Array radar double gain precise sensing method and system

Abstract

The application provides a kind of array radar double gain precision sensing method and system, it is related to radar signal processing technical field, solve the technical problem that present technique cannot consider detection sensitivity and precision parameter distinguishability under the premise of eliminating common interference parameter coupling.The method specifically includes: obtaining the multichannel complex data formed by at least two radar channels;Using at least one spatial processing configuration to process the multichannel complex data, form coherent observation and spatial manifold relationship observation;Construct a joint statistical model, and keep cross-covariance when sharing original samples;Establish a joint measurement model, and eliminate conditions for interference parameters to obtain effective condition information;Based on the sensitivity index of coherent observation and the quality index of effective condition information, select or update the spatial processing configuration;Under the condition of meeting the preset constraint condition, target parameters are estimated by using coherent observation and spatial manifold relationship observation jointly.The application is used in radar sensing.

Description

Technical Field

[0001] This application relates to the field of radar signal processing technology, and in particular to a dual-gain precision sensing method and system for array radar. Background Technology

[0002] In array radar signal processing, multi-channel coherent synthesis is typically used to obtain array gain to improve detection performance, while precise parameters such as target angle, near-field position, and scattering structure are extracted using the relative relationships between channels. However, existing engineering processing often lumps these two types of gains together into a single array gain index, primarily designing spatial weights around maximizing the output signal-to-noise ratio, neglecting the potential reduction in angle measurement sensitivity caused by high-gain beams and the strong coupling between interference parameters such as common scattering amplitude and channel errors and target parameters. Furthermore, when coherent observations and relational observations share the same original samples, treating them as independent observations leads to redundant weighting and underestimation of estimation errors. Therefore, existing technologies suffer from the technical problem of failing to balance detection sensitivity and the discernibility of precise parameters while eliminating the coupling of common interference parameters. Summary of the Invention

[0003] This application provides a dual-gain precision sensing method and system for array radar, which solves the technical problem that existing technologies cannot balance detection sensitivity and precision parameter identifiability while eliminating the coupling of common interference parameters.

[0004] To achieve the above objectives, this application adopts the following technical solution: In a first aspect, a dual-gain precision sensing method for array radar is provided, comprising: acquiring multi-channel complex data formed by at least two radar channels; processing the multi-channel complex data using at least one spatial processing configuration to form coherent observations and spatial manifold relation observations; using coherent observations to improve the effective signal-to-noise ratio or coherent accumulation capability; using spatial manifold relation observations to preserve array spatial manifold information; constructing a joint statistical model of coherent observations and spatial manifold relation observations, and preserving cross-covariance when the two share the original samples; establishing a joint measurement model including target parameters and interference parameters, and performing conditional elimination on the interference parameters to obtain effective conditional information of the target parameters; the interference parameters include at least one of common scattering complex amplitude, channel parameters, clutter parameters, and calibration parameters; selecting or updating the spatial processing configuration based on the sensitivity index of coherent observations and the quality index of effective conditional information; and, under the condition of satisfying preset constraints, jointly estimating the target parameters using coherent observations and spatial manifold relation observations.

[0005] The above scheme decomposes the array gain into two types of observations: coherent sensitivity and spatial manifold relationship, and constructs a joint statistical model that preserves cross covariance, thus avoiding double weighting caused by shared samples. By performing conditional elimination on interference parameters such as common scattering complex amplitude, coupling factors unrelated to the target state are removed, and effective conditional information that truly reflects the identifiability of target parameters is obtained. This drives spatial configuration selection and joint estimation, thereby improving the accuracy and robustness of precision sensing while ensuring detection sensitivity.

[0006] In conjunction with the first aspect above, in one possible implementation, multi-channel complex data is processed using at least one spatial processing configuration to form coherent observations and spatial manifold relationship observations, including: coherently combining multi-channel complex data using a first spatial processing configuration to obtain coherent observations; processing multi-channel complex data using at least two second spatial processing configurations to obtain multiple complex output signals; generating spatial manifold relationship observations based on at least one of the normalized complex ratio, conjugate product, amplitude-phase difference, normalized vector, cross spectrum, or coherence coefficient among the multiple complex output signals; the spatial manifold relationship observations retain information related to the target spatial state, near-field curvature, scattering structure, micro-motion, or relative state of array channels by eliminating or normalizing common scattering complex amplitude, common gain, or common phase.

[0007] In conjunction with the first aspect mentioned above, one possible implementation involves constructing a joint statistical model of coherent observations and spatial manifold relation observations, while preserving the cross covariance when both share the original samples. This includes: converting coherent observations and spatial manifold relation observations into augmented real vectors containing real and imaginary parts; constructing a complete joint covariance matrix based on the augmented real vectors, including the covariance of coherent observations, the covariance of relation observations, and the cross covariance; obtaining the complete joint covariance matrix through at least one of analytical error propagation, sample covariance, resampling, Monte Carlo, or joint likelihood, and performing positive definiteness, contraction estimation, eigenvalue lower bound protection, or condition number monitoring on the complete joint covariance matrix; and inflating or reducing the weight of the relation observation covariance when the denominator amplitude of the spatial manifold relation observation is lower than a preset amplitude threshold, a phase jump occurs, the number of samples is insufficient, or the complete joint covariance matrix is ​​ill-conditioned.

[0008] In conjunction with the first aspect mentioned above, one possible implementation involves establishing a joint measurement model that includes target parameters and interference parameters, and performing conditional elimination on the interference parameters to obtain valid conditional information for the target parameters. This includes: constructing a joint Fisher information matrix for the target parameters and interference parameters based on the joint statistical model; performing Schur complement operation on the interference parameters in the joint Fisher information matrix to obtain valid conditional information for the target parameters; or projecting the whitened target parameter Jacobian matrix onto the orthogonal complement space of the column space of the interference parameter Jacobian matrix to obtain valid conditional information for the target parameters. The valid conditional information retains only the variation information of the target parameters that cannot be explained by the degrees of freedom of the interference parameters.

[0009] In conjunction with the first aspect above, in one possible implementation, selecting or updating the spatial processing configuration based on the sensitivity index of coherent observations and the quality index of effective condition information includes: determining a set of candidate spatial configurations; each candidate spatial configuration in the set of candidate spatial configurations satisfying preset constraints; the preset constraints include at least one of the following: minimum coherent signal-to-noise ratio, minimum complex response amplitude, interference suppression, clutter suppression, phase fidelity, hardware dynamic range, transmit power, bandwidth, and configuration switching continuity; for each candidate spatial configuration in the set of candidate spatial configurations, calculating the corresponding sensitivity index of coherent observations and the quality index of effective condition information; the quality index includes at least one of the following: minimum eigenvalue, determinant, trace, posterior covariance, or dangerous ambiguity direction distance of effective condition information; and selecting from the set of candidate spatial configurations the candidate spatial configuration that maximizes the quality index and satisfies the preset threshold of the sensitivity index as the updated spatial processing configuration.

[0010] In conjunction with the first aspect above, in one possible implementation, the calculation process of the dangerous fuzzy direction distance includes: determining a dangerous fuzzy direction set; the dangerous fuzzy direction set includes at least one of angle grid neighbor points, distance period subpeaks, velocity folding states, scattering center exchange states, image sidelobe states, or near-field mirror states; for each candidate target state in the dangerous fuzzy direction set, calculating the profile statistical distance between the real target state and the candidate target state, while allowing the interference parameters to be adjusted; determining the minimum value of the profile statistical distance as the dangerous fuzzy direction distance; the profile statistical distance is used to quantify the distinguishability between the real target state and the candidate target state after eliminating the influence of interference parameters.

[0011] In conjunction with the first aspect mentioned above, in one possible implementation, the method further includes an incremental information verification and automatic degradation step. This step includes: calculating the first effective condition information using only coherent observations, and the second effective condition information using both coherent observations and spatial manifold relationship observations; determining an incremental information index based on the first and second effective condition information; the incremental information index being the difference between the two or a normalized ratio; triggering an automatic degradation mechanism when the incremental information index is below a preset information threshold, the covariance condition number of the joint statistical model is greater than a preset condition number threshold, the denominator amplitude of the spatial manifold relationship observation is below a preset amplitude threshold, or the model residual exceeds a preset residual threshold; the automatic degradation mechanism includes: closing spatial manifold relationship observations, reducing the weight of spatial manifold relationship observations, or expanding the covariance of spatial manifold relationship observations, causing the system to degrade to a processing mode using only coherent observations.

[0012] In conjunction with the first aspect mentioned above, in one possible implementation, the automatic degradation mechanism further includes state machine control logic, which includes: maintaining the processing mode of using only coherent observations when the output signal-to-noise ratio of coherent observations meets a preset detection threshold; maintaining the low-rate parameter tracking mode and requesting the generation of a new spatial processing configuration when the output signal-to-noise ratio of coherent observations does not meet the preset detection threshold and the relative phase of the spatial manifold relation observations remains continuous; and re-entering the joint precision mode of using both coherent observations and spatial manifold relation observations when the new spatial processing configuration is adopted, and the recalculated incremental information index recovers to a level higher than the preset information threshold and the covariance condition number of the joint statistical model recovers to a level lower than the preset condition number threshold.

