Passive rfid fiber optic distribution frame port positioning method and system

Abstract

The present application relates to the technical field of optical fiber communication, and more particularly to a passive RFID optical fiber distribution frame port positioning method and system. Independent response components in multi-label mixed signals are extracted by compressed sensing sparse reconstruction, and signal demixing with statistical independence constraint is performed to obtain pure response and spatial azimuth. A virtual array is constructed according to physical aperture and phase coherence, and spatial spectrum estimation is performed to obtain spatial distribution imaging. A space-time event graph is constructed by imaging results, response time domain trajectory, port layout and historical record. Port attribution confidence evaluation is generated by counterfactual inference and causal path tracking to realize accurate positioning. The accuracy and anti-interference ability of port positioning in a multi-label environment are improved.

Description

Technical Field

[0001] This invention relates to the field of optical fiber communication technology, and in particular to a passive RFID optical fiber distribution frame port positioning method and system. Background Technology

[0002] In the field of fiber optic patch panel port management, passive RFID technology is increasingly being applied to the identification of tags and ports. The conventional approach involves deploying multiple passive RFID tags on the patch panel, each associated with a specific port, and acquiring the tag's response signal through reader antenna scanning. Existing systems mostly employ positioning methods based on Signal Strength Indication (RSSI) or phase difference ranging, combined with multi-antenna arrays or mobile readers to achieve spatial resolution. Some solutions utilize time-division multiple access or frequency-division multiple access mechanisms to handle scenarios with multiple tags responding simultaneously, relying on threshold comparison or triangulation algorithms to estimate tag positions.

[0003] When the port density of a patch panel is high and the tag spacing is small, severe aliasing of multi-tag response signals can easily occur in both the time and frequency domains. Traditional anti-collision algorithms (such as Q-value algorithms) can read tags sequentially, but they are time-consuming and cannot handle misreading or missed readings caused by overlapping synchronization signals. More importantly, RSSI-based positioning is greatly affected by environmental multipath effects and antenna polarization mismatch, and the actual positioning error often exceeds the physical spacing between ports, making it impossible to reliably distinguish adjacent ports. Even with multi-antenna arrays, the spatial angular resolution is insufficient to meet the precise positioning requirements of high-density patch panels due to limitations in the number of array elements and physical aperture.

[0004] Existing methods lack effective utilization of the dynamic evolution information of tag signals. In actual operation and maintenance, fiber optic distribution frames frequently involve patch cord insertion and removal, causing instantaneous fluctuations in amplitude, phase, or delay in tag response signals as port status changes. Conventional systems rely solely on single scan results for location estimation, easily misinterpreting transient interference or tag drift as changes in port affiliation. Furthermore, the lack of a correlation analysis mechanism between historical operation records and real-time signal characteristics makes it impossible to trace the causal relationship between tag signal anomalies and port operations, resulting in low reliability and confidence of positioning results, especially with a significantly increased error rate in high-frequency operation and maintenance scenarios. Summary of the Invention

[0005] This invention provides a passive RFID fiber optic patch panel port positioning method and system, which can solve the problems in the prior art.

[0006] A first aspect of the present invention provides a passive RFID fiber optic patch panel port positioning method, comprising:

[0007] The initial coherent RFID data is processed by sparse signal reconstruction based on compressed sensing. The independent tag response components in the multi-tag aliasing signal are extracted by solving sparse constraints in the overcomplete dictionary space.

[0008] The independent tag response components are demixed under statistical independence constraints to obtain the pure response signal representation of each tag and the spatial azimuth angle estimate in the array antenna coordinate system;

[0009] Based on the pure response signal representation and the spatial azimuth angle estimation, combined with the physical aperture parameters of the read / write device array and the signal phase coherence, the equivalent signal representation of the virtual extended array is constructed through the synthetic aperture principle. The spatial spectrum of the equivalent signal representation is estimated to obtain the spatial distribution imaging result of the tag on the patch panel plane.

[0010] The spatial distribution imaging results, the temporal evolution trajectory represented by the pure response signal, the port physical layout of the fiber optic distribution frame, and the historical maintenance operation records are collectively constructed into a spatiotemporal event diagram.

[0011] Counterfactual inference and causal path tracing are performed on the spatiotemporal event map. By identifying the causal chain between tag response changes and port operations, a port attribution confidence assessment that integrates spatial imaging evidence and temporal causal evidence is generated. Based on the port attribution confidence assessment, the target port location result corresponding to each tag is determined.

[0012] The initial coherent RFID data undergoes sparse signal reconstruction processing based on compressed sensing. Within an overcomplete dictionary space, sparse constraints are used to extract the independent tag response components from the multi-tag aliasing signal, including:

[0013] An overcomplete atom dictionary adapted to the response characteristics of passive RFID tags is constructed. By analyzing the time-frequency joint representation of the coherent RFID initial data, time-frequency atom clusters reflecting the transient characteristics of the tag response are extracted. The time-frequency atom clusters are expanded by tensor product with the modulation envelope prototype reflecting the steady-state characteristics of the tag response to form the overcomplete dictionary space with transient acquisition capability and steady-state characterization capability.

[0014] A sparse representation optimization problem is established in the overcomplete dictionary space. By introducing a hybrid penalty term combining L1 norm sparse constraints and total variational regularization constraints, the coherent RFID initial data is sparsely decomposed and solved. The L1 norm sparse constraints are used to promote the concentration of atomic selection in the sparse solution, and the total variational regularization constraints are used to maintain the smooth continuity of sparse coefficients in adjacent time steps. The sparse representation optimization problem is solved iteratively by the alternating direction multiplier method to obtain the sparse coefficient vector corresponding to each passive RFID tag and its atomic index set in the overcomplete dictionary space.

[0015] Using the coefficient values ​​in the sparse coefficient vector as weights, the time-frequency atoms in the atom index set and the modulation envelope prototype are reconstructed by weighted linear combination to generate an independent tag response component corresponding to each passive RFID tag.

[0016] The signal demixing of the independent tag response components under statistical independence constraints yields a clean response signal representation for each tag and a spatial azimuth estimate in the array antenna coordinate system, including:

[0017] The fourth-order cumulant tensor is calculated on the independent tag response components to capture the non-Gaussianity and high-order statistical correlation characteristics in the independent tag response components. The fourth-order cumulant tensor is then jointly diagonalized to obtain the initial value of the unmixing matrix that reflects the statistical independence of the passive RFID tag response.

[0018] Starting with the initial value of the unmixing matrix, adaptive iterative updates are performed by introducing a joint optimization objective of minimizing mutual information constraint and maximizing non-Gaussianity constraint. The mutual information minimization constraint reduces the statistical dependence between the separated signal components, and the non-Gaussianity maximization constraint enhances the independence discrimination ability of the separated signal components, thus obtaining the converged optimal unmixing matrix.

[0019] The optimal unmixing matrix is ​​applied to the independent tag response components, and the pure response signal representation corresponding to each passive RFID tag is separated by linear transformation.

[0020] The phase difference sequence between each element of the array antenna is extracted from the pure response signal. A geometric mapping relationship between the phase difference sequence and the spatial azimuth angle of the tag is established. The geometric mapping relationship is solved using the maximum likelihood estimation criterion to obtain the spatial azimuth angle estimate of each passive RFID tag in the array antenna coordinate system.

[0021] Based on the pure response signal representation and the spatial azimuth angle estimation, combined with the physical aperture parameters of the read / write device array and the signal phase coherence, the equivalent signal representation of the virtual extended array is constructed using the synthetic aperture principle, including:

[0022] Based on the physical aperture parameters of the read / write device array, the spatial position coordinates and element spacing distribution characteristics of each array element in the read / write device array are extracted, and the spatial geometric relationship of each passive RFID tag relative to the read / write device array is determined by combining the spatial azimuth angle estimation.

[0023] Calculate the phase correlation coefficient between the pure response signal and different array elements at different times, extract the signal time period and array element combination that meet the phase coherence condition, and establish the phase coherence criterion matrix;

[0024] Based on the effective signal components selected by the phase coherence criterion matrix, and combined with the spatial geometric relationship and the movement trajectory information of the reading and writing device array, the signals collected by the reading and writing device array at different spatial positions are phase-compensated and coherently superimposed using the synthetic aperture principle. The phase compensation is calculated based on the spatial distance difference between the passive RFID tag and each element of the reading and writing device array. The coherent superposition is a weighted accumulation of the phase-compensated signals according to the spatial position mapping relationship.

[0025] The equivalent element response sequence of the virtual extended array is generated by the phase compensation and the coherent superposition. The spatial sampling density of the equivalent element response sequence is higher than the physical element density of the read / write device array, and the equivalent element response sequence is organized into the equivalent signal expression of the virtual extended array.

[0026] Spatial spectrum estimation is performed on the equivalent signal representation to obtain the spatial distribution imaging results of the tag in the patch panel plane, including:

[0027] By performing autocorrelation operation on the equivalent signal representation in the time domain, a covariance matrix reflecting the signal correlation between each equivalent array element of the virtual extended array is obtained.

