Artificial intelligence-based community security situation awareness system and method

Abstract

The application relates to the field of artificial intelligence and multi-modal security monitoring technology, and discloses a community security situation awareness system and method based on artificial intelligence, which comprises a perception baseline construction module, an adaptive scheduling module, a feature extraction module, a divergence verification module and a situation judgment module; the perception baseline construction module synchronously extracts an underground optical phase change sequence and a ground channel state information complex response matrix; the adaptive scheduling module calculates an instantaneous expected angle of arrival based on a time delay difference, controls an antenna array to realize directional focusing of radio frequency energy, the feature extraction module extracts a microseismic joint feature set representing gravity load and a Doppler frequency band feature set representing spatial volume, the divergence verification module converts the two types of feature sets into a probability density distribution space matrix and quantifies a space-time topology divergence, and the situation judgment module outputs security situation early warning information according to the space-time topology divergence, thereby solving the problems of a perception blind area of a single physical sensor and identification misjudgment.

Description

Technical Field

[0001] This invention relates to the fields of artificial intelligence and multimodal security monitoring technology, specifically to a community security situation awareness system and method based on artificial intelligence. Background Technology

[0002] Conventional community security monitoring systems typically deploy ground-based wireless radio frequency sensor networks and underground physical detection equipment independently. When faced with dynamic intrusion targets, traditional wireless radio frequency sensor networks often rely on a traversal global polling scanning mechanism to capture the target's trajectory. This conventional global polling scanning mechanism not only generates significant system communication delays but also causes limited radio frequency energy to be dispersed in the background clutter environment without targets, resulting in severely insufficient sensing accuracy when extracting the micro-Doppler scattering cross-section features of physical targets.

[0003] In the process of introducing multiple physical sensors to try to improve the perception dimension, the existing security monitoring system has obvious limitations at the underlying data fusion level. The vibration signal generated by the underground microseismic physical field and the electromagnetic wave signal generated by the ground space radio frequency physical field belong to completely different physical modes. The existing fusion technology has failed to effectively eliminate the dimensional differences of different modal data at the original feature level, which makes it impossible for the system to uniformly map discrete physical sensing data to a continuous set of spatial coordinate grids, making it difficult to establish a mathematical comparison benchmark for heterogeneous sensing data in a unified spatiotemporal coordinate system.

[0004] Traditional security monitoring systems that rely on single physical feature judgment logic have serious perception blind spots when dealing with complex intrusion scenarios. When physical targets are in environments with deep spatial obstruction or adopt disguised infiltration behavior that shields electromagnetic detection, ground radio frequency detection equipment is very prone to missed detection. Furthermore, in the event of multiple physical targets moving closely together to prevent tailgating, ground radio frequency detection equipment often misjudges multiple entities as a single target due to the overlapping radio frequency sensing volumes of the physical targets. Traditional systems cannot accurately quantify and identify the degree of contradiction in the distribution of underlying features between different modalities, ultimately leading to a significant reduction in the reliability and accuracy of community security situational awareness. Summary of the Invention

[0005] To address the shortcomings of existing technologies, this invention provides an artificial intelligence-based community security situation awareness system and method, which solves the problems of excessive communication latency and insufficient feature extraction accuracy caused by the reliance on global polling scanning in existing wireless radio frequency sensor networks.

[0006] To achieve the above objectives, the present invention provides the following technical solution: The first aspect of the present invention provides an artificial intelligence-based community security situation awareness system, including a perception baseline construction module, an adaptive scheduling module, a feature extraction module, a divergence verification module, and a situation determination module.

[0007] The sensing baseline construction module is configured to simultaneously construct a heterogeneous continuous physical sensing baseline. It injects probe light pulses into the underground single-mode communication fiber via a phase-sensitive optical time-domain reflectometer (OTDR) and extracts the optical phase change sequence based on the local geometric deformation caused by microseismic stress waves and the refractive index perturbation caused by the elasto-optic effect. Simultaneously, the module extracts the complex response matrix of the channel state information under the orthogonal frequency division multiplexing (OFDM) architecture from the terrestrial wireless access point. This complex response matrix records the instantaneous amplitude attenuation factor and electromagnetic wave propagation delay of the multipath propagation path.

[0008] The adaptive scheduling module performs cross-correlation calculations based on the optical phase change sequence generated by adjacent fiber demodulation channels when receiving the same microseismic stress wave to extract the time delay difference. It then calculates the instantaneous expected angle of arrival (Angle of Arrival) of the microseismic source direction by combining the Rayleigh surface wave propagation phase velocity and generates an antenna steering vector. The adaptive scheduling module embeds a cross-domain front-end controller, which converts the instantaneous expected Angle of Arrival (Angle of Arrival) into spatial coordinate data format and encapsulates the highest priority execution flag in the beam scheduling command's frame header structure. This beam scheduling command is injected into the first position of the medium access control layer (MAP) buffer queue of the terrestrial wireless access point, allowing the MAP to directly set the underlying transmission state of the physical layer RF transceiver hardware. The baseband processing unit configured within the terrestrial wireless access point obtains the environmental RF signal sample covariance matrix by calculating the expected value of the product of the channel state information complex response matrix and its conjugate transpose. This covariance is then combined with the minimum variance distortionless response beamforming criterion to update the transmit weight vector of the multi-input multi-output antenna array, controlling the dynamic alignment of the RF main lobe null direction with the spatial coordinate grid calculated from the instantaneous expected Angle of Arrival.

