identified, the estimated evidence matrix of adjacent spatiotemporal grids is fused to obtain the confidence information of multi-source heterogeneous data; If the confidence level is lower than the preset confidence level threshold, the estimated evidence matrix is ​​updated based on the backup data; Based on the updated estimated evidence matrix and a weighted fusion algorithm, the dynamic population monitoring results of the spatiotemporal grid unit are output.

[0045] It should be noted that this embodiment provides a comprehensive dynamic population monitoring method, the core of which lies in improving the accuracy and reliability of monitoring results through multi-source data fusion and a multi-level verification mechanism. In this embodiment, raw monitoring data is first received from multiple independent data sources, which typically include mobile communication base stations, IoT sensing devices, public transportation systems, and image acquisition devices, to ensure data diversity and complementarity. Subsequently, the raw data is preprocessed, including invalid data removal, duplicate data deduplication, timestamp alignment and conversion, geographic location mapping, and identity information anonymization, forming a standardized multi-source heterogeneous data set. Based on this, the monitoring area is divided into several grid units according to preset spatiotemporal grid rules, and corresponding multi-source data is extracted for each grid unit. Multiple evidence vectors reflecting population size are generated through a pre-trained estimation model, and then combined into an estimation evidence matrix. Then, a consistency threshold comparison is used to determine whether there are short-term data anomalies within the matrix. If anomalies are found, further analysis is conducted based on a continuity threshold to determine whether it is a sustainable event, and the weight coefficients of the estimation model are dynamically adjusted accordingly. If no anomalies are found, the evidence matrices of adjacent spatiotemporal grids are fused to assess the confidence level of the current data. If the confidence level is lower than a preset threshold, backup data is used to correct and update the evidence matrix. Finally, the population monitoring results are dynamically calculated based on the estimated evidence matrix of the current spatiotemporal grid unit using a weighted fusion algorithm. This embodiment ensures the robustness and adaptability of the monitoring process by providing a complete closed loop from data input, anomaly detection, adaptive adjustment to result output.

[0046] According to an embodiment of the present invention, based on a preset data preprocessing logic, multi-source heterogeneous data is obtained from multi-source monitoring data, specifically including: The multi-source monitoring data is subjected to invalid data removal and duplicate data deduplication operations. Based on timestamp alignment rules, time alignment is performed on the deduplicated multi-source monitoring data; Based on the location-unified mapping rules, spatial unification is performed on the time-aligned multi-source monitoring data; Irreversible cryptographic hashing is applied to data involving identity information to achieve data anonymization. The processed data is standardized and stored according to its spatiotemporal structure to form the multi-source heterogeneous data.

[0047] It should be noted that this embodiment provides a data preprocessing operation logic to ensure the availability and consistency of multi-source data through standardized operations. In this embodiment, firstly, monitoring data obtained from different sources is cleaned, removing invalid data with incorrect formats, missing data, or obviously illogical data, and duplicate data records are deduplicated to ensure the uniqueness and accuracy of the data source. Secondly, the time information of all data is standardized and converted based on a unified timestamp rule, for example, aligning timestamps from different time zones or formats to UTC time, to facilitate subsequent time series analysis. Spatially, location information in various types of data, such as base station cell IDs, device GPS coordinates, and transit station numbers, is converted to a unified geographic coordinate system through geographic location mapping rules, enabling all data to be accurately mapped to a preset spatiotemporal grid. Then, to further protect personal privacy, identifiers in the data that may involve personal identification, such as mobile phone IMSI, device MAC address, and transit card number, are subjected to irreversible encrypted hashing, making the original information unrecoverable while ensuring the consistency of the same identifier across different data sources. Finally, the processed data is classified and stored according to its spatiotemporal attributes, forming a structured multi-source heterogeneous data set, which provides high-quality input for the subsequent generation of the evidence matrix.

