A method and system for multi-target radio monitoring

Abstract

This invention relates to the field of radio monitoring technology, specifically to a multi-target radio monitoring method and system. The method includes: constructing a unified time-frequency energy expression through multi-node time-consistent fusion; reconstructing the target trajectory using dynamic segmentation and bicontinuous constraints driven by the energy change rate; constructing behavioral primitive vectors; jointly modeling the behavior of multiple targets, extracting stable cooperative relationships and forming a weighted cooperative structure graph; using structural entropy to characterize the overall electromagnetic environment organization state; driving adaptive scheduling of monitoring resources with the stability of the cooperative structure as the core, dynamically generating sampling frequency, monitoring frequency band, and time resolution and issuing them for execution; using the historical stable cooperative structure as a baseline, calculating standardized deviations at the edge level, node level, and global level, identifying abnormal targets and types, and selectively triggering dynamic corrections of the cooperative model and monitoring strategy. This invention forms a closed loop of perception, analysis, scheduling, and optimization, achieving long-term refined monitoring and efficient resource utilization.

Description

Technical Field

[0001] This invention relates to the field of radio monitoring technology, and specifically to a multi-target radio monitoring method and system. Background Technology

[0002] With the rapid development and widespread application of radio communication technology, the number of various wireless terminals and systems has surged, and the electromagnetic environment has become a complex system characterized by dense targets, diverse behaviors, and overlapping spectra. The coexistence and interaction of multiple targets in time, frequency, and spatial dimensions have transformed the electromagnetic space from a relatively simple set of signals into a dynamic system with complex relationships.

[0003] Chinese invention patent CN118713735B discloses a multi-target radio monitoring method and system, comprising: generating a monitoring area data container on a cloud data server based on the location information of the monitoring area; dividing the monitoring area into regions based on the data collection range of the UAV, obtaining a sequence of monitoring sub-regions, and sending it to the monitoring area data container; obtaining the initial number of task UAVs based on the distance between the UAV control center and the monitoring area; generating monitoring sub-tasks corresponding to the initial number of task UAVs on the cloud data server based on the monitoring area sequence and the initial number of task UAVs, and obtaining a monitoring sub-task sequence based on the monitoring sub-region sequence; the task UAVs executing the monitoring sub-task sequence and UAV takeoff time until the monitoring data collection of the monitoring area is completed, thus completing the command and dispatch of the UAVs.

[0004] However, accurately extracting the behavioral profile of each radio target from massive, multi-dimensional continuous monitoring data, and further revealing the intrinsic relationships among multiple targets in terms of spectrum usage, temporal rhythms, and power changes, has become an important direction for the development of current radio monitoring technology. At the same time, an adaptive match needs to be achieved between the limited monitoring resources and the dynamic changes in the electromagnetic environment to improve monitoring efficiency and response capabilities. Therefore, constructing a method that can deeply perceive the behavior of multiple targets, uncover stable cooperative relationships, and dynamically adjust monitoring strategies according to environmental evolution is of great significance for promoting the leap of electromagnetic space monitoring technology from data acquisition to intelligent cognition and supporting spectrum management and frequency usage order supervision. Summary of the Invention

[0005] The purpose of this invention is to address the problems existing in the background art by proposing a multi-target radio monitoring method and system.

[0006] The technical solution of this invention: a multi-target radio monitoring method, comprising the following specific implementation steps:

[0007] S1. A unified time-frequency energy expression is constructed by multi-node time consistency fusion. The target trajectory is reconstructed by using the energy change rate to drive dynamic segmentation and frequency-time dual continuity constraints, and behavioral primitive vectors are constructed.

[0008] S2. Jointly model multi-objective behavior, extract stable cooperative relationships and form a weighted cooperative structure diagram, and use structural entropy to characterize the overall electromagnetic environment organization state.

[0009] S3. Driven by the stability of the collaborative structure, the monitoring resources are adaptively scheduled, and the sampling frequency, monitoring frequency band and time resolution are dynamically generated and executed.

[0010] S4. Using the historically stable collaborative structure as a baseline, calculate the standardized deviation at the edge level, node level, and global level to identify abnormal targets and types.

[0011] S5. In response to the anomaly identification results, selectively trigger the dynamic correction of the collaborative model and monitoring strategy to form a closed loop of perception, analysis, scheduling and optimization.

[0012] Preferably, step S1 specifically includes:

[0013] The original observation signals collected by several broadband sensing nodes in the monitoring area are time-aligned by introducing a time offset compensation model to obtain time-aligned observation signals.

[0014] By combining the signal-to-noise ratio of each node and the weighting coefficients set for spatial coverage reliability, the time-aligned observation signals are subjected to short-time Fourier transform and weighted fusion to form a continuous and stable joint time-frequency energy observation matrix.

[0015] Based on the joint time-frequency energy observation matrix, the average energy of each frequency band within a unit time window is calculated using a sliding time window, and an adaptive change threshold is constructed by calculating the energy change rate and historical fluctuation variance.

[0016] When the rate of energy change exceeds the adaptive change threshold, it is determined that a new behavior segment has begun, thus realizing the dynamic division of behavior segments;

[0017] For the energy segments obtained by dynamic division, frequency continuity constraint functions and time continuity constraint functions are constructed respectively, and the comprehensive correlation strength between segments is calculated accordingly.

[0018] Two behavior segments with a comprehensive correlation strength greater than a set threshold are assigned to the same target trajectory and continuously spliced ​​together to form a continuous energy evolution path that is independent of each target.

[0019] After obtaining the target trajectory, frequency band usage density, power fluctuation entropy, time occupancy rhythm, average transmit power, and power change variance are extracted for each target to construct a unified behavior primitive vector.