[0009] The feature extraction module uses a sliding time window to perform a short-time Fourier transform on the optical phase change sequence, converting it into a time-frequency distribution matrix. It then uses the cutoff frequency parameter to perform frequency band segmentation to separate low-frequency and high-frequency components. It extracts the first energy spectral density sequence mapping the gravity load distribution and the discrete peak time interval data mapping the gait period, integrating them to generate a joint microseismic feature set. Simultaneously, the feature extraction module combines the instantaneous radial velocity of the physical target, the electromagnetic wavelength of the radio frequency signal, and the spatial incident angle to calculate the micro-Doppler frequency shift generated by the radio frequency subcarrier within the beam focusing region. It then aggregates the micro-Doppler frequency shift sets fed back from the multipath propagation links of the radio frequency signal within the spatial coordinate grid to construct a Doppler band feature set.

[0010] The divergence verification module uses bilinear spatial interpolation and threshold segmentation algorithms to convert the joint microseismic feature set and Doppler band feature set into the spatial matrix of the target probability density distribution of microseismic inversion and the spatial matrix of the target probability density distribution of radio frequency inversion, respectively. After injecting a preset system background smoothing factor into the matrix elements of the target probability density distribution spatial matrix of radio frequency inversion, the divergence verification module uses relative entropy calculation rules to quantify the spatiotemporal topological divergence between the two types of probability density distribution spatial matrices.

[0011] The situation assessment module outputs security situation warning information based on the comparison between the spatiotemporal topological divergence calculation results and the preset background threshold. When the spatial matrix of the target probability density distribution of microseismic inversion contains high-value features of continuous periodic gait signals and the matrix element values ​​of the spatial matrix of the target probability density distribution of radio frequency inversion are lower than the radio frequency background noise limit, under the condition that the spatiotemporal topological divergence is greater than the preset first divergence threshold, the situation assessment module determines that the physical target is in a camouflaged state shielded from electromagnetic detection or in a deep spatial obstruction environment, and generates an implicit intrusion warning command. When the situation assessment module resolves two independent step frequency phase parameters with spatial separation characteristics based on the spatial matrix of the target probability density distribution of microseismic inversion and the spatial matrix of the target probability density distribution of radio frequency inversion is fitted to a single connected radio frequency volume envelope, under the condition that the temporal gradient change rate corresponding to the spatiotemporal topological divergence is greater than the preset second dynamic fluctuation threshold, the situation assessment module determines that multiple physical targets have moved closely together and generates an anti-tailgating intrusion warning command.

[0012] A second aspect of the present invention provides an artificial intelligence-based community security situation awareness method, applied to the artificial intelligence-based community security situation awareness system provided in the first aspect of the present invention. The artificial intelligence-based community security situation awareness method includes: S100, the sensing baseline construction module continuously and synchronously extracts the optical phase change sequence and the complex response matrix of the channel state information; S200, when the change amplitude of the optical phase change sequence exceeds the preset background threshold, the adaptive scheduling module calculates the instantaneous expected angle of arrival of the micro-vibration source direction based on the propagation phase velocity of the micro-vibration signal and the spatial time delay difference of the adjacent fiber demodulation channels and generates the antenna steering vector. The cross-domain front-end controller updates the transmit weight vector of the multi-input multi-output antenna array based on the antenna steering vector. S300, the feature extraction module extracts the Doppler frequency band feature set within the adjusted spatial coordinate grid, and the feature extraction module simultaneously extracts the joint microseismic feature set; S400, the divergence verification module performs feature fusion operations in a unified spatiotemporal coordinate system to calculate the spatiotemporal topological divergence between the spatial matrices of two types of probability density distributions; The S500 situation determination module determines the detection result as a safe passage, a hidden intrusion, or an anti-tailgating abnormal state based on the comparison between the spatiotemporal topological divergence and the preset background threshold, and outputs the corresponding warning command and tracking trajectory to the terminal through the communication interface.

[0013] This invention provides a community security situation awareness system and method based on artificial intelligence. It has the following beneficial effects: 1. This invention synchronously extracts the complex response matrix of underground optical phase change sequence and ground channel state information through the sensing baseline construction module, and calculates the instantaneous expected angle of arrival of microseismic source direction by the adaptive scheduling module based on the spatial delay difference of the fiber demodulation channel. The cross-domain front-end controller generates antenna steering vector based on the instantaneous expected angle of arrival to update the transmit weight vector of the multi-input multi-output antenna array. This cross-mode hardware-level linkage mechanism controls the directional focusing of radio frequency energy on the spatial coordinate grid where the target is located, avoiding the communication delay caused by the global polling scan of conventional radio frequency networks, and improving the extraction accuracy of the micro-Doppler scattering cross-section features of physical targets.

[0014] 2. This invention utilizes a feature extraction module to separate the first energy spectral density sequence mapping gravity load distribution and the discrete peak time interval data mapping biomechanical gait from optical signals. Simultaneously, it extracts the Doppler frequency band feature set of the three-dimensional spatial volume of the mapped physical target based on the radio frequency echo signal. The divergence verification module maps the discrete physical sensing data of different dimensions to a continuous spatial coordinate grid set and converts them into the spatial matrix of probability density distribution of the target obtained by microseismic inversion and the spatial matrix of probability density distribution of the target obtained by radio frequency inversion. Feature transformation and normalization processing eliminate the dimensional differences between the underground microseismic physical field and the ground spatial radio frequency physical field at the original feature level, and establishes a mathematical comparison benchmark for heterogeneous sensing data in a unified spatiotemporal coordinate system.

