Target trajectory tracking method based on mobile phone signaling space-time big data

Abstract

The invention relates to the technical field of high-precision positioning, in particular to a target trajectory tracking method based on mobile phone signaling space-time big data, which comprises the following steps that: network side equipment generates a network service triggering condition based on dynamic parameters, and dynamically adjusts an early warning service area of an electronic fence; a service area is mapped to a cellular network topology, and a high-precision positioning technology is combined, so that an early warning service range is finely determined; mobile phone signaling data are managed through network mobility, and an existing mobile terminal set and a potential mobile terminal set are predicted and identified; analyzing historical signaling data in the mobile terminal set, constructing a personalized user portrait, and associating the personalized user portrait with specific personnel; differentiated position-related early warning information is generated and sent, and visual early warning output is generated. Finally, the dynamic early warning of the electronic fence, the refinement of the service range and the association of personnel are realized.

Description

Technical Field

[0001] The present invention relates to the field of high-precision positioning technology, and in particular to a target trajectory tracking method based on mobile phone signaling spatiotemporal big data. Background Art

[0002] Tracking the location of large-scale mobile targets in specific physical environments (such as areas affected by flash floods) is a key current technology application. Existing technologies generally use a positioning method that combines electronic fences with mobile terminal radio signals. This method monitors changes in physical quantities in the external environment. When a specific threshold is reached, the location of all mobile targets within a pre-defined electronic fence is automatically identified and determined. This is the current mainstream technical approach for achieving large-scale target tracking.

[0003] However, existing technical solutions have significant defects in positioning accuracy and tracking continuity. The electronic fences used to define the positioning range are static and have rough accuracy. Existing electronic fences are mostly based on fixed administrative or historical data. Their boundaries cannot be adjusted with real-time changes in the external environment, making it difficult to accurately match specific areas where location measurement is required (such as river valleys and low-lying areas), resulting in deviations in the positioning range of the target group. The system only performs "snapshot" location recognition at the moment of event triggering and cannot achieve continuous tracking. It cannot measure the position of new targets that enter the area after the event is triggered, nor can it confirm whether the target has left the area. The output of its positioning results has a simplification problem, that is, it provides homogeneous location coordinates to all targets within the circled range, and fails to adapt according to the target's movement characteristics or group distribution attributes, resulting in the final obtained location data lacking specificity and validity.

[0004] To this end, the present invention proposes a target trajectory tracking method based on mobile phone signaling spatiotemporal big data. Summary of the Invention

[0005] The purpose of the present invention is to provide a target trajectory tracking method based on mobile phone signaling spatiotemporal big data, which realizes dynamic, refined and personnel-related fence warning by dynamically generating electronic fences linked to risks and combining trajectory prediction with user profiling.

[0006] To achieve the above object, the present invention provides the following technical solutions:

[0007] A target trajectory tracking method based on mobile phone signaling spatiotemporal big data, comprising:

[0008] The network-side device receives and integrates the dynamic parameters of the monitoring site to generate network event trigger conditions;

[0009] Comparing the network event triggering condition with the preset multi-level warning threshold, and determining the corresponding radius with the geographical coordinates of the monitoring site as the center based on the comparison result, to generate a dynamic electronic fence;

[0010] Mapping the dynamic electronic fence to a cellular network topology of a wireless communication network to determine a dynamic base station set;

[0011] Based on a hybrid high-precision positioning algorithm, continuous mobile phone signaling data obtained from the dynamic base station set is processed to identify the existing mobile terminal set located within the dynamic electronic fence; historical movement trajectory information in the mobile phone signaling data is analyzed to predict and identify the potential mobile terminal set that will enter the dynamic electronic fence within a preset warning time window;

[0012] Analyzing historical signaling data of each mobile terminal in the existing mobile terminal set and the potential mobile terminal set, and constructing a personalized user profile based on the analysis results;

[0013] Differentiated warning information is generated based on the personalized user portrait and sent to each mobile terminal; and a visual warning output is generated on the network side device.

[0014] Preferably, the network side device receives and fuses the dynamic parameters of the monitoring site to generate network event triggering conditions, including: the network side device collects real-time risk dynamic parameters including real-time rainfall data, water level data and flow rate data through the external parameter monitoring station; receives forecast risk dynamic parameters including short-term rainfall forecast in the future time period, fuses the real-time risk dynamic parameters and the forecast risk dynamic parameters and calculates them to generate network event triggering conditions that can reflect the superposition effect of current risks and future potential risks.

[0015] Preferably, the process of generating a dynamic electronic fence includes: dynamically adjusting a preset benchmark threshold based on previous rainfall data of the area to determine the preset multi-level warning threshold; the multi-level warning threshold includes multiple warning levels arranged in ascending order according to the network event level; comparing the network event triggering condition with multiple warning levels to determine the highest network event level reached by the network event triggering condition; and based on a preset mapping relationship that maps multiple levels to different radius values, determining the value of the radius corresponding to the highest network event level; and generating a circular geographical area according to the radius with the geographical coordinates of the monitoring site as the center as a dynamic electronic fence; wherein the value of the radius increases with the increase of the highest warning level, and the boundary of the electronic fence will extend outward in a stepped manner to form a concentric multi-layer warning belt.

[0016] Preferably, the hybrid high-precision positioning algorithm includes: collecting signal strength, signal arrival time difference and arrival angle parameters in mobile phone signaling data; in a multi-base station scenario, performing triangulation calculation on at least one of the signal strength, signal arrival time difference and arrival angle parameters to determine the coordinates of the mobile terminal; in a single base station scenario, based on the signal strength and the geographical characteristics of a base station, the offset position of the mobile terminal is calculated; for 4G and 5G networks, enhanced cell ID positioning technology is used, and the position reference signal measurement of the serving base station and the detection of adjacent base stations are combined to determine the position coordinates of the mobile terminal in the wireless communication network.

[0017] Preferably, the mobile phone signaling data includes: a communication record generated by the wireless resource control interaction between the user mobile terminal and the base station in the dynamic base station set, the communication record containing a timestamp, base station cell number, signal strength, signal arrival time difference, arrival angle parameter, periodic location update signal as mobility management event signaling, switching request signal, call establishment signal and data service request signal.