[0048] According to an embodiment of the present invention, the step of extracting the multi-source heterogeneous data based on a preset spatiotemporal grid partitioning rule and obtaining the estimation evidence matrix of the same spatiotemporal grid unit according to a pre-trained estimation model specifically includes: Based on preset latitude and longitude coordinates, the monitoring area is divided into geographic subgrids of fixed size; Based on a preset time slice, combined with the geographic sub-grid, a spatiotemporal grid unit is formed; For each spatiotemporal grid cell, each type of data is input into the corresponding pre-trained estimation model to obtain the corresponding estimated value representing the population size. All population estimates are combined into a multidimensional evidence vector to obtain the estimation evidence matrix of the spatiotemporal grid cell.

[0049] It should be noted that this embodiment provides a process for spatiotemporal grid division and evidence matrix generation to transform abstract data into a concrete spatial population distribution representation. In this embodiment, firstly, the entire monitoring area is divided into multiple fixed-size geographic sub-grids based on a preset latitude and longitude coordinate range, each sub-grid representing an independent spatial unit. Simultaneously, the continuous time axis is divided into time slices of equal length, thereby expanding the geographic sub-grids into spatiotemporal three-dimensional grid units, which serve as the basic analytical unit for population monitoring. Secondly, for each spatiotemporal grid unit, data falling within the unit's range is extracted from preprocessed multi-source heterogeneous data, and each type of data is input into its corresponding pre-trained estimation model. The estimation models for each type of data are trained based on historical data and can output an estimated population size represented by that type of data based on the input features. Then, multiple estimates from different data sources are combined into a multi-dimensional evidence vector, where each dimension represents evidence provided by an independent data source. Finally, the evidence vectors from all time slices within the same spatiotemporal unit are organized sequentially to form an estimation evidence matrix. Among them, the estimated evidence matrix serves as the basis for subsequent consistency comparison and data fusion, reflecting the integration and expression of multi-source data in the spatiotemporal dimension.

[0050] According to an embodiment of the present invention, the comparison based on a preset consistency threshold, when a short-term data anomaly is identified in the estimated evidence matrix, is then compared based on a preset continuity threshold to obtain anomaly event information, specifically including: Calculate the pairwise dissimilarity between each evidence vector in the estimated evidence matrix of the current spatiotemporal grid cell; Determine whether the degree of difference is lower than a preset consistency threshold; If so, then it is determined that there is no short-term data anomaly; If not, mark the current evidence vector as short-term anomalous evidence; Extract the historical evidence vector of a preset number of time slices prior to the time slice in which the short-term anomaly evidence is located; Based on short-term anomaly evidence and historical evidence vectors, the changing trend and fluctuation amplitude are obtained; Based on the comparison between the changing trend and the jitter amplitude and a preset continuity threshold, it is determined whether it conforms to the preset continuity rule. If the conditions are met, the change is determined to be caused by a real event, and information on a persistent abnormal event is generated. If the condition does not meet the requirements, the change is determined to be transient data interference, and transient abnormal event information is generated.

[0051] It should be noted that this embodiment provides a logic for consistency comparison and anomaly event determination, which distinguishes real events from data noise through multi-level verification. In this embodiment, firstly, the difference degree between each pair of evidence vectors in the estimation evidence matrix of the current spatiotemporal grid cell is calculated. This difference degree reflects the consistency of different data sources in population estimation. Subsequently, the difference degree is compared with a preset consistency threshold: if all difference degrees are below the threshold, the data is determined to be normal, and the process enters the adjacent grid fusion stage; if there is a difference degree exceeding the threshold, the corresponding evidence vector is marked as short-term anomaly evidence. To analyze the cause of short-term anomalies, historical evidence vectors from several time slices before the time slice where the short-term anomaly evidence is located are extracted. By analyzing the changing trend and jitter amplitude of the anomaly evidence and the historical sequence, it is determined whether it conforms to the preset continuity rules. Based on the continuity threshold comparison, if the changing trend is stable and the amplitude is within a reasonable range, it is determined that the anomaly is caused by a sustainable real event, and sustainable anomaly event information is generated, such as data changes caused by crowd gathering; if the change is drastic and does not conform to historical patterns, it is determined to be instantaneous data interference, and instantaneous anomaly event information is generated, such as data anomalies caused by equipment failure. This embodiment achieves precise identification of anomaly types by combining temporal and spatial logic, providing a basis for decision-making in subsequent weight adjustment and data correction.