[0015] 3. This invention relies on a divergence verification module combined with relative entropy calculation rules to quantify the spatiotemporal topological divergence between the spatial matrix of the probability density distribution of targets retrieved from microseismic inversion and the spatial matrix of the probability density distribution of targets retrieved from radio frequency inversion. The situation determination module determines the degree of contradiction in the distribution of the underlying features of physical targets based on the calculated values ​​of the spatiotemporal topological divergence. Combining the unhidden mass load characteristics of underground microseismic detection and the three-dimensional volume perception advantages of ground radio frequency detection, the system directly identifies the camouflage and infiltration behavior of physical targets shielding electromagnetic detection through threshold comparison. When the radio frequency sensing volume overlaps, it relies on the spatial independence of underground physical load to determine the anti-tailgating event of multiple physical targets moving close together, thus solving the perception blind zone defect of a single physical sensor in a deep spatial occlusion environment. Attached Figure Description

[0016] Figure 1 This is a system architecture diagram of the present invention; Figure 2 This is a flowchart of the method of the present invention. Detailed Implementation

[0017] The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0018] Please see Figures 1-2 This invention provides an artificial intelligence-based community security situation awareness system and method. The artificial intelligence-based community security situation awareness system is built upon existing underground communication optical cables and terrestrial wireless radio frequency coverage networks within the community. The artificial intelligence-based community security situation awareness system includes: The sensing baseline construction module is configured to synchronously construct a heterogeneous continuous physical sensing baseline. The sensing baseline construction module injects probe light pulses into the underground communication single-mode fiber by accessing a phase-sensitive optical time-domain reflectometer to obtain the optical phase change sequence under the corresponding spatiotemporal coordinates. At the same time, it extracts the complex response matrix of the channel state information under the orthogonal frequency division multiplexing architecture through the ground wireless access point. The adaptive scheduling module connects to the sensing baseline construction module. The adaptive scheduling module embeds a cross-domain front-end controller. The cross-domain front-end controller communicates with the phase-sensitive optical time-domain reflectometer and the ground wireless access point respectively. The adaptive scheduling module calculates the instantaneous expected angle of arrival of the microseismic source direction based on the spatial time delay difference of adjacent fiber demodulation channels and generates an antenna steering vector. The cross-domain front-end controller updates the transmit weight vector of the multi-input multi-output antenna array based on the antenna steering vector and controls the dynamic alignment of the RF main lobe null direction with the spatial coordinate grid obtained by calculating the instantaneous expected angle of arrival. The feature extraction module is connected to the sensing baseline construction module. The feature extraction module performs time-frequency domain transformation on the optical phase change sequence to extract a microseismic joint feature set that characterizes the distribution of gravity load and gait frequency band. At the same time, based on the micro-Doppler frequency shift obtained by the radio frequency subcarrier in the beam focusing area, it extracts a Doppler frequency band feature set that characterizes the three-dimensional spatial volume of the target. The divergence verification module is connected to the feature extraction module. The divergence verification module converts the joint microseismic feature set and the Doppler frequency band feature set into normalized probability density distribution space matrices respectively, and uses the relative entropy variant operation rules to quantify the spatiotemporal topological divergence between the two types of probability density distribution space matrices. The situation assessment module is connected to the divergence verification module. The situation assessment module outputs security situation warning information based on the comparison between the spatiotemporal topological divergence calculation results and the preset background threshold.

[0019] The workflow of the AI-based community security situation awareness system is executed collaboratively by the aforementioned functional modules. After system startup, the perception baseline construction module continuously and synchronously extracts the optical phase change sequence and the complex response matrix of channel state information. When the change amplitude of the optical phase change sequence exceeds the preset background threshold, the adaptive scheduling module calculates the instantaneous expected angle of arrival of the ground target based on the propagation phase velocity of the microseismic signal and the spatial delay difference of adjacent fiber demodulation channels. The cross-domain front-end controller adjusts the weights of the multi-input multi-output antenna array of the ground wireless access point accordingly. The feature extraction module extracts the Doppler band feature set in the adjusted spatial coordinate grid and simultaneously extracts the joint microseismic feature set. The divergence verification module performs feature fusion operations in a unified spatiotemporal coordinate system to calculate the spatiotemporal topological divergence between the two types of probability density distribution spatial matrices. Finally, the situation determination module determines the detection result as a safe passage, covert intrusion, or anti-tailgating anomaly state based on the comparison between the spatiotemporal topological divergence and the preset background threshold, and outputs corresponding warning commands and tracking trajectories to the terminal via the communication interface.

[0020] The working principle of the system of this invention will be explained in detail below, in conjunction with the specific execution logic of each functional module: The sensing baseline construction module establishes a connection with the underground single-mode fiber optic communication network deployed in the community via a phase-sensitive optical time-domain reflectometer (OTDR), and controls the ORT to continuously inject a center wavelength of [missing value] into the underground single-mode fiber optic communication network. High coherence probe light pulses, when transmitted inside underground single-mode optical fibers, generate Rayleigh scattering light based on the inhomogeneous structure of the fiber core medium.

[0021] When there are physical activities such as human activity or vehicle movement on the surface of the community, the gravity and mechanical impact exerted by these activities on the surface medium generate micro-seismic stress waves that are transmitted underground. These micro-seismic stress waves penetrate the geological medium and act on the outer sheath and core structure of the underground single-mode optical fiber, forcing the fiber to produce axial strain and elastic-optical effect. The axial strain and elastic-optical effect directly change the physical interference conditions of Rayleigh scattered light.

[0022] The sensing baseline construction module receives Rayleigh scattered light propagating backward along the underground single-mode communication fiber and converts it into an electrical signal using a photodetector. The sensing baseline construction module then demodulates the electrical signal to extract the spatial coordinates of the underground single-mode communication fiber. And the time is Optical phase change sequence Optical phase change sequence The mathematical relationship between them satisfies the following operational formula: ; In the above calculation formula, Represents the background refractive index of the fiber core. This represents the spatial demodulation gauge length set at the optical receiver of the phase-sensitive optical time-domain reflectometer. This represents the local geometric deformation caused by the action of microseismic stress waves in underground single-mode optical fibers for communication. This represents the amount of refractive index disturbance in the fiber core medium caused by the elasto-optic effect.

[0023] Optical phase change sequence The sensing baseline construction module extracts the optical phase change sequence from the continuous fluctuation waveform data of the surface physical target active payload in the time domain. The data is stored and transmitted to the adaptive scheduling module as a priori physical data reference for triggering the operation of the adaptive scheduling module.