[0028] The covariance matrix is ​​decomposed into eigenvalues. Based on the jump positions in the eigenvalue amplitude sequence, the eigenvector space is divided into a signal subspace and a noise subspace. The jump positions are determined by calculating the amplitude ratio between adjacent eigenvalues. The signal subspace is spanned by eigenvectors associated with eigenvalues ​​before the jump position, and the noise subspace is spanned by eigenvectors associated with eigenvalues ​​after the jump position.

[0029] A spatial search grid is established in the patch panel plane, and a corresponding direction guidance vector is constructed for each grid node in the spatial search grid. The direction guidance vector reflects the phase response relationship of each equivalent array element when the virtual extended array receives the position signal from the grid node.

[0030] Calculate the orthogonality measure between the directional steering vector and the noise subspace, and construct a spatial spectrum function by evaluating the projection amplitude of the directional steering vector onto the noise subspace. The spatial spectrum function produces a sharp peak at the true location of the passive RFID tag.

[0031] Peak detection is performed on the spatial spectral function, and the spatial search grid coordinates corresponding to the peak positions that exceed the preset detection threshold are mapped to the spatial position estimation of the passive RFID tag on the patch panel plane, and organized into the spatial distribution imaging result of the tag on the patch panel plane.

[0032] The spatial distribution imaging results, the temporal evolution trajectory represented by the pure response signal, the port physical layout of the fiber optic distribution frame, and historical maintenance operation records are collectively constructed into a spatiotemporal event diagram, including:

[0033] The spatial coordinates of the passive RFID tags and the spatial coordinates of the ports are extracted from the spatial distribution imaging results and the physical layout of the fiber optic distribution frame ports, and mapped to tag spatial nodes and port spatial nodes in the spatiotemporal event graph, respectively.

[0034] The pure response signal representation is segmented in the time domain, and the signal response feature changes of each passive RFID tag in different time windows are extracted. A time-series dependency matrix is ​​established by calculating the correlation of the signal response feature changes between adjacent time windows. Time edges are constructed between tag space nodes based on the time-series dependency matrix.

[0035] Extract the timestamps and operation types of port operation events from historical maintenance operation records, align the timestamps of the port operation events with the signal response change moments in the time-domain evolution trajectory represented by the pure response signal, and establish a causal relationship between port operation events and tag signal response changes within a preset time window by calculating the time difference.

[0036] Based on the causal relationship, a causal edge is constructed between the port space node and the label space node. The starting point of the causal edge is the port space node, and the ending point is the label space node. The causal edge carries the port operation type and time difference as edge attributes. The time edge and the causal edge are combined to form the complete spatiotemporal event graph.

[0037] Counterfactual inference and causal path tracing are performed on the spatiotemporal event graph. By identifying the causal chain between tag response changes and port operations, a port attribution confidence assessment that integrates spatial imaging evidence and temporal causal evidence is generated. Based on the port attribution confidence assessment, the target port location result corresponding to each tag is determined, including:

[0038] Counterfactual inference is performed on each causal edge in the spatiotemporal event graph. A counterfactual scenario is constructed by removing port operation events. The difference measure between the expected state of the label signal response change in the counterfactual scenario and the actual observed label signal response change is calculated. The causal strength of the port operation event on the label signal response change is quantified based on the difference measure.

[0039] Based on the temporal and causal edges in the spatiotemporal event graph, a set of causal paths from port space nodes to label space nodes is constructed. The path propagation strength is calculated for each causal path in the set of causal paths. The path propagation strength is jointly determined by the causal strength of each edge on the path and the temporal dependency.

[0040] The spatial distance between each tag and each port is extracted from the spatial distribution imaging results as spatial imaging evidence, and the causal strength and the path propagation strength are used as temporal causal evidence. The port attribution confidence is calculated by weighted fusion of the spatial imaging evidence and the temporal causal evidence.

[0041] For each passive RFID tag, the port attribution confidence score is calculated by traversing all port space nodes, and the port corresponding to the maximum port attribution confidence score is selected as the target port location result corresponding to the tag.

[0042] A second aspect of the present invention provides a passive RFID fiber optic patch panel port positioning system, comprising:

[0043] The sparse reconstruction unit is used to perform sparse signal reconstruction processing on coherent RFID initial data based on compressed sensing. It extracts the independent tag response components in the multi-tag aliasing signal by solving sparse constraints in the overcomplete dictionary space.

[0044] The demixing estimation unit is used to demix the independent tag response components under statistical independence constraints to obtain the pure response signal representation of each tag and the spatial azimuth angle estimate in the array antenna coordinate system.

[0045] The spatial imaging unit is used to construct an equivalent signal expression of the virtual extended array based on the pure response signal representation and the spatial azimuth angle estimation, combined with the physical aperture parameters and signal phase coherence of the read / write device array, and to perform spatial spectrum estimation on the equivalent signal expression to obtain the spatial distribution imaging result of the tag on the patch panel plane.

[0046] The event graph unit is used to construct a spatiotemporal event graph by combining the spatial distribution imaging results, the temporal evolution trajectory represented by the pure response signal, the port physical layout of the fiber optic distribution frame, and historical maintenance operation records.

[0047] The causal localization unit is used to perform counterfactual inference and causal path tracing on the spatiotemporal event map. By identifying the causal chain between tag response changes and port operations, it generates a port attribution confidence assessment that integrates spatial imaging evidence and temporal causal evidence. Based on the port attribution confidence assessment, it determines the target port localization result corresponding to each tag.

[0048] A third aspect of the present invention provides an electronic device, comprising:

[0049] processor;

[0050] Memory used to store processor-executable instructions;

[0051] The processor is configured to invoke instructions stored in the memory to execute the aforementioned method.

[0052] A fourth aspect of the present invention provides a computer-readable storage medium having stored thereon computer program instructions that, when executed by a processor, implement the aforementioned method.

[0053] By employing compressed sensing and sparse constraint reconstruction, multi-label aliased signals are separated from an overcomplete dictionary, significantly suppressing mutual interference between labels. Even with densely packed labels or weakened signals, the response components of each independent label can still be accurately extracted. The signal demixing process under statistical independence constraints effectively eliminates the influence of environmental clutter and multipath effects, obtaining a clean response signal and high-precision spatial azimuth estimation, laying a reliable foundation for subsequent positioning.

[0054] By combining the physical aperture of the read / write device array with the signal phase coherence, a virtual extended array is constructed using the synthetic aperture principle. This effectively increases the aperture and refines spatial sampling, and then spatial spectrum estimation is used to obtain an image of the tag distribution within the patch panel plane. This imaging method overcomes the limitations of physical aperture, achieving super-resolution localization of tag positions at subwavelength scales. It can clearly distinguish tags on adjacent ports, avoiding misjudgments caused by angular ambiguity in traditional methods.

[0055] Spatial imaging results, the temporal evolution trajectory of pure response signals, port physical layout, and historical maintenance records are collectively constructed into a spatiotemporal event map, comprehensively integrating spatial characteristics and temporal series information. Based on this event map, counterfactual inference and causal path tracing are performed to identify the causal chain between tag response changes and port operations, effectively eliminating false positives caused by simple signal fluctuations or external interference, and generating a port attribution confidence assessment that integrates spatial and causal evidence. The final location result not only relies on spatial location but also undergoes causal logic verification, significantly improving the accuracy and robustness of port attribution determination. It is particularly suitable for the dynamic environment of frequent plugging and unplugging operations in fiber optic distribution frames, reducing the need for manual verification and ensuring operational efficiency and reliability. Attached Figure Description

[0056] Figure 1 A flowchart illustrating the passive RFID fiber optic patch panel port positioning method;

[0057] Figure 2 A schematic diagram of the process for obtaining spatial distribution imaging results. Detailed Implementation

[0058] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of 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.

[0059] The technical solution of the present invention will be described in detail below with reference to specific embodiments. These specific embodiments can be combined with each other, and the same or similar concepts or processes may not be described again in some embodiments.

[0060] Figure 1 This is a flowchart illustrating the passive RFID fiber optic patch panel port positioning method according to an embodiment of the present invention, as shown below. Figure 1 As shown, the method includes:

[0061] The initial coherent RFID data is processed by sparse signal reconstruction based on compressed sensing. The independent tag response components in the multi-tag aliasing signal are extracted by solving sparse constraints in the overcomplete dictionary space.

[0062] The independent tag response components are demixed under statistical independence constraints to obtain the pure response signal representation of each tag and the spatial azimuth angle estimate in the array antenna coordinate system;

[0063] Based on the pure response signal representation and the spatial azimuth angle estimation, combined with the physical aperture parameters of the read / write device array and the signal phase coherence, the equivalent signal representation of the virtual extended array is constructed through the synthetic aperture principle. The spatial spectrum of the equivalent signal representation is estimated to obtain the spatial distribution imaging result of the tag on the patch panel plane.

[0064] The spatial distribution imaging results, the temporal evolution trajectory represented by the pure response signal, the port physical layout of the fiber optic distribution frame, and the historical maintenance operation records are jointly constructed into a spatiotemporal event diagram.

[0065] Counterfactual inference and causal path tracing are performed on the spatiotemporal event map. By identifying the causal chain between tag response changes and port operations, a port attribution confidence assessment that integrates spatial imaging evidence and temporal causal evidence is generated. Based on the port attribution confidence assessment, the target port location result corresponding to each tag is determined.