[0052] According to an embodiment of the present invention, dynamically adjusting the weight coefficients of the estimation model based on abnormal event information specifically includes: In response to the information on the persistent anomalous events, the weight coefficient of the estimation model corresponding to the short-term anomalous evidence is increased; In response to the instantaneous anomalous event information, the weight coefficient of the estimation model corresponding to the short-term anomalous evidence is reduced, or an alternative value for the short-term anomalous evidence is generated based on the historical evidence vector using a pre-trained time series prediction model.

[0053] It should be noted that this embodiment provides a specific process for dynamically adjusting the weights of the estimation model based on the type of anomalous event. An adaptive strategy optimizes the contribution of multi-source data in the fusion process to improve the accuracy of the final result. In this embodiment, when sustainable anomalous event information is generated, it indicates that the anomalous evidence provided by a certain data source is caused by a real event rather than noise interference. Therefore, the system will increase the weight coefficient of the estimation model corresponding to this anomalous evidence, giving it a higher proportion in the subsequent weighted fusion algorithm, thus more accurately reflecting the actual population change trend. When it is determined to be transient anomalous event information, it means that the data source may have experienced a temporary failure or been subject to random interference. In this case, the system will adopt two strategies: first, directly reduce the weight coefficient of the corresponding estimation model to reduce the impact of unreliable evidence on the fusion result; second, call the pre-trained time series prediction model to generate a reasonable replacement value based on the historical evidence vector sequence of the data source to replace the current anomalous evidence value, thereby maintaining the integrity and consistency of the evidence matrix. The weight adjustment mechanism provided in this embodiment is not a one-time operation, but is embedded in the continuous monitoring process. It iteratively optimizes the benchmark weights of various estimation models in combination with historical accuracy records, thereby forming an intelligent fusion system with self-learning and improvement capabilities, ensuring that the population monitoring model can adapt to changes and evolution of the data environment in the long term.

[0054] According to an embodiment of the present invention, if no abnormal event information is identified, the step of fusing the estimated evidence matrices of adjacent spatiotemporal grids to obtain confidence information of multi-source heterogeneous data specifically includes: In response to the absence of short-term data anomalies, the estimated evidence matrix of adjacent spatiotemporal grid cells in the same time slice is obtained; Calculate the spatial distribution deviation between the current grid estimation evidence matrix and the estimation evidence matrices of adjacent grids; Based on the deviation, calculate the confidence information of the multi-source heterogeneous data in the current grid.

[0055] It should be noted that this embodiment provides a process for assessing data confidence through spatial consistency verification under anomaly-free conditions, using the continuity patterns of geospatial data to provide auxiliary criteria for data reliability. In this embodiment, when the estimated evidence matrix within a specific spatiotemporal grid cell passes the consistency threshold comparison and no short-term data anomalies are found, the estimated evidence matrices of multiple adjacent spatiotemporal grid cells at the same time slice are further obtained. These adjacent grids typically have similar population distribution characteristics or change patterns in space. By calculating the deviation of the current grid evidence matrix from the evidence matrices of each adjacent grid in spatial distribution, the consistency between the current data and the surrounding environment is quantified. If the deviation is low, it indicates that the population estimate of the current grid is consistent with the surrounding area, and the data has high confidence. If the deviation is high, it suggests that the current data may have potential biases or unidentified anomalies. Finally, based on the deviation calculation results, a preset confidence assessment model is used to generate the overall confidence information of the multi-source heterogeneous data of the current grid. The confidence level information will serve as an important input for subsequent decision nodes, determining whether further verification or correction of the current evidence matrix is ​​needed, thereby enhancing the system's ability to perceive potential risks without triggering anomaly handling procedures.

