A security protection method and system based on communication-navigation-remote control fusion technology

Abstract

This invention discloses a security protection method and system based on communication-navigation-remote sensing fusion technology. By constructing a multi-link communication system, an integrated indoor and outdoor high-precision positioning mechanism, and a multi-modal intelligent sensing and collaborative mechanism, it achieves adaptive selection and redundant transmission of communication links in environments with weak networks or obstructions. It combines multi-source information such as GNSS, inertial navigation, geomagnetic, and remote sensing image backbone maps to achieve continuous positioning across scenarios. Furthermore, it integrates inertial navigation, video, sound, and environmental sensor data to generate abnormal behavior recognition results, thereby forming a full-process, all-scenario security protection capability for all personnel. It effectively overcomes the problems of easy communication interruption, inaccurate positioning, and unreliable abnormal identification in existing technologies, and is applicable to multiple application scenarios such as crowd security protection.

Description

Technical Field

[0001] This invention relates to the field of artificial intelligence and multi-source fusion positioning and sensing technology, and proposes a security protection method and system utilizing "communication-navigation-remote control" fusion technology. Background Technology

[0002] With the improvement of urbanization and the acceleration of population mobility, the safety risks faced by the public in daily travel, work and production, campus and community activities, cultural tourism and outdoor sports are becoming more diversified and frequent. In complex environments, people may get lost, fall accidentally or suffer sudden illness, stay for a long time, cross boundaries into dangerous areas, deviate from the path, or encounter emergencies. At the same time, with the increasing demand for emergency response and rapid rescue, the need for real-time safety protection, location services and risk warnings for all people is becoming more and more urgent.

[0003] To address this challenge, technologies such as location tracking, abnormal behavior identification, and remote communication have become crucial supports for personnel safety. However, existing methods still suffer from problems such as unstable communication, inaccurate guidance, and lack of intelligent remote control in complex environments: communication stability is limited in weak network or obstructed scenarios such as mountainous areas, indoors, tunnels, and under trees; positioning accuracy is prone to decline in obstructed and multipath scenarios; and the ability to actively identify and perceive anomalies such as falls, lingering, boundary crossings, and path deviations is relatively weak. Summary of the Invention

[0004] This invention makes full use of communication, navigation, and remote sensing technologies, and proposes a security protection method and system based on communication-navigation-remote sensing fusion technology. By constructing multi-link communication protection, indoor and outdoor integrated high-precision positioning, and multi-modal intelligent perception integrated collaboration, it provides all-round, all-weather intelligent security protection for all personnel.

[0005] The technical solution of this invention is a security protection method based on communication-navigation-remote control fusion technology, comprising the following steps:

[0006] A multi-link communication strategy module is constructed to obtain the current communication link status, determine message priority and transmission mechanism by calculating the business importance index, classify alarm levels, and coordinate the indoor and outdoor integrated high-precision positioning mechanism module and the multimodal intelligent sensing and anomaly identification mechanism module.

[0007] Construct an integrated indoor and outdoor high-precision positioning mechanism module to obtain positioning results and location reliability;

[0008] Construct a multimodal intelligent perception and anomaly recognition mechanism module to identify abnormal events and obtain the confidence level of abnormal events;

[0009] Establish a rule base for associating personnel, location, and time. Dynamically compare the personnel to be protected with the rule base to calculate the corresponding risk score. Combine the current communication link status, current location confidence, and abnormal event confidence to weight and correct the credibility factor in the business importance index. When the risk score reaches or exceeds the corresponding threshold, trigger the risk event and send an alarm.

[0010] Based on alarm notifications, the multi-link communication strategy module, the indoor and outdoor integrated high-precision positioning mechanism module, and the multimodal intelligent sensing and anomaly recognition mechanism module were optimized.

[0011] Furthermore, constructing the multi-link communication strategy module includes the following sub-steps:

[0012] Step 1.1: Establish a multi-link communication set. This involves constructing a set of multiple communication links, including cellular communication links, satellite communication links or short message communication links, low-power wide-area communication links, and emergency broadcast channels, to form a communication foundation that is adaptable to different scenarios and network conditions.

[0013] Step 1.2: Construct a link selection function and implement adaptive primary / backup switching. Estimate the link quality, time delay, communication cost, and power consumption of various communication links in real time, and establish a comprehensive link score.

[0014]

[0015] Where Q is the link quality, L is the time delay, C is the communication cost, P is the power consumption, and w Q w represents the link quality weight. L For time delay weights, w C As a weight for communication costs, w P Power consumption weight;

[0016] Step 1.3, establish message priority and reliable transmission mechanism: classify messages to be sent according to the business importance index, including at least SOS distress messages, abnormal alarm messages and regular business messages, and manage queues in the priority order of SOS > abnormal alarm messages > regular business messages; for SOS distress messages and critical alarm messages, adopt the fragmentation and retransmission window mechanism, and perform concurrent redundant transmission on multiple links;

[0017] Step 1.4: When a weak network or obstructed environment is detected, causing a decrease in communication quality, a data downsampling and compression strategy is triggered. The data sampling and uploading frequency is dynamically adjusted based on the positioning results and the confidence level of abnormal events, so as to achieve coordinated control between the indoor and outdoor integrated high-precision positioning mechanism module and the multimodal intelligent sensing and anomaly recognition mechanism module.

[0018] Furthermore, the business importance index is calculated using the following formula:

[0019] I=S×U×Ctx×Conf

[0020] Where S represents severity, U represents urgency, Conf represents credibility factor, and Ctx represents scenario coefficient;

[0021] Alarm levels are categorized based on the importance of the business: Level I0: I≥40, Level I1: 25≤I<40, Level I2: 10≤I<25, Level I3: I<10.