[0013] In conjunction with the first aspect mentioned above, in one possible implementation, under the condition of satisfying preset constraints, the target parameters are jointly estimated using coherent observations and spatial manifold relationship observations. This includes: using at least one algorithm among maximum likelihood estimation, Bayesian estimation, Gauss-Newton method, Kalman filtering, factor graph inference, sparse optimization, or low-rank optimization to jointly estimate the target parameters; when there are multiple targets or extended targets, setting independent target states and scattering amplitudes for each target, and sharing channel parameters, calibration parameters, or clutter parameters among multiple targets or extended targets; handling target aliasing through at least one of joint data association, sparse reconstruction, or low-rank estimation, and using the scattering center exchange state as the dangerous ambiguity direction for separation evaluation.

[0014] Secondly, a dual-gain precision sensing system for array radar is provided, comprising: a multi-channel data interface for acquiring multi-channel complex data formed by at least two radar channels; an observation generation module for processing the multi-channel complex data using at least one spatial processing configuration to form coherent observations and spatial manifold relationship observations, wherein coherent observations are used to improve the effective signal-to-noise ratio or coherent accumulation capability, and spatial manifold relationship observations are used to preserve array spatial manifold information; a joint statistical modeling module for constructing a joint statistical model of coherent observations and spatial manifold relationship observations, and preserving cross-covariance when the two share original samples; a conditional elimination module for establishing a joint measurement model including target parameters and interference parameters, and performing conditional elimination on the interference parameters to obtain effective conditional information of the target parameters, wherein the interference parameters include at least one of common scattering complex amplitude, channel parameters, clutter parameters, and calibration parameters; a configuration design module for selecting or updating the spatial processing configuration based on the sensitivity index of coherent observations and the quality index of effective conditional information; and a joint estimation module for jointly estimating the target parameters using coherent observations and spatial manifold relationship observations under preset constraints.

[0015] The technical solution provided in this application solves the problem of repeated weighting of deterministic transformations caused by sharing original samples by clearly distinguishing between coherent sensitivity gain and spatial manifold information gain, and constructing a complete joint statistical model that preserves cross-covariance. This ensures the consistency between the estimated covariance and the actual error. By conditionally eliminating common scattering complex amplitude, channels, clutter, and calibration parameters, coupling interference unrelated to the target state is removed, ensuring that the obtained effective conditional information can truly reflect the identifiability of the target parameters, rather than just the signal-to-noise ratio level. Based on this, the selection of spatial processing configuration is driven by the quality index of the effective conditional information, and optimization is performed under the premise of meeting hard constraints such as minimum coherent output. This effectively avoids the problem of decreased angle measurement sensitivity or falling into deep null traps caused by traditional methods simply pursuing high gain. At the same time, the introduction of incremental information verification and automatic degradation mechanism can smoothly degrade to coherent processing mode when there is no incremental information or ill-conditioned covariance in relational observations, preventing the pollution of precise estimation by false information gain and significantly improving the adaptability and reliability of the system in complex electromagnetic environments and variable target scenarios.

[0016] It should be understood that the descriptions of technical features, technical solutions, beneficial effects, or similar language in this application do not imply that all features and advantages can be achieved in any single embodiment. Rather, it is understood that the description of a feature or beneficial effect means that a specific technical feature, technical solution, or beneficial effect is included in at least one embodiment. Therefore, the descriptions of technical features, technical solutions, or beneficial effects in this specification do not necessarily refer to the same embodiment. Furthermore, the technical features, technical solutions, and beneficial effects described in this embodiment can be combined in any suitable manner. Those skilled in the art will understand that embodiments can be implemented without one or more specific technical features, technical solutions, or beneficial effects of a particular embodiment. In other embodiments, additional technical features and beneficial effects may be identified in specific embodiments that do not embody all embodiments. Attached Figure Description

[0017] Figure 1 A flowchart illustrating a dual-gain precision sensing method for array radar provided in this application embodiment; Figure 2 A flowchart illustrating another dual-gain precision sensing method for array radar provided in this application embodiment; Figure 3 A flowchart illustrating another dual-gain precision sensing method for array radar provided in this application embodiment; Figure 4 A flowchart illustrating another dual-gain precision sensing method for array radar provided in this application embodiment; Figure 5 This is a schematic diagram of a dual-target space configuration provided in an embodiment of this application; Figure 6 A flowchart illustrating another dual-gain precision sensing method for array radar provided in this application embodiment; Figure 7 A schematic diagram of minimum statistical distance for a dangerous ambiguous direction provided in an embodiment of this application; Figure 8 A flowchart illustrating a dual-gain precision sensing method for array radar provided in this application embodiment; Figure 9 A schematic diagram illustrating the flow of an incremental information verification and automatic degradation state machine provided in an embodiment of this application; Figure 10 A flowchart illustrating another dual-gain precision sensing method for array radar provided in this application embodiment; Figure 11 This is a system architecture diagram of an array radar dual-gain precision sensing system provided in an embodiment of this application. Detailed Implementation

[0018] In the description of this application, unless otherwise stated, " / " means "or," for example, A / B can mean A or B. The "and / or" in this document is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A alone, A and B simultaneously, and B alone. Furthermore, "at least one" means one or more, and "multiple" means two or more. The terms "first," "second," etc., do not limit the quantity or order of execution, and "first," "second," etc., do not necessarily imply differences.

[0019] It should be noted that, in this application, the terms "exemplary" or "for example" are used to indicate that something is being described as an example, illustration, or illustration. Any embodiment or design described as "exemplary" or "for example" in this application should not be construed as being more preferred or advantageous than other embodiments or design solutions. Specifically, the use of terms such as "exemplary" or "for example" is intended to present the relevant concepts in a concrete manner.

[0020] Example 1: like Figure 1 As shown, this application provides a dual-gain precision sensing method for array radar. This method decouples the array gain into two types of observations: coherent sensitivity and spatial manifold information. It then performs joint optimization while eliminating common interference coupling to balance detection performance and parameter identification accuracy. The specific steps of this method include S101 to S106: S101. Acquire multi-channel complex data formed by at least two radar channels.

[0021] Among them, multi-channel complex data refers to the baseband digital signal sequence output by the radar receiver front end after down-conversion, analog-to-digital conversion and necessary preprocessing.

[0022] In this embodiment of the application, the IQ data of each channel of the array antenna is acquired in real time through the radio frequency front-end and the digital acquisition unit.

[0023] Optionally, the multi-channel data model satisfies the following formula:

[0024] in, For the first Multi-channel complex data at each observation time For the number of channels, Let be the target parameter vector. It is a vector composed of common scattering complex amplitude, channel parameters, clutter parameters, calibration parameters, and other interference parameters. Configuration for space processing For prediction models that include array manifold, transmission coding, propagation delay, target scattering, channel response, and clutter, To account for multi-channel noise and unmodeled errors, Its covariance matrix. The actual quantization weights, actual subarray connections, actual transmit codes, and corresponding version channel calibration parameters are all used as... The components are involved in the prediction.

[0025] S102. Process multi-channel complex data using at least one spatial processing configuration to form coherent observations and spatial manifold relationship observations.

[0026] Among them, coherent observation is used to improve the effective signal-to-noise ratio or coherent accumulation capability, while spatial manifold relation observation is used to preserve array spatial manifold information. Although both originate from the same set of data, they carry complementary physical information. Spatial processing configuration refers to the weighting, combination, or transformation rules applied to multi-channel data.

[0027] In the embodiments of this application, a first set of weights is used to synthesize multi-channel data in phase to output coherent observations with high signal-to-noise ratio for detection and tracking. At the same time, a second set or more sets of different weights, subarray divisions or encoding methods are used to obtain multiple complex outputs, and relational observations are constructed based on the ratios, conjugate products or phase differences between these outputs.

[0028] Based on the above steps, the functional decoupling of array gain is realized, enabling the system to simultaneously obtain high-sensitivity energy indicators for maintaining stable tracking and high-resolution structural indicators for improving parameter estimation accuracy, thus resolving the contradiction that detection and identification performance are difficult to balance under a single observation mode.

[0029] S103. Construct a joint statistical model of coherent observations and spatial manifold relation observations, and retain the cross covariance when the two share the original samples.

[0030] Among them, the joint statistical model is a complete mathematical description of the two types of observational uncertainty and their correlation. Cross covariance specifically refers to the statistical correlation term generated because coherent observations and related observations originate from the same set of random samples. Ignoring this term will lead to an incorrect assessment of the amount of information.

[0031] In the embodiments of this application, coherent observations and relational observations are concatenated into augmented real vectors, and a complete joint covariance matrix containing their respective self-covariances and cross-covariances is constructed. This matrix truly reflects the joint probability distribution characteristics of the two types of observations under finite sample conditions.

[0032] It should be noted that existing technologies often assume that different observations are independent of each other and set the cross term to zero. However, in this application, since the relational observations are often derived from coherent observations or their underlying samples through nonlinear transformation, there is a definite functional coupling between the two. Preserving the cross covariance is a key measure to prevent information from being weighted repeatedly and to avoid the estimated variance being underestimated artificially.

[0033] Based on the above steps, a complete statistical description framework that conforms to physical reality was established, ensuring statistical consistency in the subsequent information fusion process, eliminating false precision gains caused by model mismatch, and improving the credibility of precision sensing results.

[0034] S104. Establish a joint measurement model that includes target parameters and interference parameters, and perform conditional elimination on the interference parameters to obtain the effective conditional information of the target parameters.