[0018] Preferably, the prediction and identification of the potential mobile terminal set includes: applying a spatiotemporal clustering algorithm to perform real-time processing on the mobile phone signaling data to filter out abnormal data of ping-pong switching; the abnormal data of ping-pong switching refers to high-frequency, repetitive switching signaling generated by a mobile terminal between the cells of two adjacent base stations in a short period of time; analyzing the base station switching signaling and residence time as network mobility events to identify the initial target population; applying a Markov model to process the historical movement trajectory information of each mobile terminal in the initial target population to predict the future movement trajectory probability distribution of each mobile terminal; and based on the future movement trajectory probability distribution, calculating the probability of each mobile terminal entering the dynamic electronic fence within the preset warning time window, and determining the mobile terminals with probabilities higher than a preset threshold as the potential mobile terminal set.

[0019] Preferably, the personalized user portrait includes: extracting mobility characteristics and network residence time information from historical signaling data; associating the anonymous identification of each mobile terminal with the desensitized user database on the network operator side to obtain a preset anonymous age label containing multiple age groups; fusing the mobility characteristics with the anonymous age label to generate the personalized user portrait containing age structure and mobility type.

[0020] Preferably, the differentiated warning information includes: selecting a corresponding SMS template from the warning SMS template library according to the age structure and population mobility reflected by the anonymized age tag in the user portrait; filling in the SMS template using the geographical coordinates, dynamic parameters and information of the preset warning time window of the monitoring site; and sending a key attention notification to the preset emergency contact when the anonymized age tag in the user portrait indicates that the mobile terminal belongs to the preset old and young groups.

[0021] Preferably, the visual warning output includes: using a high-precision electronic map as a carrier to overlay and display the boundaries of a dynamic electronic fence; generating a dynamic population heat map, which represents the density of mobile terminals through different colors, and marks the peak points in the core area in red; generating a bar chart showing the number of people and age distribution within the fence, and the bar chart supports interactive operations such as click query, area screening and time backtracking.

[0022] Compared with the prior art, the present invention has the following beneficial effects:

[0023] 1. By integrating real-time risk dynamic parameters with predicted risk dynamic parameters, the present invention takes the risk source (monitoring site) as the core anchor point and constructs a dynamic electronic fence whose radius changes in real time with the risk level. The fence can adaptively expand or contract according to the actual evolution of the risk, achieving accurate matching of the warning range and the actual danger zone.

[0024] 2. By introducing a warning time window mechanism and combining it with trajectory prediction based on a Markov model, this invention not only identifies individuals currently within the danger zone, but also proactively identifies individuals at risk who are about to enter the zone in the future. This design expands warnings from a static "point in time" to a dynamic "time period," ensuring that all potentially exposed individuals within the risk period are promptly and comprehensively covered.

[0025] 3. This invention analyzes mobile phone signaling data to construct user profiles that reflect mobility characteristics and age structure, enabling differentiated push notifications for early warning information. By matching different profiles (e.g., permanent residents vs. mobile tourists) and sending warning messages with varying content and guidance, and by triggering additional emergency notifications for specific groups, the efficiency, understanding, and effectiveness of early warning information are significantly improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] Figure 1 This is an overall flow chart of a target trajectory tracking method based on mobile phone signaling spatiotemporal big data of the present invention;

[0027] Figure 2 A logical diagram generated for a dynamic electronic fence according to an embodiment of the present invention;

[0028] Figure 3 Schematic diagram of a method for dynamically identifying and covering risk groups based on spatiotemporal evolution according to an embodiment of the present invention;

[0029] Figure 4 A schematic diagram of a mechanism for generating differentiated warning information based on user profiles according to an embodiment of the present invention;

[0030] Figure 5 The figure is a flowchart of the interaction between the early warning platform and the operator in a specific scenario of an embodiment of the present invention. DETAILED DESCRIPTION

[0031] In order to make the objectives, technical solutions and advantages of the present invention clearer, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention, rather than to limit the present invention. Other embodiments obtained by those skilled in the art based on the ideas in this specification without creative work shall fall within the scope of protection of the present invention.

[0032] Reference Figures 1 to 5 , this embodiment provides a target trajectory tracking method based on mobile phone signaling spatiotemporal big data for flash flood disaster warning, which is executed by a network-side device.

[0033] Example 1:

[0034] Reference Figure 1 , a target trajectory tracking method based on mobile phone signaling spatiotemporal big data, comprising:

[0035] The network-side device receives and integrates the dynamic parameters of the monitoring site to generate network event trigger conditions;

[0036] Comparing the network event triggering condition with the preset multi-level warning threshold, and determining the corresponding radius with the geographical coordinates of the monitoring site as the center based on the comparison result, to generate a dynamic electronic fence;

[0037] Mapping the dynamic electronic fence to a cellular network topology of a wireless communication network to determine a dynamic base station set;

[0038] Based on a hybrid high-precision positioning algorithm, continuous mobile phone signaling data obtained from the dynamic base station set is processed to identify the existing mobile terminal set located within the dynamic electronic fence; historical movement trajectory information in the mobile phone signaling data is analyzed to predict and identify the potential mobile terminal set that will enter the dynamic electronic fence within a preset warning time window;

[0039] Analyzing historical signaling data of each mobile terminal in the existing mobile terminal set and the potential mobile terminal set, and constructing a personalized user profile based on the analysis results;

[0040] Differentiated warning information is generated based on the personalized user portrait and sent to each mobile terminal; and a visual warning output is generated on the network side device.

[0041] Furthermore, the network side device receives and integrates the dynamic parameters of the monitoring site to generate a network event trigger condition including:

[0042] Network-side equipment establishes data communication with multiple external parameter monitoring stations located in flash flood-prone areas via a standardized data interface. This network-side equipment periodically collects data (e.g., every five minutes) to obtain a set of dynamic risk parameters defined as real-time. This set of parameters consists of three components: real-time rainfall data measured by the monitoring station's rain gauge, water level data measured by the water level gauge, and flow velocity data measured by the current meter.