[0022] Alternatively, alarm levels can be classified based on severity and urgency. If S≥5 and U≥4, it is directly classified as level I0.

[0023] If S≥4 and U≥4, the minimum classification is I1.

[0024] Furthermore, the construction of an integrated indoor and outdoor high-precision positioning mechanism module includes the following sub-steps:

[0025] Step 2.1, scene recognition, including indoor scenes, open outdoor scenes, and occluded outdoor scenes;

[0026] Step 2.2, Scene-specific positioning strategy selection and fusion positioning; Based on the scene identified in Step 2.1, determine the observation source and external constraints used for fusion positioning: In indoor scenes, select a combination of inertial navigation data and geomagnetic data to achieve indoor positioning; In open outdoor scenes, select GNSS positioning results and fuse inertial navigation data to achieve continuous high-precision positioning; In obstructed outdoor scenes, select inertial navigation calculation results as the main method, and introduce a backbone map of the passable area constructed offline by remote sensing imagery / map as an external constraint to restrict the trajectory to the passable area; After completing the selection of observation source and constraint, a unified fusion positioning process is used for estimation: Establish a state space model and an observation model, and set a noise model and covariance matrix for inertial navigation drift, GNSS observation error, geomagnetic matching error and map constraint error; The noise model parameters and constraint weights are adaptively adjusted according to the scene identification results, satellite geometric accuracy factor and matching residual, thereby correcting unreasonable position solutions and outputting positioning confidence;

[0027] Step 2.3: Output the positioning results and perform trajectory quality assessment; based on the positioning results, output positioning information including position coordinates, velocity information and corresponding confidence scores, and calculate trajectory continuity indicators, including positioning point missing rate, number of breakpoints, maximum breakpoint duration, and jump distance between adjacent points. Perform segmented quality assessment and mark abnormal segments on the trajectory, and perform trajectory evidence storage and playback.

[0028] Furthermore, the indoor and outdoor integrated high-precision positioning mechanism module also outputs the confidence level of the positioning results to the multimodal intelligent perception and anomaly recognition mechanism module in real time, which is used to weight the location confidence level of the abnormal event recognition results; the trajectory quality assessment results are fed back to the multi-link communication strategy module, which is used to advance the priority of alarm links or enable backup links when the positioning quality deteriorates.

[0029] Furthermore, the construction of the multimodal intelligent perception and anomaly recognition mechanism module includes the following sub-steps:

[0030] Step 3.1, Multimodal data acquisition and preprocessing: Inertial navigation data is acquired through inertial sensors, and environmental measurement data is acquired through authorized public cameras, home cameras, environmental sound sensors, and temperature sensors. The data of each type are filtered and denoised, time-synchronized, feature-extracted and normalized to provide a unified spatiotemporal reference and usable features for subsequent identification.

[0031] Step 3.2, Modal Recognizer Construction: Based on inertial navigation data, gait analysis, peak acceleration detection, and posture change modeling algorithms are used to identify abnormal behavior; based on image data, target detection, human posture estimation, and behavior recognition algorithms are used to determine the duration of abnormal behavior; based on sound and environmental data, model recognition methods are used to detect breathing sounds, abnormal noise, and high-temperature environmental risks, and preliminary recognition results and corresponding confidence levels are obtained for each modality.

[0032] Step 3.3, Multimodal Fusion and Abnormal Event Judgment: Input the single-modal recognition results obtained in Step 3.2 into the multimodal fusion module, and use a judgment method that combines modal deep learning network and weighted confidence fusion model to perform joint analysis on the same target or the same time in the time series, generate a comprehensive confidence score and classification result for abnormal behavior protection events, and output spatiotemporal labels associated with abnormality type, severity and location information.

[0033] Furthermore, by collecting historical trajectories, activity frequencies, stay patterns, and abnormal event records, a dynamic user profile is formed for each individual, recording their typical travel time, common activity areas, and behavioral characteristics. The real-time location trajectory and behavior recognition results are dynamically compared with the user profile and rule base. When it is found that an individual deviates from the normal route, crosses the electronic fence, appears in an abnormal area during an abnormal time period, stays for a significantly longer period than normal, or falls with a high degree of confidence, the values ​​of severity S and urgency U are determined, and the risk score Rbase=S×U×Ctx is calculated by combining the scenario coefficient Ctx.

[0034] The confidence factor Conf is obtained by weighting and adjusting the confidence level based on the communication confidence level Conf_com, the location confidence level Conf_loc, and the abnormal event confidence level Conf_evt. The calculation formula is as follows:

[0035] Conf=clip(w_e×Conf_evt+w_l×Conf_loc+w_c×Conf_com,0,1)

[0036] w_e+w_l+w_c=1

[0037] Where w_e, w_l, and w_c are weights, and clip is a cutoff function that takes values ​​in the range of 0-1.

[0038] Furthermore, triggering risk events and sending alarms includes: requesting the indoor and outdoor integrated high-precision positioning mechanism module to activate the high-frequency positioning mode to improve the spatiotemporal tracking accuracy of the target; calling the constructed multi-link communication strategy module to push alarm information, associated trajectories, and identification results to family member terminals, community or school management terminals, and emergency command terminals to achieve multi-role linkage response; and for extremely high-risk abnormal events, simultaneously notifying the multimodal intelligent perception and anomaly recognition mechanism module to increase the perception sampling frequency or enable more camera perspectives.