[0066] In one optional implementation, the initial coherent RFID data undergoes sparse signal reconstruction processing based on compressed sensing. Extracting the independent tag response components from the multi-tag aliasing signal by solving sparse constraints within an overcomplete dictionary space includes:

[0067] An overcomplete atom dictionary adapted to the response characteristics of passive RFID tags is constructed. By analyzing the time-frequency joint representation of the coherent RFID initial data, time-frequency atom clusters reflecting the transient characteristics of the tag response are extracted. The time-frequency atom clusters are expanded by tensor product with the modulation envelope prototype reflecting the steady-state characteristics of the tag response to form the overcomplete dictionary space with transient acquisition capability and steady-state characterization capability.

[0068] A sparse representation optimization problem is established in the overcomplete dictionary space. By introducing a hybrid penalty term combining L1 norm sparse constraints and total variational regularization constraints, the coherent RFID initial data is sparsely decomposed and solved. The L1 norm sparse constraints are used to promote the concentration of atomic selection in the sparse solution, and the total variational regularization constraints are used to maintain the smooth continuity of sparse coefficients in adjacent time steps. The sparse representation optimization problem is solved iteratively by the alternating direction multiplier method to obtain the sparse coefficient vector corresponding to each passive RFID tag and its atomic index set in the overcomplete dictionary space.

[0069] Using the coefficient values ​​in the sparse coefficient vector as weights, the time-frequency atoms in the atom index set and the modulation envelope prototype are reconstructed by weighted linear combination to generate an independent tag response component corresponding to each passive RFID tag.

[0070] When handling passive RFID fiber optic patch panel port positioning tasks, simultaneous responses from multiple tags can lead to severe aliasing in the raw signals collected by the reader / writer, meaning that the initial coherent RFID data contains superimposed responses from different tags. To accurately separate the independent response components of each tag from this aliased data, it is necessary to construct an overcomplete atomic dictionary that can precisely characterize the response features of passive RFID tags.

[0071] The response signal of passive RFID tags is typically determined by two types of characteristics: one is the transient response characteristic generated by the tag at the moment of excitation, which manifests as local abrupt changes and energy accumulation in the time-frequency domain; the other is the steady-state modulation characteristic exhibited by the tag during the stable reflection phase, which manifests as a specific amplitude modulation envelope shape. For these two types of characteristics, the corresponding time-frequency atom clusters and modulation envelope prototypes are extracted respectively, and then fused through tensor product expansion to construct an overcomplete dictionary space that combines transient capture capability and steady-state representation capability. Specifically, short-time Fourier transform or wavelet transform is performed on the coherent RFID initial data to obtain the joint time-frequency representation of the signal. From this, local structures with significant energy concentration characteristics in the time-frequency plane are identified, and these local structures are extracted as time-frequency atom clusters. Simultaneously, by modeling the modulation schemes of known tag types, a set of modulation envelope prototypes reflecting different modulation depths, modulation frequencies, and envelope shapes is obtained. After expanding the time-frequency atom cluster and the modulation envelope prototype by tensor product, the total number of atoms in the resulting overcomplete dictionary space far exceeds the signal dimension, thus providing sufficient representational redundancy for subsequent sparse decomposition and ensuring that the response signal of each label can find a sparse and accurate representation in this dictionary space.

[0072] After constructing the incomplete dictionary space, the problem of separating multi-tag aliased signals is transformed into a sparse representation optimization problem for solution. Let the observation vector of the initial coherent RFID data be... The complete dictionary matrix is The sparse coefficient matrix to be found is Then the sparse representation optimization problem can be formulated as follows: Under the constraint of data fidelity, a hybrid penalty term is introduced to achieve dual constraints of sparsity and smoothness. The hybrid penalty term consists of two parts: an L1 norm sparsity constraint term. With total variational regularization constraint term ,in The L1 norm penalty coefficient, The total variational regularization penalty coefficients are both positive real hyperparameters that need to be adaptively adjusted based on the signal-to-noise ratio and tag density. The L1 norm sparsity constraint aims to concentrate non-zero coefficients in the solution onto a small number of atoms, ensuring the concentration of atom selection and thus achieving a sparse representation of each tag's response. The total variational regularization constraint, by penalizing the difference in sparse coefficients between adjacent time steps, maintains the smooth continuity of coefficients on the time axis, avoiding drastic fluctuations in the coefficient sequence due to noise interference, thereby improving the physical rationality of the reconstructed signal.

[0073] The complete optimization objective function can be expressed as follows: ,in Represents the sparse coefficient matrix The total variational value calculated along the time dimension is the sum of the absolute values ​​of the coefficient differences between adjacent time steps. Since this optimization problem simultaneously includes two types of non-smooth convex regularized terms: the L1 norm and the total variational value, the Alternating Direction Multiplier Method (ADMM) is used for iterative solution. ADMM decomposes the original problem into multiple independently solvable subproblems by introducing auxiliary variables. Each iteration sequentially updates the sparse coefficient variables, auxiliary variables, and dual variables. Each subproblem has a closed-form solution or can be efficiently solved using a soft thresholding operator, thus significantly reducing computational complexity while ensuring convergence accuracy. During iteration, a residual convergence threshold and a maximum number of iterations are set as stopping conditions. When both the original residual and the dual residual are below the preset threshold, the algorithm is considered to have converged, and the final sparse coefficient matrix is ​​output.

[0074] After the ADMM iterative solution is completed, a sparse coefficient vector corresponding to each passive RFID tag and its set of activated atom indices in the overcomplete dictionary space can be obtained. The value of each non-zero element in the sparse coefficient vector represents the contribution weight of the corresponding atom to the tag's response signal, while the atom index set identifies the specific time-frequency atom and modulation envelope prototype combination activated by the tag's response. Using the coefficient values ​​in the sparse coefficient vector as weights, a weighted linear combination is performed on all activated time-frequency atoms and modulation envelope prototypes in the atom index set to reconstruct the independent tag response component corresponding to that tag. Repeating the above weighted linear combination reconstruction process for all tags ultimately yields the independent tag response component corresponding to each passive RFID tag, achieving effective separation of multi-tag aliasing signals.

[0075] It is worth noting that in practical fiber optic distribution frame scenarios, the port density is high, and the tag response frequencies on adjacent ports are similar, leading to severe aliasing. In this case, the redundancy design of the overcomplete dictionary space is particularly critical. It is necessary to ensure that the mutual correlation between dictionary atoms (i.e., dictionary coherence) is sufficiently low to guarantee the uniqueness and stability of the sparse decomposition results. When constructing the overcomplete dictionary, uniform sampling and refined design of the center frequency, time window width, and modulation envelope prototype parameters of the time-frequency atoms can be used to ensure that dictionary atoms are distributed as evenly as possible in the time-frequency feature space, thereby reducing dictionary coherence. Furthermore, and The value of has a significant impact on the reconstruction quality. A cross-validation strategy can be used to perform a grid search within a certain parameter range to select the optimal parameter combination that balances reconstruction error and sparsity, thus adapting to changes in signal characteristics under different patch panel scenarios. After the above processing, the independent tag response components extracted from the initial coherent RFID data will serve as inputs for subsequent signal demixing and spatial azimuth estimation steps, laying the foundation for ultimately achieving accurate port positioning.

[0076] In one optional implementation, the signal demixing of the independent tag response components under statistical independence constraints to obtain the pure response signal representation for each tag and the spatial azimuth angle estimate in the array antenna coordinate system includes:

[0077] The fourth-order cumulant tensor is calculated on the independent tag response components to capture the non-Gaussianity and high-order statistical correlation characteristics in the independent tag response components. The fourth-order cumulant tensor is then jointly diagonalized to obtain the initial value of the unmixing matrix that reflects the statistical independence of the passive RFID tag response.

[0078] Starting with the initial value of the unmixing matrix, adaptive iterative updates are performed by introducing a joint optimization objective of minimizing mutual information constraint and maximizing non-Gaussianity constraint. The mutual information minimization constraint reduces the statistical dependence between the separated signal components, and the non-Gaussianity maximization constraint enhances the independence discrimination ability of the separated signal components, thus obtaining the converged optimal unmixing matrix.

[0079] The optimal unmixing matrix is ​​applied to the independent tag response components, and the pure response signal representation corresponding to each passive RFID tag is separated by linear transformation.

[0080] The phase difference sequence between each element of the array antenna is extracted from the pure response signal. A geometric mapping relationship between the phase difference sequence and the spatial azimuth angle of the tag is established. The geometric mapping relationship is solved using the maximum likelihood estimation criterion to obtain the spatial azimuth angle estimate of each passive RFID tag in the array antenna coordinate system.

[0081] After obtaining the independent tag response components after sparse reconstruction, it is necessary to further eliminate the residual statistical coupling in multi-tag aliasing scenarios, thereby extracting an independent and pure response signal representation for each passive RFID tag. Addressing the prevalent non-Gaussian distribution characteristics of passive RFID tag response signals, a fourth-order cumulant tensor is used as the carrier of higher-order statistical features. Compared to the covariance matrix method relying solely on second-order statistics, the fourth-order cumulant can capture non-Gaussian structural information such as kurtosis and skewness of the signal distribution, offering a fundamental advantage in distinguishing multiple tag responses with similar power spectra but different statistical distributions.

