Intelligent community building perimeter intelligent protection system and intrusion detection method

Abstract

This invention discloses an intelligent perimeter protection system and intrusion detection method for smart communities, relating to the field of smart community security technology. Based on threat situation information generated by a threat situation assessment module, the system identifies target areas with excessive threat levels and calculates a sensing demand gap vector. It then filters neighboring and pattern-matching candidate resources from a sensing resource pool, generating a temporary focusing cluster configuration scheme that includes requisitioned nodes and reconfigured functional roles. The temporary focusing cluster execution module sends reconfiguration commands to synchronously switch the functional roles of each requisitioned sensing node, forming a multi-node collaborative temporary focusing cluster to perform enhanced sensing tasks on the target area. After the threat is resolved, the system automatically releases resources and restores its original working mode. This achieves on-demand flow of sensing resources, functional reconfiguration, and cluster collaboration, enabling the system to possess self-organizing protection capabilities when facing dynamically changing intrusion threats and improving the global response speed of community building perimeter protection.

Description

Technical Field

[0001] This invention relates to the field of smart community security technology, specifically to a smart community building perimeter intelligent protection system and intrusion detection method. Background Technology

[0002] With the deepening of smart community construction, building perimeter security is gradually evolving from single video surveillance to multi-sensor collaborative perception. Existing community building perimeter protection systems typically deploy various sensing devices such as cameras, radar, infrared beams, and vibration fiber optics. Each sensing device operates independently according to its preset fixed functions. For example, cameras only undertake video acquisition tasks, radar is only responsible for moving target detection, and acoustic sensors are only used for listening to abnormal sounds. The data collected by each sensing device is uploaded to the central management platform through wired or wireless links, and the central management platform performs simple threshold judgments and linkage control based on a preset rule engine.

[0003] However, the functions of existing sensing devices are deeply tied to their hardware. When deployed, a single sensing device is assigned a fixed working mode and cannot dynamically adjust its functional role according to changes in the real-time threat situation. When a high-threat event occurs in an area, cameras in adjacent areas cannot automatically switch from wide-area inspection mode to narrow-area tracking mode to support the fine identification needs of high-threat areas. Furthermore, there is a lack of a unified management mechanism among sensing resources. Each sensing device operates as an isolated node, and the system cannot perceive the currently available sensing capabilities from a global perspective. It is also difficult to perform optimal resource scheduling and functional reconstruction when multiple threats occur concurrently.

[0004] Therefore, there is an urgent need for an intelligent protection system that can manage heterogeneous sensing devices in a unified manner and reorganize and reconstruct the sensing function cluster according to the real-time threat situation. Summary of the Invention

[0005] The purpose of this invention is to overcome the shortcomings of existing technologies and provide an intelligent perimeter protection system and intrusion detection method for smart communities. Based on threat situation information generated by a threat situation assessment module, this system identifies target areas with excessive threat levels and calculates the sensing demand gap vector. It then filters neighboring and pattern-matching candidate resources from the sensing resource pool to generate a temporary focusing cluster configuration scheme that includes requisitioned nodes and reconfigured functional roles. The temporary focusing cluster execution module sends reconfiguration commands to synchronously switch the functional roles of each requisitioned sensing node, forming a multi-node collaborative temporary focusing cluster to perform enhanced sensing tasks on the target area. After the threat is resolved, resources are automatically released and the original working mode is restored. This achieves on-demand flow of sensing resources, functional reconfiguration, and cluster collaboration, enabling the system to possess self-organizing protection capabilities when facing dynamically changing intrusion threats. It improves the global response speed and multimodal collaborative accuracy of community building perimeter protection and is suitable for high-level security protection needs in complex scenarios such as large high-rise residential communities and park building complexes.

[0006] To solve the above-mentioned technical problems, the present invention provides the following technical solution: On one hand, a smart community building perimeter intelligent protection system, the system comprising:

[0007] The sensing resource module is used to deploy various heterogeneous sensing nodes around the perimeter of community buildings. Each heterogeneous sensing node has a current working mode and a set of switchable working modes.

[0008] The resource virtualization middleware module is used to treat the sensing capabilities of each heterogeneous sensing node as a unified description of the capability resource item, and to aggregate the capability resource items of all heterogeneous sensing nodes to form a sensing resource pool.

[0009] The threat situation assessment module is used to fuse and process multi-source data returned by the perception resource module to generate threat situation information covering all areas of the community building perimeter in real time.

[0010] The resource orchestration module is used to determine the first region as the target region when the threat level value of the first region is continuously exceeded within a preset time window. It calculates the perception demand gap vector of the target region, filters candidate resource items from the perception resource pool that are spatially adjacent and whose switchable working mode sets match the perception demand gap vector, and generates a temporary focus cluster configuration scheme that includes a list of requisitioned nodes and the reconstruction function role corresponding to each requisitioned perception node in the list of requisitioned nodes.

[0011] The temporary focus cluster execution module is used to send a reconstruction instruction containing the target working mode and the effective timestamp to the requisitioned nodes, so that each requisitioned sensing node synchronously switches to the corresponding reconstruction function role and forms a temporary focus cluster. After the threat level value of the target area is less than the preset release threshold and remains so for a preset period of time, a release instruction is sent to each requisitioned sensing node, so that each requisitioned sensing node returns to the working mode before requisition.

[0012] Furthermore, the heterogeneous sensing nodes include multiple types such as camera nodes, radar nodes, acoustic array nodes, and light curtain nodes;

[0013] The capability resource items include node identifier, spatial location coordinates, current working mode, set of switchable working modes, detection range parameters, accuracy parameters, and current occupancy status field.

[0014] Furthermore, the set of switchable working modes in the capability resource item includes wide-area inspection mode, narrow-area tracking mode, cooperative cross-positioning mode, spectrum detection mode, and sleep mode;

[0015] The wide-area inspection mode is a working mode in which the sensing node performs a cyclic scan of the allocated wide-angle field of view at a preset cycle.

[0016] The narrow-field tracking mode is a working mode in which the sensing node continuously locks onto a specified target at a high frame rate and narrow field of view.

[0017] The collaborative cross-positioning mode is a working mode in which two or more sensing nodes perform time-synchronized angle-of-arrival joint calculation to determine the spatial location of the target.

[0018] The spectrum detection mode is a working mode in which the sensing node performs directional detection of electromagnetic signals in a preset frequency band.

[0019] The sleep mode is a low-power standby mode in which the sensing node disables the active detection function and only maintains the control link listening.

[0020] Furthermore, the threat situation information includes a gridded area index, the corresponding area's threat level value, and threat evolution trend prediction information, and the threat situation information is output to the resource orchestration module;

[0021] When generating the threat situation information, the threat situation assessment module uses an attention-weighted graph network to perform gridded threat probability calculation on the perimeter of the community buildings;

[0022] The attention-weighted graph network uses each gridded area of ​​the community building perimeter as graph nodes, the spatial adjacency relationship between adjacent gridded areas and the detection overlap relationship of heterogeneous sensing nodes as graph edges, and the preprocessed event stream returned by each heterogeneous sensing node as node input features. Through a multi-head attention mechanism, the event features of multiple heterogeneous sensing nodes in the same gridded area are weighted and aggregated to output the threat probability value of each gridded area.

[0023] The threat situation assessment module also calculates threat evolution trend prediction information based on the threat probability value sequence of multiple consecutive time steps using the exponentially weighted moving average method.