[0056] It is worth mentioning that if the confidence level information is lower than a preset confidence threshold, the estimated evidence matrix is ​​updated based on the backup data, specifically including: When the confidence information is lower than a preset confidence threshold, the current fusion result is determined to be unreliable, and a backup data source is activated. Based on historical or backup data, backup data is extracted. Based on the preset correction rules, the corresponding evidence vector of the current spatiotemporal grid cell is corrected or replaced according to the backup data; The fusion result of the updated evidence matrix is ​​recalculated and its confidence level is reassessed until the requirements are met or the manual verification mechanism is triggered.

[0057] It should be noted that this embodiment provides a control logic for enabling a backup data source to correct the evidence matrix when the confidence level is insufficient, thereby improving the reliability of the fusion result by introducing a backup data source. In this embodiment, when the evaluation finds that the confidence level of the data in the current spatiotemporal grid cell is lower than a preset confidence level threshold, it is determined that the credibility of the fusion result obtained based on the existing evidence matrix is ​​insufficient, and a backup data source intervention mechanism needs to be activated. First, backup data is extracted from a historical database or a dedicated backup platform. The backup data usually comes from a third-party authoritative platform or a highly reliable data source that has been independently verified, and has high accuracy and stability. Subsequently, according to preset correction rules, the backup data is compared and analyzed with the corresponding evidence vector in the current evidence matrix: if there is a significant difference between the backup data and a certain evidence vector, the vector is corrected or directly replaced; if the backup data provides a completely new evidence dimension, it can be added to the matrix as a new evidence vector. After the correction is completed, the weighted fusion result of the updated evidence matrix is ​​recalculated and its confidence level is evaluated again. If the confidence level still does not meet the requirements, the above correction process is repeated or the manual verification mechanism is finally triggered, thus forming a closed-loop correction process with multi-level protection to ensure the robustness and accuracy of the output results.

[0058] It is worth mentioning that the receipt of monitoring data from multiple independent sources specifically includes: Receive signaling data from a mobile communication base station, wherein the signaling data includes at least an anonymous user identifier, a timestamp, and a base station identifier; Receive detection data uploaded by IoT sensing devices deployed within the monitoring area, wherein the detection data includes at least device identifier, signal strength, and acquisition time point; Receive card-swipe passage record data from the public transportation system, wherein the record data includes at least the anonymous card number, entry / exit identifier, and timestamp; Receive and analyze the population statistics output by image acquisition devices deployed in key areas.

[0059] It should be noted that this embodiment provides the types and content of the multi-source monitoring data received, which is used to clearly define the scope and format of data input to ensure the comprehensiveness and standardization of the monitoring basis. First, the first type of data received comes from signaling data from mobile communication base stations. This signaling data includes at least anonymized user identifiers, precise timestamps, and base station cell identifiers, reflecting the real-time location and movement trajectory of mobile phone users. Second, the second type of data received comes from IoT sensing devices deployed in the monitoring area, including but not limited to detection data uploaded by devices such as Wi-Fi probes and Bluetooth sniffers. This detection data includes at least anonymized device identifiers, signal strength indicators, and data collection time points, used to capture the distribution of people carrying smart terminal devices. Third, the third type of data received comes from card-swiping records of public transportation systems, including anonymous card numbers, entry and exit identifiers, and corresponding timestamps for subways, buses, and other transportation, providing key information about people traveling via public transportation. Finally, the fourth type of data received is output by image acquisition devices deployed in key areas after intelligent recognition and analysis. This mainly consists of statistical data on the number of people, without any personal identification information, supplementing crowd density and distribution information through visual perception. These multi-source data undergo format verification and security audits during the input stage to ensure data compliance and usability, laying a solid foundation for subsequent processing.

[0060] It is worth mentioning that it also includes: Real-time monitoring of data input interfaces ensures that all input monitoring data has undergone compliant anonymization processing. Sampling checks were conducted on intermediate data states at each stage of data preprocessing, evidence matrix generation, and fusion analysis to determine the risk of leakage of original personal information. Record all data access, model calls, and weight adjustment operations to form a complete operation log for auditing and tracing.