[0043] Network-side devices also receive forecast risk dynamic parameters from meteorological data service agencies through another data interface. These parameters specifically cover the short-term rainfall forecast for the next one to two hours within the monitoring area. After obtaining these two types of parameters, the network-side devices use a multi-factor weighted superposition algorithm to calculate the real-time risk dynamic parameters and the forecast risk dynamic parameters. This algorithm assigns preset weight coefficients to the measured values ​​and the forecast values ​​at different time steps. The result of this calculation generates a single quantitative indicator, namely the network event trigger condition. The value of this indicator represents the combined effect of the current monitored risk and the future forecast risk.

[0044] In a specific embodiment, the multi-factor weighted superposition algorithm is implemented through the following logical steps: the network-side device normalizes the received real-time risk dynamic parameters (such as real-time rainfall, water level, flow rate) and forecast risk dynamic parameters (such as forecast rainfall for the next hour). This processing process converts the measured or forecast value of each parameter into a dimensionless risk index between 0 and 1 by comparing it with the safety benchmark value and historical extreme value set for the region, where 0 represents no risk and 1 represents the highest historical risk; the network-side device will preset a weight coefficient for the risk index of each parameter based on the historical data analysis of flash flood disasters in the region and the focus of the emergency plan. For example, in a typical scenario, the weight of real-time rainfall can be set to 0.4, the water level to 0.3, the future forecast rainfall to 0.2, and the flow rate to 0.1, and the sum of all weight coefficients is ensured to be 1; finally, the network-side device calculates a final, quantified comprehensive risk score by multiplying the normalized risk index of each parameter by its corresponding weight coefficient and accumulating all the product results. This comprehensive risk score is the "network event triggering condition", and its value intuitively reflects the cumulative effect of current and future potential risks.

[0045] As a preferred embodiment, in order to solve the problem of sparse monitoring stations in some mountainous areas or data interruption due to bad weather, this method also includes a hydrological parameter modeling and inference mechanism, which is triggered when the real-time risk dynamic parameters are unavailable or the confidence level is lower than a preset threshold. Specifically, when the real-time risk dynamic parameters are unavailable or the confidence level is lower than the preset threshold, the mechanism is triggered, and the mechanism includes: calling a hydrological model for a specific river basin that is pre-trained based on historical data; using wide-area estimated rainfall data from meteorological radar and the previous soil moisture index dynamically calculated based on previous rainfall as inputs to the hydrological parameter model; inferring a qualitative water level risk level as a supplement to the real-time risk dynamic parameters; and assigning a confidence weight lower than the measured data to the risk component inferred by the model when integrating and calculating the network event triggering conditions.

[0046] Specifically, the hydrological parameter model utilizes a conceptual model structure based on the Antecedent Precipitation Index (API). During the model training phase, historical daily rainfall data for the target basin over the past five years and corresponding water level data for key sections were imported. Through regression analysis, a response curve between API values ​​and cross-section water levels was established.

[0047] The calculation of the aforementioned soil moisture index (ASMI) is based on a time-decayed weighted accumulation principle. The network-side device obtains the daily rainfall in the target area over a continuous period of time (e.g., fifteen days) in the past. When calculating the cumulative value, the rainfall closer to the current time is given a higher weight, while the weight of the rainfall in the past is gradually reduced over time. This decreasing weight strategy can simulate the natural evaporation and infiltration process of water in the soil over time, thereby deriving a quantitative index that can dynamically reflect the current soil saturation level.

[0048] When this mechanism is triggered, network-side equipment feeds rainfall data estimated by weather radar and the calculated ASMI value into a pre-trained API model. The model's direct output is a virtual saturation index ranging from 0 to 1. This index is mapped to a qualitative water level risk level using a set of fixed thresholds (e.g., below 0.4 for "low risk," 0.4-0.7 for "medium risk," and above 0.7 for "high risk"). The system also assesses an uncertainty based on the quality of the radar data, ultimately outputting a virtual water level estimate with a clear uncertainty interval, such as "medium risk (water level estimated between 2.5 and 3.5 meters)."

[0049] When the measured data of the monitoring station exists, but the confidence level is lower than the preset threshold due to abnormal fluctuations in the values, the risk level calculated by the model will be used as supplementary data. When the trigger conditions of network events are integrated and calculated, the weight of the measured data is 0.7, while the weight of the model-calculated data is 0.3; when the measured data of the monitoring station is completely interrupted and unavailable, the water level risk parameters calculated by the model will be used as alternative data. At this time, its weight will be temporarily increased to 0.9 to ensure the continuity of the early warning process.

[0050] The hydrological parameter modeling and inference mechanism, by utilizing the wide-area coverage of radar data and the inference capability of historical data models, solves problems such as single point failures, data transmission interruptions, or insufficient layout density that may occur at monitoring sites, thereby enhancing the robustness of the entire early warning method and business continuity in complex environments.

[0051] This embodiment transforms the traditional early warning model, which relies solely on currently measured thresholds, by integrating real-time hydrological parameters with future weather forecasts. It enables the generation of network event trigger conditions to combine the dual characteristics of "truthful reflection of the current situation" and "forward-looking prediction of the future." This allows for early identification of risks and triggering of early warnings, creating a valuable window of time for disaster prevention preparations.

[0052] Further, refer to Figure 2 The process of generating a dynamic electronic fence includes:

[0053] The network-side device retrieves and analyzes the target area's historical meteorological database, obtaining cumulative rainfall data for the region over the past period. Based on this data, it calculates a saturation index representing the current soil moisture. The network-side device stores a set of pre-set baseline thresholds, which define thresholds corresponding to different network event levels (e.g., four levels: blue, yellow, orange, and red). Based on the calculated soil saturation index, the network-side device dynamically adjusts these thresholds, lowering each level when the soil saturation index is high and raising each level when it is low. This creates a set of pre-set, multi-level warning thresholds appropriate for the current environmental context.