[0020] Preferably, step S2 specifically includes:

[0021] Normalization transformation is performed on the behavior primitive vector of each target, including standardizing the frequency band usage density according to the start and end frequencies of the monitoring frequency band, centering the average transmit power according to the mean power and power standard deviation of all targets, and performing periodic scale mapping and peak normalization on the time occupancy rhythm.

[0022] After standardization, a cooperative coupling strength function between two targets is defined. By calculating the frequency coupling strength, time coupling strength, and energy coupling strength, and then weighting and fusing the three, the comprehensive coupling strength of the target pair within the sliding time window is obtained.

[0023] By using a sliding time window, the mean of the cooperative strength and the cooperative volatility of each target pair within the window are statistically analyzed, and a stable cooperative criterion is constructed. When the stable cooperative index exceeds a set threshold, it is determined that there is a stable cooperative relationship between the target pairs.

[0024] All target pairs that meet the stable cooperation conditions are constructed into a weighted cooperative network graph, where nodes represent radio targets, edge weights represent stable cooperation indices, and network structure entropy is calculated to characterize the complexity of the overall cooperative network.

[0025] Preferably, step S3 specifically includes:

[0026] Construct a stable collaborative weight matrix among nodes, and statistically analyze the weighted average of collaborative strength and collaborative fluctuation of each node pair within a continuous time window, thereby constructing an overall collaborative structure stability index.

[0027] For each target node, a comprehensive activity index is calculated based on its cooperative connection strength, frequency band occupancy ratio, and power fluctuation entropy.

[0028] Based on the overall collaborative structure stability index and the comprehensive activity index of the nodes, the sampling frequency of each target is dynamically adjusted, the high-occupancy frequency band of the nodes is selected as the key monitoring frequency band, and the sliding time window length is adjusted to match the current time resolution requirements.

[0029] The generated monitoring strategy, which includes sampling frequency, key monitoring frequency bands, and sliding time window length, is sent to each sensing node for execution in real time. After collecting new data, the target behavior primitive vector and collaborative structure are recalculated, and the stability index and node weights are continuously updated.

[0030] Preferably, step S4 specifically includes:

[0031] When the cooperative structure is within the stable range determined by the stability index, the cooperative strength of each target pair in multiple time periods is statistically averaged and its standard deviation is calculated to construct a historical reference model that includes historical stable cooperative baseline and historical cooperative fluctuation baseline.

[0032] Within the current monitoring period, the latest behavioral profile is obtained and the current collaboration matrix is ​​reconstructed based on the updated sampling strategy.

[0033] The standardization difference between the current collaboration strength in the collaboration matrix and the historical stable collaboration baseline is calculated to obtain the edge-level deviation of each target pair.

[0034] The node-level deviation index of each target node is formed by aggregating all edge-level deviations, and the global structure deviation index is formed by aggregating all edge-level deviations.

[0035] Based on the distribution characteristics of node-level deviation index, edge-level deviation degree, and global structure deviation index, anomalies are classified as single-node deviation type, collaborative fracture type, or group linkage type, and a comprehensive anomaly level index is constructed to classify the degree of impact of anomalies.

[0036] Preferably, step S5 specifically includes:

[0037] After confirming the anomaly, the anomaly impact propagation coefficient of the abnormal node in the cooperative network is calculated. The anomaly impact propagation coefficient is used to characterize the degree of influence of the abnormal node on all its neighboring nodes.

[0038] If the proportion of the anomaly propagation coefficient in most neighboring nodes is lower than the set threshold, it is determined to be a local fluctuation anomaly, and only the monitoring accuracy of that node and its neighboring nodes is improved.

[0039] If the anomaly propagation coefficient accounts for a high proportion of several neighboring nodes, it is determined to be a structural reorganization anomaly, triggering the updating of the collaborative baseline, adjusting the weight of nodes in resource scheduling, increasing the sampling density of relevant frequency bands, and storing the current anomaly collaborative mode in the anomaly mode library.

[0040] Preferably, step S5 further includes:

[0041] The identified abnormal nodes and abnormal edge information are combined with the node's comprehensive activity index. By integrating the historical stable collaborative baseline and the latest abnormal collaborative deviation, and introducing the neighbor anomaly level, the collaborative correction weight of each node is calculated.

[0042] The collaborative correction weight is used as a priority factor in subsequent monitoring resource scheduling and baseline correction to achieve dynamic feedback and integration of information on the impact of anomalies.

[0043] Preferably, step S5 further includes:

[0044] The historical stable collaborative baseline is updated by weighted fusion using the calculated node collaborative correction weights.

[0045] During the update process, the latest collaborative behavior of abnormal nodes is introduced into the new collaborative baseline with corresponding correction weights, while the historical baseline information of stable nodes is kept undisturbed, thereby achieving controllable dynamic correction and stable evolution of the multi-objective collaborative structure.

[0046] Preferably, step S5 further includes:

[0047] Based on the updated collaborative baseline and the node's comprehensive activity index, a short-time autoregressive prediction model is used to predict the collaborative strength of nodes in future monitoring periods.

[0048] The sampling frequency and monitoring frequency band selection of nodes are dynamically adjusted based on the predicted coordination strength, so that the monitoring resources can be proactively adjusted to follow the evolution trend of the target coordination structure, and the adjusted strategy is used as the basis for the next monitoring cycle.

[0049] The technical solution of the present invention: a multi-target radio monitoring system, used to execute the above-mentioned multi-target radio monitoring method, comprising:

[0050] The multi-source sensing and behavior primitive construction module is used to continuously acquire radio signals from multiple nodes in a time sequence, and transform the raw acquired data into behavior primitive vectors for each target through signal preprocessing and feature decomposition.

[0051] The collaborative structure analysis module is used to calculate the collaborative strength between objectives by taking behavioral primitive vectors as input, screen stable collaborative relationships to construct a multi-objective weighted collaborative structure diagram, and generate collaborative stability indicators.