[0039] Further optimizations based on alarm notifications include:

[0040] Step 5.1: Receipt and record the handling process after the alarm information is pushed, and store the receipt information in association with the communication link quality, positioning accuracy and multimodal recognition results used in the current alarm process;

[0041] Step 5.2: Based on the event samples recorded in Step 5.1, adaptively update the weight parameters of the link selection function in the multi-link communication strategy module, the positioning strategy selection in the indoor and outdoor integrated high-precision positioning mechanism module, and the multi-modal fusion weights and the judgment thresholds of each abnormal event in the multi-modal intelligent perception and anomaly recognition mechanism module.

[0042] This invention also provides a security protection system based on communication-navigation-remote control technology, including a processor and a memory. The memory is used to store program instructions, and the processor is used to call the program instructions in the memory to execute a security protection method based on communication-navigation-remote control technology as described in the above technical solution.

[0043] The advantages of this invention lie in its proposal of an intelligent safety protection method based on the fusion of communication, navigation, and remote sensing technologies. By constructing a multi-link communication guarantee, integrated indoor and outdoor high-precision positioning, and a multi-modal intelligent sensing integrated collaborative mechanism, it provides comprehensive, all-weather intelligent safety protection for people, including the elderly and children. This method achieves multi-link communication, integrated indoor and outdoor high-precision positioning, and intelligent behavior perception through the integrated collaboration of communication, navigation, and remote sensing, applying all-time and all-space safety monitoring and location services. Attached Figure Description

[0044] Figure 1 This is a flowchart of an embodiment of the present invention.

[0045] Figure 2 This is the indoor positioning result in an embodiment of the present invention.

[0046] Figure 3 This is the outdoor positioning result in an embodiment of the present invention.

[0047] Figure 4 This is the visual perception result in an embodiment of the present invention.

[0048] Figure 5 This is the IMU sensing result in an embodiment of the present invention. Detailed Implementation

[0049] The technical solution of the present invention will be further described below with reference to the accompanying drawings and embodiments.

[0050] like Figure 1 As shown, the technical solution provided by this embodiment of the invention is a security protection method based on the fusion technology of "communication-navigation-remote control", which includes the following steps:

[0051] Step 1, construct the multi-link communication strategy module (Tong), including the following sub-steps,

[0052] Step 1.1: Establish a multi-link communication set. By constructing a set of multiple communication links, including but not limited to cellular communication links (4G / 5G), satellite communication links or short message communication links, low-power wide-area communication links, and emergency broadcast channels, a communication foundation adaptable to different scenarios and network conditions can be formed.

[0053] Step 1.2: Construct the link selection function and implement adaptive primary / backup switchover. Real-time estimates are performed on the link quality, time delay, communication cost, and power consumption of various communication links to establish a comprehensive link score.

[0054]

[0055] Where Q is the link quality (i.e., the probability of successful transmission in the near future, which can be obtained by dividing the number of successful deliveries by the number of transmissions), L is the time delay (which can be determined by message confirmation, i.e., the time from when the terminal sends a message to when the server confirms receipt), C is the communication cost (estimated according to a traffic-based billing model, such as cellular cost per MB, satellite cost per duration / message / traffic, and for WiFi / self-organizing networks, it can be approximated as extremely low cost), P is the power consumption (obtained through experimental calibration by adding the energy consumption of each transmission and reception for each link), and w Q w represents the link quality weight. L For time delay weights, w C As a weight for communication costs, w P The power consumption weight can be obtained through deep learning. Based on historical communication performance (information recorded and periodically reported by the terminal and server, i.e., communication logs automatically accumulated during system operation, including delivery success rate, packet loss rate, retransmission count, timeout count, disconnection count, and handover count) and handling feedback (whether the pushed message or the task to be done has been effectively handled, such as information confirmed / acted, completed, false alarm, no need for assistance, accidental trigger, I am safe, etc.), the weight is learned and adaptively updated online, thereby realizing automatic selection of primary and backup links in different environments and enabling concurrent redundant transmission of multiple links when necessary.

[0056] Step 1.3: Establish message prioritization and reliable transmission mechanisms. Messages to be sent are categorized according to their importance, including at least SOS distress messages, abnormal alarm messages, and regular business messages. Queues are managed according to the priority order: SOS > Alarm > Regular. For SOS and critical alarm messages, a fragmentation and retransmission window mechanism is adopted, and concurrent redundant transmission is performed on multiple links. A confirmation receipt mechanism ensures reachability and reliability under complex network conditions.

[0057] The formula for calculating business importance is as follows:

[0058] I=S×U×Ctx×Conf

[0059] Where S represents severity: impact on personal safety (1-5), U represents urgency: how quickly to reach / handle the situation (1-5), Conf represents credibility factor: whether the algorithm is reliable (0-1), and Ctx represents scenario coefficient: dangerous areas, nighttime, being alone, weak network, etc. (1-k).

[0060] Alarm messages with a business importance ≥ I1 are critical alarm messages, which require immediate closed-loop processing and demand higher delivery rates and lower latency. If necessary, multi-link concurrent redundant sending and strong acknowledgment mechanisms should be enabled.

[0061] I0 Emergency Life Safety: Must be done as quickly as possible, cost sacrifices are acceptable (multi-link redundancy / repeated transmission / strong acknowledgment); I1 High Priority Security: Strong real-time and strong reliability (master / backup switchover + critical field redundancy; bit rate reduction when necessary but without losing critical fields); I2 General Business: Can be sent in real-time but not with mandatory compression (weak network triggers compression / frequency reduction); I3 Low Priority: Can be delayed, can be compressed, can be uploaded in batches.

[0062] The importance of the business is classified into levels: Level I0: I≥40, Level I1: 25≤I<40, Level I2: 10≤I<25, Level I3: I<10.