[0024] Furthermore, when generating the temporary focusing cluster configuration scheme, the resource orchestration module assigns a narrow-domain tracking mode as the narrow-domain tracking role of the reconstructing functional role to at least one reconstructed sensing node in the reconstructed node list, and assigns a cooperative cross-location mode as the cooperative cross-location role of the reconstructing functional role to at least another reconstructed sensing node, so that the temporary focusing cluster simultaneously possesses both a narrow-domain tracking role and a cooperative cross-location role, and the narrow-domain tracking role and the cooperative cross-location role work together to form a continuous tracking and spatial positioning cross-linking of the intruding target within the target area.

[0025] Furthermore, when filtering candidate resource items, the resource orchestration module sets a first filtering rule and a second filtering rule;

[0026] The first filtering rule is as follows: select the sensing nodes whose current occupancy status field is idle, sort them according to the spatial distance between their spatial location coordinates and the geometric center of the target area, and select the first m sensing nodes with the smallest spatial distance as candidate resource items, where m is the minimum number of nodes that meet the minimum coverage requirement of the sensing demand gap vector.

[0027] The second filtering rule is as follows: when the number of sensing nodes that meet the first filtering rule is insufficient to cover the sensing demand gap vector, from the sensing nodes whose current occupancy status field is occupied, sort them according to the priority of the tasks performed by the occupied sensing nodes, and sequentially select sensing nodes with low priority tasks, and add the selected sensing nodes to the candidate resource items until the number of candidate resource items meets the minimum requirement to cover the sensing demand gap vector.

[0028] Furthermore, the resource orchestration module calculates the perceived demand gap vector of the target area in the following way:

[0029] The set of full-coverage sensing capabilities required for the target area is taken as the demand capability set. The set of required capabilities It includes the detection modality dimension, spatial coverage dimension, positioning accuracy dimension, and sampling refresh rate dimension required for effective protection of the target area;

[0030] The set of sensing capabilities actually provided by the sensing nodes that are currently deployed and operational in the target area is taken as the existing capability set. ;

[0031] Calculate the perceived demand gap vector For demand capability set With existing capabilities The difference set, i.e. ,in, The modal gap coefficient is a ratio of the number of modal types in demand detection that are not covered by existing sensing nodes to the total number of demand modal types. The spatial coverage gap coefficient represents the percentage of angles within the required spatial angle range that are not covered by the detection range of existing sensing nodes. The positioning accuracy gap coefficient represents the normalized difference between the required positioning accuracy and the highest positioning accuracy achievable by existing sensing nodes. The sampling refresh rate gap coefficient represents the normalized difference between the required sampling refresh rate and the highest sampling refresh rate achievable by existing sensing nodes. This difference occurs when the sensing demand gap vector... When the value of any gap coefficient exceeds the preset gap tolerance threshold of the corresponding dimension, the resource orchestration module determines that there is a perception demand gap in the target area of ​​the corresponding dimension.

[0032] Furthermore, in the temporary focusing cluster execution module, when the requisitioned sensing nodes designated as cooperative cross-location roles within the temporary focusing cluster are the first acoustic array node and the second acoustic array node, the first acoustic array node and the second acoustic array node use a joint time difference of arrival (TDOA) calculation method to locate the sound source of the intruding target within the target area in the cooperative cross-location mode. The calculation steps of the joint TDOA calculation method are as follows:

[0033] A local coordinate system is established with the spatial coordinates of the first acoustic array node as the origin, and the spatial coordinates of the second acoustic array node are... ,in The baseline distance between the first acoustic array node and the second acoustic array node;

[0034] The first acoustic array node records the arrival time of the acoustic signal emitted by the intruding target. The second acoustic array node records the arrival time of the acoustic signal emitted by the intruding target. Calculate the time difference of arrival ;

[0035] Based on the speed of sound signal propagation in air Spatial coordinates of the intrusion target sound source Satisfies the hyperbolic equation: Combined with the angle of arrival of the acoustic signal measured by the first acoustic array node and the angle of arrival of the acoustic signal measured by the second acoustic array node Establish the equations for the direction lines: , ;

[0036] By solving the simultaneous equations of the hyperbola and the direction line equations, the unique spatial coordinates of the sound source of the intruding target can be calculated. Acoustic cross-location of intrusion targets within the target area is completed;

[0037] Within the temporary focusing cluster, the camera sensing node designated as the narrow-domain tracking role adjusts the gimbal pointing according to the calculated unique spatial position coordinates to perform visual locking and tracking of the intrusion target.

[0038] On the other hand, the method for detecting intrusion at building perimeters in smart communities includes the following specific steps:

[0039] S100, Resource Virtualization: The resource virtualization middleware module collects the status information and capability descriptions of each heterogeneous sensing node in the sensing resource module, updates the capability resource items corresponding to each heterogeneous sensing node, and aggregates all capability resource items to form a sensing resource pool.

[0040] S200 Threat Situation Generation: The threat situation assessment module receives pre-processed data from heterogeneous sensing nodes in the sensing resource module, performs fusion processing on the pre-processed data, and generates threat situation information covering all areas of the community building perimeter in real time, and calibrates the threat level values ​​of each gridded area.

[0041] S300, Focus Decision and Solution Generation: The resource orchestration module polls the threat situation information. When it is identified that the threat level value of the first region continuously exceeds the preset threat threshold within a preset time window, the first region is determined as the target region. The sensing demand gap vector of the target region is calculated, the candidate sensing node set is matched, the list of requisitioned nodes without resource conflicts is determined, and a temporary focus cluster configuration scheme containing requisitioned nodes and reconstructed functional roles is generated.

[0042] S400, Cluster Construction and Functional Reconstruction: The temporary focused cluster execution module sends a reconstruction instruction containing the target working mode, mode parameters and effective timestamp to each requisitioned sensing node in the requisitioned node list. Each requisitioned sensing node synchronously switches to the corresponding reconstruction function role at the effective timestamp, forming a temporary focused cluster.

[0043] S500, Collaborative Enhanced Perception: Each requisitioned perception node within the temporary focusing cluster performs collaborative perception tasks according to its reconstructed functional role, continuously tracks, spatially locates, and identifies the behavior of intrusion targets within the target area, and reports the collaborative perception results to the threat situation assessment module to update the threat situation information;

[0044] S600, Resource Release: When the threat situation assessment module detects that the threat level value of the target area is less than the preset release threshold and remains so for a preset period of time, it generates a release command, which is sent to each requisitioned sensing node via the temporary focusing cluster execution module. Each requisitioned sensing node is restored to its working mode before requisition and updates the current occupied status field to idle status.

[0045] Compared with existing technologies, this smart community building perimeter intelligent protection system and intrusion detection method have the following advantages:

[0046] I. This invention unifies heterogeneous sensing nodes deployed around the community perimeter, such as camera nodes, radar nodes, acoustic array nodes, and light curtain nodes, into capability resource items that include node identifier, spatial coordinates, current working mode, switchable working mode set, detection range parameters, accuracy parameters, and current occupancy status fields through a resource virtualization middleware module. These resources are then aggregated into a sensing resource pool. Based on the threat situation information generated in real time by the threat situation assessment module, the resource orchestration module automatically identifies target areas with excessive threat levels, calculates the sensing demand gap vector, and filters candidate resource items from the sensing resource pool according to spatial proximity and pattern matching degree to generate a temporary focusing cluster configuration scheme. This mechanism enables sensing nodes in idle or low-priority tasks to be requisitioned immediately and given new reconfiguration functions, realizing the on-demand flow and elastic scheduling of sensing resources.