[0024] While constructing the underground microseismic physical baseline, the sensing baseline construction module simultaneously initiates the construction of the ground radio frequency spatial physical baseline. The sensing baseline construction module establishes a communication connection with wireless access points distributed in the community's ground space. The sensing baseline construction module controls the wireless access points to extract the underlying radio frequency link status data under the orthogonal frequency division multiplexing communication architecture. The baseband processing unit configured inside the wireless access point analyzes the amplitude attenuation and phase shift caused by the radio frequency carrier passing through the ground space medium during the transmission and reception of radio frequency detection data packets.

[0025] The sensing baseline construction module obtains the complex response matrix of the channel state information output by the physical layer of the baseband processing unit configured inside the wireless access point. Channel state information complex response matrix Records the channel fading characteristics of radio frequency signals caused by the multipath propagation environment consisting of buildings and ground media within the community environment, and the complex response matrix of the channel state information. The mathematical relationship between them satisfies the following operational formula: ; In the above calculation formula, This represents the radio frequency subcarrier frequency currently used by the wireless access point's transceiver link. This represents the total number of independent multipath propagation paths in the ground space environment. This represents the index of a single independent propagation path out of the total number of multipath propagation paths. Representing the The radio frequency signal of the multipath propagation path is in time. Instantaneous amplitude decay factor at time Representing the The propagation delay of radio frequency signals in space via multiple path propagation.

[0026] Complex response matrix of channel state information extracted by the sensing baseline building module Reflecting the background state of the electromagnetic wave scattering cross section within the ground space, the sensing baseline construction module uses the complex response matrix of channel state information. Synchronously transmitted to the adaptive scheduling module and feature extraction module, the complex response matrix of channel state information. The initial environmental RF signal sample covariance data of the multi-input multi-output antenna array is provided. The feature extraction module utilizes the displacement of the physical target in the complex response matrix of the channel state information in subsequent processing. The target's three-dimensional spatial volume is calculated using the radio frequency subcarrier micro-Doppler frequency shift component generated by the mapping.

[0027] The sensing baseline construction module continuously extracts optical phase change sequences. During operation, the optical phase change sequence Real-time transmission to the adaptive scheduling module, when the optical phase change sequence When the change exceeds the preset background threshold, the adaptive scheduling module determines that there is a physical target intrusion event in the community surface area. The adaptive scheduling module simultaneously intercepts the physical data of multiple fiber demodulation channels associated within the spatial coordinate area that triggered the preset background threshold condition.

[0028] The adaptive scheduling module processes the optical phase change sequence generated by adjacent first and second fiber demodulation channels when receiving the same micro-vibration stress wave. By performing cross-correlation calculations, the adaptive scheduling module extracts the time delay difference between the arrival of the micro-seismic stress wave at the first fiber demodulation channel and the second fiber demodulation channel. Time delay difference The propagation time difference of microseismic stress waves from the seismic source to different fiber optic physical sensing nodes is mapped, and the adaptive scheduling module retrieves the pre-set Rayleigh surface wave propagation phase velocity under the current geological medium of the community. Rayleigh surface wave propagation phase velocity Provides reference dynamic parameters for the propagation of microseismic stress waves in the Earth's surface medium.

[0029] The adaptive scheduling module is based on the time delay difference. and Rayleigh surface wave propagation phase velocity A wavefield vector solution model is constructed, and the adaptive scheduling module calculates the instantaneous expected angle of arrival of the source of the microseismic stress wave relative to the line connecting the first and second fiber demodulation channels. Instantaneous expected angle The mathematical relationship between them satisfies the following operational formula: ; In the above calculation formula, This represents the spatial location index of the first fiber demodulation channel. This represents the spatial location index of the second fiber demodulation channel. This represents the physical linear spacing between the first fiber demodulation channel and the second fiber demodulation channel.

[0030] Instantaneous expected angle of arrival The spatial incident direction of the microseismic stress wave is defined, and the adaptive scheduling module is based on the instantaneous expected angle of arrival. Establish the spatial pointing vector of the seismic source that generates microseismic stress waves, and the adaptive scheduling module will determine the instantaneous expected angle of arrival. And a spatial pointing vector input cross-domain front-end controller, which utilizes the instantaneous expected angle of arrival. Guide the wireless access point to perform radio frequency energy beam focusing on the spatial coordinate grid of the physical target.

[0031] The cross-domain front-end controller receives the instantaneous expected angle of arrival from the adaptive scheduling module. and based on instantaneous expected angle of arrival Established antenna steering vector The cross-domain front-end controller will instantaneously expect to reach the angle The data is converted to a spatial coordinate data format compatible with the underlying wireless network communication protocol. The cross-domain front-end controller establishes a direct control plane communication link with the baseband processing unit configured inside the wireless access point, and sends data carrying the spatial coordinate data format and antenna steering vector to the baseband processing unit through the underlying hardware interface. Beam scheduling instructions.

[0032] The beam scheduling command embeds the radio frequency subcarrier allocation weights and corresponding antenna steering vectors based on the spatial coordinate grid of the physical target. The cross-domain front-end controller encapsulates the highest priority execution flag in the frame header structure of the beam scheduling command and directly injects the beam scheduling command into the first position of the media access control layer buffer queue of the wireless access point. The cross-domain front-end controller controls the wireless access point to bypass channel contention and polling scanning workflow. The cross-domain front-end controller directly sets the underlying transmission state of the physical layer RF transceiver hardware through the media access control layer.

[0033] The baseband processing unit configured inside the wireless access point parses the beam scheduling command and extracts the antenna steering vector. The baseband processing unit synchronously retrieves the complex response matrix of the channel state information extracted by the sensing baseline construction module in the initial state. The baseband processing unit calculates the complex response matrix of the channel state information. Complex response matrix with channel state information The expected value of the product of the conjugate transposes of the matrix is ​​used to obtain the covariance matrix of the environmental radio frequency signal samples corresponding to the ground space region. Environmental radio frequency signal sample covariance matrix Characterizes the background electromagnetic interference distribution and static multipath channel characteristics of physical targets before they enter the spatial coordinate grid.