[0035] The interference parameters include at least one of the common scattering complex amplitude, channel parameters, clutter parameters, and calibration parameters. Conditional elimination is a mathematical projection operation designed to remove components that can be explained by the interference parameters from the total information, thereby separating out the effective information that purely reflects the changes in the target parameters.

[0036] In this embodiment, a joint Fisher information matrix containing the target state and various interference factors is constructed. By using Schur complement operation or orthogonal projection method, non-target degrees of freedom such as common scattering amplitude and channel amplitude and phase error are removed from the information space. The resulting effective condition information only contains the target feature changes that cannot be fitted by the interference parameters.

[0037] Based on the above steps, a pure target information metric is obtained after removing interference from environmental and non-ideal factors. This provides an objective and robust performance benchmark for evaluating and designing spatial processing configurations, avoiding inflated system performance due to neglecting parameter coupling.

[0038] S105. Based on the sensitivity index and quality index of coherent observation and effective condition information, select or update the spatial processing configuration.

[0039] Among them, sensitivity indicators usually refer to the output signal-to-noise ratio or detection probability, which are used to ensure basic detection capabilities; quality indicators refer to the minimum feature value, determinant, or dangerous ambiguity direction distance of effective condition information, which are used to quantify the identifiability of precise parameters; configuration selection is the process of optimizing the latter under the premise of meeting the threshold of the former.

[0040] In this embodiment, the candidate space configuration set is traversed, and a feasible subset that meets the minimum detection signal-to-noise ratio requirement is first selected. Then, the configuration that maximizes the quality index of effective condition information in the subset is selected as the current working mode, thereby achieving a Pareto optimal balance between detection and identification performance.

[0041] Based on the above steps, intelligent closed-loop control of space resource allocation is realized, enabling the radar to dynamically adjust its "observation angle" according to the real-time environment and mission requirements, and maintain optimal parameter extraction capability at all times while ensuring that the target is not lost.

[0042] S106. Under the condition of satisfying the preset constraints, estimate the target parameters by jointly using coherent observations and spatial manifold relationship observations.

[0043] Among them, the preset constraints are the concretization of engineering implementation and physical security boundaries, including but not limited to hardware dynamic range, transmit power limit, sidelobe level, phase continuity and minimum complex response amplitude, etc. They are hard prerequisites for configuration selection and estimation algorithm execution. Joint estimation refers to the execution process of optimally solving the angle, distance, velocity or micro-motion parameters of the target based on the constructed complete statistical model and effective information measurement, using algorithms such as maximum likelihood, Bayesian inference or weighted least squares.

[0044] In this embodiment, coherent and relational observations verified by conditional elimination are input into a unified estimation algorithm framework. The joint covariance matrix is ​​used as the optimal weighting coefficient. Under the configuration that satisfies engineering constraints, the estimated value of the target parameter and its posterior uncertainty are output, thus completing the final transformation from data to information.

[0045] It should be noted that the joint estimation here is not a simple data splicing, but a deep integration of the complementary advantages of the two observations: coherent observations provide high-precision distance / velocity benchmarks and energy references, while relational observations provide high-resolution angle / structure constraints. The two correct each other in the sense of statistical optimality and converge together to the global optimal solution.

[0046] Based on the above steps, the final implementation of dual-gain precision sensing was completed, transforming the statistical advantages and information purity accumulated in the previous steps into actual target state estimation accuracy, and achieving a simultaneous improvement in detection sensitivity and parameter identification capability.

[0047] Based on the above technical solution, the traditional single-gain mode of array radar is upgraded to a dual-gain cooperative mode of coherent sensitivity and spatial manifold information. By preserving the cross covariance, the information of shared samples is avoided from being weighted repeatedly. By eliminating interference parameters, effective information that truly reflects the identifiability of the target is obtained. This information drives the adaptive optimization and joint estimation of spatial configuration, thereby significantly improving the radar's ability to perceive the precise state of the target and its environmental adaptability while ensuring detection power.

[0048] Example 2: In one possible implementation of the embodiments of this application, combined with Figure 1 ,like Figure 2 As shown, S102 and S103 can be specifically implemented through S201 to S207, which are explained in detail below: S201. Using the first spatial processing configuration, coherent synthesis of multi-channel complex data is performed to obtain coherent observations.

[0049] The first spatial processing configuration refers to the weighted vector or beamforming weight designed based on maximizing the signal energy or signal-to-noise ratio in the target direction. Its output directly reflects the absolute intensity and phase reference of the target echo.

[0050] In this embodiment, the weight of the main beam pointing in the expected direction of the target is used to perform weighted summation on the multi-channel complex data. This process realizes the acquisition of array gain, provides a highly sensitive energy reference for subsequent detection, range Doppler estimation and tracking, and also serves as a reference signal source for constructing relationship observation.

[0051] Alternatively, coherent observations satisfy the following formula:

[0052] in, For spatial configuration a, coherent sensitivity observations, This is the complex weight vector corresponding to this configuration, with the superscript indicating the conjugate transpose.

[0053] Based on the above steps, a high signal-to-noise ratio energy dimension observation was obtained, which not only met the basic detection requirements, but also provided a stable phase and amplitude reference for subsequent extraction of relative structure information.

[0054] S202. Process the multi-channel complex data using at least two second-space processing configurations to obtain multiple complex output signals.

[0055] Among them, the weighting method with spatial resolution differences between the two second spatial processing configurations, such as offset beam, subarray synthesis, weights corresponding to different MIMO coding or frequency subbands, aims to sample the spatial response of the target from different perspectives or dimensions.

[0056] In the embodiments of this application, multiple sets of second spatial processing configurations are applied in parallel or in a time-division manner to generate a set of complex output signal sequences containing rich spatial difference information. Although these signals may be lower than the main beam in absolute level, their relative relationships carry fine features such as the angle of the target's deviation from the beam center, near-field curvature, or scatterer distribution.

[0057] Based on the above steps, diverse signal samples carrying spatial manifold information were obtained, providing the necessary physical basis for constructing relative observables that can be isolated from common interference.

[0058] S203. Based on at least one of the normalized complex ratio, conjugate product, amplitude-phase difference, normalized vector, cross spectrum, or coherence coefficient among multiple complex output signals, generate spatial manifold relationship observations.

[0059] Among them, spatial manifold relationship observation retains information related to the target's spatial state, near-field curvature, scattering structure, micro-motion, or array channel relative state by eliminating or normalizing common scattering complex amplitude, common gain, or common phase. It is a relative quantity derived through nonlinear transformation. Its core mathematical feature is that the numerator and denominator contain the same common factors (such as target RCS, transmit power, path loss, etc.). These factors are canceled out by ratio or conjugate operation, thereby achieving decoupling of the target's absolute intensity.

[0060] In the embodiments of this application, the multiple complex output signals obtained are combined in pairs or normalized as a whole to construct dimensionless or normalized feature quantities that depend only on array geometry, target position and scattering characteristics.

[0061] Optionally, normalized complex ratio and normalized conjugate product Satisfy the following formula:

[0062]

[0063] in, and These represent the coherent output complex signals at time k under configurations a and b, respectively. This indicates the conjugate operation. This represents the modulo operation. Normalized conjugate product. The amplitude effect was eliminated by taking the phase portion of the conjugate product. It is a complex ratio.

[0064] Relationship observation can also employ inter-beam amplitude and phase difference, subarray ratio, cross spectrum, coherence coefficient, closed-loop phase, normalized element vector, and their equivalent statistics. For a vector composed of L configured outputs... A normalized relation observation is represented as:

[0065] in, Denotes the Euclidean norm. To prevent regularization constants with denominators of zero, This represents the normalized relational observation vector.

[0066] Based on the above steps, it is possible to successfully extract pure spatial structure information that is immune to target intensity fluctuations and environmental disturbances, thus solving the problem of strong targets masking weak features or weak targets being unable to be precisely measured.

[0067] S204. Convert coherent observations and spatial manifold relation observations into augmented real vectors containing real and imaginary parts.

[0068] Since the real and imaginary parts of complex observations are often correlated, and the nonlinear transformation of the relational observations may cause the probability distribution on the complex plane to no longer be a rotationally symmetric circular Gaussian distribution, the augmented real vector is a standard mathematical tool for mapping the non-circular or asymmetric statistical properties of the complex field to the real field.

[0069] In the embodiments of this application, after concatenating the coherent observation complex scalar or vector with the relational observation complex vector, a joint observation representation vector is obtained. The real and imaginary parts are taken respectively and arranged in order to form a real-valued column vector with doubled dimension.

[0070] Optionally, augmented real vectors Satisfy the following formula:

[0071] in, For a joint complex observation vector containing coherent and relational observations, and These represent the operations of taking the real part and the imaginary part, respectively. This indicates transpose. Where, , For coherent observation vectors, is the spatial manifold relation observation vector.

[0072] It should be noted that if the complex covariance matrix is ​​used directly without augmentation, the pseudo-covariance information may be lost, resulting in information loss or estimation bias in the observation and processing of non-circular signals or relationships after nonlinear transformation.

[0073] Based on the above steps, a complete statistical expression framework for the real number field was established, laying a mathematical foundation for accurately characterizing the joint uncertainty of two types of heterogeneous observations.

[0074] S205. Based on augmented real vectors, construct a complete joint covariance matrix including coherent observation covariance, relational observation covariance, and cross covariance.

[0075] The complete joint covariance matrix is ​​a block matrix. Its diagonal blocks describe the fluctuation characteristics of various observations themselves, while the off-diagonal blocks, i.e., the cross covariance, accurately quantify the statistical dependence caused by sharing the same set of original snapshot data.