[0047] Second, when generating a temporary focusing cluster configuration scheme, the resource orchestration module assigns specific reconstruction function roles to the requisitioned sensing nodes, enabling the temporary focusing cluster to simultaneously possess a narrow-field tracking role and a cooperative cross-positioning role. The narrow-field tracking role is undertaken by the camera sensing node that switches to a high frame rate narrow field of view mode to continuously visually lock onto the intruding target. The cooperative cross-positioning role is undertaken by the acoustic array node that switches to a joint time difference of arrival calculation mode or the radar node that switches to a cooperative mode to calculate the spatial coordinates of the intruding target. The two roles establish a point-to-point data channel within the temporary focusing cluster through a cooperative parameter set, realizing direct sharing and cooperative processing of sensing data within the cluster, forming a cross-linking mechanism for tracking and positioning, and improving the stability of continuous tracking of the intruding target and the accuracy of multi-angle spatial positioning.

[0048] Other advantages, objectives and features of the invention will be set forth in part in the description which follows, and in part will be apparent to those skilled in the art from the following examination or study, or may be learned from the practice of the invention. Attached Figure Description

[0049] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the accompanying drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are merely some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without any creative effort.

[0050] Figure 1 Data flow diagram for a smart community building perimeter intelligent protection system;

[0051] Figure 2 A modular framework diagram of a smart community building perimeter intelligent protection system.

[0052] Figure 3 A flowchart illustrating the steps of a smart community building perimeter intrusion detection method. Detailed Implementation

[0053] To better understand the above technical solutions, a detailed description of the solutions will be provided below in conjunction with the accompanying drawings and specific embodiments. Obviously, the described embodiments are merely some, not all, of the embodiments of the present invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0054] To address the issues in existing technologies such as the deep integration of sensing device functions with hardware, the lack of a unified management mechanism among sensing resources, and the inability of the system to dynamically adjust sensing function configurations based on real-time threat situations, this invention provides a smart community building perimeter intelligent protection system and intrusion detection method. This system unifies the resources of various heterogeneous sensing nodes deployed along the community perimeter, forming a sensing resource pool. The resource orchestration module dynamically filters candidate resource items and generates a temporary focus cluster configuration scheme based on threat situation information generated in real-time by the threat situation assessment module. The temporary focus cluster execution module sends reconstruction instructions to the requisitioned sensing nodes, enabling each requisitioned sensing node to synchronously switch to its corresponding reconstruction function role and form a temporary focus cluster. These clusters collaboratively execute enhanced sensing tasks for the target area. Once the threat is resolved, the resources are automatically released and the system returns to its original working mode, thereby achieving on-demand flow of sensing resources, functional reconstruction, and cluster collaboration.

[0055] It should be noted that the specific values, node numbers, device models, spatial coordinates, time windows, threshold parameters, and scenario examples described in this embodiment are all illustrative and intended only to help understand the technical solution and working principle of this invention. They do not constitute any limitation on the scope of protection of this invention. In actual deployment and application, those skilled in the art can adaptively adjust and configure the above-mentioned exemplary parameters according to factors such as the actual scale of the community building perimeter, security protection level requirements, specific technical specifications of heterogeneous sensing nodes, and network communication conditions. Such adjustments and configurations do not depart from the spirit and principles of this invention and should be included within the scope of protection of this invention.

[0056] like Figure 1 and Figure 2 As shown, Figure 1 This is a data flow diagram of a smart community building perimeter intelligent protection system according to an embodiment of the present invention. Figure 1 This is a modular framework diagram of a smart community building perimeter intelligent protection system according to an embodiment of the present invention. The system generally includes: a sensing resource module, a resource virtualization middleware module, a threat situation assessment module, a resource orchestration module, and a temporary focus cluster execution module.

[0057] In the aforementioned sensing resource module, various heterogeneous sensing nodes are deployed around the perimeter of community buildings. In this embodiment, the heterogeneous sensing nodes include multiple types such as camera nodes, radar nodes, acoustic array nodes, and light curtain nodes. Each heterogeneous sensing node has a current working mode and a set of switchable working modes. The current working mode is the working mode executed by the sensing node at the current moment, and the set of switchable working modes is the collection of all working modes that the sensing node can switch into under the support of hardware capabilities and software configuration.

[0058] In one specific embodiment, the camera nodes include, but are not limited to, fixed bullet cameras, dome cameras, and panoramic fisheye cameras. The set of switchable operating modes for the camera nodes includes wide-area inspection mode, narrow-area tracking mode, cooperative cross-positioning mode, and sleep mode. In wide-area inspection mode, the camera node performs preset position patrol scans within its allocated wide-angle field of view at preset intervals, transmitting the acquired panoramic video stream back to the threat situation assessment module in real time. In narrow-area tracking mode, the camera node receives the target spatial coordinates from the resource orchestration module, adjusts the azimuth and pitch angles of the gimbal, and continuously locks onto and tracks the designated target with a high frame rate and narrow field of view. In cooperative cross-positioning mode, the camera node synchronizes time with the acoustic array node or radar node, participating in multi-node joint spatial positioning calculations based on its own calibrated spatial position coordinates.

[0059] Radar nodes include, but are not limited to, millimeter-wave radar and ultra-wideband radar. The set of switchable operating modes for radar nodes includes wide-area inspection mode, narrow-area tracking mode, cooperative cross-location mode, and dormant mode. In wide-area inspection mode, the radar node transmits frequency-modulated continuous waves at a preset period to detect the range, velocity, and azimuth of moving targets within its perimeter. In narrow-area tracking mode, the radar node increases the pulse repetition frequency to perform high-data-rate tracking of a designated target. In cooperative cross-location mode, two or more radar nodes cross-beam cover the same target area, determining the target's two-dimensional or three-dimensional spatial coordinates through joint angle-of-arrival calculation.

[0060] Acoustic array nodes include, but are not limited to, linear or area arrays composed of multiple microphone units. The switchable operating modes of the acoustic array nodes include wide-area inspection mode, cooperative cross-location mode, spectrum detection mode, and sleep mode. In wide-area inspection mode, the acoustic array node continuously monitors the ambient sound field; when it detects an abnormal sound signal exceeding the background noise threshold, it extracts the time-frequency characteristics of the sound signal and transmits them back. In cooperative cross-location mode, two or more acoustic array nodes perform precise time synchronization and use a joint time difference of arrival (TDOA) calculation method to spatially locate the sound source target. In spectrum detection mode, the acoustic array node performs directional reception of electromagnetic or communication signals in a preset frequency band.

[0061] Light curtain nodes include, but are not limited to, infrared beam light curtains and laser scanning light curtains. The switchable operating modes of the light curtain nodes include a wide-area inspection mode and a sleep mode. In the wide-area inspection mode, the light curtain node continuously emits infrared or laser beams, and triggers an out-of-bounds event signal when the beam is blocked.

[0062] The aforementioned resource virtualization middleware module is used to uniformly describe the sensing capabilities of each heterogeneous sensing node as capability resource items, and to aggregate the capability resource items of all heterogeneous sensing nodes to form a sensing resource pool. The capability resource item includes node identifier, spatial location coordinates, current operating mode, set of switchable operating modes, detection range parameters, accuracy parameters, and current occupancy status fields.