[0034] The baseband processing unit uses antenna steering vectors As a constraint on the desired signal propagation direction, and based on the minimum variance distortionless response beamforming criterion, an RF weighting matrix equation for a multi-input multi-output (MIMO) antenna array is established. The baseband processing unit solves the RF weighting matrix equation and outputs the transmit weight vector of the MIMO antenna array corresponding to the current spatial coordinate grid. emission weight vector The mathematical relationship between them satisfies the following operational formula: ; In the above calculation formula, Represents the covariance matrix of environmental radio frequency signal samples The inverse matrix, superscript This represents performing a conjugate transpose operation on a vector or matrix, emitting a weight vector. It includes the amplitude adjustment parameters and baseband phase offset parameters required to configure each independent RF transmitting element in a multi-input multi-output antenna array.

[0035] The baseband processing unit will solve the transmit weight vector. The transmit weight vector of the MIMO antenna array is updated based on the digital-to-analog converter hardware registers at the bottom layer of the RF transceiver link. Adjusting the electromagnetic wave interference superposition state of the transmitted signal in physical space, a multi-input multi-output antenna array generates a direction pointing to the instantaneous desired angle of arrival. The high-gain RF main lobe of the physical grid is used to construct a spatial radiation null in the direction of background clutter interference. The directional reconstruction mechanism of spatial RF energy is used to actively enhance the RF micro-Doppler scattering cross-section characteristics within the spatial coordinate grid of the physical target.

[0036] After the multi-input multi-output antenna array completes the underlying transmission status update, the baseband processing unit sends an instruction execution confirmation message back to the cross-domain front-end controller through the control plane direct communication link. The cross-domain front-end controller receives the instruction execution confirmation message and performs timestamp association assignment between the optical microseismic signal extraction clock and the radio frequency communication sampling clock at the hardware level. The cross-domain front-end controller establishes the time dimension synchronization alignment benchmark between the underground microseismic physical sensing field and the ground space radio frequency sensing field through timestamp association assignment.

[0037] The feature extraction module receives the optical phase change sequence extracted by the sensing baseline construction module under the condition of triggering a preset background threshold. And a sliding time window of preset length is used to analyze the sequence of optical phase changes in the time domain. The feature extraction module performs a short-time Fourier transform operation to convert the time-domain physical vibration waveform into a time-frequency distribution matrix that includes the frequency and time mapping dimensions.

[0038] The feature extraction module sets a cutoff frequency parameter in the time-frequency distribution matrix based on the physical difference between the slowly varying low-frequency signal generated by the static mass action of the physical target and the high-frequency broadband signal generated by the impact of the physical target. The feature extraction module uses the cutoff frequency parameter to perform frequency band segmentation on the time-frequency distribution matrix to separate the low-frequency component and the high-frequency component, and extracts the first energy spectral density sequence containing the low-frequency component. The first energy spectral density sequence maps the gravity load distribution characteristics generated when the physical target comes into contact with the surface medium.

[0039] The feature extraction module simultaneously extracts a second energy spectral density sequence containing high-frequency components. The feature extraction module performs envelope detection processing on the second energy spectral density sequence to eliminate high-frequency oscillating carrier components and extracts the discrete peak time interval data output after envelope detection processing. The discrete peak time interval data maps the biomechanical gait periodic characteristics of the physical target under continuous movement.

[0040] The feature extraction module integrates the first energy spectral density sequence and discrete peak time interval data, and defines the resulting data combination as the microseismic joint feature set. The feature extraction module will combine the microseismic joint feature set. Microseismic joint feature set stored in memory unit It forms the underground physical sensing input benchmark for subsequent spatiotemporal synchronization calculations.

[0041] After the adaptive scheduling module completes the spatial beam focusing operation of the multi-input multi-output antenna array, the feature extraction module continuously receives the radio frequency echo signal reflected by the physical target. The feature extraction module extracts the radio frequency carrier phase distortion sequence caused by the continuous displacement of the physical target's body and the subtle movements of the physical target's surface by analyzing the radio frequency echo signal.

[0042] The feature extraction module calculates the micro-Doppler frequency shift of the RF subcarrier within the beam focusing region based on the RF carrier phase distortion sequence. Micro Doppler frequency shift The mathematical relationship between them satisfies the following operational formula: ; In the above calculation formula, Represents the instantaneous radial velocity of a physical target relative to the transceiver antennas of a wireless access point. The electromagnetic wavelength representing the radio frequency signal transmitted in the current radio frequency transceiver link. This represents the spatial angle of incidence corresponding to the radio frequency signal incident on the outer surface of the physical target.

[0043] The feature extraction module aggregates the micro-Doppler frequency shifts fed back from the multipath propagation links of radio frequency signals within the spatial coordinate grid. Sets are used to construct Doppler band feature sets. Doppler band feature set Includes frequency broadening distribution data of different body parts of a physical target moving at different instantaneous radial velocities, Doppler band feature set. Based on the frequency broadening distribution data mapping, the three-dimensional spatial volume characteristics exhibited by physical targets when forming physical shielding and surface scattering effects on electromagnetic wave energy are mapped. The feature extraction module extracts the Doppler band feature set. The data is stored and synchronized to the divergence verification module as a ground-based radio frequency sensing input reference.

[0044] The divergence verification module obtains the joint microseismic feature set output by the feature extraction module. With Doppler band feature set The timestamp association value established at the hardware level of the cross-domain front-end controller during the communication interaction is retrieved, and the timestamp association value is used to analyze the joint feature set of microseismic events. With Doppler band feature set Perform time-dimension synchronization and alignment operations to construct a unified spatiotemporal mapping coordinate system. .