[0076] In this embodiment, the autocorrelation matrix of the augmented real vector is calculated and divided into four sub-blocks. Optionally, the complete joint covariance matrix... Satisfy the following formula:

[0077] in, The autocovariance matrix of coherent observations, The autocovariance matrix of the relation observations, and These are the cross-covariance matrices and their transposes between coherent and relational observations, respectively. Specifically, This reflects how the coherent output and the ratio observation derived from it change in tandem as the noise in the raw data increases.

[0078] Based on the above steps, a complete statistical model that conforms to physical facts was constructed, avoiding the problem of information duplication due to neglecting the commonality of data, and ensuring the credibility of the precision sensing results.

[0079] S206. Obtain the complete joint covariance matrix by at least one of analytical error propagation, sample covariance, resampling, Monte Carlo, or joint likelihood, and perform positive definiteness, shrinkage estimation, eigenvalue lower bound protection, or condition number monitoring on the complete joint covariance matrix.

[0080] The choice of acquisition method depends on the system's trade-off between real-time performance and accuracy, as well as the linearity of the observation model; positive definiteness and shrinkage estimation are used to deal with matrix singularity or ill-conditioned problems caused by finite samples, and to ensure the numerical stability of subsequent inversion operations.

[0081] In this embodiment, for weakly nonlinear regions, an analytical error propagation method based on first-order Taylor expansion is used to quickly calculate the covariance; for strongly nonlinear or unknown distribution scenarios, the covariance is estimated using historical data or the sample covariance of the current snapshot, and the matrix condition number is corrected by combining Ledoit-Wolf contraction or eigenvalue truncation techniques. Optionally, the analytical error propagation approximately satisfies the following formula:

[0082] in, The augmented real covariance matrix of the original multi-channel data. For the nonlinear transformation function from raw data to joint observations, For this transformation at the current estimation point The Jacobian matrix at that location.

[0083] When nonlinearity is strong, the sample distribution is non-Gaussian, or analytical Jacobians are difficult to obtain, sample covariance can be used for estimation. Taking a group of valid samples as an example, the sample covariance satisfies the following formula:

[0084] in, For sample covariance estimation, The number of valid sample groups. For the first Groups to expand the actual observation sample, This is the sample mean. The system... Perform positive definiteness, contraction estimation, eigenvalue lower limit protection, and condition number monitoring.

[0085] It should be noted that analytical methods are computationally efficient but rely on linearization assumptions, while sample methods are highly adaptable but converge slowly and are susceptible to outliers. In practical engineering, a hybrid strategy is often adopted, which uses the analytical results as initial values ​​and a small number of samples for online correction, thus balancing real-time performance and robustness.

[0086] Based on the above steps, adaptive high-precision acquisition and numerical reinforcement of the joint covariance matrix were achieved, ensuring that the algorithm can run stably under various working conditions without numerical divergence.

[0087] S207. When the denominator amplitude of the spatial manifold relation observation is lower than the preset amplitude threshold, the phase jumps, the number of samples is insufficient, or the complete joint covariance matrix is ​​ill-conditioned, the relation observation covariance is inflated or its weight is reduced.

[0088] Among them, the inflation process refers to artificially increasing the eigenvalues ​​of the covariance sub-blocks corresponding to the relational observations, which is equivalent to reducing the signal-to-noise ratio weight of the observation in the joint estimation, so that its influence on the final result is weakened or even approaches zero.

[0089] In this embodiment, the signal strength of the denominator and the condition number of the joint covariance matrix are monitored in real time during the relationship observation generation process. Once the preset abnormality criterion is triggered, the protection mechanism is immediately activated. The system is multiplied by an expansion factor greater than 1, or the fusion weight of the relationship observation is continuously reduced according to the quality index, so that the system smoothly degenerates into a processing mode dominated by coherent observation.

[0090] Based on the above steps, an automatic fault-tolerance mechanism for relational observation failure scenarios was established, ensuring that the dual-gain precision sensing system can maintain basic functional integrity and result reliability under extreme or harsh conditions, and avoiding engineering failures caused by pursuing theoretical optimization.

[0091] Based on the above technical solutions, the abstract concept of dual observation is translated into specific signal processing operators and statistical modeling processes. The construction mechanism of normalized relational observations and the physical essence of eliminating common interference are clarified. By preserving cross covariance and handling ill-conditioned expansion, the problems of double weighting and numerical instability in the fusion of homogeneous data are solved, providing solid and reliable data and model support for subsequent configuration optimization based on effective condition information.

[0092] Example 3: In one possible implementation of the embodiments of this application, combined with Figure 1 ,like Figure 3 As shown, the above S104 can be implemented through the following S301, S302 and S303, which are explained in detail below: S301. Based on the joint statistical model, construct the joint Fisher information matrix of the target parameters and interference parameters.

[0093] The joint Fisher information matrix is ​​a statistical measure of the total amount of information provided by the observed data to the unknown parameter vector. Its construction depends on the complete joint covariance matrix and the sensitivity of the observation model to each parameter.

[0094] In this embodiment, the augmented real vector is first uniformly written as:

[0095] in, This is an augmented observational prediction model that incorporates predictions from coherent and relational observations. To observe the noise, This is the joint observation covariance matrix. Let... The symmetric positive definite square root of the target parameter Jacobian is used to whiten the residuals and Jacobian. The whitened target parameter Jacobian and the interference parameter Jacobian satisfy the following formulas respectively:

[0096] in, The target parameter Jacobian matrix after whitening. The Jacobian matrix represents the interference parameters after whitening. The symmetric positive definite square root of the inverse of the joint covariance matrix. and These are the partial derivative matrices of the observation and prediction models with respect to the target parameters and disturbance parameters, respectively.

[0097] Based on the aforementioned whitening Jacobian matrix, combined with the Fisher information matrix Satisfy the following formula:

[0098] in, This represents the self-information block of the target parameters. The self-information block representing the interference parameters. and This represents a block of mutual information between two types of parameters. and These are the target parameter Jacobian matrix and the interference parameter Jacobian matrix after whitening, respectively.

[0099] As an example, in a precision angle measurement scenario, the target parameter is the azimuth angle, and the interference parameters include the target RCS complex amplitude and channel phase error, combined with the mutual information blocks in the Fisher information matrix. A non-zero value indicates that the accuracy of angle estimation is directly limited by the identifiability of amplitude and phase errors. Without processing this coupling term, a simple assessment of angle information would be distorted.

[0100] Based on the above steps, the statistical characteristics of the observation space are mapped to the information metric of the parameter space, quantifying the global correlation strength between the target parameter and the interference parameter under the current observation configuration, and providing a precise input object for subsequent decoupling operations.

[0101] S302. Perform Schur complement operation on the interference parameters in the joint Fisher information matrix to obtain the effective condition information of the target parameters; or, project the whitened target parameter Jacobian matrix onto the orthogonal complement space of the column space of the interference parameter Jacobian matrix to obtain the effective condition information of the target parameters.

[0102] Among them, Schur complement operation and orthogonal complement projection are two mathematically equivalent conditional elimination methods. The former focuses on algebraically removing the influence of interference parameters from the information matrix, while the latter focuses on geometrically projecting the sensitivity vector of the target parameter onto a subspace orthogonal to the interference parameters.

[0103] In the embodiments of this application, any implementation method can be selected according to the requirements of computational efficiency and numerical stability. When Schur complement is used, mutual information coupling is eliminated by the generalized inverse of the self-information block of the interference parameters; when orthogonal projection is used, an orthogonal complement projection matrix of the Jacobian column space of the interference parameters is constructed and applied to the target parameter Jacobian.

[0104] Optional, valid condition information The calculation satisfies one of the following formulas:

[0105] or

[0106] in, This is the effective condition information matrix for the target parameters. This represents the Moore-Penrose generalized inverse, used to accommodate cases where the disturbance parameter is locally rank deficient. It is the identity matrix. Let be the orthogonal complementary projection matrix of the Jacobian space of the interference parameters.

[0107] Based on the above steps, the mathematical decoupling of target parameter information and interference parameter information is achieved. The resulting effective condition information matrix no longer contains any components that can be explained by interference parameters, and truly reflects the identifiable upper limit of the target parameters themselves under the current observation conditions.

[0108] S303. Determine the effective condition information and retain only the change information of the target parameter that cannot be interpreted by the degrees of freedom of the disturbed parameters.

[0109] Among them, the degree of freedom that cannot be interpreted by the interference parameter means that the change vector of the observation response caused by the change of the target parameter cannot be represented as a linear combination of the change vector of the observation response caused by the change of the interference parameter, that is, the two are linearly independent in the observation space.

[0110] In this embodiment of the application, after conditional elimination, the remaining information matrix The eigenvalue distribution directly characterizes the net discriminability of the target parameters after removing all known interference sources. This indicator serves as the sole information quality benchmark for subsequent spatial configuration optimization, ensuring that the configuration design targets the true physical resolution rather than the apparent signal-to-noise ratio.

[0111] As an example, for single-pulse angle measurement, if the target is located at the peak of the sum beam, the amplitude change caused by the angle change is zero, and the angle information depends entirely on the phase slope of the difference beam. If the system only uses the sum beam for observation, no matter how high the signal-to-noise ratio is, the effective angle information after conditional elimination is zero. This accurately reveals the physical essence that a single sum beam cannot measure angles.