[0082] For the multi-channel observation matrix composed of independent label response components, the fourth-order cumulants of the signal are extracted channel by channel, and the cross-fourth-order cumulants between channels are organized into a fourth-order tensor form. Let the observed multi-channel signal vector be... , its first The component and the first The component, the first The component, the first The fourth-order cumulant between the components is defined as: ,in This indicates the conjugate operation. Channel index. Arrange the fourth-order cumulants of all channel combinations into a tensor. ,in The number of array element channels represents the total number of channels. This tensor statistically fully characterizes the high-order correlation structure of each independent label response component, providing sufficient statistical information for the initial value estimation of the subsequent unmixing matrix.

[0083] For fourth-order cumulant tensors The goal of joint diagonalization decomposition is to find a transformation matrix. This makes it possible to exist After being expanded into several slice matrices under the influence of the joint diagonalization, each slice matrix simultaneously exhibits a diagonal structure. The physical significance of joint diagonalization is that if the statistical independence of each label response component is maintained, then under the influence of the true unmixing matrix, the slice matrices of the fourth-order cumulant tensor will be strictly diagonalized, with off-diagonal elements approaching zero. This is achieved through a joint diagonalization iterative algorithm (such as the JADE algorithm based on Jacobi rotation), gradually adjusting... By minimizing the sum of the off-diagonal elements of each slice matrix, the initial value of the unmixing matrix is ​​finally obtained. This initial value is sufficient to reflect the basic statistical independence structure of the passive RFID tag response, providing a stable starting point for subsequent fine-grained iterative optimization.

[0084] by Starting from the iteration point, a joint optimization objective function is constructed by introducing mutual information minimization constraints and non-Gaussianity maximization constraints. The role of mutual information minimization constraints is to quantitatively measure the statistical dependence between the separated signal components; the smaller the mutual information, the closer the separated components are to statistical independence. For the set of separated signal components, their mutual information can be approximated by the difference between the sum of the marginal entropies of each component and the joint entropy. In practical calculations, methods based on kernel density estimation or negative entropy approximation are used for efficient solution, avoiding the curse of dimensionality caused by directly calculating high-dimensional joint probability density. The non-Gaussianity maximization constraint uses negative entropy as a measure of non-Gaussianity. Negative entropy is defined as the difference between the entropy of the current component and the entropy of a Gaussian distribution with the same variance. The larger the negative entropy, the more the component deviates from the Gaussian distribution, and the stronger the independence discrimination ability. The mutual information minimization term and the negative entropy maximization term are weighted and combined into a joint objective function, and the unmixing matrix is ​​adaptively iteratively updated using the natural gradient descent method. After each iteration, orthogonality constraints or whitening constraints are applied to the unmixing matrix to prevent numerical degradation or component energy imbalance during the iteration process. The iteration termination condition is that the change in the Frobenius norm of the unmixing matrix in two consecutive iterations is less than a preset threshold. or the number of iterations exceeds the maximum limit. At this point, the optimal unmixing matrix after convergence is output. .

[0085] The optimal unmixing matrix Multichannel observation matrix acting on independent label response components Through linear transformation Extract the pure response signal representation corresponding to each passive RFID tag. Matrix Each row corresponds to a clean response time series of a label, and the rows statistically satisfy the independence constraint. It is important to note that independent component analysis methods suffer from permutation and scale uncertainties during the separation process; that is, the order and amplitude scale of the separated components do not correspond to the physical relationship of the original labels. To address this, the energy distribution information of each label response component obtained in the sparse reconstruction stage is used. By matching the power spectral characteristics of each clean response signal component with the power spectral characteristics of the corresponding component in the sparse reconstruction result, the component order is automatically aligned, and the amplitude of each component is normalized to eliminate the influence of scale uncertainty.

[0086] After obtaining the pure response signal representation of each tag, the phase difference sequence between array elements is extracted from the pure response signals of that tag received by each element of the array antenna. Assume the array antenna has a total of... There are 1 array element, and the distance between adjacent array elements is 1. , No. The element received the first The pure response signal of each label is Then the first The phase difference between each array element and the reference array element (taking the first array element) is ,in This indicates the operation of taking the complex argument. Averaging the phase differences over multiple moments yields a stable estimate of the phase difference. .

[0087] Establish phase difference sequence With label spatial azimuth The geometric mapping relationship between them. For a uniform linear array, the first... The theoretical phase difference between each array element and the reference array element is: ,in The carrier wavelength of the RFID signal. For the spacing between array elements, Let be the spatial azimuth angle to be estimated. Based on the maximum likelihood estimation criterion, a likelihood function is constructed to measure the degree of agreement between the measured phase difference sequence and the theoretical phase difference sequence. Under the Gaussian phase noise assumption, maximum likelihood estimation is equivalent to minimizing the following weighted sum of squared residuals: By searching the azimuth range The grid search is performed with a fine step size, and the estimation accuracy is refined by parabolic interpolation to obtain the spatial azimuth angle estimate of each passive RFID tag in the array antenna coordinate system. The final output spatial azimuth estimation result, together with the pure response signal representation, will be passed to the subsequent synthetic aperture imaging processing stage to provide accurate angular prior information for spatial distribution imaging of patch panel ports.

[0088] In one optional implementation, based on the pure response signal representation and the spatial azimuth angle estimation, and combining the physical aperture parameters of the read / write device array with the signal phase coherence, the equivalent signal representation of the virtual extended array is constructed using the synthetic aperture principle, including:

[0089] Based on the physical aperture parameters of the read / write device array, the spatial position coordinates and element spacing distribution characteristics of each array element in the read / write device array are extracted, and the spatial geometric relationship of each passive RFID tag relative to the read / write device array is determined by combining the spatial azimuth angle estimation.

[0090] Calculate the phase correlation coefficient between the pure response signal and different array elements at different times, extract the signal time period and array element combination that meet the phase coherence condition, and establish the phase coherence criterion matrix;

[0091] Based on the effective signal components selected by the phase coherence criterion matrix, and combined with the spatial geometric relationship and the movement trajectory information of the reading and writing device array, the signals collected by the reading and writing device array at different spatial positions are phase-compensated and coherently superimposed using the synthetic aperture principle. The phase compensation is calculated based on the spatial distance difference between the passive RFID tag and each element of the reading and writing device array. The coherent superposition is a weighted accumulation of the phase-compensated signals according to the spatial position mapping relationship.

[0092] The equivalent element response sequence of the virtual extended array is generated by the phase compensation and the coherent superposition. The spatial sampling density of the equivalent element response sequence is higher than the physical element density of the read / write device array, and the equivalent element response sequence is organized into the equivalent signal expression of the virtual extended array.

[0093] After obtaining the clean response signals and spatial azimuth estimates of each tag, it is necessary to further utilize the physical aperture information of the read / write device array to construct a virtual extended array, thereby overcoming the limitations of the physical array aperture and improving spatial resolution. The read / write device array typically consists of several antenna elements arranged according to a certain pattern. The spatial coordinates of each element are calibrated during installation. A local coordinate system is established with the array's geometric center as the origin, denoted as _i_. The position coordinates of each array element are ,in , The total number of physical array elements. The distribution characteristics of the spacing between adjacent array elements are obtained through statistical analysis of the Euclidean distances between all array element coordinate pairs, including minimum spacing, maximum spacing, and mean spacing. These parameters collectively describe the aperture geometry of the array. Combined with the first step... Spatial azimuth angle estimation of each label The direction vector of the tag relative to each array element can be determined in the array coordinate system, thereby establishing a complete spatial geometric relationship between the tag and the array, providing a geometric basis for subsequent phase compensation calculations.

[0094] After determining the spatial geometric relationships, it is necessary to evaluate the phase coherence of the pure response signal with different array elements at different times in order to screen out effective signal segments suitable for coherent superposition. For the first... The tag, which is in the first Individual elements, time The pure response signal at the location is denoted as Define two array elements. and During the period The phase correlation coefficient within is The calculation method involves extracting the instantaneous phase sequence of the signal within that time period and normalizing the standard deviation of the phase difference sequence. When the standard deviation of the phase difference sequence is lower than a preset coherence threshold... When the element combination is considered to satisfy the phase coherence condition during that time period, the judgment results for all element pairs and all time periods are summarized to construct a phase coherence criterion matrix. The row index corresponds to the element pair number, and the column index corresponds to the time period number. The matrix elements take values ​​of 0 or 1, representing whether the coherence condition is not met or is met, respectively. The construction of the phase coherence criterion matrix ensures that subsequent synthetic aperture processing uses only data segments with stable signal phases and clear physical meanings, avoiding errors introduced by phase abrupt changes or signal interruptions.

[0095] After selecting valid signal components based on the phase coherence criterion matrix, and combining spatial geometric relationships with the movement trajectory information of the read / write device array, the core steps of synthetic aperture processing—phase compensation and coherent superposition—are performed. When the read / write device array moves along a predetermined trajectory, at different spatial positions... ( , The signal collected at (total number of trajectory sampling points) is equivalent to observing the same target from different locations. For the th Each label has an estimated real spatial location denoted as _ . Estimated by spatial azimuth Determined together with distance estimation. At trajectory location. Place, No. The propagation path length of the signal received by each array element is ,in Represents the Euclidean norm. (Based on reference position) Taking the first element at the location as the reference, the second... The trajectory position, the first The path difference of each element relative to the reference is The corresponding phase correction is ,in This is the carrier wavelength of the RFID signal. After phase compensation, the... The trajectory position, the first The compensation signal for each array element is ,in To be at the trajectory position The first one collected at the site The first label in the The pure response signal of each array element.