[0045] Divergence verification module in spatiotemporal mapping coordinate system Internal extraction of microseismic joint feature set The included first energy spectral density sequence data was used to combine the microseismic joint feature set using a bilinear spatial interpolation algorithm. Mapping discretely distributed physical sensing data to a spatiotemporal coordinate system Inner continuous spatial coordinate grid set In the process, normalization is performed on the data generated by the mapping to generate the spatial matrix of the probability density distribution of the target microseismic inversion target. Microseismic inversion target probability density distribution spatial matrix The physical target is characterized by the quantitative existence probability generated by the action of surface microseismic loads on specific spatial coordinate nodes.

[0046] The divergence verification module is synchronized in the spatiotemporal mapping coordinate system. Internal extraction of Doppler frequency band feature set The included frequency broadening distribution data, based on the positive correlation between the frequency broadening distribution data and the three-dimensional spatial volume of the physical target, utilizes a threshold segmentation algorithm to calculate the continuous spatial coordinate range covered by the radio frequency scattering cross section. The characteristic values ​​corresponding to this continuous spatial coordinate range are then normalized to generate the spatial matrix of the radio frequency inversion target probability density distribution. Radio frequency inversion target probability density distribution space matrix The quantized existence probability of the same physical target generated at the same spatial coordinate node based on the electromagnetic wave volume scattering effect.

[0047] Microseismic inversion target probability density distribution space matrix With radio frequency inversion target probability density distribution space matrix The internal matrix elements are all constrained to the real number range of zero to one. The divergence verification module eliminates the difference in signal dimensions between the underground microseismic physical field and the ground space radio frequency physical field at the original extraction level through normalization processing, and transforms the spatial matrix of the probability density distribution of the target from the microseismic inversion. With radio frequency inversion target probability density distribution space matrix The data is written into the computation buffer as a dimensionless input variable for subsequent spatiotemporal topological divergence measurement operations.

[0048] The divergence verification module reads the spatial matrix of the probability density distribution of the target obtained from the microseismic inversion from the computation buffer. With radio frequency inversion target probability density distribution space matrix and in a continuous set of spatial coordinate grids A spatial topology mapping model is constructed based on this.

[0049] The divergence verification module uses relative entropy calculation rules to quantify the degree of conflict between the physical feature distributions of the underground physical sensing field and the ground radio frequency sensing field at the same spatial coordinate node, and performs a logarithmic operation to retrieve the target probability density distribution spatial matrix via radio frequency inversion. The matrix elements are injected with a pre-defined system background smoothing factor to avoid zero values ​​in the denominator of the calculation, thereby calculating the spatial matrix of the probability density distribution of the target inversion in microseismic analysis. With radio frequency inversion target probability density distribution space matrix Spatiotemporal topological divergence between Spatiotemporal topological divergence The mathematical relationship between them satisfies the following operational formula: ; In the above calculation formula, Represents a set of spatial coordinate grids The internal discrete two-dimensional spatial coordinate nodes, and the spatial matrix of the probability density distribution of the target in microseismic inversion. As the numerator of the baseline observation distribution participating in the logarithmic operation, the spatial matrix of the radio frequency inversion target probability density distribution... The denominator term used in logarithmic operations as a reference contrast distribution.

[0050] Spacetime topological divergence The numerical calculation quantifies the degree of contradiction between the physical distribution of physical targets on the surface microseismic physical characteristics and spatial radio frequency physical characteristics. The divergence verification module will convert the spatiotemporal topological divergence... The calculated numerical values ​​and the spatial matrix of the target probability density distribution obtained from the microseismic inversion The matrix element data contained therein is transmitted to the situation determination module as a logical judgment input parameter. The situation determination module receives the logical judgment input parameter and uses it as the calculation benchmark for performing community security situation classification processing and condition comparison.

[0051] The situation assessment module constructs a multi-branch decision logic tree based on the received logical decision input parameters and extracts the spatial matrix of the probability density distribution of the target from the microseismic inversion. The number of characteristic peaks in the data, and simultaneously the spatiotemporal topological divergence. The system compares the data with a preset first divergence threshold and a preset second dynamic fluctuation threshold, respectively, and outputs security situation warning information based on the comparison results.

[0052] When the situation assessment module detects the spatial matrix of the probability density distribution of the microseismic inversion target... It contains high-value features of continuous periodic gait signals and the spatial matrix of the probability density distribution of the target retrieved by radio frequency inversion. When the values ​​of matrix elements are below the radio frequency noise floor limit, the situation assessment module confirms the existence of a physical target within the same spatiotemporal coordinate node. If the spatiotemporal topological divergence... If the divergence exceeds the preset first divergence threshold, the situation determination module determines that the physical target is in a camouflaged state that shields electromagnetic detection or in a deep spatial occlusion environment, and generates a hidden intrusion warning command based on the determination result.

[0053] When the situation assessment module is based on the target probability density distribution space matrix retrieved from microseismic inversion The two independent step-frequency phase parameters with spatial separation characteristics were analyzed, and the spatial matrix of the probability density distribution of the target obtained by radio frequency inversion was obtained. When fitting a single connected radio frequency volume envelope under the same spatiotemporal mapping coordinate system, the situation assessment module confirms that multiple physical targets are moving closely together within the spatial coordinate grid. The situation assessment module calculates the spatiotemporal topological divergence of adjacent time frames. The difference is used to obtain the time gradient change rate. If the time gradient change rate is greater than the preset second dynamic fluctuation threshold, the situation determination module excludes the sensing overlap caused by the overlap of radio frequency volume in the surface space based on the spatial independence distribution characteristics of the underground microseismic physical load, and generates an anti-tailgating intrusion warning command based on the determination result.