[0054] The calculation of the saturation index is achieved through a time decay model that better reflects the actual soil water storage capacity. The network-side equipment automatically obtains the daily rainfall data of the target area for the past 15 consecutive days. During the calculation, a higher weight is given to the more recent rainfall to simulate the natural evaporation and infiltration process of water. For example, the previous day's rainfall may be counted with a weight of 100%, while the weight of rainfall 15 days ago is attenuated to a smaller value. By summing these 15 weighted rainfall values, the network-side equipment can derive a quantitative "saturation index" that can dynamically reflect the current disaster-bearing capacity of the underlying surface. The dynamic adjustment of the threshold based on this index is automatically executed according to a preset, clear graded adjustment rule table. For example, the rule table can be set as follows: if the saturation index is lower than the "dry" threshold (such as 50), the soil disaster-bearing capacity is considered strong and the warning threshold remains at the baseline; if the index is between the "dry" and "saturated" thresholds, the soil is considered semi-saturated, and the trigger thresholds of all warning levels are lowered by 15% based on the baseline value; if the index is higher than the "saturated" threshold, the soil is considered to be unable to effectively absorb more precipitation, and all trigger thresholds are lowered by 30% based on the baseline value, thereby achieving precise and scientific warning.

[0055] The network event triggering condition is compared with a plurality of dynamically adjusted warning level thresholds, and the comparison is performed from the highest level (red) downwards until the highest network event level that the network event triggering condition can reach is determined.

[0056] The network side device has a preset mapping table that maps multiple levels to different radius values, such as Figure 2As shown, a specific example of the mapping relationship table is: a blue warning corresponds to a radius r1, such as 1 km; a yellow warning corresponds to a radius r2, such as 3 km; an orange warning corresponds to a radius r3, such as 5 km; and a red warning corresponds to a radius r4, such as 10 km, where the radius values ​​satisfy the relationship r4>r3>r2>r1. Based on the determined highest network event level, the network-side device queries and determines the value of the radius corresponding to the highest network event level from the mapping relationship table; finally, the network-side device uses the precise geographic coordinates of the monitoring site as the center and generates a circular geographic area on the geographic information system (GIS) layer based on the radius. This area is defined as a dynamic electronic fence; under this mechanism, the value of the radius increases with the increase of the highest warning level, and the boundary of the electronic fence will extend outward in a step-by-step manner, forming a concentric multi-layer warning belt with a radius that increases from the inside to the outside.

[0057] As a preferred implementation, to address potential model delays or emergencies requiring human intervention during early warning decisions, this method also provides a hybrid fence update triggering mechanism combining "scheduled updates + manual triggering." Under this mechanism, in addition to executing updates triggered by real-time data, the network-side device automatically re-executes the entire fence generation process at a fixed interval (e.g., every 30 minutes) to reflect the slowly evolving risk accumulation process. Furthermore, the network-side device provides an emergency management interface that allows authorized users to forcibly update or expand the dynamic electronic fence in an emergency by directly selecting an alert level or entering a radius value based on on-site observations or other non-modeled intelligence.

[0058] The introduction of a hybrid fence update trigger mechanism increases timeliness and decision-making flexibility. Scheduled updates ensure that warning status does not become obsolete due to data not reaching thresholds. Manual triggering provides the necessary human-computer interaction channel, allowing human expert judgment and real-time intelligence to be integrated into the warning process, greatly enhancing emergency response capabilities in complex and unexpected situations.

[0059] This embodiment dynamically adjusts the warning threshold based on previous rainfall, so that the triggering of the warning is more in line with the actual disaster-bearing capacity of the underlying surface, avoiding the warning lag or false alarm that may be caused by static thresholds, and significantly improving the accuracy and scientific nature of the warning.

[0060] Furthermore, the hybrid high-precision positioning algorithm includes:

[0061] The network-side device has built-in execution logic for adaptively selecting positioning technologies and fusing positioning results based on the real-time network environment. When processing each positioning request, the network-side device first diagnoses the mobile terminal's signal environment, including evaluating the number of base stations with which it can effectively communicate, signal quality, and whether specific positioning signaling (such as TDOA or AOA) is supported. Based on this diagnosis, the network-side device follows a preset priority strategy (for example, TDOO / AOA > signal strength fingerprint library matching > enhanced cell ID) to select and execute the optimal positioning method. If the highest-priority method fails due to unmet conditions or the returned confidence level falls below a preset threshold, the network-side device automatically switches to a backup method with the next highest priority to ensure the success rate of positioning. When multiple technologies simultaneously return valid positioning results, the network-side device initiates a confidence-based weighted fusion mechanism, which assigns higher weight to positioning results with a smaller estimated error radius. The final coordinate is calculated by taking a weighted average of the coordinates of each positioning result.

[0062] Specifically, the network-side device first collects the positioning-related parameters contained in each piece of mobile phone signaling data, mainly including signal strength, signal arrival time difference, and arrival angle; the algorithm automatically adapts to different positioning strategies according to the network environment in which the mobile terminal is located. In a multi-base station scenario, that is, when the mobile terminal can communicate with multiple base stations at the same time, the network-side device performs triangulation calculations on at least one of the collected signal strength, signal arrival time difference, and arrival angle parameters, and determines the geographic coordinates of the mobile terminal through geometric solution; in a single-base station scenario, that is, when the mobile terminal is only connected to one base station, the network-side device calculates the offset position of the mobile terminal relative to the base station based on the received signal strength and combined with the known geographic features of the base station (for example, a pre-modeled signal propagation attenuation model specific to the cell).

[0063] For mobile terminals in 4G and 5G network environments, network-side equipment uses enhanced cell ID positioning technology. The positioning process combines the measurement of the location reference signal of the serving base station and the detection information of adjacent base stations. By integrating these multi-dimensional network measurement data, the network-side equipment can determine the more precise location coordinates of the mobile terminal in the wireless communication network.

[0064] As a preferred implementation method, in order to further improve positioning accuracy and anti-interference capabilities in complex environments, this method also integrates a fusion mechanism of fingerprint library and differential positioning. Under this mechanism, the network-side device pre-collects and builds a "signal strength-position" fingerprint library. This fingerprint library records the historical signal strength distribution characteristics from multiple base stations in a gridded geographical area (for example, a grid with a unit of 100 meters); when performing real-time positioning, the network-side device matches the collected real-time signal strength with the fingerprint library using the K nearest neighbor algorithm to obtain a preliminary position estimate; at the same time, differential positioning technology is introduced, using the known precise relative position relationship between base stations, and calculating the distance difference between the mobile terminal and multiple base stations through the signal arrival time difference, and then combining the geometric algorithm to solve the precise coordinates.