[0052] The adaptive monitoring strategy generation module is used to generate an adaptive sampling strategy for each target based on the collaborative stability index and node activity, and then send it to the collection terminal for execution.

[0053] The collaborative deviation and anomaly identification module is used to compare the real-time collected target behavior data with the historical collaborative baseline, calculate the multi-scale collaborative deviation, and perform anomaly classification and grading.

[0054] The Cooperative Baseline Correction and Long-Term Evolution module is used to dynamically update the cooperative baseline based on the anomaly identification results, predict the future evolution trend of the cooperative structure, and optimize the sampling strategy for the next cycle.

[0055] Compared with the prior art, the above-mentioned technical solution of the present invention has the following beneficial technical effects:

[0056] This invention designs a multi-target radio monitoring method and system. By constructing a multi-level abstract model from raw signals to behavioral primitives and then to cooperative structures, it achieves deep perception and accurate analysis of complex electromagnetic environments, significantly improving the intelligence level of radio monitoring in multi-target scenarios. First, multi-source sensing signals are transformed into unified behavioral primitive vectors, completing the leap from raw data to structured information and providing a standardized foundation for target behavior characterization. Based on this, by constructing a three-dimensional coupling strength function of frequency, time, and energy, stable cooperative relationships among multiple targets are extracted and a weighted cooperative structure diagram is formed. The structural entropy is used to quantify the overall electromagnetic environment's organizational state, transforming the originally discrete signal set into a measurable and evolvable cooperative structure model, providing a new analytical dimension for global situational awareness. An adaptive monitoring mechanism driven by the stability of the cooperative structure realizes the monitoring of resources. Dynamic matching with the electromagnetic environment significantly improves resource utilization efficiency while ensuring precise observation of key targets, enabling the system to autonomously adjust sampling strategies according to environmental changes. A multi-scale deviation analysis method based on historical collaborative baselines identifies anomalous behaviors layer by layer from the edge level, node level to the global level, and accurately classifies and grades them according to their impact range and propagation characteristics, effectively distinguishing between occasional interference and structural evolution, avoiding false alarms and missed alarms common in traditional threshold detection. A closed-loop correction mechanism after anomaly triggering allows the collaborative model to continuously optimize along with environmental evolution, and the knowledge accumulation of the anomaly pattern library provides support for long-term stable operation. This invention forms a complete closed loop of perception, analysis, scheduling, identification, and correction, possessing adaptive, self-learning, and self-optimization capabilities, enabling long-term precise monitoring in complex and variable electromagnetic environments, and providing reliable technical support for applications such as spectrum management, interference investigation, and target tracking. Attached Figure Description

[0057] Figure 1 This is a flowchart of a multi-target radio monitoring method proposed in this invention;

[0058] Figure 2 This is a system architecture diagram of a multi-target radio monitoring system proposed in this invention. Detailed Implementation

[0059] Example 1, as Figure 1 As shown, the multi-target radio monitoring method proposed in this invention includes the following specific implementation steps:

[0060] S1. This involves constructing target behavior primitives based on multi-source continuous sensing. A unified time-frequency energy expression is formed through multi-node time-consistent fusion. On this basis, a dynamic segmentation mechanism driven by the energy change rate is used to identify behavior boundaries. Then, target-level trajectory reconstruction is completed through dual continuity constraints of frequency and time. Finally, the continuous trajectory is transformed into a behavior primitive vector containing frequency band usage density, power fluctuation characteristics, and time rhythm indicators, achieving a structured mapping from the original electromagnetic signal to a stable behavior expression layer. The specific implementation process is as follows:

[0061] S11. By constructing a multi-source sensing node time consistency mechanism, the broadband signals collected by each node are uniformly time-aligned, and weighted fusion is performed based on the node signal-to-noise ratio and coverage reliability to form a continuous and stable joint time-frequency energy observation matrix, specifically:

[0062] M broadband sensing nodes are deployed within the monitoring area. Each node acquires broadband electromagnetic signals continuously over a period of time. The raw observation signal of the m-th node is represented as... ;

[0063] Due to the distributed deployment across multiple nodes, time drift exists. To avoid mismatch in target behavior, a time offset compensation model is introduced: ;

[0064] Then, a joint observation matrix is ​​constructed: ;

[0065] Where m represents the sensing node number, m=1,2,…,M; t represents the continuous time variable; This represents the original observation signal of the m-th node, that is, the complex baseband signal received by the m-th node; This represents the time calibration offset of the m-th node; This represents the time-aligned observation signal; Represents frequency variables; This represents the time-frequency energy distribution after fusion; Represents the short-time Fourier transform operator; This represents the set node weight coefficient, which is related to the node signal-to-noise ratio and spatial coverage reliability;

[0066] It should be noted that the time calibration offset of the m-th node The difference between the built-in high-stability clock of each node and the master clock is estimated and corrected by periodic synchronization pulses.

[0067] S12. Based on the fused time-frequency energy matrix, the average energy of the frequency band is calculated using a sliding time window. An adaptive threshold is constructed using the energy change rate and historical fluctuation variance to achieve dynamic division of the behavior segment, so that the behavior boundary is automatically determined by the signal evolution trend. Specifically:

[0068] For joint observation matrix Perform energy flow decomposition and define a unit time window. Internal frequency band Average energy: ;

[0069] Introducing the energy change rate function: ;

[0070] ;

[0071] when Exceeding the adaptive change threshold When this occurs, it is determined to be the start of a new action segment;

[0072] in, Indicates the starting point of the k-th time window; This represents the k-th frequency segment; This represents the average energy of the k-th frequency segment within the k-th time window; Indicates the rate of energy change; This indicates the set sensitivity adjustment coefficient; This indicates the historical energy standard deviation for that frequency band;