[0063] In addition, the severity and urgency are classified into levels. If S≥5 and U≥4, it is directly judged as level I0.

[0064] If S≥4 and U≥4, the minimum classification is I1.

[0065] If Conf is less than 0.4, but the total value of S×U×Ctx is very high, I1 will still be pushed, but it will be marked as pending confirmation;

[0066] When I0 triggering (to avoid emergency flooding), at least one of the following must also be satisfied:

[0067] Serious incidents fall under the category of life-threatening incidents (e.g., falls, SOS calls, entering high-risk fences, etc.):

[0068] U=5 and a risk evolution trend was detected (increased depth of crossover, increased deviation distance, or rapid increase in duration).

[0069] Step 1.4: Implement a weak network downsampling and compression strategy. When a weak network or obstructed environment is detected, causing a decline in communication quality, the data downsampling and compression strategy is triggered. Key data such as positioning data and anomaly identification results are prioritized for retention and encoding compression, while non-critical or redundant data is transmitted at a reduced frequency to ensure the timeliness and continuity of critical information in bandwidth-constrained scenarios. During this process, the data sampling and upload frequency can be dynamically adjusted based on the positioning accuracy requirements provided in Step 2 and the severity of the anomalies output in Step 3, achieving coordinated control of the "navigation" and "remote" modules by the communication strategy.

[0070] Step 2, construct an integrated indoor and outdoor high-precision positioning mechanism module (guide), including the following sub-steps,

[0071] Step 2.1, Scene Recognition. The scene in which the user is located is identified based on indicators such as GNSS signal availability and positioning deviation: when GNSS signal is sufficient and positioning deviation is small, it is determined to be an open outdoor scene; when GNSS signal is present but positioning deviation is large, it is determined to be an outdoor scene obscured by trees or buildings; when GNSS signal is severely attenuated or cannot be calculated normally, it is determined to be an indoor scene.

[0072] Step 2.2, Scene-Specific Positioning Strategy Selection and Fusion Positioning. Based on the scene identified in Step 2.1, the observation source and external constraints for fusion positioning are determined: In indoor scenes, inertial navigation data is selected and combined with geomagnetic fingerprint matching to achieve indoor positioning; in open outdoor scenes, GNSS positioning results are selected and fused with inertial navigation data to achieve continuous high-precision positioning; in obstructed outdoor scenes, inertial navigation estimation results are selected as the main method, and a backbone map of the passable area constructed offline by remote sensing imagery / map is introduced as an external constraint to restrict the trajectory to the passable area to achieve positioning. After completing the selection of observation source and constraints, a unified fusion positioning process is used for estimation: a state-space model and an observation model are established, and a noise model and covariance matrix are set for inertial navigation drift, GNSS observation error, geomagnetic matching error, and map constraint error; the noise model parameters and constraint weights are adaptively adjusted based on scene identification results, satellite geometric accuracy factors, and matching residuals, thereby correcting unreasonable location solutions and outputting positioning reliability.

[0073] Step 2.3: Output positioning results and trajectory quality assessment. Based on the positioning results, output positioning information including location coordinates, velocity information, and corresponding confidence levels. Calculate trajectory continuity indicators, including positioning point missing rate, number of breakpoints, maximum breakpoint duration, and jump distance between adjacent points. Perform segmented quality assessment and mark abnormal segments on the trajectory, and store and replay the trajectory to support path tracking, search and rescue decisions, and post-event review analysis for missing persons incidents. Simultaneously, output the confidence level of the positioning results in real time to the multimodal anomaly recognition module in Step 3 to weight the behavior recognition results with location confidence, thus constraining and enhancing the judgment of remote sensing. The trajectory quality assessment results can also be fed back to Step 1 to proactively increase the priority of alarm links or activate backup links when positioning quality deteriorates.

[0074] The segmented quality assessment includes:

[0075] (1) Trajectory segmentation: A new trajectory segment is opened when any of the following changes occur:

[0076] Scene recognition changes (switching between open outdoor, occluded, and indoor environments);

[0077] The localization solution method has changed (similar to scene recognition);

[0078] Changes in communication links;

[0079] The sudden change in velocity / acceleration / heading exceeds the threshold;

[0080] (2) Quality assessment: The following parameters are used to calculate the quality characteristics of each trajectory segment, and the output quality level (Excellent / Medium / Poor) is determined:

[0081] Positioning uncertainty: horizontal accuracy, etc.;

[0082] Continuity: percentage of missing points, number of breakpoints, and maximum breakpoint duration;

[0083] Reasonableness: The jump distance between adjacent points, the speed / acceleration are reasonable, the GNSS movement speed, the degree of deviation between the trajectory and the road, etc.;

[0084] The abnormal paragraph markers include:

[0085] Drift / Jump Point Abnormality: Unreasonable large jumps occur within the segment, or the instantaneous speed exceeds the upper limit;

[0086] Low-quality anomaly: Too few satellites, no solution value;

[0087] Intermittent abnormalities: prolonged absence, intermittent localization;

[0088] Scene inconsistency anomaly: If the scene is judged to be indoors but has good GNSS quality, or if the scene is judged to be outdoor but has poor GNSS quality.

[0089] Map constraint anomaly: Persistently located inside buildings / on water / in restricted areas.

[0090] Step 3, construct a multimodal intelligent sensing and anomaly recognition mechanism module (remote sensing), including the following sub-steps,

[0091] Step 3.1, Multimodal Data Acquisition and Preprocessing. Inertial navigation data is acquired through the device's built-in inertial sensors, and environmental measurement data is collected through public authorized cameras, home cameras, ambient sound sensors, temperature sensors, etc. The various types of data are filtered and denoised, time-synchronized, feature-extracted, and normalized to provide a unified spatiotemporal reference and usable features for subsequent identification.