[0054] The situation determination module integrates implicit intrusion warning commands and anti-tailgating intrusion warning commands, and generates a situation tracking trajectory map by combining the continuous spatial coordinate sequence of physical targets in the spatiotemporal mapping coordinate system over time. The situation determination module outputs the situation tracking trajectory map and the corresponding warning commands to the security terminal display device through the communication bus, and then drives the external security linkage subsystem to perform fixed-point physical intervention operations against abnormal physical targets through the security terminal display device.

[0055] Corresponding to the above-described embodiment of the AI-based community security situation awareness system, this embodiment of the invention also provides an AI-based community security situation awareness method. This method is applied to the aforementioned system, and its specific execution logic corresponds one-to-one with the operating principles of each functional module of the aforementioned system. Specifically, the method includes the following steps: S100, the sensing baseline construction module continuously and synchronously extracts the optical phase change sequence and the complex response matrix of the channel state information; S200, when the change amplitude of the optical phase change sequence exceeds the preset background threshold, the adaptive scheduling module calculates the instantaneous expected angle of arrival of the micro-vibration source direction based on the propagation phase velocity of the micro-vibration signal and the spatial time delay difference of the adjacent fiber demodulation channel and generates the antenna steering vector. The cross-domain front-end controller updates the transmit weight vector of the multi-input multi-output antenna array based on the antenna steering vector. S300, the feature extraction module extracts the Doppler frequency band feature set within the adjusted spatial coordinate grid, and the feature extraction module simultaneously extracts the microseismic joint feature set; S400, the divergence verification module performs feature fusion operations in a unified spatiotemporal coordinate system to calculate the spatiotemporal topological divergence between the spatial matrices of two types of probability density distributions; S500, the situation determination module determines the detection result as a safe passage, a hidden intrusion, or an anti-tailgating abnormal state based on the comparison relationship between the spatiotemporal topological divergence and the preset background threshold, and outputs the corresponding warning command and tracking trajectory to the terminal through the communication interface.

[0056] The specific implementation details, data processing flow, mathematical solution model, and corresponding cross-modal physical field coupling relationship of each step in the above-mentioned AI-based community security situation awareness method have been described in detail and completely in the operation logic of the aforementioned system embodiment. Those skilled in the art can implement the method without ambiguity based on the description of the aforementioned system embodiment. In order to avoid redundancy in the specification, it will not be repeated here.

[0057] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A community security situation awareness system based on artificial intelligence, characterized in that, include: The sensing baseline construction module acquires the optical phase change sequence of underground communication single-mode optical fiber through a phase-sensitive optical time-domain reflectometer, and extracts the complex response matrix of channel state information under the orthogonal frequency division multiplexing architecture through a ground wireless access point; An adaptive scheduling module, connected to the sensing baseline construction module, has an embedded cross-domain front-end controller. It calculates the instantaneous expected angle of arrival based on the spatial delay difference of adjacent fiber demodulation channels and generates an antenna steering vector. The cross-domain front-end controller updates the transmit weight vector of the multi-input multi-output antenna array based on the antenna steering vector and the complex response matrix of the channel state information, and controls the radio frequency main lobe to align with the spatial coordinate grid calculated by the instantaneous expected angle of arrival. The feature extraction module, connected to the sensing baseline construction module, performs time-frequency domain transformation on the optical phase change sequence to extract a joint microseismic feature set, and extracts a Doppler band feature set based on the micro-Doppler frequency shift of the radio frequency subcarrier in the beam focusing region; The divergence verification module, connected to the feature extraction module, converts the microseismic joint feature set and the Doppler frequency band feature set into a probability density distribution space matrix and quantifies the spatiotemporal topological divergence. The situation determination module is connected to the divergence verification module and outputs security situation warning information based on the comparison relationship between the spatiotemporal topological divergence and the preset background threshold.

2. The community security situation awareness system based on artificial intelligence according to claim 1, characterized in that, The sensing baseline construction module receives Rayleigh scattered light generated when the probe light pulse propagates inside the underground single-mode optical fiber and propagates backward along the underground single-mode optical fiber. It then uses a photodetector to convert the Rayleigh scattered light into an electrical signal. The sensing baseline construction module demodulates the electrical signal to extract the optical phase change sequence, which includes local geometric deformation and refractive index perturbation. The sensing baseline construction module transmits the extracted optical phase change sequence to the adaptive scheduling module as a priori physical data reference to trigger the operation of the adaptive scheduling module.

3. The community security situation awareness system based on artificial intelligence according to claim 1, characterized in that, The sensing baseline construction module is further configured to control the terrestrial wireless access point to extract the underlying radio frequency link state data under the orthogonal frequency division multiplexing architecture. The baseband processing unit configured inside the terrestrial wireless access point analyzes the amplitude attenuation and phase shift caused by the radio frequency carrier crossing the ground space medium. The baseband processing unit extracts the index of a single independent propagation path from the total number of multipath propagation paths. The baseband processing unit generates the complex response matrix of the channel state information based on the instantaneous amplitude attenuation factor and electromagnetic wave propagation delay corresponding to the single independent propagation path.

4. The community security situation awareness system based on artificial intelligence according to claim 1, characterized in that, The adaptive scheduling module calculates the instantaneous expected angle of arrival based on the spatial time delay difference between adjacent fiber demodulation channels. Specifically, the adaptive scheduling module performs cross-correlation operation on the optical phase change sequence generated by the adjacent first and second fiber demodulation channels when receiving the same micro-vibration stress wave to extract the time delay difference. The adaptive scheduling module retrieves the Rayleigh surface wave propagation phase velocity. Based on the time delay difference, the Rayleigh surface wave propagation phase velocity, and the physical straight-line distance between the first and second fiber demodulation channels, the adaptive scheduling module calculates the instantaneous expected angle of arrival of the source of the micro-vibration stress wave relative to the line connecting the first and second fiber demodulation channels.