[0065] The introduction of a fingerprint library and differential positioning fusion mechanism significantly enhances the reliability and accuracy of positioning. The fingerprint library provides a reliable baseline positioning in areas with stable signals, while differential technology demonstrates strong anti-interference capabilities in urban scenes with complex signals. The combination of the two enables high-precision and highly robust positioning in all scenarios.

[0066] This embodiment integrates multiple positioning technologies to form a hybrid algorithm that can adaptively switch according to real-time network conditions. This design ensures that no matter what signal coverage scenario the terminal is in, it can use the current optimal method for positioning, ensuring the universality of positioning services and the accuracy in different scenarios.

[0067] Furthermore, the mobile phone signaling data includes:

[0068] In a communication network, it is defined as a collection of communication records generated on the network side when a user mobile terminal interacts with a base station within the network for wireless resource control. The continuous mobile phone signaling data processed by the network side equipment refers to a subset of records whose base station cell numbers belong to the dynamic base station set and are screened in real time from the above-mentioned continuously generated communication record collection.

[0069] Each of the communication records specifically includes: the timestamp of the interaction, the base station cell number to which the terminal is connected, the downlink pilot signal strength received by the terminal, the signal arrival time difference and arrival angle parameters required for positioning solution, and a series of specific control signaling as mobility management event signaling. These signaling specifically include periodic location update signals for maintaining network attachment status, switching request signals triggered when the terminal moves between different cell coverage areas, call establishment signals generated when making voice calls, and data service request signals generated when accessing Internet services.

[0070] This embodiment reveals the data foundation that positioning and trajectory analysis rely on by clearly defining the specific composition of mobile phone signaling data. It summarizes and defines the multi-type wireless interaction records at the bottom layer of the network, providing standardized and information-rich data input for upper-layer algorithms such as high-precision positioning, mobility analysis, and risk prediction.

[0071] Further, refer to Figure 3 , predicting and identifying the potential mobile terminal set, including:

[0072] The network device continuously analyzes the base station handover signaling sequence of a single mobile terminal within a sliding time window (e.g., 60 seconds). In a specific embodiment, the execution logic of the sequence pattern matching algorithm is as follows: the network device maintains a fixed-length base station cell ID sequence for each terminal in real time, containing its most recent N handover records. The algorithm continuously checks this sequence for predefined subsequence patterns that characterize ping-pong handovers, such as "A→B→A" or "A→B→A→B" (where A and B are the IDs of adjacent cells). To ensure filtering accuracy and avoid misjudging normal rapid movement, the algorithm also sets a time threshold. Only when the total duration of a complete ping-pong handover pattern (e.g., "A→B→A") is within a very short time (e.g., 30 seconds) is it considered an invalid ping-pong handover event. Once a successful ping-pong handover is identified, all handover signaling within the pattern is marked as abnormal data and removed from the subsequent construction of the terminal's valid movement trajectory.

[0073] After data preprocessing, network equipment analyzes base station handover signaling (mobility events) and the duration of mobile devices' stays in each cell to identify the initial target population. The network equipment selects mobile terminals that meet both of the following conditions: first, they are undergoing continuous base station handovers and have a short stay in each cell (e.g., less than 10 minutes); second, the movement vector formed by their handover sequence points in the general direction of the dynamic electronic fence.

[0074] The network side device applies the Markov model to process the historical movement trajectory information of each mobile terminal in the initial target population and predict the probability distribution of the future movement trajectory of each mobile terminal. Figure 3 As shown at the T-1 (past) time point, this is based on the historical movement trajectory of a potential mobile terminal.

[0075] Specifically, in order to achieve accurate mapping from network location to geographic location, the network-side device will divide the warning area and the surrounding geographic space into a series of fixed-size grids (for example, 200 meters × 200 meters), and each grid is defined as an independent geographic "location state"; then, the network-side device will process the historical mobility trajectory information of the terminal over a period of time (for example, one month), that is, a time series consisting of the base station cell IDs it has visited; by parsing the base station cell ID and its signal characteristics at each time point into geographic coordinates and mapping them to the above-defined grid state, the network-side device can learn and count the user's movements from The frequency of one grid state transitioning to another adjacent grid state is calculated, thereby constructing a state transition probability matrix that describes its movement habits; when an early warning is triggered, this model can be used by network-side devices to start from the user's current grid state and, through iterative calculation, deduce the probability of the user appearing in each possible grid state within the future early warning window (such as 2 hours); finally, to obtain a single indicator for decision-making, the network-side device will accumulate the predicted probabilities of all grid states covered by the dynamic electronic fence, thereby deriving a clear and quantified "total probability of entering the danger zone", and use this as a basis for identifying potential risk groups.

[0076] The network side device calculates the probability of each mobile terminal entering the dynamic electronic fence within the preset warning time window based on the probability distribution of the future motion trajectory, and determines the mobile terminals with probabilities higher than a preset threshold as the potential mobile terminal set. Figure 3 The T0 (warning trigger) time point of the dynamic electronic fence is used to intuitively demonstrate this process: when the dynamic electronic fence (T0) is generated, the network-side device generates a predicted future motion trajectory probability distribution pointing to the inside of the fence for potential mobile terminals that are still outside the fence, and at the same time identifies existing mobile terminals that are already inside the fence.

[0077] Specifically, the network-side device maps the geographical range of the dynamic electronic fence to a group of base station cell IDs, and then traverses the probability distribution of the terminal's future movement trajectory, and accumulates the probabilities of all mobile terminals within a preset warning time window (for example, the next 2 hours) whose locations fall within this group of base station cell IDs. The network-side device determines that the mobile terminals whose accumulated probability is higher than a preset threshold (for example, 70%) are members of the potential mobile terminal set. Figure 3 The T+1 (future) time point in the figure indicates a state in the subsequent time evolution: the original potential mobile terminal has moved into the original boundary of the dynamic electronic fence (T0); at the same time, the boundary of the dynamic electronic fence has expanded to a new position, thereby including the existing mobile terminals at the time T0 and some newly covered mobile terminals.