[0063] In a specific embodiment, taking a PTZ spherical camera node, numbered PTZ-07, located on the perimeter wall of the eastern section of the community as an example, the specific content of the capability resource items generated by the resource virtualization middleware module for this camera node includes: the value of the node identifier field is PTZ-07; the value of the spatial location coordinate field is represented in the community's unified coordinate system as (X=128.5, Y=36.2, Z=4.8); the value of the current working mode field is wide-area inspection mode; the value of the switchable working mode set field is a set of four enumerated items including wide-area inspection mode, narrow-area tracking mode, cooperative cross-positioning mode, and sleep mode; the value of the detection range parameter field includes a maximum detection distance of 200 meters, a horizontal field of view range of 0°-360°, and a vertical field of view range of -20°-+60°; the value of the accuracy parameter field includes a horizontal rotation accuracy of 0.01° and a vertical rotation accuracy of 0.01°; and the value of the current occupancy status field is idle.

[0064] Taking an acoustic array node, numbered AU-03, located in the northern section of the community fence, as an example, the specific content of the capability resource items generated by the resource virtualization middleware module for this acoustic array node is as follows: the node identifier field has the value AU-03; the spatial location coordinate field has the value (X=96.3, Y=41.7, Z=1.5); the current working mode field has the value of wide-area inspection mode; the switchable working mode set field has the value of a set of four enumerated items including wide-area inspection mode, cooperative cross-positioning mode, spectrum detection mode, and sleep mode; the detection range parameter field has the value of an effective listening radius of 80m and a frequency response range of 20Hz to 20kHz; the accuracy parameter field has the value of a sound source positioning accuracy of 0.5° and the number of array microphones of 8; and the current occupancy status field has the value of idle.

[0065] The resource virtualization middleware module is also responsible for periodically collecting the status information and capability descriptions of each heterogeneous sensing node, and updating the capability resource items corresponding to each heterogeneous sensing node. The typical collection period is 1s-5s. When the working mode of a certain sensing node changes or its current occupancy status changes, the resource virtualization middleware module immediately triggers an incremental update to synchronize the changed capability resource items to the global resource view of the sensing resource pool.

[0066] The aforementioned sensing resource pool contains capability resource items of all registered heterogeneous sensing nodes within the perimeter of community buildings, constituting a global resource view at the current moment. The global resource view is a four-dimensional data table, with the four dimensions being spatial location, working mode, capability parameter, and occupancy status. The resource orchestration module performs multi-dimensional retrieval of the global resource view through a query interface.

[0067] The aforementioned threat situation assessment module is used to fuse and process the multi-source data returned by the perception resource module, and generate threat situation information covering all areas of the community building perimeter in real time. The threat situation information includes a gridded area index, the corresponding area's threat level value, and threat evolution trend prediction information, and outputs the threat situation information to the resource orchestration module.

[0068] In one specific embodiment, the threat situation assessment module first divides the perimeter of the community buildings into gridded areas of uniform size, and assigns a unique gridded area index to each gridded area.

[0069] The threat situation assessment module receives preprocessed data from each heterogeneous sensing node in the sensing resource module. The preprocessed data consists of structured event information extracted by each sensing node after preliminary signal processing at the edge. The preprocessed data transmitted by the camera node includes motion detection bounding box area, target type classification results, and motion trajectory identifiers. The preprocessed data transmitted by the radar node includes target distance, azimuth angle, radial velocity, and radar cross section estimation. The preprocessed data transmitted by the acoustic array node includes anomalous acoustic signal arrival direction estimation, spectral feature vector, and signal duration. The preprocessed data transmitted by the light curtain node includes beam obstruction status and obstruction duration.

[0070] In the process of generating threat situation information, the threat situation assessment module uses an attention-weighted graph network to perform gridded threat probability estimation on the perimeter of community buildings. The attention-weighted graph network uses each gridded area of ​​the community building perimeter as a graph node, the spatial adjacency relationship between adjacent gridded areas and the detection overlap relationship of heterogeneous sensing nodes as graph edges, and the preprocessed event streams returned by each heterogeneous sensing node as node input features. Through a multi-head attention mechanism, the event features of multiple heterogeneous sensing nodes in the same gridded area are weighted and aggregated to output the threat probability value of each gridded area.

[0071] In this embodiment of the invention, the attention-weighted graph network is used to estimate the threat probability of each gridded area around the perimeter of community buildings. The process includes a graph structure construction stage, a node input feature construction stage, a multi-head attention aggregation stage, and a threat probability output stage, specifically:

[0072] Construct a graph structure and define the graph. ,in, For a set of nodes, each node A gridded area corresponding to the perimeter of the community buildings, with nodes The number corresponds one-to-one with the gridded area index of the gridded area. The graph is a set of edges, containing two types of edges. The first type is spatial adjacency edges; if two gridded regions share a boundary or a vertex in space, a spatial adjacency edge is established between the nodes corresponding to these two gridded regions. The second type is detection overlap edges; if the detection range of a heterogeneous sensing node simultaneously covers two or more gridded regions, a detection overlap edge is established between the nodes corresponding to the two or more covered gridded regions. Each edge is assigned a weight. The weight of a spatial adjacency edge is 1, and the weight of a detection overlap edge is the normalized value of the accuracy parameter of the heterogeneous sensing node. The normalized value of the accuracy parameter is calculated as the ratio of the value of the positioning accuracy parameter field of the heterogeneous sensing node to the highest value of the positioning accuracy parameter field among all heterogeneous sensing nodes, resulting in a normalized value between 0 and 1. The graph structure is updated when the deployment location of the heterogeneous sensing nodes changes or the gridded region division is adjusted.

[0073] Construct node input features for a graph Each node in Collect the node The preprocessed event streams transmitted from all heterogeneous sensing nodes within the corresponding gridded area are analyzed within the most recent time window. The time window duration is set to 2 seconds. Four feature components are extracted from the preprocessed event stream transmitted by each heterogeneous sensing node: the first feature component is the number of events, i.e., the total number of valid events detected by the node within the time window; the second feature component is the event type distribution, represented by a one-hot encoded vector. Each dimension of the vector corresponds to an event type; if an event of that type occurs within the time window, the corresponding dimension is set to 1, otherwise it is set to 0. Event types include motion detection events, target classification events, abnormal sound signal detection events, and beam obstruction events; the third feature component is the event duration, i.e., the proportion of the total duration of valid events within the time window to the total time window duration; the fourth feature component is the event credibility score, which is the average credibility score of all valid events within the time window. The event credibility score is calculated by each sensing node based on signal quality when generating preprocessed events. The four feature components extracted from each heterogeneous sensing node within the gridded area are then concatenated sequentially according to the sensing node type to form the node's identity. The node input feature vector If there are no heterogeneous sensing nodes in a certain gridded region, then the node input feature vector of the corresponding node in that gridded region is a zero vector.