[0096] The coherent superposition process weights and accumulates the compensated signals according to their spatial location mapping. Weighting coefficients. Based on the signal amplitude stability and phase coherence criterion matrix The following criteria are jointly determined: a signal component participates in superposition only when the corresponding coherence criterion is 1; otherwise, the weight is reset to zero. Among the components that satisfy the coherence condition, the weight is proportional to the mean of the signal amplitude within that time period to enhance the contribution of signals with higher signal-to-noise ratios. The superposition yields the [number of components]. The combined response signal of each tag The signal exhibits coherent enhancement from different spatial locations in the time domain, equivalent to the observation results of a virtual array with a larger aperture on the target.

[0097] The virtual extended array equivalent element response sequence generated through phase compensation and coherent superposition has a significantly higher spatial sampling density than the physical element density of the read / write device array. Specifically, the physical array has a total of Each array element, while synthetic aperture processing will The trajectory position and The combination mapping of physical array elements is as follows: A virtual array element, The value is determined by the product of the number of effective trajectory sampling points and the number of physical array elements. However, by removing redundant positions and performing uniform resampling, the final equivalent array elements are uniformly distributed in space, with a spacing approximately equal to that of the physical array element spacing. times, of which The equivalent aperture expansion factor is determined by the ratio of the actual moving distance to the physical aperture length. The equivalent element response sequence of the virtual extended array is arranged according to the spatial coordinates of the virtual elements, forming an equivalent signal matrix. Its rows correspond to virtual matrix element indices, and its columns correspond to time sampling points. This matrix is ​​the first... The equivalent signal representation of the virtual extended array of each label. After unifying and organizing the equivalent signal representations of all labels, they are input into the subsequent spatial spectrum estimation process. By applying a high-resolution spectral estimation algorithm, a fine spatial distribution imaging result of each tag is obtained on the patch panel plane, providing a high-precision spatial basis for port positioning.

[0098] In practical engineering implementation, the movement trajectory of the read / write device array can be achieved through manual handheld movement, guide rail sliding, or robotic arm scanning. The coordinates of the trajectory sampling points are recorded in real time by an inertial navigation unit or optical positioning system to ensure the accuracy of the location information. Coherence threshold. The value of is typically adaptively adjusted based on the noise level of the channel environment. In low-noise equipment room environments, it can be set to a smaller value to retain more effective signal components, while in high-noise scenarios, it should be appropriately relaxed to avoid excessive filtering of effective signals. The increased spatial sampling density of the equivalent array element response sequence allows subsequent spatial spectrum estimation to be performed on a finer spatial grid, thereby improving the spatial resolution of port positioning to the millimeter level, meeting the engineering requirements for accurate differentiation of adjacent ports in high-density fiber optic distribution frames.

[0099] In one optional implementation, spatial spectrum estimation is performed on the equivalent signal representation to obtain the spatial distribution imaging results of the tag in the patch panel plane, including:

[0100] By performing autocorrelation operation on the equivalent signal representation in the time domain, a covariance matrix reflecting the signal correlation between each equivalent array element of the virtual extended array is obtained.

[0101] The covariance matrix is ​​decomposed into eigenvalues. Based on the jump positions in the eigenvalue amplitude sequence, the eigenvector space is divided into a signal subspace and a noise subspace. The jump positions are determined by calculating the amplitude ratio between adjacent eigenvalues. The signal subspace is spanned by eigenvectors associated with eigenvalues ​​before the jump position, and the noise subspace is spanned by eigenvectors associated with eigenvalues ​​after the jump position.

[0102] A spatial search grid is established in the patch panel plane, and a corresponding direction guidance vector is constructed for each grid node in the spatial search grid. The direction guidance vector reflects the phase response relationship of each equivalent array element when the virtual extended array receives the position signal from the grid node.

[0103] Calculate the orthogonality measure between the directional steering vector and the noise subspace, and construct a spatial spectrum function by evaluating the projection amplitude of the directional steering vector onto the noise subspace. The spatial spectrum function produces a sharp peak at the true location of the passive RFID tag.

[0104] Peak detection is performed on the spatial spectral function, and the spatial search grid coordinates corresponding to the peak positions that exceed the preset detection threshold are mapped to the spatial position estimation of the passive RFID tag on the patch panel plane, and organized into the spatial distribution imaging result of the tag on the patch panel plane.

[0105] like Figure 2 As shown, the method includes:

[0106] After completing the equivalent signal matrix of the virtual extended array After construction, spatial spectrum estimation is performed to obtain the spatial distribution imaging results of the tags in the patch panel plane. The first step of spatial spectrum estimation is to perform autocorrelation operation on the equivalent signal representation in the time domain to construct the covariance matrix. Specifically, for the virtual extended array... Signal observations of equivalent array elements at continuous-time sampling points will result in an equivalent signal matrix. Each row is considered as a temporal sampling sequence of the corresponding array element. The mean of all temporal sampling points is calculated, and the statistical expectation of the outer product is obtained, yielding a sequence of size [value missing]. covariance matrix Since the actual number of sampling points is limited, a finite-time average is used instead of the statistical expectation, i.e. ,in The total number of time-domain sampling points, indicated by the superscript. This represents the conjugate transpose. Covariance matrix. The diagonal elements reflect the signal power of each equivalent array element itself, while the off-diagonal elements reflect the signal correlation between different equivalent array elements. Its structure contains directional information about the spatial location of the tag.

[0107] For covariance matrix Perform eigenvalue decomposition to obtain Each eigenvalue and its corresponding eigenvector is assigned a value. The eigenvalues ​​are then arranged in descending order of amplitude to form an eigenvalue amplitude sequence. In multi-label scenarios, the eigenvalues ​​of the effective signal components are significantly larger than those of the noise components, with a clear amplitude jump between them. This can be addressed by calculating the amplitude ratio between adjacent eigenvalues. (in Sort by index of feature values. Find the position of the maximum value in the ratio sequence as the jump position. Before the jump position (i.e. The eigenvectors associated with the eigenvalues ​​of a signal span the signal subspace. After the jump position (i.e.) The eigenvectors associated with the eigenvalues ​​of ) span the noisy subspace. The signal subspace and noise subspace are theoretically orthogonal, and this orthogonality forms the core basis for subsequent spatial spectrum estimation. It should be noted that when the actual number of tags is unknown, the adaptive partitioning method relying on the jump in amplitude ratio can avoid the error caused by manually pre-setting the number of tags, and has good adaptability to scenarios where the number of tags changes dynamically in patch panel port operations.

[0108] A spatial search grid is established on the patch panel plane, and the physical dimensions of the patch panel are defined according to a preset grid resolution. (Unit: millimeters) Uniformly discretized to form a set of two-dimensional grid nodes covering the entire patch panel. Each grid node... ( (For each grid node index) represents a candidate label location on the patch panel plane. Based on the spatial coordinates of each equivalent array element in the virtual extended array, the propagation distance from the node to each equivalent array element is calculated. This allows for the estimation of the phase delay generated when each equivalent array element receives a signal from the node's location, thus constructing the corresponding direction steering vector. Direction guidance vector It is a length of A complex vector, whose first... Each component ( For equivalent matrix element index, ) indicates the first Each equivalent array element comes from the node When calculating the phase response of the directional signal, the geometric relationship between the actual three-dimensional spatial coordinates of the equivalent array elements and the normal direction of the patch panel plane must be fully considered to ensure that the steering vector accurately reflects the true phase relationship of the signal propagating from the patch panel plane to each equivalent array element.

[0109] Calculate the direction guidance vector With noise subspace Orthogonality measures between them, thereby constructing the spatial spectral function Orthogonality is measured by evaluating the magnitude of the projection of the steering vector onto the noise subspace: when the candidate location... When the candidate position coincides with a real label position, the corresponding steering vector theoretically lies within the signal subspace and is completely orthogonal to the noise subspace, with the projection amplitude approaching zero. When the candidate position deviates from the real label position, the steering vector is not completely orthogonal to the noise subspace, and the projection amplitude is not zero. Based on this, the spatial spectral function is defined as the reciprocal of the projection amplitude, i.e. The denominator term This represents the projected energy of the guide vector onto the noise subspace. When the candidate position matches the true label position, this denominator approaches zero, and the spatial spectrum function... This produces sharp peaks, the sharpness of which is determined by the equivalent aperture of the virtual extended array. The larger the aperture, the narrower the peak value, the higher the spatial resolution, and the stronger the ability to distinguish densely arranged tags on adjacent ports.