5. The community security situation awareness system based on artificial intelligence according to claim 1, characterized in that, The cross-domain front-end controller converts the instantaneous expected angle of arrival into a spatial coordinate data format. The cross-domain front-end controller establishes a direct control plane communication link with the baseband processing unit configured inside the terrestrial wireless access point. The cross-domain front-end controller sends a beam scheduling command carrying the spatial coordinate data format and the antenna steering vector to the baseband processing unit. The cross-domain front-end controller encapsulates the highest priority execution flag bit in the frame header structure of the beam scheduling command. The cross-domain front-end controller injects the beam scheduling command into the first position of the medium access control layer buffer queue of the terrestrial wireless access point and directly sets the underlying transmission state of the physical layer radio frequency transceiver hardware through the medium access control layer.

6. The community security situation awareness system based on artificial intelligence according to claim 5, characterized in that, The baseband processing unit solves the transmit weight vector according to the beam scheduling command. Specifically, the baseband processing unit obtains the environmental radio frequency signal sample covariance matrix by calculating the expected value of the product of the complex response matrix of the channel state information and the conjugate transpose of the complex response matrix of the channel state information. The baseband processing unit uses the antenna steering vector as a constraint condition for the desired signal spatial propagation direction. Based on the minimum variance distortionless response beamforming criterion and combined with the inverse matrix of the environmental radio frequency signal sample covariance matrix, the baseband processing unit establishes a radio frequency weight solution matrix equation for the multi-input multi-output antenna array. The baseband processing unit solves the radio frequency weight solution matrix equation and outputs the transmit weight vector.

7. The community security situation awareness system based on artificial intelligence according to claim 1, characterized in that, The feature extraction module performs time-frequency domain transformation on the optical phase change sequence to extract a joint microseismic feature set. Specifically, the feature extraction module uses a sliding time window to perform a short-time Fourier transform operation on the optical phase change sequence in the time domain to convert it into a time-frequency distribution matrix. The feature extraction module uses a cutoff frequency parameter to perform frequency band segmentation on the time-frequency distribution matrix to separate low-frequency and high-frequency components. The feature extraction module extracts a first energy spectral density sequence containing the low-frequency components. The feature extraction module simultaneously extracts a second energy spectral density sequence containing the high-frequency components and performs envelope detection processing on the second energy spectral density sequence to extract discrete peak time interval data. The feature extraction module integrates the first energy spectral density sequence and the discrete peak time interval data to generate the joint microseismic feature set.

8. The community security situation awareness system based on artificial intelligence according to claim 1, characterized in that, The feature extraction module extracts a Doppler band feature set based on the micro-Doppler frequency shift of the radio frequency subcarrier in the beam focusing region. Specifically, the feature extraction module receives the radio frequency echo signal reflected by the physical target and extracts the phase distortion sequence of the radio frequency carrier. The feature extraction module calculates the micro-Doppler frequency shift generated by the radio frequency subcarrier in the beam focusing region by combining the instantaneous radial motion velocity of the physical target, the electromagnetic wave wavelength of the radio frequency signal, and the spatial incident angle. The feature extraction module aggregates the micro-Doppler frequency shift set fed back by the multipath propagation link of the radio frequency signal in the spatial coordinate grid to construct the Doppler band feature set.

9. The community security situation awareness system based on artificial intelligence according to claim 1, characterized in that, The divergence verification module converts the joint microseismic feature set and the Doppler band feature set into a probability density distribution space matrix and quantifies the spatiotemporal topological divergence. Specifically, the divergence verification module converts the joint microseismic feature set and the Doppler band feature set into a microseismic inversion target probability density distribution space matrix and a radio frequency inversion target probability density distribution space matrix, respectively. Before performing logarithmic operations, the divergence verification module injects a preset system background smoothing factor into the matrix elements of the radio frequency inversion target probability density distribution space matrix. The divergence verification module calculates the spatiotemporal topological divergence between the microseismic inversion target probability density distribution space matrix and the radio frequency inversion target probability density distribution space matrix using relative entropy calculation rules. When the spatial matrix of the microseismic inversion target probability density distribution contains high-value features of continuous periodic gait signals and the matrix element values ​​of the spatial matrix of the radio frequency inversion target probability density distribution are lower than the radio frequency background noise limit, under the condition that the spatiotemporal topological divergence is greater than the preset first divergence threshold, the situation determination module determines that the physical target is in a camouflaged state shielded from electromagnetic detection or in a deep spatial occlusion environment, and generates a hidden intrusion warning command based on the determination result.

10. A community security situation awareness method based on artificial intelligence, applied to the community security situation awareness system based on artificial intelligence as described in any one of claims 1 to 9, characterized in that, Artificial intelligence-based methods for community security situation awareness include: S100, the sensing baseline construction module continuously and synchronously extracts the optical phase change sequence and the complex response matrix of the channel state information; S200, when the change amplitude of the optical phase change sequence exceeds the preset background threshold, the adaptive scheduling module calculates the instantaneous expected angle of arrival of the micro-vibration source direction based on the propagation phase velocity of the micro-vibration signal and the spatial time delay difference of the adjacent fiber demodulation channel and generates the antenna steering vector. The cross-domain front-end controller updates the transmit weight vector of the multi-input multi-output antenna array based on the antenna steering vector. S300, the feature extraction module extracts the Doppler frequency band feature set within the adjusted spatial coordinate grid, and the feature extraction module simultaneously extracts the microseismic joint feature set; S400, the divergence verification module performs feature fusion operations in a unified spatiotemporal coordinate system to calculate the spatiotemporal topological divergence between the spatial matrices of two types of probability density distributions; S500, the situation determination module determines the detection result as a safe passage, a hidden intrusion, or an anti-tailgating abnormal state based on the comparison relationship between the spatiotemporal topological divergence and the preset background threshold, and outputs the corresponding warning command and tracking trajectory to the terminal through the communication interface.

Images