[0061] It should be noted that this embodiment provides a mechanism for privacy and security auditing throughout the entire process. By embedding compliance control mechanisms at each stage of data processing, it ensures personal information security and system compliance. In this embodiment, firstly, the data input interface is monitored in real time to ensure that all incoming monitoring data has undergone compliant anonymization processing before entering the processing pipeline. Any data stream containing raw personal information will be intercepted and trigger an alarm. Then, during data processing, the intermediate states of data at each stage—data preprocessing, evidence matrix generation, and fusion analysis—are periodically sampled and checked. A rule engine and risk model are used to determine if there is a risk of raw personal information leakage. If a potential risk is found, the process is immediately interrupted and a corrective procedure is initiated. Simultaneously, the system records all data access operations, model call processes, and weight adjustment actions, forming a structured and complete operation log. These logs contain key information such as timestamps, operation types, involved data identifiers, and execution results, providing a complete basis for post-event auditing and source tracing analysis. This constructs a compliant and reliable security governance mechanism.

[0062] A third aspect of the present invention provides a computer-readable storage medium comprising a dynamic population monitoring method program based on multi-source heterogeneous data, wherein when the dynamic population monitoring method program based on multi-source heterogeneous data is executed by a processor, it implements the steps of the dynamic population monitoring method based on multi-source heterogeneous data as described in any of the preceding claims.

[0063] In summary, this invention provides a dynamic population monitoring method, system, and medium based on multi-source heterogeneous data. First, it receives multi-source monitoring data from mobile communication base stations, IoT sensing devices, public transportation systems, and image acquisition devices. This data undergoes deduplication, spatiotemporal unification, and anonymization preprocessing. Then, data is extracted according to spatiotemporal grid units and input into a pre-trained estimation model to generate an estimation evidence matrix. Next, a dual comparison mechanism using consistency and continuity thresholds is employed to identify short-term data anomalies and sustainable events, dynamically adjusting model weights or generating alternative evidence values. When no anomalies are found, adjacent grid evidence matrices are fused to calculate confidence levels; when the confidence level is below the threshold, backup data is used to correct the evidence matrix. Finally, a weighted fusion algorithm outputs the dynamic population monitoring results. This improves the accuracy, reliability, and real-time performance of population monitoring, while multi-level verification and privacy protection mechanisms ensure data security and system robustness.

[0064] If the aforementioned functions are implemented as software functional modules and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this invention, or the part that contributes to the prior art, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0065] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A dynamic population monitoring method based on multi-source heterogeneous data, characterized in that, The method includes: The system receives monitoring data from multiple independent sources, specifically including: receiving signaling data from mobile communication base stations, the signaling data including at least an anonymous user identifier, timestamp, and base station identifier; receiving detection data uploaded by IoT sensing devices deployed within the monitoring area, the detection data including at least a device identifier, signal strength, and acquisition time; receiving card-swipe passage record data from public transportation systems, the record data including at least an anonymous card number, entry / exit identifier, and timestamp; and receiving population statistics output by image acquisition devices deployed in key areas after identification and analysis. Based on the preset data preprocessing logic, multi-source heterogeneous data is obtained from multi-source monitoring data; Based on the preset spatiotemporal grid division rules, the multi-source heterogeneous data of each spatiotemporal grid cell are extracted, and the estimation evidence matrix of the same spatiotemporal grid cell is obtained according to the pre-trained estimation model. Specifically, this includes: for each spatiotemporal grid cell, inputting each type of data into the corresponding pre-trained estimation model to obtain the corresponding population size estimate; and combining all the population size estimates into a multi-dimensional evidence vector to obtain the estimation evidence matrix of the spatiotemporal grid cell. Based on a preset consistency threshold, the difference between each pair of evidence vectors in the estimated evidence matrix of each spatiotemporal grid cell is compared. When a short-term data anomaly is identified in the estimated evidence matrix, the change trend and jitter amplitude calculated from the historical evidence vectors of the short-term anomaly evidence are compared based on a preset continuity threshold to obtain the anomaly event information. The weight coefficients of the estimation model are dynamically adjusted based on the abnormal event information. If no abnormal event information is identified, the deviation of the current grid estimation evidence matrix from the adjacent grid estimation evidence matrix in spatial distribution is calculated, and the confidence information of the multi-source heterogeneous data is obtained based on the deviation. If the confidence level is lower than the preset confidence level threshold, the estimated evidence matrix is ​​updated based on the backup data; Based on the adjusted estimation model weight coefficients or the updated estimation evidence matrix, and using a weighted fusion algorithm, the dynamic population monitoring results of the spatiotemporal grid unit are output.