[0112] Based on the above steps, a precision perception evaluation system with "effective condition information" as its core was established, correcting the cognitive bias of substituting energy gain for information gain, and providing theoretical support and engineering criteria for radar systems to achieve robust and reliable precision parameter estimation in complex electromagnetic environments.

[0113] Based on the above technical solution, the abstract concept of parameter decoupling is transformed into a computable joint Fisher information matrix and Schur complement / orthogonal projection operation, which enables the system to quantitatively evaluate and eliminate the false contributions of interference factors such as common scattering amplitude and channel error to precision sensing.

[0114] Example 4: In one possible implementation of the embodiments of this application, combined with Figure 1 ,like Figure 4 As shown, the above S105 can be implemented through the following S401 and S402, which are explained in detail below: S401. Determine the candidate space configuration set, where each candidate space configuration in the candidate space configuration set satisfies the preset constraints.

[0115] The preset constraints are hard boundaries that ensure the basic functions and hardware security of the radar system, including at least one of the following: minimum coherent signal-to-noise ratio, minimum complex response amplitude, interference suppression, clutter suppression, phase fidelity, hardware dynamic range, transmit power, bandwidth, and configuration switching continuity. The candidate configuration space set is the search space consisting of all feasible configurations that simultaneously satisfy the above constraints, rather than a complete set containing all theoretical possibilities.

[0116] In this embodiment, threshold values ​​for various constraints are first set according to the current task requirements and environmental conditions. For example, the minimum coherent signal-to-noise ratio is set to 3dB above the detection threshold to ensure a stable false alarm rate, and the minimum complex response amplitude is set to the lower limit of the receiver's linear region to prevent numerical singularities caused by excessively small relational observation denominators. Then, the predefined configuration library is traversed or a configuration sequence is generated through a real-time optimization algorithm, and each configuration is checked to see if it falls within the feasible region. Only compliant configurations are retained as candidates for subsequent optimization.

[0117] Optionally, a set of candidate space configurations Satisfy the following formula:

[0118] in, Indicates configuration The coherent output signal-to-noise ratio at the following conditions. This is the minimum signal-to-noise ratio threshold. This represents the minimum complex response magnitude in the target neighborhood. The minimum amplitude threshold, Indicates the sidelobe or clutter level. This is the upper limit of sidelobe suppression. Indicates transmit power or hardware load. This refers to the upper limit of power or load. This indicates the difference between the current configuration and the previous configuration. Discontinuity of switching between them To switch the smoothness threshold.

[0119] Optionally, the coherent output signal-to-noise ratio satisfies: ,in For the target power, The target direction array response vector. For space configuration The corresponding complex weight vector, This represents the covariance matrix of interference and noise. The output signal-to-noise ratio (SNR) measures the combined effect of the coherent branch's energy harvesting efficiency and interference / noise suppression capability.

[0120] It should be noted that this "constraint-first, optimization-later" strategy avoids the risk of the system operating in deep zero traps, saturation regions, or transiently unstable regions due to blindly pursuing theoretical optimality in traditional design. Figure 5 As shown, only configurations located on the Feasible Frontier are eligible for selection, while configurations that, although containing a large amount of information, violate security constraints are strictly eliminated, ensuring the engineering robustness of the precision perception algorithm.

[0121] Based on the above steps, a safety barrier connecting theoretical performance and engineering practice is constructed, ensuring that subsequent configuration optimization is always carried out within a controllable and reliable range, preventing performance crashes caused by model mismatch or sudden environmental changes.

[0122] S402. For each candidate spatial configuration in the candidate spatial configuration set, calculate the corresponding sensitivity index of coherent observation and the quality index of effective condition information, and select the candidate spatial configuration that maximizes the quality index and meets the preset threshold of the sensitivity index from the candidate spatial configuration set as the updated spatial processing configuration.

[0123] Sensitivity metrics quantify the configuration's ability to collect target energy, typically represented by the output signal-to-noise ratio (SNR). Quality metrics quantify the configuration's ability to distinguish precise target parameters, including at least one of the following: minimum eigenvalue of effective conditional information, determinant, trace, posterior covariance, or distance to the dangerous ambiguity direction. The essence of configuration selection is to maximize parameter discriminability while meeting the sensitivity baseline.

[0124] In this embodiment, for each candidate configuration selected, its dual indicators are calculated in parallel or serially; then, a constraint optimization strategy is adopted to regard all configurations that meet the preset sensitivity threshold as the qualified set, and the configuration that maximizes the quality index is selected from the qualified set as the current working mode.

[0125] Optionally, optimal configuration The choice of satisfies the following formula:

[0126] in, For the comprehensive quality evaluation function, the effective condition information matrix can be taken. The smallest eigenvalue Logarithmic determinant The reciprocal of the posterior covariance trace Or the direction and distance of danger are unclear Alternatively, a weighted combination of the above indicators or the results of a multi-objective Pareto ranking can be used. This represents the parameter variable that maximizes the function value.

[0127] It should be noted that this mechanism corrects the cognitive bias that "high gain equals high precision." For example... Figure 5 As shown, although the signal-to-noise ratio of configuration 6 is slightly lower than that of configuration 3, its spatial information discernibility is significantly higher. Therefore, it is selected first when the minimum SNR constraint is met. This shows that in some scenarios, moderately sacrificing peak gain in exchange for a steeper single-pulse slope or a larger fuzzy direction distance can significantly improve the overall precision sensing performance of the system, realizing a design paradigm shift from "energy center" to "information center".

[0128] Based on the above steps, intelligent closed-loop control of space resource allocation is realized, enabling the radar to dynamically adjust its observation angle according to the real-time environment and mission requirements, and to maintain optimal parameter extraction capability while ensuring that the target is not lost, thus solving the problem of difficulty in balancing detection and identification performance.

[0129] In some embodiments, to further enhance the targeting and anti-ambiguity of configuration selection, the dangerous ambiguity direction distance is introduced as a key quality indicator. This is a local enhancement and supplement to the global measurement of effective condition information, such as... Figure 6 As shown, the calculation process for the distance in the hazard ambiguity direction includes S601 and S602: S601. Determine the set of dangerous fuzzy directions, which includes at least one of the following: angle grid neighbor points, distance period subpeaks, velocity folding state, scattering center exchange state, image sidelobe state, or near-field mirror state.

[0130] Among them, dangerous fuzzy directions refer to the set of candidate states that are most likely to be confused with the true target state under a specific observation configuration. These states are "dangerous" because they are very similar to the true state in the observation space and are easily masked by noise, interference, or model errors, leading to serious misestimation or spurious correlation.

[0131] In the embodiments of this application, a set of dangerous ambiguity directions is dynamically constructed based on the current radar system, waveform parameters, and prior target information. For example, for a digital beamforming system, adjacent points of the angle search grid are included in the set to assess the risk of discretization error; for a pulse Doppler radar, range ambiguity subpeaks and velocity folding states are included in the set to assess the risk of periodic ambiguity; for high-resolution imaging, scattering center exchange states and image sidelobe positions are included in the set to assess the risk of structural confusion.

[0132] It should be noted that the definition of this set is not fixed, but rather adaptively adjusted according to the task stage and environment. In the coarse search stage, the focus is on a large range of grid neighbors and secondary peaks; in the fine search stage, the focus is on minute scattering center exchanges and near-field mirror states. This dynamic focusing mechanism ensures that computational resources are always concentrated on the most critical source of ambiguity, avoiding indiscriminate traversal of all possible states and significantly improving real-time performance.

[0133] Based on the above steps, a risk perception mechanism for specific physical scenarios was established, transforming the abstract "identifiability" into a defensive capability targeting specific error-prone points, thus giving the configuration design a clear tactical orientation.

[0134] S602. For each candidate target state in the dangerous fuzzy direction set, under the condition that the interference parameters are allowed to be adjusted, calculate the profile statistical distance between the real target state and the candidate target state, and determine the minimum value of the profile statistical distance as the dangerous fuzzy direction distance.

[0135] In this context, the profile statistical distance refers to the measure of the remaining, non-eliminable differences between two comparison states in the observation space after optimal matching adjustments have been made to allow interference parameters (such as common scattering amplitude and channel phase) for each state. This distance reflects the pure, essential distinguishability between the two states after eliminating all similarities that can be explained by interference parameters.

[0136] In this embodiment of the application, for each candidate state in the danger set, a constrained minimization problem is solved, that is, to find the combination of interference parameters that minimizes the difference between the predicted observations of the two states, and the minimum difference value is taken as the profile statistical distance between the two states; then the minimum value of the distances of all candidate states is taken as the danger ambiguity direction distance of the current configuration.

[0137] Optionally, profile statistical distance Distance from the direction of danger Satisfy the following formula:

[0138]

[0139] in, and These represent the true target state and the candidate fuzzy state, respectively. and For each of the corresponding interference parameter vectors, Configure the current space processing. The inverse of the joint observation covariance matrix is ​​used as a weighting metric. Index for the set of dangerous and ambiguous directions. For a set of fuzzy directions of danger, This represents the squared Mahalanobis distance.