[0074] Multi-head attention aggregation, for each attention head Input feature vectors to the nodes respectively Perform a linear transformation to obtain the query vector. Key vector Sum value vector The formula for calculating the linear transformation is: ,in, , , The first The query weight matrix, key weight matrix, and value weight matrix for each attention head are all dimensional. , Input feature vectors for nodes The dimension. For each node compute nodes with neighboring nodes Attention weights between neighbors This includes detecting spatial adjacent edges and overlapping edges and nodes. All directly connected nodes, and nodes The attention weights are calculated using a scaled dot product attention mechanism: first, the node... query vector with neighboring nodes key vector The dot product, then divided by the scaling factor. Scaling is performed, followed by normalization using a flexible maximum function to obtain the th... Each attention node For neighboring nodes attention weights Attention weight The calculation formula is: ,in, Represents a node The set of neighboring nodes, including nodes The dot symbol itself represents the vector dot product operation. Input feature vectors for nodes The dimension. Based on the calculated attention weights. For neighboring nodes value vector Perform a weighted summation to obtain the first... Each attention node The output feature vector ,Will The output feature vector of each attention head The concatenated feature vector is obtained by concatenating the features vectors, and then the concatenated feature vectors are combined with the output weight matrix. Multiplication followed by a linear transformation yields the output feature vector of the multi-head attention aggregation layer. ,in, To output the weight matrix, the dimension is... Concat represents the vector concatenation operation.

[0075] Threat probability output, which is the output feature vector of the multi-head attention aggregation layer. The input to the feedforward neural network layer undergoes a nonlinear transformation. This layer comprises two fully connected layers and one activation function layer. The activation function is a linear rectified function. The output of the feedforward neural network layer is mapped to the interval between 0 and 1 using a sigmoid function to obtain the threat probability value for that gridded region. The formula for calculating the sigmoid function is: ,in, This represents the output value of the feedforward neural network layer. For nodes Threat probability value for the corresponding gridded region at the current time step. The value ranges from 0 to 1. The closer the value is to 1, the higher the probability that there is an intrusion threat in the gridded area.

[0076] Through the above four stages of processing, the attention-weighted graph network completes the calculation of threat probability values ​​for each gridded region from the preprocessed event streams fed back from each heterogeneous sensing node. The threat situation assessment module calculates threat evolution trend prediction information based on the threat probability value sequence over multiple consecutive time steps using an exponentially weighted moving average method. The formula for calculating the threat evolution trend prediction value is as follows: ,in This is a smoothing coefficient, ranging from 0 to 1; in this embodiment, it is set to 0.3. This is the predicted threat evolution trend value from the previous time step. The threat level value is a weighted composite value of the threat probability value and the predicted threat evolution trend value, calculated using the following formula: ,in The weighting coefficient is 0.6 in this embodiment.

[0077] Threat situation information is organized into a mapping table between gridded area indexes and the corresponding threat level values ​​and threat evolution trend prediction information for each gridded area, and is output to the resource orchestration module in real time via message queues or shared memory interfaces.

[0078] The resource orchestration module described above is used to determine the first region as the target region when the threat level value of the first region is continuously exceeded within a preset time window. It calculates the perception demand gap vector of the target region, selects candidate resource items from the perception resource pool that are spatially adjacent and whose switchable working mode sets match the perception demand gap vector, and generates a temporary focus cluster configuration scheme that includes a list of requisitioned nodes and the reconstruction function role corresponding to each requisitioned perception node in the list of requisitioned nodes.

[0079] In one specific embodiment, the preset threat threshold is 0.7, and the preset time window is 5 seconds. The resource orchestration module reads the threat situation information output by the threat situation assessment module in a polling manner, and iterates through the threat level values ​​of each gridded area one by one. When it is found that the threat level value of the first area remains above 0.7 for 5 consecutive seconds, the resource orchestration module determines the first area as the target area.

[0080] After determining the target area, the resource orchestration module calculates the sensing demand gap vector for that area. The calculation method is as follows: the required full-coverage sensing capability set for the target area is defined as the demand capability set R. The demand capability set R includes the detection modality dimension, spatial coverage dimension, positioning accuracy dimension, and sampling refresh rate dimension required for effective protection of the target area. The actual sensing capability set provided by the sensing nodes currently deployed and operational in the target area is defined as the existing capability set C. The composition of the existing capability set C corresponds one-to-one with the four dimensions of the demand capability set R, including the existing detection modality set. Existing space coverage Current positioning accuracy and existing sampling refresh rate Four components.

[0081] In a specific scenario, the existing detection mode set This is the union of the detection modes provided by all sensing nodes deployed and operational in the target area at the current moment. If there is one camera node and one radar node operational in the target area, and the camera node provides a visual detection mode and the radar node provides a radio frequency detection mode, then... Visual detection mode, radio frequency detection mode Detect modal notch coefficient ,in, For demand capability set The set of demand detection modes in the middle, The total number of modal types in the demand detection modal set. This refers to the number of modal types that are in the demand detection modal set but do not belong to the existing detection modal set, i.e., the number of modal types not covered by existing sensing nodes.

[0082] In a specific scenario, the existing spatial coverage By determining the complete spatial angular range of the target area, the field of view coverage of each sensing node in operation within the target area is obtained one by one. The field of view coverage of all sensing nodes is then projected onto the complete spatial angular range of the target area in a union manner. The proportion of the sum of angles covered by the union projection to the complete spatial angular range of the target area is calculated, along with the spatial coverage gap coefficient. ,in, For demand capability set The required spatial coverage rate refers to the spatial coverage value required for effective protection of the target area. The current spatial coverage rate refers to the proportion of the spatial angular range actually covered by all currently operational sensing nodes within the target area. hour, A value of 0 indicates that there is no spatial coverage gap.

[0083] In a specific scenario, the existing positioning accuracy The highest value of the positioning accuracy parameter is selected from all operational sensing nodes within the target area. The positioning accuracy parameter is the accuracy parameter field in the capability resource item of each sensing node, and the positioning accuracy gap coefficient is also considered. ,in, For demand capability set The required positioning accuracy is expressed in meters, representing the positioning error. Given the current positioning accuracy, the highest value of the positioning accuracy parameter among all operational sensing nodes within the target area is selected. Since a smaller positioning error indicates higher accuracy, the reciprocal of the positioning accuracy is used in the calculation of the normalized difference. hour, A value of 0 indicates that the current positioning accuracy already meets the requirements.

[0084] In a specific scenario, the existing sampling refresh rate The highest value of the sampling refresh rate parameter is selected from all operational sensing nodes within the target area. This parameter originates from the refresh rate field within the accuracy parameter or detection range parameter field of each sensing node's capability resource item. (Sampling refresh rate gap coefficient) ,in, For demand capability set The required sampling refresh rate, Given the existing sampling refresh rate, take the highest value of the sampling refresh rate parameter of all sensing nodes in working state within the target area. hour, A value of 0 indicates no sampling refresh rate gap.

[0085] The perceived demand gap vector G is calculated as the difference between the demand capacity set R and the existing capacity set C, i.e. ,in, The modal gap coefficient is a ratio of the number of modal types in demand detection that are not covered by existing sensing nodes to the total number of demand modal types. The spatial coverage gap coefficient represents the percentage of angles within the required spatial angle range that are not covered by the detection range of existing sensing nodes. The positioning accuracy gap coefficient represents the normalized difference between the required positioning accuracy and the highest positioning accuracy achievable by existing sensing nodes. The sampling refresh rate gap coefficient represents the normalized difference between the required sampling refresh rate and the highest sampling refresh rate achievable by existing sensing nodes. This difference occurs when the sensing demand gap vector... When the value of any gap coefficient exceeds the preset gap tolerance threshold of the corresponding dimension, the resource orchestration module determines that there is a perception demand gap in the target area of ​​the corresponding dimension.