2. The dynamic population monitoring method based on multi-source heterogeneous data according to claim 1, characterized in that, Based on preset data preprocessing logic, multi-source heterogeneous data is obtained from multi-source monitoring data, specifically including: The multi-source monitoring data is subjected to invalid data removal and duplicate data deduplication operations. Based on timestamp alignment rules, time alignment is performed on the deduplicated multi-source monitoring data; Based on the location-unified mapping rules, spatial unification is performed on the time-aligned multi-source monitoring data; Irreversible cryptographic hashing is applied to data involving identity information to achieve data anonymization. The processed data is standardized and stored according to its spatiotemporal structure to form the multi-source heterogeneous data.

3. The dynamic population monitoring method based on multi-source heterogeneous data according to claim 1, characterized in that, Based on a preset spatiotemporal grid partitioning rule, the multi-source heterogeneous data of each spatiotemporal grid cell are extracted, and the estimation evidence matrix of the same spatiotemporal grid cell is obtained according to a pre-trained estimation model. The method also includes: Based on preset latitude and longitude coordinates, the monitoring area is divided into geographic subgrids of fixed size; Based on a preset time slice, combined with the geographic sub-grid, a spatiotemporal grid unit is formed.

4. The dynamic population monitoring method based on multi-source heterogeneous data according to claim 3, characterized in that, Based on a preset consistency threshold, the differences between each pair of evidence vectors in the estimated evidence matrix of each spatiotemporal grid cell are compared. When a short-term data anomaly is identified in the estimated evidence matrix, the trend and fluctuation amplitude calculated from the historical evidence vectors of the short-term anomaly are compared based on a preset continuity threshold to obtain anomaly event information, specifically including: Calculate the pairwise dissimilarity between each evidence vector in the estimated evidence matrix of the current spatiotemporal grid cell; Determine whether all the stated differences are below a preset consistency threshold; If so, then it is determined that there is no short-term data anomaly; If not, mark the current evidence vector as short-term anomalous evidence; Extract the historical evidence vector of a preset number of time slices prior to the time slice in which the short-term anomaly evidence is located; Based on short-term anomaly evidence and historical evidence vectors, the changing trend and fluctuation amplitude are obtained; Based on the comparison between the changing trend and the jitter amplitude and a preset continuity threshold, it is determined whether it conforms to the preset continuity rule. If the conditions are met, the change is determined to be caused by a real event, and information on a sustainable abnormal event is generated. If the condition does not meet the requirements, the change is determined to be transient data interference, and transient abnormal event information is generated.

5. A dynamic population monitoring method based on multi-source heterogeneous data according to claim 4, characterized in that, The step of dynamically adjusting the weight coefficients of the estimation model based on abnormal event information specifically includes: In response to the information on the persistent anomalous events, the weight coefficient of the estimation model corresponding to the short-term anomalous evidence is increased; In response to the instantaneous anomalous event information, the weight coefficient of the estimation model corresponding to the short-term anomalous evidence is reduced, or an alternative value for the short-term anomalous evidence is generated based on the historical evidence vector using a pre-trained time series prediction model.

6. The dynamic population monitoring method based on multi-source heterogeneous data according to claim 1, characterized in that, If no abnormal event information is identified, the deviation of the current grid estimation evidence matrix from the adjacent grid estimation evidence matrices in spatial distribution is calculated, and the confidence information of the multi-source heterogeneous data is obtained based on the deviation, specifically including: In response to the absence of short-term data anomalies, the estimated evidence matrix of adjacent spatiotemporal grid cells in the same time slice is obtained; Calculate the spatial distribution deviation between the current grid estimation evidence matrix and the estimation evidence matrices of adjacent grids; Based on the deviation, calculate the confidence information of the multi-source heterogeneous data in the current grid.