[0140] It should be pointed out that, as Figure 7 As shown, the dangerous ambiguity direction distance differs from simple Euclidean distance or Fisher information distance without elimination; it explicitly considers the "masking" effect of interference parameters. If two states can become identical by adjusting their RCS or channel phase, their profile statistical distance is zero, indicating that they are absolutely indistinguishable under the current configuration. Conversely, a larger distance indicates that even under the worst interference parameter coupling, the two still maintain sufficient separability. Using this metric to drive configuration selection can directly reduce the probability of misjudgment of the most easily confused states and improve the system's confidence in complex environments.

[0141] Based on the above technical solutions, a configuration-driven framework of "constraint satisfaction + performance optimization + risk defense" was constructed. This not only ensures that the system operates within the safety boundary, but also compensates for the deficiency of the global information quantity index being insensitive to local error-prone points by introducing the micro-discriminability index of dangerous ambiguity direction distance. This enables the radar to avoid ambiguity and confusion risks to the greatest extent while ensuring detection sensitivity, and achieves a substantial improvement in precision sensing performance.

[0142] Example 5: In one possible implementation of the embodiments of this application, combined with Figure 1 ,like Figure 8 As shown, this application also includes incremental information verification and automatic degradation steps, which can be implemented through the following steps S801 and S802, and are described in detail below: S801. Calculate the first effective condition information when using only coherent observations, and the second effective condition information when using coherent observations and spatial manifold relationship observations together, and determine the incremental information index based on the two.

[0143] Among them, the incremental information index is used to quantify the amount of new independent information provided by spatial manifold relation observations relative to coherent observations. Its core purpose is to verify whether relation observations truly improve the identifiability of target parameters, rather than creating the illusion of information duplication simply because more data dimensions are introduced.

[0144] In the embodiments of this application, the conditional elimination method is used to calculate the effective conditional information matrix for the sub-model containing only coherent observations and the complete joint model containing two observations, respectively, and then the difference or normalized ratio between the two is obtained.

[0145] Optionally, incremental information matrix Ratio of normalized incremental information Satisfy the following formula:

[0146]

[0147] in, This represents the first effective conditional information matrix when only coherent observations are used. This represents the second effective conditional information matrix when using two types of observations together. Represents the trace operation of a matrix. To prevent regularization constants where the denominator is zero. When When it approaches zero, it indicates that the relationship observation did not provide substantial new information; when When the value is significantly greater than zero, it indicates that the relational observation effectively improves the accuracy of parameter estimation.

[0148] Based on the above steps, an observation effectiveness evaluation mechanism based on information quality rather than simply signal energy was established, providing an objective and quantitative basis for subsequent adaptive degradation and ensuring the statistical consistency of precision sensing results.

[0149] S802. When the incremental information index is lower than the preset information threshold, the covariance condition number of the joint statistical model is greater than the preset condition number threshold, the denominator amplitude of the spatial manifold relationship observation is lower than the preset amplitude threshold, or the model residual exceeds the preset residual threshold, the automatic degradation mechanism is triggered.

[0150] Among them, the automatic degradation mechanism is a defensive control logic that dynamically adjusts the observation fusion strategy based on the real-time information quality. Its triggering conditions cover four typical abnormal scenarios: insufficient information content (low incremental index), numerical instability (large condition number), low signal-to-noise ratio (low denominator amplitude), and model mismatch (large residual).

[0151] In this embodiment, the system monitors the above four indicators in parallel during each processing cycle. If any indicator exceeds its limit, a degradation procedure is initiated. The automatic degradation mechanism includes: disabling the spatial manifold relation observations, reducing the weight of the spatial manifold relation observations, or inflating the covariance of the spatial manifold relation observations, thus degrading the system to a processing mode that uses only the coherent observations. The degradation action is not a simple "on / off" switch, but rather achieves a continuous and smooth transition by adjusting the weights of relation observations in the joint estimation. Optionally, the adaptive weight adjustment mechanism satisfies the following formula:

[0152] in, For the first The effective covariance matrix of the relationship observed at any given time. Covariance is the fundamental theory for observing relationships. As an adaptive weighting factor, The lower bound of the weight is used to prevent division by zero. This represents the current proportion of incremental information. and These are the threshold boundaries for complete degradation and complete enable, respectively. This means truncating the value within the [0,1] interval. When information quality deteriorates, Automatic reduction is equivalent to increasing the covariance of the relational observations, causing its contribution to the joint estimation to decrease continuously until it approaches zero.

[0153] Furthermore, the automatic degradation mechanism also includes state machine control logic for orderly switching and recovery between different perception modes.

[0154] Among them, the state machine control logic defines the transition rules between the joint precision mode, the coherent observation mode only, the low-rate parameter tracking mode and the reconfiguration mode. It is designed to cope with more complex systemic anomalies or environmental changes than single weight adjustment, and to ensure that the radar can maintain a minimum functional integrity under various extreme conditions.

[0155] In the embodiments of this application, such as Figure 9 As shown, the state machine contains four core states and their transition conditions: when the signal-to-noise ratio of the coherent observation output meets the detection threshold and the incremental information is sufficient, the system is in joint precision mode; when the incremental information is insufficient but the coherent output is normal, it degenerates to coherent observation-only mode to maintain stable tracking; when the coherent output also does not meet the detection threshold, but the relative phase of the spatial manifold relationship observation remains continuous, it enters low-rate parameter tracking mode and synchronously requests the generation of a new spatial processing configuration; when the incremental information and covariance condition number are restored to normal after the new configuration is adopted, it re-enters joint precision mode.

[0156] Optionally, the joint mode enables the criterion. Satisfy the following logic: When and and and hour, In other cases 0. Among them This indicates the activation of the joint precision mode. 0 indicates exiting the joint precision mode. The proportion of incremental information. For incremental information threshold, For the joint covariance condition number, For condition number thresholds, The signal-to-noise ratio of the coherent output. This is the minimum signal-to-noise ratio threshold. The distance to the dangerous, ambiguous direction is unknown. This is the distance threshold. When... When using two types of observations for joint precise estimation; when When 0 and coherent output is sufficient, it degenerates into a coherent observation-only mode.

[0157] It should be noted that the "low-rate parameter tracking mode" is a key design feature that distinguishes this embodiment from conventional degradation strategies. When the target is briefly obscured or in deep fading, coherent energy may be completely lost, but the phase continuity in relational observations can often be maintained for a longer period. In this situation, the system does not directly declare target loss, but instead utilizes the inertia of relational observations to maintain a low-update-rate track and actively searches for a better configuration (such as switching frequencies or beam pointing). Once the signal is recovered, high-precision tracking can be seamlessly resumed, significantly reducing reacquisition time and the probability of track breakage.

[0158] Based on the above technical solution, the abstract information quality metric is transformed into specific weight adjustment and state transition actions, which not only solves the problem of false gain in the fusion of data from the same source, but also endows the system with self-healing ability and continuous perception resilience in harsh electromagnetic environments and dynamic occlusion scenarios, ensuring reliability and practicality in actual engineering applications.

[0159] Example 6: In one possible implementation of the embodiments of this application, combined with Figure 1 ,like Figure 10 As shown, the above S106 can be specifically implemented through the following S1001, S1002 and S1003, which are explained in detail below: S1001. Use at least one of the following algorithms to jointly estimate the target parameters: maximum likelihood estimation, Bayesian estimation, Gauss-Newton method, Kalman filtering, factor graph inference, sparse optimization, or low-rank optimization.

[0160] The choice of the joint estimation algorithm depends on the linearity of the observation model, the statistical characteristics of noise, and the real-time requirements. Its core input is the constructed complete joint covariance matrix and the effective conditional information after conditional elimination, rather than the original observation data.

[0161] In the embodiments of this application, for precision sensing tasks with strong nonlinearity, the Gauss-Newton method or the Levenberg-Marquardt algorithm is preferred. The effective conditional information matrix is ​​used as an approximation of the Hessian matrix, and the inverse of the joint covariance is used as a weighting metric to iteratively search for the optimal solution of the target parameters. For dynamic tracking scenarios, extended Kalman filtering or factor graph inference is used to incorporate the effective conditional quality index into the adaptive adjustment of process noise or factor node confidence. For imaging or structured sensing tasks, sparse regularization or low-rank constraints are introduced to enhance the uniqueness of the solution by utilizing the structural priors provided by spatial manifold relation observations.

[0162] Optionally, a weighted least squares objective with prior knowledge. Satisfy the following formula:

[0163] in, and These are the estimated values ​​of the target parameter and the disturbance parameter, respectively. To augment the observed vectors, For observation and prediction models, The inverse of the joint covariance matrix. and As a priori value, and This is the corresponding prior accuracy matrix.

[0164] When the disturbance parameters can be quickly solved for given target parameters, the variable projection method can be used, satisfying the following: ,in The above is the weighted least squares objective function. This is the conditionally optimal solution for the disturbance parameters given the target parameters.

[0165] Based on the above steps, the abstract effective condition information is transformed into specific numerical solution dynamics, enabling the estimation algorithm to directly utilize the purified target information resources. This significantly improves the estimation accuracy and robustness in complex environments while ensuring convergence speed.

[0166] S1002. When there are multiple targets or extended targets, set independent target states and scattering amplitudes for each target, and share channel parameters, calibration parameters or clutter parameters among multiple targets or extended targets.

[0167] Among them, the parameter sharing mechanism is a classification modeling strategy based on physical attributes: the target state (position, velocity) and scattering amplitude depend on the electromagnetic characteristics and motion state of the individual target and are independent; while the channel amplitude and phase error, array calibration coefficient and background clutter statistical characteristics are inherent attributes of the radar system and the environment, which have the same impact on all target echoes and are common.