[0086] The gap tolerance threshold can be configured according to the security level requirements of different communities. In one embodiment, the detection modal gap coefficient... The gap tolerance threshold is set to 0, meaning that any modal absence will trigger resource requisition; the spatial coverage gap coefficient The notch tolerance threshold is set to 0.1, indicating that a 10% angular blind zone is allowed; the positioning accuracy notch coefficient... The notch tolerance threshold is set to 0.2; the sampling refresh rate notch coefficient is... The gap tolerance threshold is set to 0.15.

[0087] After calculating the perceived demand gap vector, the resource orchestration module filters candidate resource items from the perceived resource pool. The filtering process sets a first filtering rule and a second filtering rule.

[0088] The first screening rule is as follows: select the sensing nodes whose current occupation status field is idle, sort them according to the spatial distance between their spatial coordinates and the geometric center of the target area, and select the first m sensing nodes with the smallest spatial distance as candidate resource items. m is the minimum number of nodes that meet the minimum coverage requirement of the sensing demand gap vector. The minimum coverage requirement is that at least one candidate sensing node can fill each sensing dimension with a gap.

[0089] The second screening rule is as follows: When the number of sensing nodes that meet the first screening rule is insufficient to cover the sensing demand gap vector, from the sensing nodes whose current occupancy status field is "occupied," the sensing nodes performing tasks in the "occupied" state are sorted according to their priority. Sensing nodes with lower priority tasks are selected sequentially and added to the candidate resource items until the number of candidate resource items meets the minimum requirement for covering the sensing demand gap vector. The task priority order from highest to lowest is: continuous tracking of confirmed intrusion targets is the highest priority; focused surveillance of high-threat areas is the second priority; auxiliary surveillance of low-threat areas is the third priority; and routine wide-area inspection tasks are the lowest priority.

[0090] After identifying candidate resources, the resource orchestration module generates a temporary focus cluster configuration scheme. This scheme includes a list of requisitioned nodes and a reconstruction function role corresponding to each requisitioned sensing node. When generating the temporary focus cluster configuration scheme, the resource orchestration module assigns a narrow-domain tracking mode as the narrow-domain tracking role for at least one requisitioned sensing node in the requisitioned node list, and assigns a cooperative cross-location mode as the cooperative cross-location role for at least another requisitioned sensing node, ensuring that the temporary focus cluster simultaneously possesses both a narrow-domain tracking role and a cooperative cross-location role. The narrow-domain tracking role and the cooperative cross-location role work together to form a continuous tracking and spatial positioning link for intruding targets within the target area.

[0091] In a specific scenario, the target area is grid area 12 in the eastern district of the community, with a threat level of 0.85 and a duration exceeding 5 seconds. The resource orchestration module calculates the perceived demand gap vector G = { =0.5, =0.1, =0.6, =0.2}. Wherein, =0.5 indicates a lack of high-precision tracking modes. A value of 0.6 indicates severely insufficient positioning accuracy. The resource orchestration module queries the global resource view of the sensing resource pool and finds that among the idle nodes closest to the geometric center of the target area, the PTZ-07 camera node has a narrow-field tracking mode, and the AU-03 and AU-04 acoustic array nodes have a cooperative cross-positioning mode. The resource orchestration module adds PTZ-07, AU-03, and AU-04 to the list of requisitioned nodes, assigns the reconstruction function role of PTZ-07 to the narrow-field tracking role, and assigns the reconstruction function role of AU-03 and AU-04 to the cooperative cross-positioning role. It then generates a temporary focusing cluster configuration scheme and sends it to the temporary focusing cluster execution module.

[0092] The aforementioned temporary focus cluster execution module is used to send a reconstruction instruction containing the target working mode and effective timestamp to the requisitioned nodes, so that each requisitioned sensing node can synchronously switch to the corresponding reconstruction function role and form a temporary focus cluster. After the threat level value of the target area is less than the preset release threshold and remains so for a preset period of time, a release instruction is sent to each requisitioned sensing node, so that each requisitioned sensing node can return to the working mode before requisition.

[0093] In one specific embodiment, the preset release threshold is set to 0.3, and the typical value for the preset time period is 10 seconds. After receiving the temporary focus cluster configuration scheme sent by the resource orchestration module, the temporary focus cluster execution module parses the node identifier and corresponding reconstruction function role of each requisitioned sensing node in the requisitioned node list. The temporary focus cluster execution module generates reconstruction instructions for the requisitioned sensing nodes. The reconstruction instructions include the target working mode, mode parameters, and effective timestamp. The target working mode is the working mode that the requisitioned sensing node needs to switch to, taken from the working mode corresponding to the reconstruction function role. The mode parameters are the operating parameters required by the requisitioned sensing node in the target working mode, and the effective timestamp is used to ensure that multiple requisitioned sensing nodes switch functions at the same time.

[0094] When the target's operating mode is narrow-field tracking mode, the mode parameters include the initial spatial coordinates of the target to be tracked, the target address for tracking data reporting, and the tracking data reporting frequency. When the target's operating mode is cooperative cross-location mode, the mode parameters include the cooperative cross-location task identifier, the node identifiers and network addresses of other requisitioned sensing nodes that need to cooperate with it, the network address of the time synchronization reference source, and the target address for cooperative data reporting.

[0095] The reconstruction instructions sent by the temporary focus cluster execution module to each requisitioned sensing node also include a set of coordination parameters. These parameters include the temporary focus cluster identifier, the node role number of the requisitioned sensing node within the temporary focus cluster, a list of target node identifiers requiring coordination, and the target address for reporting coordination data. After switching to the reconstruction function role, each requisitioned sensing node establishes a point-to-point data channel with other requisitioned sensing nodes within the temporary focus cluster based on the coordination parameter set, enabling direct intra-cluster sharing and collaborative processing of sensing data.

[0096] Each requisitioned sensing node synchronously switches to its corresponding reconstruction function role upon the arrival of its effective timestamp. The synchronous switching relies on the network time protocol for global clock synchronization. After each node completes the switching, it sends a switching confirmation message to the temporary focus cluster execution module. Once the temporary focus cluster execution module receives the switching confirmation messages from all nodes in the requisitioned node list, it confirms that the temporary focus cluster has been successfully constructed.

[0097] After the temporary focusing cluster is established, each requisitioned sensing node within the temporary focusing cluster performs collaborative sensing tasks according to its reconstructed functional role. In a specific embodiment, when the requisitioned sensing node designated as the collaborative cross-location role is the first acoustic array node and the second acoustic array node, the first acoustic array node and the second acoustic array node use the joint time difference of arrival (TDOA) calculation method to locate the sound source of the intruding target within the target area in the collaborative cross-location mode. The calculation steps of the joint TDOA calculation method are as follows:

[0098] A local coordinate system is established with the spatial coordinates of the first acoustic array node as the origin, and the spatial coordinates of the second acoustic array node are... ,in The baseline distance between the first and second acoustic array nodes is used, and the first acoustic array node records the arrival time of the acoustic signal emitted by the intruding target. The second acoustic array node records the arrival time of the acoustic signal emitted by the intruding target. Calculate the time difference of arrival ;

[0099] Based on the speed of sound signal propagation in air Spatial coordinates of the intrusion target sound source Satisfies the hyperbolic equation: ;

[0100] Combined with the angle of arrival of the acoustic signal measured by the first acoustic array node Angle of arrival of the acoustic signal measured by the second acoustic array node Establish the equations for the direction lines: , ;

[0101] By solving the simultaneous equations of the hyperbola and the direction line equations, the unique spatial coordinates of the sound source of the intruding target can be calculated. To complete the acoustic cross-location of intruding targets within the target area;

[0102] Within the temporary focus cluster, the camera sensing nodes designated as narrow-domain tracking roles adjust the azimuth and pitch angles of the gimbal based on the calculated unique spatial position coordinates to visually lock onto and track the intruding target.