[0092] Step 3.2, Modal Recognizer Construction. Based on inertial navigation data, algorithms such as gait analysis, peak acceleration detection, and posture change modeling are used to identify behaviors such as falls, prolonged stillness, sudden stops, and abnormal acceleration. Based on image data, target detection, human posture estimation, and behavior recognition algorithms are used to make preliminary judgments on the time of falls, lingering, running, and abnormal wandering. Based on sound and environmental data, model recognition methods are used to detect environmental risks such as breathing sounds, abnormal noise, and high temperatures, obtaining preliminary recognition results and corresponding confidence levels for each modality.

[0093] Step 3.3, Multimodal Fusion and Abnormal Event Judgment. The single-modal recognition results obtained in Step 3.2 are input into the multimodal fusion module. A judgment method combining, but not limited to, multimodal deep learning networks and weighted confidence fusion models, is used to jointly analyze the same target or time period over time. This generates comprehensive confidence scores and classification results for guardianship events such as falls, lingering, path deviations, and abnormal nighttime outings. The module outputs spatiotemporal labels associated with the abnormality type, severity, and location information. The multimodal fusion module sets fusion weights for each modality output. These weights are adaptively determined and normalized based on the availability, stability, and historical accuracy statistics of the corresponding modality within a sliding time window. A judgment threshold is set for each abnormal event type to map the comprehensive confidence level to whether the abnormality is established / not established or a risk level. Furthermore, an intermodal conflict metric is defined to characterize the degree of disagreement among different modalities regarding the same event (e.g., calculated from the difference in confidence levels or weighted variance of each modality), and compared with a conflict threshold. When the defined intermodal conflict metric exceeds the conflict threshold, secondary confirmation is triggered, the sampling frequency is increased, or more modal data is requested. The fusion weights for each modality output can be updated online based on the event samples recorded in step 5.1. When the overall confidence level is low or there is conflict between modalities, the fusion module can call the trajectory continuity and velocity information provided in step 2 for cross-validation, and if necessary, request step 2 to increase the positioning update frequency to perform secondary confirmation of the abnormal state. After confirming a high-risk anomaly, the sending priority of relevant alarm messages is increased and multi-link redundancy is enabled through step 1.

[0094] (1) Fall cross-validation:

[0095] If the IMU / imagery gives a high confidence level for a fall, but the trajectory shows that the movement is still ongoing and the speed is consistently greater than the threshold, it is judged as a conflict, and the fall confidence level is reduced or marked as "pending confirmation".

[0096] The IMU / imagery provides a high confidence level for the fall, with the trajectory showing instantaneous deceleration / speed close to 0 followed by a stop, and good trajectory continuity, thus improving the overall confidence level.

[0097] (2) Retention cross-validation:

[0098] Image recognition shows that if the trajectory continuously decreases in speed within a small area and the dwell time exceeds the threshold, it is confirmed as a dwelling.

[0099] Image recognition shows lag, but the trajectory has many breakpoints and poor continuity, which reduces the credibility and turns it into a suspected lag.

[0100] (3) Path offset cross-validation:

[0101] If the trajectory segment has a low quality score or contains abnormal jump points, deviance events will be downgraded or reconfirmed (to avoid false alarms caused by GNSS drift).

[0102] If the trajectory is continuous and the deviation from the preset route / geofence continues to exceed the threshold;

[0103] (4) Unusual nighttime outings

[0104] If the time exceeds the nighttime window and the trajectory shows leaving the residential / safe area and continuous movement exceeding the threshold, then abnormal nighttime outings are established.

[0105] If the trajectory quality is low or the drift is significant, trigger a second confirmation (increase the positioning frequency / request more modalities).

[0106] Step 4: Construct a risk assessment and rule triggering mechanism, including the following sub-steps,

[0107] Step 4.1: Establish the Person-Location-Event Rule Module. Based on the daily life routines, common travel routes, activity areas, and monitoring needs of the protected individuals, configure rules including electronic fence rules (such as areas like home, school, nursing home, park, and community), walking route rules (such as routes for going out and returning home), time rules (such as nighttime outing restrictions), and behavior rules (such as dwell time thresholds and fall recognition thresholds), and construct a "Person-Location-Event" association rule library.

[0108] Step 4.2: Implement dynamic user profiling and risk assessment. By collecting historical trajectories, activity frequencies, dwell patterns, and abnormal event records over a long period, dynamic user profiles are created for each individual, recording their typical travel times, common activity areas, and behavioral characteristics. During system operation, real-time location trajectories and behavior recognition results are dynamically compared with user profiles and rule bases. When an individual deviates from their usual route, crosses electronic fences, appears in abnormal areas during abnormal times, stays for significantly longer than normal, or experiences a high-confidence fall, a severity parameter S and a urgency parameter U are determined. These parameters are used to calculate a risk score (S×U) and are used as a key business indicator. The basic input for risk assessment is to determine the severity parameter S and urgency parameter U when an anomaly is triggered. Combined with the scenario coefficient Ctx, the basic risk score Rbase=S×U×Ctx is calculated. In the process of calculating business importance, the current communication link status provided in step 1, the location confidence provided in step 2, and the anomaly event confidence provided in step 3 are comprehensively considered. The confidence factor is weighted to obtain conf, thereby obtaining the business importance index I=Rbase×conf, which is used for alarm classification and communication scheduling, and thus realizes joint risk assessment and alarm classification based on multi-source information of "communication-navigation-remote control".