[0168] In this embodiment, when constructing a multi-target joint measurement model, the parameter vector to be estimated is divided into a private parameter group and a common parameter group. Each target has independent position, velocity, and complex scattering coefficient variables, while all targets share the same set of channel gain phase vector, calibration bias vector, and clutter covariance parameters. In the joint likelihood function or cost function, the common parameters appear only once but are referenced by the prediction models of all targets, while the private parameters are updated independently.

[0169] Based on the above steps, by using physically accurate parameter structure modeling, the degree of freedom of the estimation problem is significantly reduced while ensuring multi-target resolution capability, effectively alleviating the problems of parameter coupling and excessive dimensionality under limited observation resources.

[0170] S1003. Target aliasing is handled by at least one of joint data association, sparse reconstruction or low-rank estimation, and separation degree is evaluated by using the scattering center exchange state as the dangerous ambiguity direction.

[0171] Among them, target aliasing refers to the state in which multiple targets occupy the same or adjacent resolution cells in the range-Doppler-angle space, resulting in the superposition of observation signals that are difficult to separate; the scattering center exchange state specifically refers to the confusion situation in which the synthetic observation is statistically highly similar to the real state after the scattering center positions of two or more targets are interchanged and the scattering amplitude is adjusted accordingly.

[0172] In this embodiment, a joint data association algorithm (such as JPDA or MHT) is first used to establish an association hypothesis between measurements and tracks, or a sparse reconstruction algorithm is used in the imaging scene to search for the best matching combination of scattering centers in an overcomplete dictionary. Then, for each possible association or reconstruction result, the profile statistical distance between it and the candidate state of "scattering center exchange" is explicitly calculated. This distance calculation allows for optimal matching adjustment of interference parameters (including the scattering amplitude of the exchanged target), retaining only spatial structural differences that cannot be eliminated by amplitude redistribution as a distinguishable metric. If this distance is below a preset threshold, it is determined that the target cannot be reliably separated under the current configuration, triggering a configuration update process or outputting an indistinguishable identifier.

[0173] Based on the above technical solutions, the system-level estimation efficiency is improved through parameter sharing mechanism. By incorporating the exchange of scattering centers into the danger ambiguity direction assessment system, the reliability and anti-aliasing ability of multi-target resolution are enhanced, and the applicability of precision sensing methods in complex and dense scenes is improved.

[0174] Example 7: This embodiment analyzes the computational complexity of the dual-gain precision sensing method for array radar, providing a theoretical basis and optimization direction for engineering implementation. In practical radar systems, real-time performance is one of the key performance indicators. The dual-gain architecture involves computationally intensive steps such as multi-configuration evaluation, joint statistical modeling, and conditional elimination. Therefore, it is necessary to quantify the computational load of each step and clarify the optimization space.

[0175] In this embodiment of the application, under the condition that covariance decomposition is reusable, the main complexity of a single configuration evaluation can be expressed as:

[0176] in, The candidate space configuration number represents the total number of optional space processing modes maintained by the system. To augment the observation dimension, let represent the dimension of the joint observation vector after concatenating coherent and relational observations. Dimension of the target parameters, representing the number of precise parameters (such as angles and distances) to be estimated. The dimension of the interference parameters represents the number of nuisance parameters (such as scattering complex amplitude and channel phase error) that require conditional elimination. The term corresponds to the product operation of the Jacobian matrix and the observation covariance. The term corresponds to the matrix inversion or decomposition operation in the Schur complement elimination process.

[0177] Example 8: like Figure 11 As shown in the figure, this application provides an array radar dual-gain precision sensing system, including a multi-channel data interface 1101, an observation generation module 1102, a joint statistical modeling module 1103, a conditional elimination module 1104, a configuration design module 1105, and a joint estimation module 1106.

[0178] The multi-channel data interface 1101 is used to acquire multi-channel complex data formed by at least two radar channels.

[0179] In this embodiment, the interface is typically composed of a high-speed analog-to-digital converter (ADC), a field-programmable gate array (FPGA), or an application-specific integrated circuit (ASIC). It integrates a digital down-conversion (DDC) unit and a synchronous buffer, and can output a calibrated IQ data stream in real time for subsequent modules to call. At the same time, it provides the observation generation module 1102 with metadata information such as the current sampling rate, local oscillator frequency, and channel calibration version.

[0180] The observation generation module 1102 is used to process multi-channel complex data using at least one spatial processing configuration to form coherent observations and spatial manifold relationship observations. The coherent observations are used to improve the effective signal-to-noise ratio or coherent accumulation capability, and the spatial manifold relationship observations are used to preserve array spatial manifold information.

[0181] In this embodiment, the module is configured to perform weighted synthesis and nonlinear transformation operations on the input multi-channel complex data according to the spatial processing configuration parameters (such as weight vector, subarray mask, encoding sequence, etc.) issued by the configuration design module 1105; the coherent branch uses a complex multiplier-adder array to achieve high-gain beamforming, while the relational branch uses dividers, conjugate multipliers and normalization units to construct ratio or phase difference characteristics. The two outputs are strictly aligned on the timestamp and then sent to the next level module.

[0182] The Joint Statistical Modeling Module 1103 is used to construct a joint statistical model of coherent observations and spatial manifold relationship observations, and to preserve the cross covariance when the two share the original samples.

[0183] In this embodiment, the module receives the augmented real vector output by the observation generation module 1102, calculates the joint covariance matrix through the analytical error propagation engine or the online sample statistics unit, and explicitly stores the cross covariance block in memory instead of setting it to zero; at the same time, it has a built-in positive definite correction and condition number monitoring unit. When it detects that the matrix is ​​ill-conditioned or the relation observation denominator is too small, it automatically triggers the inflation factor injection or weight decay operation, and outputs a numerically stable effective covariance matrix for downstream use.

[0184] The condition elimination module 1104 is used to establish a joint measurement model that includes target parameters and interference parameters, and to perform condition elimination on the interference parameters to obtain the effective condition information of the target parameters. The interference parameters include at least one of common scattering complex amplitude, channel parameters, clutter parameters and calibration parameters.

[0185] In this embodiment, the module constructs a joint Fisher information matrix of target parameters and interference parameters based on the currently selected spatial processing configuration and target priors, and eliminates degrees of freedom such as common scattering amplitude and channel error through hardware-accelerated Schur complement operation unit or orthogonal projection operator; the output effective condition information matrix is ​​not only used for subsequent joint estimation, but also fed back to the configuration design module 1105 as a quality index, forming an information-driven closed loop.

[0186] The configuration design module 1105 is used to select or update the spatial processing configuration based on the sensitivity index and quality index of the coherent observation and the effective condition information.

[0187] Among them, the configuration design module 1105 is the decision-making center of the system. It is responsible for searching and selecting the spatial processing configuration that optimizes the precision sensing performance under the premise of meeting engineering constraints, so as to achieve a dynamic balance between detection and identification.

[0188] In this embodiment, the module maintains a candidate configuration library or online optimizer, and receives in real time effective information quality indicators (such as minimum eigenvalue, dangerous ambiguity direction distance) from the condition elimination module 1104 and sensitivity indicators (such as output signal-to-noise ratio) from the observation generation module 1102; it filters out configurations that violate hardware security or performance limits through the constraint checking unit, and then selects the optimal configuration from the remaining feasible set according to the preset objective function, and sends the configuration parameters to the observation generation module 1102 and the joint estimation module 1106 to perform switching.

[0189] The joint estimation module 1106 is used to jointly estimate the target parameters using coherent observations and spatial manifold relation observations, under the condition of satisfying preset constraints.

[0190] Among them, the joint estimation module 1106 is the final output port of the system, which is responsible for fusing dual observation data and effective information models to calculate the target's precise state parameters such as angle, distance, and velocity, as well as their uncertainties.

[0191] In this embodiment, the module receives the observation data and covariance matrix confirmed by the configuration design module 1105, and iteratively solves the problem using algorithms such as maximum likelihood, Kalman filtering, or factor graph. In multi-target scenarios, it automatically enables parameter sharing mechanisms and aliasing processing strategies, and uses the scattering center exchange distance as the separation criterion to output a confidence-weighted target state estimate. At the same time, it continuously monitors the model residuals and incremental information indicators to provide feedback signals for the automatic degradation mechanism.

[0192] Although this application has been described herein in conjunction with various embodiments, those skilled in the art, by reviewing the accompanying drawings, disclosure, and appended claims, will understand and implement other variations of the disclosed embodiments in carrying out the claimed application. In the claims, the word "comprising" does not exclude other components or steps, and "a" or "an" does not exclude multiple instances. A single processor or other unit can implement several functions listed in the claims. While different dependent claims may recite certain measures, this does not mean that these measures cannot be combined to produce good results.

[0193] Although this application has been described in conjunction with specific features and embodiments, it is obvious that various modifications and combinations can be made thereto without departing from the spirit and scope of this application. Accordingly, this specification and drawings are merely exemplary illustrations of this application as defined by the appended claims, and are considered to cover any and all modifications, variations, combinations, or equivalents within the scope of this application. Clearly, those skilled in the art can make various alterations and modifications to this application without departing from the spirit and scope of this application. Thus, if such modifications and modifications of this application fall within the scope of the claims of this application and their equivalents, this application is also intended to include such modifications and modifications.