[0078] This embodiment applies a probability model to the preprocessed signaling data to perform trajectory prediction, identifies potential people about to enter the risk area, realizes the transition from passive response to active intervention, gains a longer and valuable time window for emergency response, and significantly improves the effectiveness of early warning.

[0079] Further, refer to Figure 4 , the personalized user portrait includes:

[0080] For each mobile terminal in the identified existing mobile terminal set and potential mobile terminal set, the network side device performs the following operations: the network side device extracts mobility characteristics and network residence time information from its historical signaling data. As a specific implementation method, the network side device analyzes the signaling data of the terminal over a long period of time in the past (for example, three months), and determines its permanent residence area by identifying the base station location where it resides for a long time at night, and combines its daytime activity range and base station switching frequency to determine its mobility type as permanent, commuting or mobile, etc.

[0081] In this embodiment, the rule logic for the determination includes:

[0082] Determining "permanent" users: The network-side equipment first analyzes the terminal's base station location at night (for example, from 22:00 to 06:00 the next day) in the past three months. If there is a stable core area of ​​residence (for example, more than 80% of nighttime signaling appears in a base station group within a radius of 1 km), and the distance between the centroid of its daytime activity range and the nighttime core area is less than a close-range threshold (for example, 5 kilometers), the terminal is determined to be "permanent".

[0083] Determining "commuter" users: If a terminal does not meet the "permanent residence" condition, but the network-side device identifies two stable, separate activity centers (one that meets the nighttime residence characteristics and the other that meets the daytime working hours residence characteristics), and the distance between the two centers is greater than a long-distance threshold (for example, 10 kilometers), and there is a regular daily round-trip trajectory, then the terminal is determined to be a "commuter type."

[0084] Determining "mobile" users: If a terminal meets neither the "resident" nor the "commuter" conditions, for example, its nighttime location is very scattered with no obvious pattern, or its activity trajectory changes greatly from day to day, then the terminal is determined to be "mobile". Such users usually correspond to tourists or temporary workers.

[0085] At the same time, the network-side device associates the anonymous identifier of each mobile terminal (for example, a temporarily generated, irreversible hash value for this warning task) with a desensitized user database on the network operator side, which does not contain any information that can directly identify an individual. To ensure the technical feasibility and user privacy of the association process, this operation is completed through a preset one-way query security interface. Specifically, the network-side device sends the temporary anonymous identifier of each mobile terminal in this task to the data security domain on the network operator side. The network operator-side system completes the matching in an internal security environment and only returns the age group label corresponding to the identifier (for example, "elderly" or "youthful"), without returning any original information such as mobile phone number and name that can directly identify an individual.

[0086] Through this association operation, the network-side device obtains a preset anonymous age label bound to the terminal, which contains multiple age groups (for example, 0-20 years old, 20-30 years old, 30-40 years old, 40-50 years old, 50-60 years old, and over 60 years old); the network-side device integrates the mobility characteristics (that is, the determined mobility type) with the anonymous age label to generate a structured data record, which is the personalized user portrait containing age structure and mobility type.

[0087] This embodiment combines anonymous demographic characteristics (age) with dynamic behavioral characteristics (mobility) through multi-dimensional profiling of mobile terminals. This enables the early warning information to be transformed from indiscriminate broadcast transmission to precise push notifications for different groups of people (such as the elderly and tourists). It can provide more instructive early warning content based on their specific needs and behavioral patterns, thereby significantly improving the communication efficiency and actual effectiveness of early warning information and enhancing the pertinence and effectiveness of emergency response.

[0088] Further, refer to Figure 4 , the differentiated warning information includes:

[0089] A warning SMS template library is pre-stored in the network-side device. Each SMS template in the library is pre-associated with a specific user portrait attribute, for example, it is bound to a combination of different age structures and population mobility (i.e., mobility type); the network-side device queries and selects the corresponding SMS template in the template library based on the anonymized age label in the user portrait and the determined mobility type. For example, for a terminal with a portrait of "permanent resident" and an age group of "elderly", the system will select template A that focuses on home evacuation guidelines; for a terminal with a portrait of "mobile" (such as a tourist) and an age group of "youth", the system will select template B that contains an emergency shelter navigation link; and for a terminal with a portrait of "commuting type" and an age group of "middle-aged", the system will select template C that provides detailed evacuation route guidance.

[0090] After the template is selected, the network side device uses the real-time acquired geographic coordinates of the monitoring site, the dynamic parameters (such as current water level, warning level) and the information of the preset warning time window to fill in the preset fields in the template, thereby generating a complete and personalized warning SMS.

[0091] The network-side device fine-tunes the content of the basic SMS based on the age structure reflected by the anonymized age tag in the user profile. Specifically, this adjustment involves appending a targeted reminder at the end of the SMS. For example, if the anonymized age tag is "over 60 years old," the system will automatically add a message like, "Due to your advanced age, please prioritize your own safety and seek help from family or community workers promptly." If the age tag is "20-30 years old," the system might add a message like, "Please obtain the latest dynamic map from the official website."

[0092] In addition, the network-side device will perform an additional judgment step. When the anonymized age tag in the user portrait indicates that the mobile terminal belongs to a preset old and young group (for example, the age group is "0-20 years old" or "over 60 years old"), the network-side device will trigger a special operation; as a specific implementation method, the network-side device obtains the preset emergency contact number by querying the family package membership associated with the terminal or the emergency contact service actively set by the user, and sends a key attention notification to the number, which will remind the contact that their family is in or about to enter a risk area and requires special attention.

[0093] This embodiment not only selects warning templates based on mobility, but also makes fine adjustments to SMS content based on age tags, and adds emergency contact notifications for both the elderly and children. By providing more targeted information and leveraging social connections for supplementary reminders, it greatly improves the effective reach and understanding of warning information and the ultimate success rate of emergency response.

[0094] Furthermore, after sending the warning information to the set of existing and potential mobile terminals for the first time, the method will start a two-hour dynamic warning window. During the duration of the window, the network side device performs the following operations:

[0095] The network-side device uses a sliding time window algorithm to continuously track and identify all new users entering the dynamic electronic fence in the next two hours. The specific identification method is: the network-side device monitors the signaling data of the dynamic base station set in real time. Once it is found that a mobile terminal accesses the base station in the area for the first time after the warning is first triggered, it will be judged as a "new entrant."