[0073] S13. Reconstruct the target attribution of the dynamically divided energy segments, establish a dual-constraint model of frequency continuity and time continuity, calculate the comprehensive correlation strength between segments, and splice the energy segments that meet the continuity characteristics into the same target trajectory to form a multi-target parallel and independent continuous energy evolution path, specifically as follows:

[0074] After completing the dynamic segmentation, the discrete energy segments are assigned to specific radio targets, and a frequency continuity constraint function is constructed. : ;

[0075] Constructing time continuity constraint functions : ;

[0076] Define the overall correlation strength: ;

[0077] When the overall correlation strength When the value exceeds a set threshold, the two behavior segments are assigned to the same target trajectory, enabling continuous splicing of the target trajectories and forming the energy evolution path of each target: ;

[0078] in, and These represent the center frequencies of the two energy bands, respectively. and These represent the center time of the two energy segments respectively; Parameters representing frequency continuity scale; Parameters representing the scale of time continuity; Let represent the set of trajectories for the i-th target;

[0079] S14. After constructing the target trajectory, extract features such as frequency band density, power distribution entropy, time occupancy rhythm, and power fluctuation statistics for each target to construct a unified behavioral primitive vector. This achieves a parameterized transformation from continuous trajectory to structured behavioral expression, providing standardized input for subsequent collaborative structure analysis. Specifically:

[0080] After obtaining the target trajectory, construct a behavior primitive vector for each target:

[0081] Frequency bands are defined using density functions: ;

[0082] Define power fluctuation entropy: ;

[0083] Define time-occupied rhythm indicators: ;

[0084] Ultimately, the target behavior primitive vector is formed. : ;

[0085] in, This represents the proportion of the i-th radio target's usage in frequency f; Indicates the total duration of target observation; This indicates an indicator function, which takes a value of 1 when the i-th radio target occupies frequency f at time t, and a value of 0 otherwise. This represents the probability that the target power falls into the k-th interval; Indicates the dispersion of power variation; The variable represents the time interval; N represents the total number of target launch events. This represents a decision function. It takes the value 1 when the expression within the parentheses is close to 0 (the error is less than the set tolerance), and 0 otherwise. Indicates the i-th radio target with periodicity. The probability of repeated firing; Indicates average transmit power; This represents the variance of power variation.

[0086] S2. Based on the radio target behavior primitive vectors output in step S1, the behavior of multiple targets is jointly modeled in a unified scale space. By constructing a three-dimensional coupling strength function of frequency, time, and energy, stable cooperative relationships are extracted and a weighted cooperative structure diagram is formed. At the same time, structural entropy is used to characterize the overall electromagnetic environment's organizational state, realizing the leap from single-target behavior representation to multi-target cooperative structure modeling. The specific implementation process is as follows:

[0087] S21. Perform scale unification and physical normalization on the target behavior primitive vectors generated in step S1, including frequency band usage density normalization, power feature centering, and time rhythm function peak normalization, so that targets with different transmission power levels and frequency band ranges are in a comparable feature space. Specifically:

[0088] Obtain the primitive vectors of each target behavior constructed in step S1. Introduce normalization transformation:

[0089] ; ;

[0090] Periodic scaling via time rhythm function: ;

[0091] in, This represents the normalized occupancy percentage of the i-th radio target at frequency f; and These represent the start and end frequencies of the corresponding monitoring frequency bands; This represents the normalized average transmit power; This represents the average power of all targets in the current monitoring area. Indicates the standard deviation of power; This indicates the normalized time occupancy rhythm;

[0092] S22. Based on standardized features, a multi-dimensional coupling strength function is constructed, which weights and fuses frequency sharing, time rhythm synchronization, and power proximity to form a comprehensive coordination strength index between target pairs. This enables a multi-dimensional characterization of resource sharing, complementarity, or synchronization behavior. Specifically:

[0093] After standardization, define the cooperative coupling strength function between the two objectives i and j:

[0094] ;

[0095] Define frequency coupling strength ;

[0096] Define temporal coupling strength ;

[0097] Define energy coupling strength ;

[0098] in, Indicates the length of the sliding time window for the analysis; and These represent the start and end times of the current sliding window, respectively. , and These represent the weighting coefficients for the frequency, time, and energy dimensions, respectively. This represents a small constant to prevent the denominator from being zero; It represents the combined coupling strength between target i and target j at time t, and is used to measure the degree of coordination between the two targets in terms of frequency, time and power;

[0099] S23. By statistically analyzing the mean and volatility of synergy strength through a sliding time window, a stable synergy criterion is constructed. Synergy relationships with long durations and small fluctuations are selected to distinguish between occasional overlaps and genuine structural synergies, ensuring the continuity and reliability of relationships entering the structural modeling stage. Specifically:

[0100] Define the mean collaborative strength within a sliding time window W: ;

[0101] Define co-variance: ;

[0102] Constructing stable collaborative criteria: ;

[0103] When stable coordination indicators When the threshold is exceeded, it is determined to be a stable collaborative relationship;

[0104] in, Indicates the stability adjustment coefficient; Indicates a stable and coordinated indicator;

[0105] S24. The selected stable cooperative relationships are constructed into a weighted cooperative network graph, where nodes represent radio targets, edge weights represent stable cooperative strength, and the network structure entropy is calculated to characterize the overall organizational complexity. This transforms the electromagnetic environment from a discrete signal set into an evolvable cooperative structure model, providing a global perspective for subsequent monitoring resource scheduling. Specifically:

[0106] Construct a weighted graph of all objective pairs that satisfy the stable cooperation condition: ;

[0107] Define network structure entropy: ; ;

[0108] in, This represents the collaborative structure diagram at time t; V represents the target node set. Represents the set of collaborative edges; Represents the set of edge weights, corresponding to the stability and coordination index. ; It represents the cooperative structure entropy, which is used to measure the complexity of cooperative networks; Let i represent the set of neighbors of node i; This represents the relative activity of node i in the collaborative network, that is, the proportion of the node's contribution to the overall collaborative structure.