[0103] In the aforementioned collaborative sensing process, the acoustic array nodes directly send the calculated spatial coordinates to the camera sensing nodes via point-to-point data channels. The camera sensing nodes can obtain the target's real-time coordinates without going through a central platform, thus achieving a low-latency response from acoustic localization to visual tracking. The visual tracking results from the camera sensing nodes are simultaneously transmitted back to the temporary focusing cluster execution module and the threat situation assessment module to update the threat level values ​​for the corresponding areas in the threat situation information.

[0104] When the threat situation assessment module detects that the threat level value of the target area has dropped below the preset disarmament threshold and remains below it for a preset period of time, it sends a threat disarmament notification to the resource orchestration module. The preset disarmament threshold is 0.3, and the preset period is 10 seconds. Upon receiving the threat disarmament notification, the resource orchestration module generates a release command, which is sent to each requisitioned sensing node via the temporary focus cluster execution module. After receiving the release command, each requisitioned sensing node stops its current collaborative sensing task, reverts to its pre-requisition working mode, and updates its current occupancy status field to idle. The resource virtualization middleware module synchronously updates the global resource view of the sensing resource pool.

[0105] like Figure 3 As shown, the present invention also provides a method for detecting intrusion into the perimeter of a smart community building, applicable to the aforementioned intelligent perimeter protection system for smart communities. The specific steps of this method are as follows:

[0106] S100, Resource Virtualization: The resource virtualization middleware module collects the status information and capability descriptions of each heterogeneous sensing node in the sensing resource module, updates the capability resource items corresponding to each heterogeneous sensing node, and aggregates all capability resource items to form a sensing resource pool.

[0107] S200 Threat Situation Generation: The threat situation assessment module receives pre-processed data from heterogeneous sensing nodes in the sensing resource module, performs fusion processing on the pre-processed data, and generates threat situation information covering all areas of the community building perimeter in real time, and calibrates the threat level values ​​of each gridded area.

[0108] S300, Focus Decision and Solution Generation: The resource orchestration module polls threat situation information. When it is found that the threat level value of the first region continuously exceeds the preset threat threshold within a preset time window, the first region is determined as the target region. The sensing demand gap vector of the target region is calculated, the candidate sensing node set is matched, the list of requisitioned nodes without resource conflicts is determined, and a temporary focus cluster configuration scheme containing the requisitioned nodes and reconstruction function roles is generated.

[0109] S400, Cluster Construction and Functional Reconstruction: The temporary focused cluster execution module sends a reconstruction instruction containing the target working mode, mode parameters and effective timestamp to each requisitioned sensing node in the requisitioned node list. Each requisitioned sensing node synchronously switches to the corresponding reconstruction function role at the effective timestamp, forming a temporary focused cluster.

[0110] S500, Collaborative Enhanced Perception: Each requisitioned perception node within the temporary focus cluster performs collaborative perception tasks according to its reconstructed functional role, continuously tracks, spatially locates, and identifies the behavior of intrusion targets within the target area, and reports the collaborative perception results to the threat situation assessment module to update the threat situation information.

[0111] S600, Resource Release: When the threat situation assessment module detects that the threat level value of the target area is less than the preset release threshold and remains so for a preset period of time, it generates a release command, which is sent to each requisitioned sensing node via the temporary focus cluster execution module. Each requisitioned sensing node is restored to the working mode before requisition and the current occupied status field is updated to idle status.

[0112] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention in any way. Although the present invention has been disclosed above with reference to preferred embodiments, it is not intended to limit the present invention. Any person skilled in the art can make some modifications or alterations to the above-disclosed technical content to create equivalent embodiments without departing from the scope of the present invention. Any simple modifications, equivalent changes and alterations made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the scope of the present invention.

Claims

1. A smart community building perimeter intelligent protection system, characterized in that, The system consists of: The sensing resource module is used to deploy various heterogeneous sensing nodes around the perimeter of community buildings. Each heterogeneous sensing node has a current working mode and a set of switchable working modes. The resource virtualization middleware module is used to treat the sensing capabilities of each heterogeneous sensing node as a unified description of the capability resource item, and to aggregate the capability resource items of all heterogeneous sensing nodes to form a sensing resource pool. The threat situation assessment module is used to fuse and process multi-source data returned by the perception resource module to generate threat situation information covering all areas of the community building perimeter in real time. The resource orchestration module is used to determine the first region as the target region when the threat level value of the first region is continuously exceeded within a preset time window. It calculates the perception demand gap vector of the target region, filters candidate resource items from the perception resource pool that are spatially adjacent and whose switchable working mode sets match the perception demand gap vector, and generates a temporary focus cluster configuration scheme that includes a list of requisitioned nodes and the reconstruction function role corresponding to each requisitioned perception node in the list of requisitioned nodes. The temporary focus cluster execution module is used to send a reconstruction instruction containing the target working mode and the effective timestamp to the requisitioned nodes, so that each requisitioned sensing node synchronously switches to the corresponding reconstruction function role and forms a temporary focus cluster. After the threat level value of the target area is less than the preset release threshold and remains so for a preset period of time, a release instruction is sent to each requisitioned sensing node, so that each requisitioned sensing node returns to the working mode before requisition.