[0096] To avoid disruption caused by duplicate alerts, network-side equipment maintains and updates a "list of users who have been warned" in real time. For each identified "new user," the network-side equipment first checks this list to ensure the user has not already been warned. It then immediately sends a graded alert message via the carrier's SMS channel. The logic behind this alert message also personalizes the content based on the user's profile.

[0097] By combining base station switching signaling and dwell time analysis, network-side equipment can also intelligently identify users who stay briefly within the window period and then quickly leave, thereby updating the list of users within the fence that require attention in real time.

[0098] Furthermore, the visual warning output is presented on an emergency management interactive interface, which dynamically reflects the real-time updated user list and status.

[0099] The interface uses a high-precision electronic map as its platform, overlaying the boundaries of the current dynamic geo-fence. A dynamic population heat map is also generated and overlaid on the interface. This heat map, derived from a real-time list of users within the fence, uses different colors (blue, green, yellow, and red) to represent the density of mobile terminals in each area of ​​the map. The system also highlights the highest density peaks within the core area in red.

[0100] In addition to the map, the interface also generates a data chart area, which shows a bar chart of the total number of people in the fence and the age distribution ratio. The bar chart is also linked to the real-time user list and supports a series of interactive operations, including click query, area filtering and time backtracking.

[0101] This example presents a multi-dimensional visualization of abstract risk ranges, dynamic population density, and specific age structures. This approach transforms complex signaling data into an intuitive situational picture, providing emergency commanders with a clear and comprehensive basis for decision-making. Its interactive support for data drilling and backtracking further facilitates precise resource scheduling and risk review, significantly improving the efficiency, scientific nature, and accuracy of emergency response.

[0102] This embodiment integrates measured and forecast data to generate a dynamic electronic fence that matches the risk level in real time, achieving a dynamic and precise warning range. Based on the trajectory prediction model, it proactively identifies existing and potential risk groups within the fence, solving the problem of delayed personnel identification. By building user portraits, it generates and pushes differentiated warning information, transforming indiscriminate warnings into precise reach. Combined with visual output, it provides intuitive decision-making support for emergency command, forming a closed-loop management from risk perception to precise response, significantly improving the timeliness, coverage and effectiveness of warnings.

[0103] Example 2:

[0104] This embodiment aims to apply the target trajectory tracking method based on mobile phone signaling spatiotemporal big data described in the present invention to a flash flood disaster smart disaster prevention and warning platform in a mountainous tourist area. The overall business interaction process formed in this embodiment in this scenario, with the warning platform as the core and interacting with the operator, is as follows: Figure 5 This embodiment is executed by a network-side device. As a preferred implementation, the network-side device is a server cluster deployed on the core side of the communication network or in a data center directly connected thereto. The server cluster serves as the execution entity of the early warning platform.

[0105] One summer afternoon, a flash flood disaster warning platform supported by the server cluster detected a sudden change in weather within its coverage area, triggering an alert. The platform first received dynamic parameters from a river monitoring station located upstream of the scenic area. The data showed a surge in rainfall and water levels. Simultaneously, an external warning data source (i.e., weather forecast) indicated heavy rainfall in the future. The server cluster immediately integrated these two sets of parameters and generated a high-risk network event trigger condition. This network event trigger condition was compared with the platform's pre-set multi-level warning thresholds to determine the warning area (a dynamic electronic fence with a 5-kilometer radius at the yellow warning level). Based on this 5-kilometer radius, the server cluster generated a dynamic electronic fence covering the core tourist area of ​​the scenic area on the platform's geographic information system, centered on the geographic coordinates of the monitoring station.

[0106] The server cluster maps the geographical range of the dynamic electronic fence to the cellular network topology of the wireless communication network, and quickly determines multiple base stations covering the area to form a dynamic base station set; the platform then begins to process the continuous mobile phone signaling data obtained from the dynamic base station set based on a hybrid high-precision positioning algorithm. Through processing, the platform identifies the existing mobile terminal set currently located within the dynamic electronic fence; at the same time, the platform also analyzes the historical movement trajectory information in the mobile phone signaling data in the area outside the fence, predicts and identifies that multiple mobile terminals in a tourist convoy heading towards the scenic area along the mountain road will enter the dynamic electronic fence within the preset warning time window, and thus identify them as a potential mobile terminal set.

[0107] The platform's early warning information push and closed-loop management are divided into two key stages:

[0108] Phase 1: Comprehensive alerts were sent to the initial targeted population. After identifying all the target terminals, the platform immediately analyzed their historical signaling data and constructed personalized user profiles. The analysis revealed that this group comprised a high proportion of tourists, with a significant number of elderly and young people. Based on these user profiles, the platform matched and generated differentiated warning content from its internal SMS template library. Most tourists received an SMS warning containing the message "Emergency Notice of Flash Flood Risk: Please immediately evacuate to higher ground along the main road." For terminals identified as elderly tourists, the platform not only sent a more accessible SMS message but also sent a focused alert to their pre-determined emergency contacts.

[0109] Phase 2: Continuous monitoring and recursive alerts within a two-hour alert window. After the initial alert, the platform immediately initiates a two-hour dynamic alert window. During this window, the system continuously monitors signaling data from the dynamic base station set. If a new mobile terminal not on the initial alert list accesses the area for the first time, it immediately creates a profile and intelligently dispatches the corresponding alert.

[0110] Finally, on the big screen of the scenic area's emergency command center, the server cluster generated a visual early warning output. The map clearly showed the yellow dynamic electronic fence boundary, the population heat map of the river valley area within the fence appeared dark red, and the bar chart on the side accurately displayed the total number of people within the fence and the proportion of the elderly and children. Through the result feedback interface, the platform continuously received the SMS sending status returned by the operator and dynamically updated the statistical data of the people reached on the big screen, providing comprehensive, real-time dynamic data support for the commanders' precise scheduling and evacuation decisions.

[0111] It should be understood that the above description is merely a preferred embodiment of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention shall be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention shall be defined by the claims and their equivalents.