[0109] S3. Based on the multi-objective collaborative structure diagram and stable collaborative weight matrix constructed in step S2, an adaptive scheduling mechanism for monitoring resources driven by the stability of the collaborative structure is established. By quantitatively analyzing the stability of the overall collaborative network, the activity level of nodes, and the evolution trend of the structure, parameters such as sampling frequency, monitoring frequency band range, and time resolution are dynamically generated, and the scheduling strategy is distributed to each monitoring node in real time for execution. At the same time, closed-loop optimization is achieved by rolling updates to the collaborative structure. The specific implementation process is as follows:

[0110] S31. By statistically analyzing the mean and volatility of each stable cooperative edge in the cooperative structure diagram, an overall cooperative stability index is constructed to characterize whether the current electromagnetic environment is in a stable phase or a structural evolution phase. Specifically:

[0111] Define the stable collaborative weight matrix between nodes and continuous time window Weighted average and volatility indicators within:

[0112] ;

[0113] ;

[0114] Define the collaborative stability index: ;

[0115] in, Indicates the average cooperation strength between node pairs; This indicates coordinated fluctuations between nodes; Indicates the overall stability of the collaborative structure; Indicates the stability adjustment coefficient; Indicates the number of collaborative edges;

[0116] S32. Based on the cooperative connection strength, frequency band occupancy ratio, and power dynamic characteristics of each node in the cooperative network, a comprehensive activity index is constructed to provide a quantitative basis for the differentiated allocation of sampling resources, so that monitoring resources are preferentially concentrated on key structural nodes, specifically:

[0117] For each node i, calculate the activity metric. :

[0118] ;

[0119] in, Represents the entropy of power fluctuations; , and These represent the weighting coefficients; This represents a comprehensive activity level indicator;

[0120] S33. Generate an adaptive monitoring strategy based on the overall collaborative stability and node activity weights, including dynamically adjusting the sampling frequency, key monitoring frequency bands, and time window length. When the collaborative structure tends to stabilize, appropriately reduce resource consumption; when structural fluctuations increase, improve time and frequency resolution to achieve a dynamic balance between monitoring accuracy and resource efficiency. Specifically:

[0121] Based on the stability of the cooperative structure and activity metrics ,include:

[0122] Adjust sampling density ;

[0123] Adjust the monitoring frequency band and select the high-occupancy frequency band of node i. : ;

[0124] Adjust the sliding time window length to: ;

[0125] in, Indicates the current frequency of node i; Indicates the basic sampling frequency; Indicates the adjustment coefficient; Indicates the base time window; Indicates the sensitivity coefficient; This indicates the adjusted sliding time window;

[0126] S34. The generated monitoring strategy is distributed to each sensing node in real time for execution. After new data is collected, the behavioral primitives and collaborative structure are recalculated, and the stability indicators and node weights are continuously updated to form a closed-loop adaptive mechanism of sensing, analysis, scheduling, and re-sensing. This enables the system to operate for a long time and continuously self-optimize as the electromagnetic environment changes. Specifically:

[0127] The generated strategy is distributed to each node for execution in real time, and the nodes adjust their sampling rate and frequency band according to their own weights; at the same time, the collaborative structure is updated through a rolling window: collecting the latest sampling data and updating the target behavior primitive vector. Recalculation of synergy strength Update stability metrics Adjust the strategy to generate parameters.

[0128] S4. Based on the dynamic scheduling of the collaborative structure formed in step S3, and using the historical stable collaborative structure as a reference baseline, a standardized deviation index for collaborative relationships is constructed. A structural comparative analysis of the current multi-objective behavior state is conducted at three scales: edge level, node level, and global level. Abnormal targets or combinations deviating from existing collaborative constraints are identified. The type and level of the abnormality are determined based on its impact range and propagation characteristics. This selectively triggers dynamic adjustments to the collaborative model and monitoring strategy, achieving anomaly identification and closed-loop optimization oriented towards structural evolution. The specific implementation process is as follows:

[0129] S41. When the cooperative structure is in a stable range, statistically average the cooperative strength over multiple time periods and calculate the fluctuation range. Construct a historical reference model including the cooperative baseline value and standard deviation parameters to characterize the stable cooperative relationship between multiple targets under normal electromagnetic conditions, forming a structural reference framework for subsequent anomaly detection. Specifically:

[0130] Defined in the stable phase (i.e.) Average collaborative weight within: ;

[0131] Simultaneously store the corresponding coordinated fluctuation baseline: ;

[0132] in, This represents the historical stable cooperation baseline, i.e., the average cooperation strength of targets i and j under normal conditions; This represents the historical co-variance baseline, i.e., within the stable range. Standard deviation; This represents the stability threshold.

[0133] S42. Based on the reconstructed collaboration matrix of the current monitoring cycle, standardized difference calculations are performed with the historical collaboration baseline to obtain the edge-level deviation. These deviations are then further aggregated to form node-level and global-level deviation indicators. This quantifies the change in the current behavioral structure relative to the historical stable structure from a multi-scale perspective, providing a quantitative basis for anomaly detection. Specifically:

[0134] Within the current monitoring period, the latest behavioral profile is obtained and the collaborative structure is reconstructed according to the sampling strategy updated in step S3, resulting in the current collaborative matrix. ;

[0135] Define one-sided deviation: ;

[0136] Further define the node-level deviation index: ;

[0137] Simultaneously define the global structure deviation index: ;

[0138] in, Indicates the strength of collaboration within the current monitoring period; This indicates the degree of cooperative deviation of the target from i and j; This represents the overall collaborative deviation index of node i; Indicates the global coordination deviation index;

[0139] S43. Based on the distribution characteristics of multi-scale deviation indicators, anomalies are classified into types, including single-node deviation, coordinated breakage, and group linkage. A comprehensive anomaly level index is constructed to classify the degree of anomaly impact, and differentiated monitoring and response strategies are matched for different anomaly scenarios. Specifically:

[0140] Anomalies are classified based on their deviation structure characteristics:

[0141] If a node However, if the deviation of its neighbors is not significant, it is judged as a single target anomaly, such as abnormal power burst or illegal frequency band use;

[0142] If some stable edges If the corresponding edge weight decreases rapidly, it is determined to be a cooperative break, which indicates a change in communication strategy or external interference.