2. The intelligent perimeter protection system for smart communities according to claim 1, characterized in that, The heterogeneous sensing nodes include multiple types such as camera nodes, radar nodes, acoustic array nodes, and light curtain nodes; The capability resource items include node identifier, spatial location coordinates, current working mode, set of switchable working modes, detection range parameters, accuracy parameters, and current occupancy status field. 3.The intelligent community building perimeter intelligent protection system according to claim 2, characterized in that, The set of switchable working modes in the capability resource item includes wide-area inspection mode, narrow-area tracking mode, cooperative cross-positioning mode, spectrum detection mode, and sleep mode; The wide-area inspection mode is a working mode in which the sensing node performs a cyclic scan of the allocated wide-angle field of view at a preset cycle. The narrow-field tracking mode is a working mode in which the sensing node continuously locks onto a specified target at a high frame rate and narrow field of view. The collaborative cross-positioning mode is a working mode in which two or more sensing nodes perform time-synchronized angle-of-arrival joint calculation to determine the spatial location of the target. The spectrum detection mode is a working mode in which the sensing node performs directional detection of electromagnetic signals in a preset frequency band. The sleep mode is a low-power standby mode in which the sensing node disables the active detection function and only maintains the control link listening. 4.The intelligent community building perimeter intelligent protection system according to claim 1, characterized in that, The threat situation information includes a gridded area index, the threat level value of the corresponding area, and the threat evolution trend prediction information, and the threat situation information is output to the resource orchestration module; When generating the threat situation information, the threat situation assessment module uses an attention-weighted graph network to perform gridded threat probability calculation on the perimeter of the community buildings; The attention-weighted graph network uses each gridded area of ​​the community building perimeter as graph nodes, the spatial adjacency relationship between adjacent gridded areas and the detection overlap relationship of heterogeneous sensing nodes as graph edges, and the preprocessed event stream returned by each heterogeneous sensing node as node input features. Through a multi-head attention mechanism, the event features of multiple heterogeneous sensing nodes in the same gridded area are weighted and aggregated to output the threat probability value of each gridded area. The threat situation assessment module also calculates threat evolution trend prediction information based on the threat probability value sequence of multiple consecutive time steps using the exponentially weighted moving average method. 5.The intelligent community building perimeter intelligent protection system according to claim 1, characterized in that, When generating the temporary focusing cluster configuration scheme, the resource orchestration module assigns a narrow-domain tracking mode as the narrow-domain tracking role of the reconstructing functional role to at least one requisitioned sensing node in the requisitioned node list, and assigns a cooperative cross-location mode as the cooperative cross-location role of the reconstructing functional role to at least another requisitioned sensing node, so that the temporary focusing cluster simultaneously has a narrow-domain tracking role and a cooperative cross-location role. The narrow-domain tracking role and the cooperative cross-location role work together to form a continuous tracking and spatial positioning link for the intruding target in the target area. 6.The intelligent community building perimeter intelligent protection system according to claim 1, characterized in that, When filtering candidate resource items, the resource orchestration module sets a first filtering rule and a second filtering rule. The first filtering rule is as follows: select the sensing nodes whose current occupancy status field is idle, sort them according to the spatial distance between their spatial location coordinates and the geometric center of the target area, and select the first m sensing nodes with the smallest spatial distance as candidate resource items, where m is the minimum number of nodes that meet the minimum coverage requirement of the sensing demand gap vector. The second filtering rule is as follows: when the number of sensing nodes that meet the first filtering rule is insufficient to cover the sensing demand gap vector, from the sensing nodes whose current occupancy status field is occupied, sort them according to the priority of the tasks performed by the occupied sensing nodes, and sequentially select sensing nodes with low priority tasks, and add the selected sensing nodes to the candidate resource items until the number of candidate resource items meets the minimum requirement to cover the sensing demand gap vector. 7.The intelligent community building perimeter intelligent protection system according to claim 1, characterized in that, The resource orchestration module calculates the perceived demand gap vector for the target area in the following way: The set of full-coverage sensing capabilities required for the target area is taken as the demand capability set. The set of required capabilities It includes the detection modality dimension, spatial coverage dimension, positioning accuracy dimension, and sampling refresh rate dimension required for effective protection of the target area; The set of sensing capabilities actually provided by the sensing nodes that are currently deployed and operational in the target area is taken as the existing capability set. ; Calculate the perceived demand gap vector For demand capability set With existing capabilities The difference set, i.e. ,in, The modal gap coefficient is a ratio of the number of modal types in demand detection that are not covered by existing sensing nodes to the total number of demand modal types. The spatial coverage gap coefficient represents the percentage of angles within the required spatial angle range that are not covered by the detection range of existing sensing nodes. The positioning accuracy gap coefficient represents the normalized difference between the required positioning accuracy and the highest positioning accuracy achievable by existing sensing nodes. The sampling refresh rate gap coefficient represents the normalized difference between the required sampling refresh rate and the highest sampling refresh rate achievable by existing sensing nodes. This difference occurs when the sensing demand gap vector... When the value of any gap coefficient exceeds the preset gap tolerance threshold of the corresponding dimension, the resource orchestration module determines that there is a perception demand gap in the target area of ​​the corresponding dimension. 8.The intelligent community building perimeter intelligent protection system according to claim 1, characterized in that, In the temporary focusing cluster execution module, when the requisitioned sensing nodes designated as cooperative cross-location roles within the temporary focusing cluster are the first acoustic array node and the second acoustic array node, the first acoustic array node and the second acoustic array node use a joint time difference of arrival (TDOA) calculation method to locate the sound source of the intruding target within the target area in the cooperative cross-location mode. The calculation steps of the joint TDOA calculation method are as follows: A local coordinate system is established with the spatial position coordinates of the first acoustic array node as the origin, and the spatial position coordinates of the second acoustic array node as wherein is a baseline distance between the first acoustic array node and the second acoustic array node; The first acoustic array node records the arrival time of the sound signal emitted by the intrusion target The second acoustic array node records the arrival time of the sound signal emitted by the intrusion target The arrival time difference is calculated ; Based on the speed of sound signal propagation in air Spatial coordinates of the intrusion target sound source Satisfies the hyperbolic equation: Combined with the angle of arrival of the acoustic signal measured by the first acoustic array node and the angle of arrival of the acoustic signal measured by the second acoustic array node Establish the equations for the direction lines: , ; By solving the simultaneous equations of the hyperbola and the direction line equations, the unique spatial coordinates of the sound source of the intruding target can be calculated. Acoustic cross-location of intruding targets within the target area is completed; Within the temporary focusing cluster, the camera sensing node designated as the narrow-domain tracking role adjusts the gimbal pointing according to the calculated unique spatial position coordinates to perform visual locking and tracking of the intrusion target.

9. The method for detecting intrusion of the perimeter of the smart community building, applicable to the intelligent protection system for the perimeter of the smart community building according to any one of claims 1-8, characterized in that, The specific steps of this method are as follows: S100, Resource Virtualization: The resource virtualization middleware module collects the status information and capability descriptions of each heterogeneous sensing node in the sensing resource module, updates the capability resource items corresponding to each heterogeneous sensing node, and aggregates all capability resource items to form a sensing resource pool. S200 Threat Situation Generation: The threat situation assessment module receives pre-processed data from heterogeneous sensing nodes in the sensing resource module, performs fusion processing on the pre-processed data, and generates threat situation information covering all areas of the community building perimeter in real time, and calibrates the threat level values ​​of each gridded area. S300, Focus Decision and Solution Generation: The resource orchestration module polls the threat situation information. When it is identified that the threat level value of the first region continuously exceeds the preset threat threshold within a preset time window, the first region is determined as the target region. The sensing demand gap vector of the target region is calculated, the candidate sensing node set is matched, the list of requisitioned nodes without resource conflicts is determined, and a temporary focus cluster configuration scheme containing requisitioned nodes and reconstructed functional roles is generated. S400, Cluster Construction and Functional Reconstruction: The temporary focused cluster execution module sends a reconstruction instruction containing the target working mode, mode parameters and effective timestamp to each requisitioned sensing node in the requisitioned node list. Each requisitioned sensing node synchronously switches to the corresponding reconstruction function role at the effective timestamp, forming a temporary focused cluster. S500, Collaborative Enhanced Perception: Each requisitioned perception node within the temporary focusing cluster performs collaborative perception tasks according to its reconstructed functional role, continuously tracks, spatially locates, and identifies the behavior of intrusion targets within the target area, and reports the collaborative perception results to the threat situation assessment module to update the threat situation information; S600, Resource Release: When the threat situation assessment module detects that the threat level value of the target area is less than the preset release threshold and remains so for a preset period of time, it generates a release command, which is sent to each requisitioned sensing node via the temporary focusing cluster execution module. Each requisitioned sensing node is restored to its working mode before requisition and updates the current occupied status field to idle status.

Images