[0109] Determining the severity of the situation (S):

[0110] Fall / Suspected drowning / Prolonged stillness and abnormal posture: S_base=5;

[0111] Entering dangerous areas (such as waterways, construction sites, and motor vehicle lanes) or crossing into high-risk areas: S_base=4;

[0112] Abnormal area appears (but not high risk) / obvious route deviation: S_base=3;

[0113] Slight deviation / slight stagnation: S_base=2;

[0114] Low-risk anomalies (statistically significant deviations only): S_base=1;

[0115] To improve feasibility and individual adaptability, the terms "significant route deviation, slight deviation, abnormal area, and abnormal delay" are determined by statistical thresholds based on the user's dynamic profile, for example:

[0116] Normal route threshold: Based on the normal route, a route corridor buffer zone (e.g., 50m) is constructed. If the deviation exceeds the buffer zone and the duration exceeds a preset time threshold, it can be determined as a significant route deviation.

[0117] Abnormal region threshold: Common activity regions can be constructed from a set of frequently visited POIs and their covered areas. Regions that are not in the set of common activity regions or are far from common regions can be identified as abnormal regions.

[0118] Abnormal travel time threshold: Typical travel time can be obtained from historical travel time statistics. Travel during low-probability travel periods can be identified as abnormal travel periods.

[0119] Abnormal Detention Threshold: Establish a detention duration distribution for common locations; exceeding the threshold indicates abnormal detention. Urgency U Determination: Characterizes the timeframe required to address or handle abnormal events, reflecting time pressure, and is independent of the severity parameter S. The system pre-sets target handling time limits Tneed for different rules or abnormal events and maps them to U level values:

[0120] If Tneed ≤ 2 minutes, U = 5;

[0121] 2 < Tneed ≤ 10 minutes, U = 4;

[0122] 10 < Tneed ≤ 30 minutes, U = 3;

[0123] 30 < Tneed ≤ 120 minutes, U = 2;

[0124] 120 < Tneed, U = 1;

[0125] Tneed is determined by the statistical threshold of the user's dynamic portrait. For example, for the event of abnormal long stay, when the stay duration exceeds the time in the user portrait, it is considered a shorter Tneed. When the stay time far exceeds the time in the user portrait, the Tneed time is further shortened. For the events of route deviation or crossing the boundary, when detecting the risk evolution trend such as the continuous increase of the deviation distance / crossing depth, the Tneed time is further shortened to improve U.

[0126] The credibility factor Conf is obtained by weighted correction of the credibility according to the communication credibility Conf_com, the location credibility Conf_loc, and the abnormal event credibility Conf_evt. The calculation formula is as follows:

[0127] Conf = clip(w_e × Conf_evt + w_l × Conf_loc + w_c × Conf_com, 0, 1)

[0128] w_e + w_l + w_c = 1

[0129] Among them, clip is a truncation function, indicating to take the value within the range of 0 - 1; Conf_com: communication credibility (0~1), indicating the credibility of "whether it can be reliably delivered / availability" of the current link.

[0130] Conf_loc: location credibility (0~1), indicating the credibility of the current location / trajectory.

[0131] Conf_evt: abnormal event credibility (0~1), indicating the credibility of the recognition results of abnormal events such as "falling / crossing the boundary / staying".

[0132] The system statistically calculates the reliability of each information source based on the alarm confirmation results / acknowledgment results within the historical window, and adjusts the corresponding weights accordingly.

[0133] Step 4.3, risk triggering, high-frequency positioning, and alarm pushing. When the risk score reaches or exceeds the corresponding threshold, a risk event is triggered: on the one hand, request the positioning module corresponding to Step 2 to enable the high-frequency positioning mode to improve the spatio-temporal tracking accuracy of the target; on the other hand, call the multi-link communication strategy constructed in Step 1 to push the alarm information, associated trajectories, and recognition results to the family terminal, community or school management terminal, and emergency command terminal to achieve multi-role linkage response. For events with extremely high risk levels, it is possible to simultaneously notify Step 3 to increase the sensing sampling frequency or enable more camera perspectives in order to obtain richer remote sensing and video information to support subsequent disposal.

[0134] Step 5: Build a disposal closed-loop and adaptive optimization mechanism, including the following sub-steps

[0135] Step 5.1, Handling Process Receipt and Recording. Record the handling process after the alarm information is pushed, including whether the family members or management personnel received the alarm, whether they checked the location information, whether they arrived at the scene, the on-site verification results, and the final nature of the incident, etc., and store the above receipt information in association with data such as the communication link quality, positioning accuracy, and multimodal recognition results used in the alarm process.

[0136] Step 5.2, Adaptive Optimization of Model and Strategy. Based on the sliding time window statistical method, using a large number of event samples recorded in Step 5.1, the weight parameters of the link selection function in Step 1, the selection of scene-specific positioning strategies in Step 2, and the multimodal fusion weights and abnormal event judgment thresholds in Step 3 are adaptively updated. By reducing false alarms and missed alarms and optimizing strategies in specific scenarios, the system gradually adapts to the characteristics of different users and environments during continuous operation, improving the overall protection effect. Specifically, when communication delays are found in alarms leading to untimely handling, the link weights and message priority strategies in Step 1 are adjusted first; when large trajectory deviations are found in missing person searches, the noise model and remote sensing constraint weights in Step 2 are adjusted first; when a large proportion of false alarms are found to be caused by unstable behavior recognition, the multimodal fusion weights and thresholds in Step 3 are adjusted first, achieving coordinated adaptive optimization of the "communication-guidance-remote" modules.

[0137] Figure 2 This is the indoor positioning result in an embodiment of the present invention. The two tracks in the figure represent the movement trajectories of the guardian in the actual shopping mall environment. Through these tracks, the user's real-time information can be obtained.