[0143] If multiple nodes simultaneously satisfy If a new high-weight edge is added, it is judged as abnormal linkage behavior;

[0144] in, , and These represent deviation thresholds, namely: single-node anomaly detection threshold, collaborative breakage detection threshold, and group linkage anomaly detection threshold.

[0145] S44. After anomaly confirmation, the impact propagation degree of the abnormal node in the cooperative network is calculated to determine whether it has triggered structural reorganization. If it is a long-term structural change, the cooperative baseline and scheduling parameters are updated. If it is a short-term fluctuation, the original structure is maintained and local monitoring is strengthened to achieve closed-loop linkage control between model correction and resource scheduling. Specifically:

[0146] After confirming the anomaly, instead of immediately reconstructing the model, an impact backtracking analysis is performed to define the anomaly impact propagation coefficient: ;

[0147] The propagation coefficient of the anomaly is used to determine whether the anomaly is a local fluctuation or structural reorganization. If it is determined to be a structural reorganization anomaly, then the following is triggered: update the collaborative baseline. Adjust node weights (feedback to step S3), increase sampling density in relevant frequency bands, and store abnormal collaborative modes in the abnormal mode library;

[0148] If the fluctuation is localized, only the monitoring accuracy will be improved, without updating the baseline;

[0149] in, This represents the set of neighboring nodes connected to target i in the current collaborative structure graph; Indicates the propagation coefficient of abnormal effects;

[0150] It should be noted that if the abnormal node... This represents a very small percentage (below the threshold) of most neighbors. This indicates that the anomaly's impact is mainly limited to individual connections and has not spread to the entire network; that is, it is a local fluctuation. Only the sampling density of the affected node and a few of its neighbors is increased, without modifying the overall cooperative baseline. If the anomaly node's... If multiple neighbors have a high percentage, or multiple nodes show a similar high percentage anomaly at the same time, it indicates that the anomaly has affected the entire collaborative structure and may lead to network reorganization. This is called a structural reorganization anomaly, which triggers collaborative baseline updates, node weight adjustments, and the anomaly pattern is included in the long-term model library.

[0151] S5. Through anomaly identification, baseline updating, collaborative structure prediction, and adaptive strategy closed-loop, the system achieves continuous and refined monitoring, efficient resource utilization, and long-term stable operation in complex electromagnetic environments. The specific implementation process is as follows:

[0152] S51. Combine the abnormal node and abnormal edge information identified in step S4 with the node activity level to calculate the collaborative correction weight for each node. Nodes with high weights are given priority in subsequent monitoring resource scheduling and baseline correction, forming a dynamic feedback node weight mechanism to achieve the integration of abnormal impact information. Specifically:

[0153] Historical stable collaborative baseline Co-deviation with the latest anomaly Integration and calculation of node weight adjustments: ;

[0154] in, This indicates that the nodes collaboratively correct the weights; Indicates node activity; This indicates the classification of neighbor anomalies. and These represent the weighting coefficients; and This represents the weighting adjustment coefficient;

[0155] S52. Using the node weights from step S51, the historical collaborative baseline is updated through weighted fusion. The latest collaborative behavior of abnormal nodes is introduced into the baseline, while the baseline of stable nodes remains undisturbed. This achieves controllable dynamic correction and stable evolution of the multi-objective collaborative structure, specifically:

[0156] ;

[0157] in, Indicates the updated collaborative baseline; Indicates the corrected fusion coefficient;

[0158] S53. Based on the updated collaborative baseline and node activity, predict the future evolution trend of the collaborative structure. Adjust the sampling frequency and monitoring frequency band by predicting the collaborative degree to achieve forward-looking adaptive strategy optimization, so that the monitoring resources actively follow the evolution of the target collaborative structure. Specifically:

[0159] Based on the updated collaborative baseline With node activity To predict the future evolution of the collaborative structure and form a basis for optimizing the monitoring strategy for the next cycle;

[0160] The short-time autoregressive prediction model (AR model) is used to predict the trend of node synergy.

[0161] ;

[0162] Adjust node sampling frequency based on predicted synergy With monitoring frequency band :

[0163] ;

[0164] ;

[0165] in, This indicates the predicted collaborative strength of nodes i and j in the next monitoring period; and These represent the AR model coefficients, which are derived from historical collaborative strength data. Indicates the sampling frequency of node i; This represents the node's basic sampling frequency, which is initially configured by the system. This represents the sampling frequency adjustment coefficient; Indicates the frequency band selection threshold;

[0166] S54. Integrate the baseline correction and prediction strategy output into the closed loop of step S3 to achieve long-term adaptive monitoring. Continuously and dynamically update node weights, sampling strategies, and the anomaly pattern library to form a complete closed loop of anomaly identification, baseline correction, strategy optimization, and monitoring feedback. This ensures the long-term stability, precision, and efficiency of the system. Specifically:

[0167] The outputs of steps S52 and S53 are used as the new baseline and adaptive strategy input for step S3; node behavior and collaborative structure are continuously monitored, and node weights are dynamically updated. Continuously record abnormal patterns and feed them back to the abnormal pattern library (step S44) to form knowledge accumulation; and automatically adjust the sampling density, frequency band selection and time resolution during long-term operation to achieve refined monitoring and efficient resource utilization.