7. A dynamic population monitoring system based on multi-source heterogeneous data, characterized in that, The system includes a memory and a processor. The memory includes a program for a dynamic population monitoring method based on multi-source heterogeneous data. When the processor executes the program for the dynamic population monitoring method based on multi-source heterogeneous data, it performs the following steps: The system receives monitoring data from multiple independent sources, specifically including: receiving signaling data from mobile communication base stations, the signaling data including at least an anonymous user identifier, timestamp, and base station identifier; receiving detection data uploaded by IoT sensing devices deployed within the monitoring area, the detection data including at least a device identifier, signal strength, and acquisition time; receiving card-swipe passage record data from public transportation systems, the record data including at least an anonymous card number, entry / exit identifier, and timestamp; and receiving population statistics output by image acquisition devices deployed in key areas after identification and analysis. Based on the preset data preprocessing logic, multi-source heterogeneous data is obtained from multi-source monitoring data; Based on the preset spatiotemporal grid division rules, the multi-source heterogeneous data of each spatiotemporal grid cell are extracted, and the estimation evidence matrix of the same spatiotemporal grid cell is obtained according to the pre-trained estimation model. Specifically, this includes: for each spatiotemporal grid cell, inputting each type of data into the corresponding pre-trained estimation model to obtain the corresponding population size estimate; and combining all the population size estimates into a multi-dimensional evidence vector to obtain the estimation evidence matrix of the spatiotemporal grid cell. Based on a preset consistency threshold, the difference between each pair of evidence vectors in the estimated evidence matrix of each spatiotemporal grid cell is compared. When a short-term data anomaly is identified in the estimated evidence matrix, the change trend and jitter amplitude calculated from the historical evidence vectors of the short-term anomaly evidence are compared based on a preset continuity threshold to obtain the anomaly event information. The weight coefficients of the estimation model are dynamically adjusted based on the abnormal event information. If no abnormal event information is identified, the deviation of the current grid estimation evidence matrix from the adjacent grid estimation evidence matrix in spatial distribution is calculated, and the confidence information of the multi-source heterogeneous data is obtained based on the deviation. If the confidence level is lower than the preset confidence level threshold, the estimated evidence matrix is ​​updated based on the backup data; Based on the adjusted estimation model weight coefficients or the updated estimation evidence matrix, and using a weighted fusion algorithm, the dynamic population monitoring results of the spatiotemporal grid unit are output.

8. A dynamic population monitoring system based on multi-source heterogeneous data according to claim 7, characterized in that, Based on preset data preprocessing logic, multi-source heterogeneous data is obtained from multi-source monitoring data, specifically including: The multi-source monitoring data is subjected to invalid data removal and duplicate data deduplication operations. Based on timestamp alignment rules, time alignment is performed on the deduplicated multi-source monitoring data; Based on the location-unified mapping rules, spatial unification is performed on the time-aligned multi-source monitoring data; Irreversible cryptographic hashing is applied to data involving identity information to achieve data anonymization. The processed data is standardized and stored according to its spatiotemporal structure to form the multi-source heterogeneous data.

9. A dynamic population monitoring system based on multi-source heterogeneous data according to claim 7, characterized in that, Based on a preset spatiotemporal grid partitioning rule, the multi-source heterogeneous data of each spatiotemporal grid cell are extracted, and the estimation evidence matrix of the same spatiotemporal grid cell is obtained according to a pre-trained estimation model. The method also includes: Based on preset latitude and longitude coordinates, the monitoring area is divided into geographic subgrids of fixed size; Based on a preset time slice, combined with the geographic sub-grid, a spatiotemporal grid unit is formed.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, The computer-readable storage medium includes a dynamic population monitoring method program based on multi-source heterogeneous data. When the dynamic population monitoring method program based on multi-source heterogeneous data is executed by a processor, it implements the steps of the dynamic population monitoring method based on multi-source heterogeneous data as described in any one of claims 1 to 6.

Images