[0110] For spatial spectral functions After performing calculations across the entire patch panel plane search grid, peak detection is executed. A preset detection threshold is set. This threshold is typically determined by multiplying the background amplitude statistics of the spatial spectral function in the unlabeled region by a signal-to-noise ratio-related amplification factor, in order to achieve a balance between effective label peak detection and noise false alarm suppression. This will satisfy... The grid nodes are considered as candidate label locations. Further filtering using local maxima removes non-maximum points, retaining the grid coordinates corresponding to the highest peak value within each local region as the estimated label location. For densely packed multi-label scenarios, if the grid distance between two adjacent peak values ​​is less than the lower limit of spatial resolution, the locations corresponding to stronger responses are retained based on peak amplitude, avoiding false label localization results due to sidelobe effects.

[0111] After peak detection, the spatial search grid coordinates corresponding to each retained peak are... Mapped to the spatial position estimation of passive RFID tags on the patch panel plane, according to the tag index. The data is processed to generate a spatial distribution imaging result containing the two-dimensional coordinates of all detected tags. This imaging result, based on the patch panel coordinate system, directly reflects the distribution of each tag on the physical plane of the patch panel, providing accurate spatial coordinate input for the subsequent construction of a spatiotemporal event map. In scenarios with high patch panel port density, the grid resolution... The selection of the port spacing and the aperture of the virtual expansion array need to be considered comprehensively. Set to 1 / 5 to 1 / 3 of the port spacing to ensure that labels on adjacent ports can be effectively distinguished in spatial spectral imaging, while avoiding a sharp increase in computational load due to excessively fine grids.

[0112] In one optional implementation, the spatial distribution imaging results, the temporal evolution trajectory represented by the pure response signal, the port physical layout of the fiber optic distribution frame, and historical maintenance operation records are collectively constructed into a spatiotemporal event graph, including:

[0113] The spatial coordinates of the passive RFID tags and the spatial coordinates of the ports are extracted from the spatial distribution imaging results and the physical layout of the fiber optic distribution frame ports, and mapped to tag spatial nodes and port spatial nodes in the spatiotemporal event graph, respectively.

[0114] The pure response signal representation is segmented in the time domain, and the signal response feature changes of each passive RFID tag in different time windows are extracted. A time-series dependency matrix is ​​established by calculating the correlation of the signal response feature changes between adjacent time windows. Time edges are constructed between tag space nodes based on the time-series dependency matrix.

[0115] Extract the timestamps and operation types of port operation events from historical maintenance operation records, align the timestamps of the port operation events with the signal response change moments in the time-domain evolution trajectory represented by the pure response signal, and establish a causal relationship between port operation events and tag signal response changes within a preset time window by calculating the time difference.

[0116] Based on the causal relationship, a causal edge is constructed between the port space node and the label space node. The starting point of the causal edge is the port space node, and the ending point is the label space node. The causal edge carries the port operation type and time difference as edge attributes. The time edge and the causal edge are combined to form the complete spatiotemporal event graph.

[0117] Coordinate information of two key spatial entities was extracted from the spatial distribution imaging results and the physical layout of the fiber optic patch panel ports. The spatial distribution imaging results provided a two-dimensional position estimate of each passive RFID tag on the patch panel plane, while the port physical layout file recorded the nominal installation coordinates of each port on the patch panel. These two types of coordinates were mapped to two types of nodes in the spatiotemporal event graph: tag space nodes carried the estimated position coordinates and unique tag identifier as node attributes, while port space nodes carried the physical installation coordinates, port number, and port type (e.g., single-mode port, multi-mode port, etc.) as node attributes. During the node initialization phase, initial confidence weights were assigned to both tag space nodes and port space nodes, providing a basis for the weighted calculations in the subsequent causal inference phase.

[0118] A sliding time window mechanism is used to slice the temporal evolution sequence of each label. Let the time window length be... The sliding step size of adjacent windows is , for the The first label in the The signal response feature vector extracted within each time window is denoted as . This feature vector contains statistics such as the mean, variance, peak-to-peak value, zero-crossing rate, and spectral centroid of the signal within the window, collectively describing the overall state of the label response during that time period. The change in signal response characteristics between adjacent time windows is defined as... This reflects the magnitude and direction of the label response drift within adjacent time periods.

[0119] The temporal dependency matrix is ​​constructed based on a correlation measure between the changes in features of different labels within the same time window. For labels... With tags In the Time series correlation coefficient at each time window Through calculation and The normalized inner product is obtained. The label pairs are obtained by averaging the correlation coefficients over all time windows. Global temporal correlation strength between This is used to fill the time-series dependency matrix. The corresponding element in. When Exceeding the preset correlation threshold At that time, in the tag space node and Establish a time edge between them, with the edge weight of the time edge taking the value of . The edge attributes also record the index of the time window with the strongest correlation, so as to locate the time interval of coordinated changes in signals when tracing causal paths later. This method of constructing time edges can capture the group linkage effect of multiple labels responding synchronously due to the operation of the same port, providing evidence of group behavior for subsequent causal inference.

[0120] The record fields are parsed in a structured manner, and each maintenance record includes an operation timestamp. The operation includes the port number, operation type (e.g., inserting fiber optic cable, removing fiber optic cable, cleaning port, tag replacement, etc.), and operator identification. The operation type is numerically encoded to form an operation type feature vector. This is used for subsequent storage and retrieval of causal edge attributes. In the time alignment step, abrupt changes are detected in the time-domain evolution sequence represented by the pure response signal to identify the set of moments where significant jumps occur in the signal response, and each jump moment is denoted as... ,in Index for transition events. Timestamp for each port operation event. With each label's transition moment Calculate time difference .when Falling within the preset causal time window Within this timeframe, it is assumed that the port operation event is related to the tag. The There is a potential causal relationship between the changes in the signal response, whereby... The effective time limit for cause and effect is usually set based on the typical response delay experience value of fiber optic patch panel operation, and is generally in the range of hundreds of milliseconds to several seconds.

[0121] The strength of a causal relationship depends not only on the magnitude of the temporal difference but also on a joint scoring method incorporating spatial distance constraints. For port operation events and label response change pairs with identified potential causal relationships, the Euclidean distance between port spatial nodes and label spatial nodes is calculated. Spatial proximity weights are obtained by weighting spatial distances using a Gaussian kernel function. ,in This is a parameter for the spatial decay scale. Simultaneously, it considers temporal differences. Also apply time proximity weights ,in This is a time decay scale parameter. It represents the overall causal relationship strength. It is given by the product of spatial proximity weight and temporal proximity weight, i.e. The larger the value, the stronger the causal explanatory power of the port operation event on the tag response change.

[0122] During the construction phase of causal edges, for each pair that satisfies the causal time window constraint and comprehensively considers the causal correlation strength... Exceeding the causal strength threshold For each port operation event and tag response change pair, a directed causal edge is established between the corresponding port space node and tag space node. The direction of the causal edge is defined as pointing from the port space node to the tag space node, expressing the causal directionality of "port operation causes tag response change". The edge attribute fields of the causal edge include: operation type encoding vector. Time difference Comprehensive causal relationship strength and the corresponding jump event index When there are multiple causal edges between the same pair of port space nodes and label space nodes (i.e., multiple operations are causally related to the label), the multiple causal edges are merged into a single aggregated causal edge. The strength of the aggregated edge is the weighted average of the strengths of all single causal associations. The edge attributes also record the number of historical associations to reflect the stability of the long-term causal coupling between the port and the label.

[0123] By integrating temporal and causal edges into a single graph structure, a complete spatiotemporal event graph is formed. This graph uses label space nodes and port space nodes as vertex sets, and temporal edges (connecting label nodes and expressing signal co-change relationships) and causal edges (connecting port nodes and label nodes and expressing operational causal relationships) as edge sets, forming a heterogeneous directed graph. Each node and edge in the graph carries multi-dimensional attributes, supporting subsequent counterfactual inference and causal path tracing algorithms for depth-first traversal and reasoning of the graph structure. The construction of the spatiotemporal event graph fully preserves spatial imaging evidence (node ​​coordinate attributes), temporal evolution evidence (temporal edges and feature change attributes), and operational history evidence (causal edges and operation type attributes), providing a multi-source fusion structured reasoning foundation for port attribution confidence assessment.

[0124] In one optional implementation, counterfactual inference and causal path tracing are performed on the spatiotemporal event graph. By identifying the causal chain between tag response changes and port operations, a port attribution confidence assessment fusing spatial imaging evidence and temporal causal evidence is generated. The target port location result for each tag is determined based on the port attribution confidence assessment, including:

[0125] Counterfactual inference is performed on each causal edge in the spatiotemporal event graph. A counterfactual scenario is constructed by removing port operation events. The difference measure between the expected state of the label signal response change in the counterfactual scenario and the actual observed label signal response change is calculated. The causal strength of the port operation event on the label signal response change is quantified based on the difference measure.

[0126] Based on the temporal and causal edges in the spatiotemporal event graph, a set of causal paths from port space nodes to label space nodes is constructed. The path propagation strength is calculated for each causal path in the set of causal paths. The path propagation strength is jointly determined by the causal strength of each edge on the path and the temporal dependency.

[0127] The spatial distance between each tag and each port is extracted from the spatial distribution imaging results as spatial imaging evidence, and the causal strength and the path propagation strength are used as temporal causal evidence. The port attribution confidence is calculated by weighted fusion of the spatial imaging evidence and the temporal causal evidence.

[0128] For each passive RFID tag, the port attribution confidence score is calculated by traversing all port space nodes, and the port corresponding to the maximum port attribution confidence score is selected as the target port location result corresponding to the tag.