Claims

1. A dual-gain precision sensing method for array radar, characterized in that, include: Acquire multi-channel complex data formed by at least two radar channels; The multi-channel complex data is processed using at least one spatial processing configuration to form coherent observations and spatial manifold relationship observations; the coherent observations are used to improve the effective signal-to-noise ratio or coherent accumulation capability; the spatial manifold relationship observations retain information related to the target spatial state, near-field curvature, scattering structure, micro-motion, or relative state of the array channels by eliminating or normalizing common scattering complex amplitude, common gain, or common phase, in order to preserve array spatial manifold information; Construct a joint statistical model of the coherent observations and the spatial manifold relation observations, and retain the cross covariance when the two share the original samples; the retention method is to concatenate the coherent observations and the spatial manifold relation observations into an augmented real vector, and construct a complete joint covariance matrix containing their respective autocovariances and cross covariances; A joint measurement model including target parameters and interference parameters is established, and conditional elimination is performed on the interference parameters to obtain effective condition information of the target parameters; the interference parameters include at least one of common scattering complex amplitude, channel parameters, clutter parameters, and calibration parameters; the effective condition information retains only the variation information of the target parameters that cannot be explained by the degrees of freedom of the interference parameters; Based on the sensitivity index of the coherent observation and the quality index of the effective condition information, the spatial processing configuration is selected or updated; the quality index includes at least one of the minimum eigenvalue, determinant, trace, posterior covariance, or dangerous ambiguity direction distance of the effective condition information. Under the condition of satisfying the preset constraints, the target parameters are jointly estimated by the coherent observation and the spatial manifold relation observation; the preset constraints include at least one of the following: minimum coherent signal-to-noise ratio, minimum complex response amplitude, interference suppression, clutter suppression, phase fidelity, hardware dynamic range, transmit power, bandwidth and configuration switching continuity.

2. The method according to claim 1, characterized in that, The process of processing the multi-channel complex data using at least one spatial processing configuration to form coherent observations and spatial manifold relation observations includes: The multi-channel complex data is coherently synthesized using the first spatial processing configuration to obtain the coherent observation; The multi-channel complex data is processed using at least two second-space processing configurations to obtain multiple complex output signals; The spatial manifold relationship observation is generated based on at least one of the normalized complex ratio, conjugate product, amplitude-phase difference, normalized vector, cross spectrum, or coherence coefficient among the plurality of complex output signals.

3. The method according to claim 2, characterized in that, The construction of the joint statistical model of the coherent observations and the spatial manifold relationship observations, while preserving the cross covariance when both share the original samples, includes: The coherent observations and the spatial manifold relation observations are converted into augmented real vectors containing real and imaginary parts; Based on the augmented real vector, a complete joint covariance matrix is ​​constructed, including the coherent observation covariance, the relational observation covariance, and the cross covariance. The complete joint covariance matrix is ​​obtained by means of at least one of analytical error propagation, sample covariance, resampling, Monte Carlo or joint likelihood, and positive definiteness, shrinkage estimation, eigenvalue lower bound protection or condition number monitoring is performed on the complete joint covariance matrix. When the denominator amplitude of the spatial manifold relation observation is lower than a preset amplitude threshold, a phase jump occurs, the number of samples is insufficient, or the complete joint covariance matrix is ​​ill-conditioned, the relation observation covariance is inflated or its weight is reduced.

4. The method according to claim 1, characterized in that, The step of establishing a joint measurement model including target parameters and interference parameters, and performing conditional elimination on the interference parameters to obtain effective conditional information for the target parameters, includes: Based on the joint statistical model, a joint Fisher information matrix of the target parameters and the interference parameters is constructed; Perform Schur complement operation on the interference parameters in the joint Fisher information matrix to obtain the effective condition information of the target parameters; or, The whitened target parameter Jacobian matrix is ​​projected onto the orthogonal complement space of the column space of the interference parameter Jacobian matrix to obtain the effective condition information of the target parameter.

5. The method according to claim 1, characterized in that, The selection or updating of the space processing configuration based on the sensitivity index of the coherent observation and the quality index of the effective condition information includes: Determine a set of candidate space configurations; each candidate space configuration in the set satisfies the preset constraints. For each candidate spatial configuration in the candidate spatial configuration set, calculate the corresponding sensitivity index of the coherent observation and the quality index of the effective condition information; From the set of candidate spatial configurations, select the candidate spatial configuration that maximizes the quality index and satisfies the preset threshold of the sensitivity index, and use it as the updated spatial processing configuration.

6. The method according to claim 5, characterized in that, The calculation process for the distance in the ambiguous danger direction includes: Determine a set of dangerous fuzzy directions; the set of dangerous fuzzy directions includes at least one of the following: angular grid neighbor points, distance period subpeaks, velocity folding states, scattering center exchange states, image sidelobe states, or near-field mirror states. For each candidate target state in the set of dangerous fuzzy directions, the profile statistical distance between the real target state and the candidate target state is calculated, while allowing the interference parameters to be adjusted. The minimum value among the profile statistical distances is determined as the danger ambiguity direction distance; the profile statistical distance is used to quantify the distinguishability between the real target state and the candidate target state after eliminating the influence of the interference parameters.

7. The method according to claim 1, characterized in that, The method further includes an incremental information verification and automatic degradation step, which includes: Calculate the first effective condition information when using only the coherent observations, and the second effective condition information when using both the coherent observations and the spatial manifold relation observations. Based on the first valid condition information and the second valid condition information, an incremental information indicator is determined; the incremental information indicator is the difference between the two or a normalized ratio. When the incremental information index is lower than the preset information threshold, the covariance condition number of the joint statistical model is greater than the preset condition number threshold, the denominator amplitude of the spatial manifold relationship observation is lower than the preset amplitude threshold, or the model residual exceeds the preset residual threshold, the automatic degradation mechanism is triggered. The automatic degradation mechanism includes: turning off the spatial manifold relation observations, reducing the weight of the spatial manifold relation observations, or expanding the covariance of the spatial manifold relation observations, so that the system degrades to a processing mode that only uses the coherent observations.

8. The method according to claim 7, characterized in that, The automatic degradation mechanism also includes state machine control logic, which includes: When the output signal-to-noise ratio of the coherent observation meets the preset detection threshold, the processing mode of using only the coherent observation is maintained. When the output signal-to-noise ratio of the coherent observation does not meet the preset detection threshold, and the relative phase of the spatial manifold relationship observation remains continuous, the low-rate parameter tracking mode is maintained, and a new spatial processing configuration is requested to be generated. When the new spatial processing configuration is adopted, the recalculated incremental information index recovers to a level higher than the preset information threshold, and the covariance condition number of the joint statistical model recovers to a level lower than the preset condition number threshold, and the joint precision mode that uses the coherent observation and the spatial manifold relationship observation is re-entered.

9. The method according to claim 1, characterized in that, The step of jointly estimating the target parameters using the coherent observations and the spatial manifold relation observations, under the condition of satisfying preset constraints, includes: The target parameters are jointly estimated using at least one of the following algorithms: maximum likelihood estimation, Bayesian estimation, Gauss-Newton method, Kalman filtering, factor graph inference, sparse optimization, or low-rank optimization. When there are multiple targets or extended targets, set independent target states and scattering amplitudes for each target, and share channel parameters, calibration parameters or clutter parameters among the multiple targets or extended targets. The target aliasing is handled by at least one of joint data association, sparse reconstruction or low-rank estimation, and the separation degree is evaluated by using the scattering center exchange state as the dangerous ambiguity direction.

10. A dual-gain precision sensing system for array radar, characterized in that, The system includes: A multi-channel data interface is used to acquire multi-channel complex data formed by at least two radar channels; An observation generation module is used to process the multi-channel complex data using at least one spatial processing configuration to form coherent observations and spatial manifold relationship observations. The coherent observations are used to improve the effective signal-to-noise ratio or coherent accumulation capability. The spatial manifold relationship observations retain information related to the target spatial state, near-field curvature, scattering structure, micro-motion, or relative state of the array channels by eliminating or normalizing common scattering complex amplitude, common gain, or common phase, thereby preserving array spatial manifold information. The joint statistical modeling module is used to construct a joint statistical model of the coherent observations and the spatial manifold relation observations, and to retain the cross covariance when the two share the original samples. The retention method is to concatenate the coherent observations and the spatial manifold relation observations into an augmented real vector and construct a complete joint covariance matrix containing their respective self-covariances and cross-covariances. The conditional elimination module is used to establish a joint measurement model including target parameters and interference parameters, and to perform conditional elimination on the interference parameters to obtain effective conditional information of the target parameters. The interference parameters include at least one of common scattering complex amplitude, channel parameters, clutter parameters, and calibration parameters. The effective conditional information only retains the variation information of the target parameters that cannot be explained by the degrees of freedom of the interference parameters. A configuration design module is used to select or update the spatial processing configuration based on the sensitivity index of the coherent observation and the quality index of the effective condition information; the quality index includes at least one of the minimum eigenvalue, determinant, trace, posterior covariance, or dangerous ambiguity direction distance of the effective condition information; The joint estimation module is used to jointly estimate the target parameters using the coherent observations and the spatial manifold relation observations, provided that preset constraints are met. The preset constraints include at least one of the following: minimum coherent signal-to-noise ratio, minimum complex response amplitude, interference suppression, clutter suppression, phase fidelity, hardware dynamic range, transmit power, bandwidth, and configuration switching continuity.

Images