Claims

1. A target trajectory tracking method based on mobile phone signaling spatiotemporal big data, characterized in that: include: The network-side device receives and integrates the dynamic parameters of the monitoring site to generate network event trigger conditions; Comparing the network event triggering condition with the preset multi-level warning threshold, and determining the corresponding radius with the geographical coordinates of the monitoring site as the center based on the comparison result, to generate a dynamic electronic fence; Mapping the dynamic electronic fence to a cellular network topology of a wireless communication network to determine a dynamic base station set; Based on a hybrid high-precision positioning algorithm, continuous mobile phone signaling data obtained from the dynamic base station set is processed to identify the existing mobile terminal set located within the dynamic electronic fence; historical movement trajectory information in the mobile phone signaling data is analyzed to predict and identify the potential mobile terminal set that will enter the dynamic electronic fence within a preset warning time window; Analyzing historical signaling data of each mobile terminal in the existing mobile terminal set and the potential mobile terminal set, and constructing a personalized user profile based on the analysis results; Generate differentiated warning information based on the personalized user portrait and send it to each mobile terminal; Generate a visual warning output on the network side device.

2. The target trajectory tracking method based on mobile phone signaling spatiotemporal big data according to claim 1 is characterized in that: The network-side device receives and integrates the dynamic parameters of the monitoring site to generate network event triggering conditions, including: The dynamic parameters include real-time risk dynamic parameters and forecast risk dynamic parameters; the network side equipment collects real-time risk dynamic parameters including real-time rainfall data, water level data and flow rate data through external parameter monitoring stations; receives forecast risk dynamic parameters including short-term rainfall forecasts in future time periods, integrates the real-time risk dynamic parameters and the forecast risk dynamic parameters and calculates them to generate network event trigger conditions that can reflect the superimposed effects of current risks and future potential risks.

3. The target trajectory tracking method based on mobile phone signaling spatiotemporal big data according to claim 1 is characterized in that: The process of generating a dynamic electronic fence includes: dynamically adjusting a preset benchmark threshold based on previous rainfall data of the area to determine the preset multi-level warning threshold; the multi-level warning threshold includes multiple warning levels arranged in ascending order according to the network event level; comparing the network event triggering condition with the multiple warning levels to determine the highest network event level reached by the network event triggering condition; and based on a preset mapping relationship that maps multiple levels to different radius values, determining the radius value corresponding to the highest network event level; and generating a circular geographical area according to the radius with the geographical coordinates of the monitoring site as the center as the dynamic electronic fence; wherein the radius value increases with the increase of the highest warning level, and the boundary of the electronic fence will extend outward in a stepped manner to form a concentric multi-layer warning belt.

4. The target trajectory tracking method based on mobile phone signaling spatiotemporal big data according to claim 1 is characterized in that: The hybrid high-precision positioning algorithm includes: collecting signal strength, signal arrival time difference and arrival angle parameters from mobile phone signaling data; in a multi-base station scenario, performing triangulation calculations on at least one of the signal strength, signal arrival time difference and arrival angle parameters to determine the coordinates of the mobile terminal; in a single-base station scenario, calculating the offset position of the mobile terminal based on the signal strength and the geographical characteristics of a base station; for 4G and 5G networks, using enhanced cell ID positioning technology, combined with the position reference signal measurement of the serving base station and the detection of adjacent base stations, to determine the position coordinates of the mobile terminal in the wireless communication network.

5. The target trajectory tracking method based on mobile phone signaling spatiotemporal big data according to claim 1 is characterized in that: The mobile phone signaling data includes: a communication record generated by the wireless resource control interaction between the user mobile terminal and the base station in the dynamic base station set, the communication record contains a timestamp, base station cell number, signal strength, signal arrival time difference, arrival angle parameter, periodic location update signal as mobility management event signaling, switching request signal, call establishment signal and data service request signal.

6. The target trajectory tracking method based on mobile phone signaling spatiotemporal big data according to claim 1 is characterized in that: The prediction and identification of the potential mobile terminal set includes: applying a spatiotemporal clustering algorithm to process the mobile phone signaling data in real time and filtering out abnormal data of ping-pong switching; the abnormal data of ping-pong switching refers to the high-frequency and repetitive switching signaling generated by a mobile terminal between the cells of two adjacent base stations in a short period of time; analyzing the base station switching signaling and residence time as network mobility events to identify the initial target population; applying a Markov model to process the historical movement trajectory information of each mobile terminal in the initial target population and predicting the future movement trajectory probability distribution of each mobile terminal; and based on the future movement trajectory probability distribution, calculating the probability of each mobile terminal entering the dynamic electronic fence within the preset warning time window, and determining the mobile terminals with probabilities higher than a preset threshold as the potential mobile terminal set.

7. The target trajectory tracking method based on mobile phone signaling spatiotemporal big data according to claim 1 is characterized in that: The personalized user profile includes: extracting mobility characteristics and network residence time information from historical signaling data; associating the anonymous identifier of each mobile terminal with the desensitized user database on the network operator side to obtain a preset anonymous age label containing multiple age groups; fusing the mobility characteristics with the anonymous age label to generate the personalized user profile including age structure and mobility type.

8. The target trajectory tracking method based on mobile phone signaling spatiotemporal big data according to claim 1 is characterized in that: The differentiated warning information includes: selecting a corresponding SMS template from a warning SMS template library based on the age structure and population mobility reflected by the anonymized age tag in the user portrait; filling the SMS template with the geographical coordinates, dynamic parameters and information of a preset warning time window of the monitoring site; and sending a key attention notification to a preset emergency contact when the anonymized age tag in the user portrait indicates that the mobile terminal belongs to a preset old and young group.

9. The target trajectory tracking method based on mobile phone signaling spatiotemporal big data according to claim 1 is characterized in that: The visual warning output includes: using a high-precision electronic map as a carrier to overlay and display the boundaries of a dynamic electronic fence; generating a dynamic population heat map, which uses different colors to represent the density of mobile terminals and marks the peak points in the core area in red; and generating a bar chart showing the number of people and age distribution within the fence. The bar chart supports interactive operations such as click query, area screening, and time backtracking.

Images