[0168] Example 2, as Figure 2 As shown, the present invention proposes a multi-target radio monitoring system, which is used to execute a multi-target radio monitoring method proposed in Embodiment 1. It includes: a multi-source sensing and behavior primitive construction module, a cooperative structure analysis module, an adaptive monitoring strategy generation module, a cooperative deviation and anomaly identification module, and a cooperative baseline correction and long-term evolution module.

[0169] The multi-source sensing and behavior primitive construction module is deployed in the monitoring area with various types of radio sensors. It is responsible for continuously acquiring the spectrum signals of each radio target in time sequence, and extracting frequency band usage, power change characteristics and time occupation information in real time. Through signal preprocessing, feature decomposition and behavior primitive construction algorithms, the raw acquired data is transformed into a behavior primitive database for each target.

[0170] The collaborative structure analysis module takes behavioral primitives as input, constructs a multi-objective collaborative structure diagram through target correlation analysis and collaborative relationship calculation, generates node behavior profiles and edge collaborative strength information, and calculates the collaborative weight matrix between nodes and the overall collaborative stability index. It explores the collaborative and constraint relationships between targets at the time, frequency and energy use levels, and provides structured information for adaptive monitoring strategies.

[0171] The adaptive monitoring strategy generation module generates an adaptive sampling strategy for each target based on the stability index of the collaborative structure and the activity of nodes, including the sampling frequency, monitoring frequency band, and time resolution. It automatically adjusts the allocation of monitoring resources through a strategy optimization algorithm, enabling the monitoring strategy to evolve in real time with changes in target behavior. The generated strategy is then distributed to the acquisition end to achieve dynamic allocation and refined management of monitoring resources.

[0172] The collaborative deviation and anomaly identification module is responsible for comparing the real-time collected target behavior data with the historical collaborative baseline, calculating the multi-scale collaborative deviation degree, including node deviation, edge deviation and global structure deviation, and classifying and grading anomalies according to the deviation structure, determining single-target anomalies, collaborative break anomalies or group linkage anomalies; at the same time, it assesses the propagation range of the anomaly impact and feeds the anomaly information back to the monitoring strategy and baseline correction module for adjusting the sampling strategy and collaborative model.

[0173] The collaborative baseline correction and long-term evolution module takes anomaly identification results and real-time collaborative structure as input, dynamically updates the collaborative baseline, corrects node collaborative parameters by combining node weights and anomaly propagation information, predicts future collaborative structure evolution trends, and optimizes the sampling strategy for the next cycle to achieve forward-looking adaptive adjustment, forming a long-term closed loop. This ensures the continuous and stable operation and monitoring accuracy of the system in complex electromagnetic environments, while accumulating an anomaly pattern knowledge base to support subsequent anomaly identification and strategy optimization.

[0174] The embodiments of the present invention have been described in detail above with reference to the accompanying drawings. However, the present invention is not limited thereto. Various changes can be made within the scope of knowledge possessed by those skilled in the art without departing from the spirit of the present invention.

Claims

1. A multi-target radio monitoring method, characterized in that, The specific implementation steps include the following: S1. A unified time-frequency energy expression is constructed through multi-node time-consistent fusion. The target trajectory is reconstructed using dynamic segmentation driven by the energy change rate and frequency-time dual continuity constraints. A behavioral primitive vector is constructed, specifically including: The original observation signals collected by several broadband sensing nodes in the monitoring area are time-aligned by introducing a time offset compensation model to obtain time-aligned observation signals. By combining the signal-to-noise ratio of each node and the weighting coefficients set for spatial coverage reliability, the time-aligned observation signals are subjected to short-time Fourier transform and weighted fusion to form a continuous and stable joint time-frequency energy observation matrix. Based on the joint time-frequency energy observation matrix, the average energy of each frequency band within a unit time window is calculated using a sliding time window, and an adaptive change threshold is constructed by calculating the energy change rate and historical fluctuation variance. When the rate of energy change exceeds the adaptive change threshold, it is determined that a new behavior segment has begun, thus realizing the dynamic division of behavior segments; For the energy segments obtained by dynamic division, frequency continuity constraint functions and time continuity constraint functions are constructed respectively, and the comprehensive correlation strength between segments is calculated accordingly. Two behavior segments with a comprehensive correlation strength greater than a set threshold are assigned to the same target trajectory and continuously spliced ​​together to form a continuous energy evolution path that is independent of each target. After obtaining the target trajectory, frequency band usage density, power fluctuation entropy, time occupancy rhythm, average transmit power and power change variance are extracted for each target to construct a unified behavior primitive vector; S2. Jointly model multi-objective behavior, extract stable cooperative relationships and form a weighted cooperative structure diagram, and use structural entropy to characterize the overall electromagnetic environment organization state. S3. Driven by the stability of the collaborative structure, the monitoring resources are adaptively scheduled, dynamically generating sampling frequency, monitoring frequency band, and time resolution, and then issuing and executing them. Specifically, this includes: Construct a stable collaborative weight matrix among nodes, and statistically analyze the weighted average of collaborative strength and collaborative fluctuation of each node pair within a continuous time window, thereby constructing an overall collaborative structure stability index. For each target node, a comprehensive activity index is calculated based on its cooperative connection strength, frequency band occupancy ratio, and power fluctuation entropy. Based on the overall collaborative structure stability index and the comprehensive activity index of the nodes, the sampling frequency of each target is dynamically adjusted, the high-occupancy frequency band of the nodes is selected as the key monitoring frequency band, and the sliding time window length is adjusted to match the current time resolution requirements. The generated monitoring strategy, which includes sampling frequency, key monitoring frequency band and sliding time window length, is sent to each sensing node for execution in real time. After collecting new data, the target behavior primitive vector and collaborative structure are recalculated, and the stability index and node weights are continuously updated. S4. Using the historically stable collaborative structure as a baseline, calculate the standardized deviation at the edge level, node level, and global level to identify abnormal targets and types. S5. In response to the anomaly identification results, selectively trigger the dynamic correction of the collaborative model and monitoring strategy to form a closed loop of perception, analysis, scheduling and optimization.