[0138] Figure 3 This is the outdoor positioning result in an embodiment of the present invention. The red trajectory represents the user's actual movement trajectory. The user experienced switching between two positioning modes: on the road, in an open outdoor environment, and after entering the school, in an outdoor obstructed environment. Both types of results accurately reflect the user's actual movement, without any deviation or inaccuracy.

[0139] Figure 4 The visual perception result in this embodiment of the invention is shown in the box, which represents the detected dynamic target. By comparing the dynamic changes of the target with the predicted changes, the user's state can be analyzed and predicted more fully.

[0140] Figure 5 The image shows the IMU sensing results in this embodiment of the invention. The right side shows the changes of the IMU accelerometer and gyroscope on the x, y, and z axes. Based on these changes, the user's behavioral state can be analyzed, as shown in the left figure. The left figure shows the statistics of the user's state analyzed using the relevant state of the IMU, including but not limited to standing, walking, going up, going down, etc.

[0141] On the other hand, embodiments of the present invention also provide a security protection system based on communication-navigation-remote control fusion technology, including a processor and a memory. The memory is used to store program instructions, and the processor is used to call the program instructions in the memory to execute a security protection method based on communication-navigation-remote control fusion technology as described in the above technical solution.

[0142] The above embodiments are described only to clearly illustrate the basic technical solution of the present invention, but the present invention is not limited to the above embodiments; those skilled in the art can make various modifications or additions to the specific embodiments described, or use similar methods to replace them, but will not depart from the spirit of the present invention or exceed the scope defined by the appended claims; any simple modifications, equivalent changes and alterations made to the embodiments based on the technical essence of the present invention fall within the protection scope of the technical solution of the present invention. In addition to all personnel, the present invention is also applicable to the safety protection of the elderly and children.

Claims

1. A security protection method based on communication-navigation-remote control fusion technology, characterized in that, include: A multi-link communication strategy module is constructed to obtain the current communication link status, determine message priority and transmission mechanism by calculating the business importance index, classify alarm levels, and coordinate the indoor and outdoor integrated high-precision positioning mechanism module and the multimodal intelligent sensing and anomaly identification mechanism module. The business importance index is calculated using the following formula: I=S×U×Ctx×Conf Where S represents severity, U represents urgency, Conf represents credibility factor, and Ctx represents scenario coefficient; Alarm levels are categorized based on the importance of the business: Level I0: I≥40, Level I1: 25≤I<40, Level I2: 10≤I<25, Level I3: I<10. Alternatively, alarm levels can be classified based on severity and urgency. If S≥5 and U≥4, it is directly classified as level I0. If S≥4 and U≥4, it must be classified as at least level I1; Construct an integrated indoor and outdoor high-precision positioning mechanism module to obtain positioning results and location reliability; Construct a multimodal intelligent perception and anomaly recognition mechanism module to identify abnormal events and obtain the confidence level of abnormal events; Establish a rule base for associating personnel, location, and time. Dynamically compare the personnel to be protected with the rule base to calculate the corresponding risk score. Combine the current communication link status, current location confidence, and abnormal event confidence to weight and correct the credibility factor in the business importance index. When the risk score reaches or exceeds the corresponding threshold, trigger the risk event and send an alarm. By collecting historical trajectories, activity frequencies, stay patterns, and abnormal event records, a dynamic user profile is formed for each individual. This profile records typical travel times, common activity areas, and behavioral characteristics. Real-time location trajectories and behavioral recognition results are dynamically compared with user profiles and rule bases. When an individual is found to deviate from the usual route, cross electronic fences, appear in abnormal areas during abnormal times, stay for significantly longer than normal, or experience a high-confidence fall, the values ​​of severity S and urgency U are determined. The risk score Rbase = S × U × Ctx is calculated by combining the scenario coefficient Ctx. The confidence factor Conf is obtained by weighting and adjusting the confidence level based on the communication confidence level Conf_com, the location confidence level Conf_loc, and the abnormal event confidence level Conf_evt. The calculation formula is as follows: Conf=clip(w_e×Conf_evt+w_l×Conf_loc+w_c×Conf_com,0,1) w_e+w_l+w_c=1 Where w_e, w_l, and w_c are weights, and clip is the cutoff function, which represents a value in the range of 0-1; Based on alarm notifications, the multi-link communication strategy module, the indoor and outdoor integrated high-precision positioning mechanism module, and the multimodal intelligent sensing and anomaly recognition mechanism module were optimized.

2. A security protection method based on communication-navigation-remote control fusion technology as described in claim 1, characterized in that: Building a multi-link communication strategy module includes the following sub-steps: Step 1.1: Establish a multi-link communication set. This involves constructing a set of multiple communication links, including cellular communication links, satellite communication links or short message communication links, low-power wide-area communication links, and emergency broadcast channels, to form a communication foundation that is adaptable to different scenarios and network conditions. Step 1.2: Construct a link selection function and implement adaptive primary / backup switching. Estimate the link quality, time delay, communication cost, and power consumption of various communication links in real time, and establish a comprehensive link score. Where Q is the link quality, L is the time delay, C is the communication cost, P is the power consumption, and w Q w represents the link quality weight. L For time delay weights, w C As a weight for communication costs, w P Power consumption weight; Step 1.3, establish message priority and reliable transmission mechanism: classify messages to be sent according to the business importance index, including at least SOS distress messages, abnormal alarm messages and regular business messages, and manage queues in the priority order of SOS > abnormal alarm messages > regular business messages; for SOS distress messages and critical alarm messages, adopt the fragmentation and retransmission window mechanism, and perform concurrent redundant transmission on multiple links; Step 1.4: When a weak network or obstructed environment is detected, causing a decrease in communication quality, a data downsampling and compression strategy is triggered. The data sampling and uploading frequency is dynamically adjusted based on the positioning results and the confidence level of abnormal events, so as to achieve coordinated control between the indoor and outdoor integrated high-precision positioning mechanism module and the multimodal intelligent sensing and anomaly recognition mechanism module.