[0129] After the spatiotemporal event graph is constructed, counterfactual inference is performed on each causal edge in the graph. The core idea of ​​counterfactual inference is: assuming that a certain port operation event never occurred, predicting the state of the tag signal response under this assumption, and then comparing the predicted state with the actually observed changes in the tag signal response to quantify the true contribution of the operation event to the response change. Specifically, for each causal edge connecting the port operation node and the tag response node in the spatiotemporal event graph, the port operation event node and all its associated edges are temporarily removed from the graph. The remaining graph structure is then used to perform counterfactual prediction of the evolution of the tag signal response to obtain the expected response state under the counterfactual scenario. .Will The difference between the observed changes in the tag signal response and the actual changes in the observed changes is measured using normalized mean squared error. ,in For the first The actual observed change in signal response of each label The expected change in response under counterfactual prediction. To prevent extremely small positive numbers with a denominator of zero. The larger the value, the more significant the contribution of the port operation event to the change in the tag signal response, i.e., the higher the causal strength; if A value close to zero indicates that even if the event is removed, the label response change is almost identical to the actual observation, suggesting a weak causal relationship. The causal strength of this causal edge, after being normalized by monotonic mapping, is denoted as . ,satisfy .

[0130] Based on the temporal and causal edges in the spatiotemporal event graph, starting from each port space node, all directed paths that can reach the target label space node are searched along the time-increasing direction to construct a set of causal paths. The path passes through multiple intermediate event nodes; for example, an operation might trigger a chain reaction in adjacent ports, indirectly affecting the signal state of the target tag. For each causal path in the set... Calculate path propagation intensity The propagation strength of a path is determined by the causal strength of each edge on the path and its temporal dependency: the causal strength of each causal edge on the path is taken. For each time edge on the path, take its temporal dependency matrix. The corresponding temporal correlation coefficient is used to multiply the intensity values ​​of all edges along the path to obtain the propagation intensity of that path. ,in Indicates the first in the path The strength value of the edge is derived from the causal edge selection. The time-series correlation coefficient is taken for the time edge. If there are multiple paths in the path set, the maximum value of the propagation intensity of all paths is taken as the comprehensive path propagation intensity from that port to the tag. When the path set is empty, let This indicates that there is no traceable causal path between the port and the label.

[0131] Spatial imaging evidence originates from spatial distribution imaging results obtained through synthetic aperture spatial spectrum estimation. The first... The estimated spatial coordinates of each label are used to calculate the Euclidean distance between the label and the physical layout coordinates of each port, denoted as . This serves as a quantitative indicator of spatial imaging evidence. The smaller the distance, the closer the tag is to the port in physical space, and the higher the probability of spatial attribution. This transforms spatial imaging evidence into a spatial attribution probability component. Gaussian decay form is used: ,in The attenuation scale parameter for spatial imaging evidence is set jointly based on the patch panel port spacing and imaging resolution to ensure that the spatial imaging evidence between adjacent ports has sufficient distinguishability.

[0132] Spatial imaging evidence components With temporal causal evidence (including causal strength) and path propagation intensity Weighted fusion is performed to calculate the port attribution confidence. The fusion formula is: ,in , , These are the spatial imaging evidence weight, causal strength weight, and path propagation strength weight, respectively, and the three satisfy the following: All values ​​are positive. The weights are adjusted based on the actual deployment environment of the patch panel: in scenarios with large port spacing and sufficient spatial imaging resolution, the weights are appropriately increased. The proportion; in scenarios with densely packed ports and limited imaging resolution, appropriately increase and The proportion of this allows temporal causal evidence to play a more dominant role in attribution determination. In practical engineering deployments, this can be verified by using historical samples of known port attributions and employing cross-validation methods. , , Adaptive calibration is performed to maximize the attribution accuracy of the fusion confidence on known samples.

[0133] For each passive RFID tag, traverse all port spatial nodes on the patch panel and calculate the port attribution confidence level for each one. This forms the confidence distribution vector for the label across all ports. The port corresponding to the highest confidence score is selected as the target port location result for this label. To further improve positioning reliability, a confidence difference check is introduced before the final output: if the difference between the maximum confidence score and the second-highest confidence score is lower than a preset confidence interval threshold... If the label is not sufficiently assigned, it will be marked as "fuzzy attribution," triggering a manual review process to avoid erroneous forced attribution when both spatial imaging evidence and temporal causal evidence are insufficient. For labels whose confidence difference meets the threshold requirement, the results are output directly. The final target port location result, along with the corresponding confidence value, is written into the maintenance record to provide data support for subsequent updates to historical operation records and iterative optimization of the spatiotemporal event graph. The entire port attribution confidence assessment process is executed in parallel on all tags on the patch panel, ultimately outputting the mapping relationship between each tag and its corresponding target port, completing the entire process of passive RFID fiber optic patch panel port location inference.

[0134] A second aspect of the present invention provides a passive RFID fiber optic patch panel port positioning system, comprising:

[0135] The sparse reconstruction unit is used to perform sparse signal reconstruction processing on coherent RFID initial data based on compressed sensing. It extracts the independent tag response components in the multi-tag aliasing signal by solving sparse constraints in the overcomplete dictionary space.

[0136] The demixing estimation unit is used to demix the independent tag response components under statistical independence constraints to obtain the pure response signal representation of each tag and the spatial azimuth angle estimate in the array antenna coordinate system.

[0137] The spatial imaging unit is used to construct an equivalent signal expression of the virtual extended array based on the pure response signal representation and the spatial azimuth angle estimation, combined with the physical aperture parameters and signal phase coherence of the read / write device array, and to perform spatial spectrum estimation on the equivalent signal expression to obtain the spatial distribution imaging result of the tag on the patch panel plane.

[0138] The event graph unit is used to construct a spatiotemporal event graph by combining the spatial distribution imaging results, the temporal evolution trajectory represented by the pure response signal, the port physical layout of the fiber optic distribution frame, and historical maintenance operation records.

[0139] The causal localization unit is used to perform counterfactual inference and causal path tracing on the spatiotemporal event map. By identifying the causal chain between tag response changes and port operations, it generates a port attribution confidence assessment that integrates spatial imaging evidence and temporal causal evidence. Based on the port attribution confidence assessment, it determines the target port localization result corresponding to each tag.

[0140] A third aspect of the present invention provides an electronic device, comprising:

[0141] processor;

[0142] Memory used to store processor-executable instructions;

[0143] The processor is configured to invoke instructions stored in the memory to execute the aforementioned method.

[0144] A fourth aspect of the present invention provides a computer-readable storage medium having stored thereon computer program instructions that, when executed by a processor, implement the aforementioned method.

[0145] This invention can be a method, apparatus, system, and / or computer program product. The computer program product may include a computer-readable storage medium having computer-readable program instructions loaded thereon for performing various aspects of the invention.

[0146] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for locating passive RFID fiber optic distribution frame ports, the method comprising: include: The initial coherent RFID data is processed by sparse signal reconstruction based on compressed sensing. The independent tag response components in the multi-tag aliasing signal are extracted by solving sparse constraints in the overcomplete dictionary space. The independent tag response components are demixed under statistical independence constraints to obtain the pure response signal representation of each tag and the spatial azimuth angle estimate in the array antenna coordinate system; Based on the pure response signal representation and the spatial azimuth angle estimation, combined with the physical aperture parameters of the read / write device array and the signal phase coherence, the equivalent signal representation of the virtual extended array is constructed through the synthetic aperture principle. The spatial spectrum of the equivalent signal representation is estimated to obtain the spatial distribution imaging result of the tag on the patch panel plane. The spatial distribution imaging results, the temporal evolution trajectory represented by the pure response signal, the port physical layout of the fiber optic distribution frame, and the historical maintenance operation records are collectively constructed into a spatiotemporal event diagram. Counterfactual inference and causal path tracing are performed on the spatiotemporal event map. By identifying the causal chain between tag response changes and port operations, a port attribution confidence assessment that integrates spatial imaging evidence and temporal causal evidence is generated. Based on the port attribution confidence assessment, the target port location result corresponding to each tag is determined.

2. The method according to claim 1, characterized in that, The initial coherent RFID data undergoes sparse signal reconstruction processing based on compressed sensing. Within an overcomplete dictionary space, sparse constraints are used to extract the independent tag response components from the multi-tag aliasing signal, including: An overcomplete atom dictionary adapted to the response characteristics of passive RFID tags is constructed. By analyzing the time-frequency joint representation of the coherent RFID initial data, time-frequency atom clusters reflecting the transient characteristics of the tag response are extracted. The time-frequency atom clusters are expanded by tensor product with the modulation envelope prototype reflecting the steady-state characteristics of the tag response to form the overcomplete dictionary space with transient acquisition capability and steady-state characterization capability. A sparse representation optimization problem is established in the overcomplete dictionary space. By introducing a hybrid penalty term combining L1 norm sparse constraints and total variational regularization constraints, the coherent RFID initial data is sparsely decomposed and solved. The L1 norm sparse constraints are used to promote the concentration of atomic selection in the sparse solution, and the total variational regularization constraints are used to maintain the smooth continuity of sparse coefficients in adjacent time steps. The sparse representation optimization problem is solved iteratively by the alternating direction multiplier method to obtain the sparse coefficient vector corresponding to each passive RFID tag and its atomic index set in the overcomplete dictionary space. Using the coefficient values ​​in the sparse coefficient vector as weights, the time-frequency atoms in the atom index set and the modulation envelope prototype are reconstructed by weighted linear combination to generate an independent tag response component corresponding to each passive RFID tag.