2. The multi-target radio monitoring method according to claim 1, characterized in that, Step S2 specifically includes: Normalization transformation is performed on the behavior primitive vector of each target, including standardizing the frequency band usage density according to the start and end frequencies of the monitoring frequency band, centering the average transmit power according to the mean power and power standard deviation of all targets, and performing periodic scale mapping and peak normalization on the time occupancy rhythm. After standardization, a cooperative coupling strength function between two targets is defined. By calculating the frequency coupling strength, time coupling strength, and energy coupling strength, and then weighting and fusing the three, the comprehensive coupling strength of the target pair within the sliding time window is obtained. By using a sliding time window, the mean of the cooperative strength and the cooperative volatility of each target pair within the window are statistically analyzed, and a stable cooperative criterion is constructed. When the stable cooperative index exceeds a set threshold, it is determined that there is a stable cooperative relationship between the target pairs. All target pairs that meet the stable cooperation conditions are constructed into a weighted cooperative network graph, where nodes represent radio targets, edge weights represent stable cooperation indices, and network structure entropy is calculated to characterize the complexity of the overall cooperative network.

3. The multi-target radio monitoring method according to claim 2, characterized in that, Step S4 specifically includes: When the cooperative structure is within the stable range determined by the stability index, the cooperative strength of each target pair in multiple time periods is statistically averaged and its standard deviation is calculated to construct a historical reference model that includes historical stable cooperative baseline and historical cooperative fluctuation baseline. Within the current monitoring period, the latest behavioral profile is obtained and the current collaboration matrix is ​​reconstructed based on the updated sampling strategy. The standardization difference between the current collaboration strength in the collaboration matrix and the historical stable collaboration baseline is calculated to obtain the edge-level deviation of each target pair. The node-level deviation index of each target node is formed by aggregating all edge-level deviations, and the global structure deviation index is formed by aggregating all edge-level deviations. Based on the distribution characteristics of node-level deviation index, edge-level deviation degree, and global structure deviation index, anomalies are classified as single-node deviation type, collaborative fracture type, or group linkage type, and a comprehensive anomaly level index is constructed to classify the degree of impact of anomalies.

4. The multi-target radio monitoring method according to claim 3, characterized in that, Step S5 specifically includes: After confirming the anomaly, the anomaly impact propagation coefficient of the abnormal node in the cooperative network is calculated. The anomaly impact propagation coefficient is used to characterize the degree of influence of the abnormal node on all its neighboring nodes. If the proportion of the anomaly propagation coefficient in most neighboring nodes is lower than the set threshold, it is determined to be a local fluctuation anomaly, and only the monitoring accuracy of that node and its neighboring nodes is improved. If the proportion of the anomaly propagation coefficient among multiple neighboring nodes is high, it is determined to be a structural reorganization anomaly, triggering the updating of the collaborative baseline, adjusting the weight of nodes in resource scheduling, increasing the sampling density of relevant frequency bands, and storing the current anomaly collaborative mode in the anomaly mode library.

5. The multi-target radio monitoring method according to claim 4, characterized in that, Step S5 also includes: The identified abnormal nodes and abnormal edge information are combined with the node's comprehensive activity index. By integrating the historical stable collaborative baseline and the latest abnormal collaborative deviation, and introducing the neighbor anomaly level, the collaborative correction weight of each node is calculated. The collaborative correction weight is used as a priority factor in subsequent monitoring resource scheduling and baseline correction to achieve dynamic feedback and integration of information on the impact of anomalies.

6. The multi-target radio monitoring method according to claim 5, characterized in that, Step S5 also includes: The historical stable collaborative baseline is updated by weighted fusion using the calculated node collaborative correction weights. During the update process, the latest collaborative behavior of abnormal nodes is introduced into the new collaborative baseline with corresponding correction weights, while the historical baseline information of stable nodes is kept undisturbed, so as to achieve controllable dynamic correction and stable evolution of the multi-objective collaborative structure.

7. The multi-target radio monitoring method according to claim 6, characterized in that, Step S5 also includes: Based on the updated collaborative baseline and the node's comprehensive activity index, a short-time autoregressive prediction model is used to predict the collaborative strength of nodes in future monitoring periods. The sampling frequency and monitoring frequency band selection of nodes are dynamically adjusted based on the predicted coordination strength, so that the monitoring resources can be proactively adjusted to follow the evolution trend of the target coordination structure, and the adjusted strategy is used as the basis for the next monitoring cycle.

8. A multi-target radio monitoring system, used to execute the multi-target radio monitoring method according to any one of claims 1 to 7, characterized in that, include: The multi-source sensing and behavior primitive construction module is used to continuously acquire radio signals from multiple nodes in a time sequence, and transform the raw acquired data into behavior primitive vectors for each target through signal preprocessing and feature decomposition. The collaborative structure analysis module is used to calculate the collaborative strength between objectives by taking behavioral primitive vectors as input, screen stable collaborative relationships to construct a multi-objective weighted collaborative structure diagram, and generate collaborative stability indicators. The adaptive monitoring strategy generation module is used to generate an adaptive sampling strategy for each target based on the collaborative stability index and node activity, and then send it to the collection terminal for execution. The collaborative deviation and anomaly identification module is used to compare the real-time collected target behavior data with the historical collaborative baseline, calculate the multi-scale collaborative deviation, and perform anomaly classification and grading. The Cooperative Baseline Correction and Long-Term Evolution module is used to dynamically update the cooperative baseline based on the anomaly identification results, predict the future evolution trend of the cooperative structure, and optimize the sampling strategy for the next cycle.

Images