3. A security protection method based on communication-navigation-remote control fusion technology as described in claim 1, characterized in that: The construction of an integrated indoor and outdoor high-precision positioning mechanism module includes the following sub-steps: Step 2.1, scene recognition, including indoor scenes, open outdoor scenes, and occluded outdoor scenes; Step 2.2, Scene-specific positioning strategy selection and fusion positioning; Based on the scene identified in Step 2.1, determine the observation source and external constraints used for fusion positioning: In indoor scenes, select the combination of inertial navigation data and geomagnetic data to achieve indoor positioning; In outdoor open scenes, select GNSS positioning results and fuse inertial navigation data to achieve continuous high-precision positioning; In outdoor occluded scenes, select inertial navigation calculation results as the main method, and introduce a backbone map of the passable area constructed offline by remote sensing imagery / map as an external constraint to restrict the trajectory to the passable area; After selecting the observation source and constraints, a unified fusion positioning process is used for estimation: a state space model and an observation model are established, and a noise model and covariance matrix are set for inertial navigation drift, GNSS observation error, geomagnetic matching error and map constraint error; the noise model parameters and constraint weights are adaptively adjusted according to the scene recognition results, satellite geometric accuracy factor and matching residual, so as to correct unreasonable position solutions and output positioning confidence. Step 2.3: Output the positioning results and perform trajectory quality assessment; based on the positioning results, output positioning information including position coordinates, velocity information and corresponding confidence scores, and calculate trajectory continuity indicators, including positioning point missing rate, number of breakpoints, maximum breakpoint duration, and jump distance between adjacent points. Perform segmented quality assessment and mark abnormal segments on the trajectory, and perform trajectory evidence storage and playback.

4. A security protection method based on communication-navigation-remote control fusion technology as described in claim 3, characterized in that: The indoor and outdoor integrated high-precision positioning mechanism module also outputs the confidence level of the positioning results to the multimodal intelligent perception and anomaly recognition mechanism module in real time, which is used to weight the location confidence level of the abnormal event recognition results; the trajectory quality assessment results are fed back to the multi-link communication strategy module, which is used to advance the priority of alarm links or enable backup links when the positioning quality deteriorates.

5. A security protection method based on communication-navigation-remote control fusion technology as described in claim 1, characterized in that: The construction of a multimodal intelligent perception and anomaly detection mechanism module includes the following sub-steps: Step 3.1, Multimodal data acquisition and preprocessing: Inertial navigation data is acquired through inertial sensors, and environmental measurement data is acquired through authorized public cameras, home cameras, environmental sound sensors, and temperature sensors. The data of each type are filtered and denoised, time-synchronized, feature-extracted and normalized to provide a unified spatiotemporal reference and usable features for subsequent identification. Step 3.2, Modal Recognizer Construction: Based on inertial navigation data, gait analysis, peak acceleration detection, and posture change modeling algorithms are used to identify abnormal behavior; based on image data, target detection, human posture estimation, and behavior recognition algorithms are used to determine the duration of abnormal behavior; based on sound and environmental data, model recognition methods are used to detect breathing sounds, abnormal noise, and high-temperature environmental risks, and preliminary recognition results and corresponding confidence levels are obtained for each modality. Step 3.3, Multimodal fusion and abnormal event judgment; The single-modal recognition results obtained in step 3.2 are input into the multimodal fusion module. A judgment method combining modal deep learning network and weighted confidence fusion model is adopted to perform joint analysis on the same target or the same time in the time series, generate a comprehensive confidence score and classification result for abnormal behavior protection events, and output spatiotemporal labels associated with abnormality type, severity and location information.

6. A security protection method based on communication-navigation-remote control fusion technology as described in claim 1, characterized in that: Triggering risk events and sending alarms includes: requesting the indoor and outdoor integrated high-precision positioning mechanism module to activate the high-frequency positioning mode to improve the spatiotemporal tracking accuracy of the target; calling the constructed multi-link communication strategy module to push alarm information, associated trajectories, and identification results to family terminal, community or school management terminal, and emergency command terminal to achieve multi-role linkage response; for extremely high-risk abnormal events, simultaneously notifying the multimodal intelligent perception and anomaly recognition mechanism module to increase the perception sampling frequency or enable more camera perspectives.

7. A security protection method based on communication-navigation-remote control fusion technology as described in claim 1, characterized in that: Optimizations based on alarm notifications include: Step 5.1: Receipt and record the handling process after the alarm information is pushed, and store the receipt information in association with the communication link quality, positioning accuracy and multimodal recognition results used in the current alarm process; Step 5.2: Based on the event samples recorded in Step 5.1, adaptively update the weight parameters of the link selection function in the multi-link communication strategy module, the positioning strategy selection in the indoor and outdoor integrated high-precision positioning mechanism module, and the multi-modal fusion weights and the judgment thresholds of each abnormal event in the multi-modal intelligent perception and anomaly recognition mechanism module.

8. A security protection system based on communication-navigation-remote control fusion technology, characterized in that: It includes a processor and a memory, the memory being used to store program instructions, and the processor being used to call the program instructions in the memory to execute a security protection method based on communication-navigation-remote control technology as described in any one of claims 1-7.

Images