3. The method according to claim 1, characterized in that, The signal demixing of the independent tag response components under statistical independence constraints yields a clean response signal representation for each tag and a spatial azimuth estimate in the array antenna coordinate system, including: The fourth-order cumulant tensor is calculated on the independent tag response components to capture the non-Gaussianity and high-order statistical correlation characteristics in the independent tag response components. The fourth-order cumulant tensor is then jointly diagonalized to obtain the initial value of the unmixing matrix that reflects the statistical independence of the passive RFID tag response. Starting with the initial value of the unmixing matrix, adaptive iterative updates are performed by introducing a joint optimization objective of minimizing mutual information constraint and maximizing non-Gaussianity constraint. The mutual information minimization constraint reduces the statistical dependence between the separated signal components, and the non-Gaussianity maximization constraint enhances the independence discrimination ability of the separated signal components, thus obtaining the converged optimal unmixing matrix. The optimal unmixing matrix is ​​applied to the independent tag response components, and the pure response signal representation corresponding to each passive RFID tag is separated by linear transformation. The phase difference sequence between each element of the array antenna is extracted from the pure response signal. A geometric mapping relationship between the phase difference sequence and the spatial azimuth angle of the tag is established. The geometric mapping relationship is solved using the maximum likelihood estimation criterion to obtain the spatial azimuth angle estimate of each passive RFID tag in the array antenna coordinate system.

4. The method according to claim 1, characterized in that, Based on the pure response signal representation and the spatial azimuth angle estimation, combined with the physical aperture parameters of the read / write device array and the signal phase coherence, the equivalent signal representation of the virtual extended array is constructed using the synthetic aperture principle, including: Based on the physical aperture parameters of the read / write device array, the spatial position coordinates and element spacing distribution characteristics of each array element in the read / write device array are extracted, and the spatial geometric relationship of each passive RFID tag relative to the read / write device array is determined by combining the spatial azimuth angle estimation. Calculate the phase correlation coefficient between the pure response signal and different array elements at different times, extract the signal time period and array element combination that meet the phase coherence condition, and establish the phase coherence criterion matrix; Based on the effective signal components selected by the phase coherence criterion matrix, and combined with the spatial geometric relationship and the movement trajectory information of the reading and writing device array, the signals collected by the reading and writing device array at different spatial positions are phase-compensated and coherently superimposed using the synthetic aperture principle. The phase compensation is calculated based on the spatial distance difference between the passive RFID tag and each element of the reading and writing device array. The coherent superposition is a weighted accumulation of the phase-compensated signals according to the spatial position mapping relationship. The equivalent element response sequence of the virtual extended array is generated by the phase compensation and the coherent superposition. The spatial sampling density of the equivalent element response sequence is higher than the physical element density of the read / write device array, and the equivalent element response sequence is organized into the equivalent signal expression of the virtual extended array.

5. The method according to claim 1, characterized in that, Spatial spectrum estimation is performed on the equivalent signal representation to obtain the spatial distribution imaging results of the tag in the patch panel plane, including: By performing autocorrelation operation on the equivalent signal representation in the time domain, a covariance matrix reflecting the signal correlation between each equivalent array element of the virtual extended array is obtained. The covariance matrix is ​​decomposed into eigenvalues. Based on the jump positions in the eigenvalue amplitude sequence, the eigenvector space is divided into a signal subspace and a noise subspace. The jump positions are determined by calculating the amplitude ratio between adjacent eigenvalues. The signal subspace is spanned by eigenvectors associated with eigenvalues ​​before the jump position, and the noise subspace is spanned by eigenvectors associated with eigenvalues ​​after the jump position. A spatial search grid is established in the patch panel plane, and a corresponding direction guidance vector is constructed for each grid node in the spatial search grid. The direction guidance vector reflects the phase response relationship of each equivalent array element when the virtual extended array receives the position signal from the grid node. Calculate the orthogonality measure between the directional steering vector and the noise subspace, and construct a spatial spectrum function by evaluating the projection amplitude of the directional steering vector onto the noise subspace. The spatial spectrum function produces a sharp peak at the true location of the passive RFID tag. Peak detection is performed on the spatial spectral function, and the spatial search grid coordinates corresponding to the peak positions that exceed the preset detection threshold are mapped to the spatial position estimation of the passive RFID tag on the patch panel plane, and organized into the spatial distribution imaging result of the tag on the patch panel plane.

6. The method according to claim 1, characterized in that, The spatial distribution imaging results, the temporal evolution trajectory represented by the pure response signal, the port physical layout of the fiber optic distribution frame, and historical maintenance operation records are collectively constructed into a spatiotemporal event diagram, including: The spatial coordinates of the passive RFID tags and the spatial coordinates of the ports are extracted from the spatial distribution imaging results and the physical layout of the fiber optic distribution frame ports, and mapped to tag spatial nodes and port spatial nodes in the spatiotemporal event graph, respectively. The pure response signal representation is segmented in the time domain, and the signal response feature changes of each passive RFID tag in different time windows are extracted. A time-series dependency matrix is ​​established by calculating the correlation of the signal response feature changes between adjacent time windows. Time edges are constructed between tag space nodes based on the time-series dependency matrix. Extract the timestamps and operation types of port operation events from historical maintenance operation records, align the timestamps of the port operation events with the signal response change moments in the time-domain evolution trajectory represented by the pure response signal, and establish a causal relationship between port operation events and tag signal response changes within a preset time window by calculating the time difference. Based on the causal relationship, a causal edge is constructed between the port space node and the label space node. The starting point of the causal edge is the port space node, and the ending point is the label space node. The causal edge carries the port operation type and time difference as edge attributes. The time edge and the causal edge are combined to form the complete spatiotemporal event graph.

7. The method according to claim 1, characterized in that, Counterfactual inference and causal path tracing are performed on the spatiotemporal event graph. By identifying the causal chain between tag response changes and port operations, a port attribution confidence assessment that integrates spatial imaging evidence and temporal causal evidence is generated. Based on the port attribution confidence assessment, the target port location result corresponding to each tag is determined, including: Counterfactual inference is performed on each causal edge in the spatiotemporal event graph. A counterfactual scenario is constructed by removing port operation events. The difference measure between the expected state of the label signal response change in the counterfactual scenario and the actual observed label signal response change is calculated. The causal strength of the port operation event on the label signal response change is quantified based on the difference measure. Based on the temporal and causal edges in the spatiotemporal event graph, a set of causal paths from port space nodes to label space nodes is constructed. The path propagation strength is calculated for each causal path in the set of causal paths. The path propagation strength is jointly determined by the causal strength of each edge on the path and the temporal dependency. The spatial distance between each tag and each port is extracted from the spatial distribution imaging results as spatial imaging evidence, and the causal strength and the path propagation strength are used as temporal causal evidence. The port attribution confidence is calculated by weighted fusion of the spatial imaging evidence and the temporal causal evidence. For each passive RFID tag, the port attribution confidence score is calculated by traversing all port space nodes, and the port corresponding to the maximum port attribution confidence score is selected as the target port location result corresponding to the tag.

8. A passive RFID fiber optic patch panel port positioning system, used to implement the method as described in any one of claims 1-7, characterized in that, include: The sparse reconstruction unit is used to perform sparse signal reconstruction processing on coherent RFID initial data based on compressed sensing. It extracts the independent tag response components in the multi-tag aliasing signal by solving sparse constraints in the overcomplete dictionary space. The demixing estimation unit is used to demix the independent tag response components under statistical independence constraints to obtain the pure response signal representation of each tag and the spatial azimuth angle estimate in the array antenna coordinate system. The spatial imaging unit is used to construct an equivalent signal expression of the virtual extended array based on the pure response signal representation and the spatial azimuth angle estimation, combined with the physical aperture parameters and signal phase coherence of the read / write device array, and to perform spatial spectrum estimation on the equivalent signal expression to obtain the spatial distribution imaging result of the tag on the patch panel plane. The event graph unit is used to construct a spatiotemporal event graph by combining the spatial distribution imaging results, the temporal evolution trajectory represented by the pure response signal, the port physical layout of the fiber optic distribution frame, and historical maintenance operation records. The causal localization unit is used to perform counterfactual inference and causal path tracing on the spatiotemporal event map. By identifying the causal chain between tag response changes and port operations, it generates a port attribution confidence assessment that integrates spatial imaging evidence and temporal causal evidence. Based on the port attribution confidence assessment, it determines the target port localization result corresponding to each tag.

9. An electronic device, characterized in that, include: processor; Memory used to store processor-executable instructions; The processor is configured to invoke instructions stored in the memory to execute the method according to any one of claims 1 to 7.

10. A computer-readable storage medium having computer program instructions stored thereon, characterized in that, When the computer program instructions are executed by the processor, they implement the method described in any one of claims 1 to 7.

Images