Low glancing angle multipath effect real-time compensation method, system and device and medium

Abstract

The invention relates to the technical field of radar target detection, in particular to a low glancing angle multipath effect real-time compensation method, system and device and a medium, and the method comprises the steps: building an augmented state vector which comprises a target motion state parameter and a pitch angle error oscillation characteristic parameter; establishing a state transition model and a nonlinear measurement model of the augmented state vector, wherein the nonlinear measurement model represents the pitch angle observed quantity as the sum of error terms formed by the target real pitch angle and the pitch angle error oscillation characteristic parameters; predicting a state prediction vector and a prediction covariance matrix at the current moment according to the state transition model and the state estimation value at the previous moment; and filtering and updating the state prediction vector and the prediction covariance matrix based on the nonlinear measurement model and the actual observed quantity of the radar at the current moment, and outputting a target state estimated value and a multipath oscillation characteristic parameter estimated value after multipath error compensation. The objective of the invention is to realize low-delay and high-precision pitch angle error compensation.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of radar target detection, in particular to a low-grazing-angle multipath effect real-time compensation method, system, device and medium. BACKGROUND

[0002] With the rapid development of low-altitude economy, efficient detection of unmanned aerial vehicles and other targets in low-altitude and ultra-low-altitude airspace has become a core requirement for ensuring airspace safety and promoting the construction of low-altitude transportation systems. As a detection equipment with all-weather and all-day working capability, radar has become a core technical means in the field of low-altitude target detection due to its advantage of continuous tracking of dynamic targets. In typical low-altitude detection scenarios such as urban near-ground airspace, deserts, and forests, radar often needs to observe targets at low grazing angles. At this time, significant multipath effects are likely to occur in the propagation process of radar electromagnetic waves. The direct wave emitted by the radar and the multipath reflected wave reflected by obstacles such as the ground, vegetation, and buildings will interfere in the spatial domain, thereby destroying the symmetry of the radar elevation beam pattern and causing the measured target elevation angle to have a nearly biased sinusoidal periodic oscillation error, which greatly reduces the target height detection accuracy and greatly restricts the application effectiveness of radar in key fields such as low-altitude surveillance and airspace control.

[0003] To address the problem of deterioration of elevation angle measurement accuracy caused by low-grazing-angle multipath effects, existing technologies mainly conduct multipath interference suppression research from two dimensions of electromagnetic environment modeling and signal processing. In the aspect of electromagnetic environment modeling, methods such as ray tracing and hybrid diffraction model are often used to construct electromagnetic wave propagation models in complex scenarios to achieve prediction and suppression of multipath signals. In the aspect of signal processing, techniques such as array antenna beamforming and environment matching filtering are used to weaken the influence of multipath interference on measurement results. However, the electromagnetic environment modeling method needs to construct a refined scene geometry and electromagnetic parameter model, which has extremely high computational complexity and poor model adaptability, and is difficult to meet the real-time detection requirements due to poor engineering portability. The array antenna beamforming scheme relies on specific multi-channel antenna hardware architecture and requires hardware modification of existing radar systems, increasing the technical implementation cost and complexity. The environment matching filtering method requires complex offline training and online matching operations on observation data, resulting in high processing delay and inability to adapt to the stable tracking requirements of high-speed dynamic targets. The existing elevation angle error compensation methods still have obvious deficiencies in real-time performance, system compatibility, and environmental adaptability. An effective technical approach that does not rely on complex modeling and special hardware and can achieve low-latency and high-precision compensation is urgently needed. SUMMARY

[0004] To achieve low-latency and high-precision elevation angle error compensation, the present application provides a low-grazing-angle multipath effect real-time compensation method and system, and the technical solutions adopted are as follows: The technical scheme of the first aspect of the application provides a low grazing angle multipath effect real-time compensation method, and the method comprises the following steps: An augmented state vector is established, and the augmented state vector comprises target motion state parameters and pitch angle error oscillation characteristic parameters; A state transition model and a nonlinear measurement model of the augmented state vector are established, wherein the nonlinear measurement model represents a pitch angle observation as a sum of a target real pitch angle and an error term formed by the pitch angle error oscillation characteristic parameters; According to the state transition model and a state estimation value at a previous time, a state prediction vector and a prediction covariance matrix at a current time are predicted; Based on the nonlinear measurement model and actual observation of the radar at the current time, the state prediction vector and the prediction covariance matrix are updated by filtering, and a target state estimation value after multipath error compensation and a multipath oscillation characteristic parameter estimation value are output.

[0005] Further, the target motion state parameters comprise position components and velocity components of the target in horizontal and vertical directions. The pitch angle error oscillation characteristic parameters comprise an amplitude, an angular frequency and a phase angle of periodic oscillation of the pitch angle caused by the multipath effect.

[0006] Further, the establishment of the state transition model specifically comprises the following steps: The position and velocity components of the target are predicted by time recursion according to a target kinematics relationship; The phase angle is updated by linear accumulation based on the angular frequency; Process noise models are respectively constructed for each state variable in the augmented state vector.

[0007] Further, the establishment of the nonlinear measurement model specifically comprises the following steps: A target real pitch angle term is calculated according to a geometric projection relationship between the target and the radar; A periodic oscillation error term representing the multipath effect is constructed according to the pitch angle error oscillation characteristic parameters; The target real pitch angle term and the periodic oscillation error term are superimposed and integrated into radar measurement noise to form the nonlinear measurement model.

[0008] Further, the filtering update of the state prediction vector and the prediction covariance matrix specifically comprises the following steps: By combining the prediction covariance matrix and a measurement noise covariance matrix, an inverse matrix of a covariance matrix of a measurement prediction residual is solved, and an optimal weighting matrix is calculated; Based on the current state prediction value, a measurement prediction result is calculated by the nonlinear measurement model, and a deviation is obtained by comparing the measurement prediction result with the actually collected radar observation data; The optimal weighting matrix is used to linearly weight the bias amount, and a weighted and adjusted compensation value is applied to the state prediction vector to complete estimation of the augmented state vector; Based on the optimal weighting matrix, the partial derivative matrix of the measurement function, and the prediction covariance matrix, the state estimation covariance matrix is updated.

[0009] Further, the method further includes noise covariance matrix initialization, specifically including: The process noise covariance matrix in the state transition process is configured to represent the uncertainty of the state prediction model; The measurement noise covariance matrix in the observation process is configured according to the measurement accuracy of the radar sensor, and is used to represent the uncertainty of the radar observation data; Based on the process noise covariance matrix and the measurement noise covariance matrix, the confidence between the prediction value and the observation value is dynamically weighted and balanced in the filtering process.

[0010] Further, the filtering update of the state prediction vector and the prediction covariance matrix further includes model linearization, specifically including: The partial derivative matrix of the state transition function with respect to the augmented state vector is calculated to obtain a linear approximation of the state transition function; The partial derivative matrix of the measurement function with respect to the augmented state vector is calculated to obtain a linear approximation of the measurement function; The linear approximations of the state transition function and the measurement function are used to perform covariance matrix prediction and update calculation of the augmented state Kalman filter.

[0011] The second aspect of the technical scheme of the present application provides a low-grazing-angle multipath effect real-time compensation system, which adopts the low-grazing-angle multipath effect real-time compensation method of the first aspect of the technical scheme of the present application, and the system comprises: A state vector construction module is configured to establish an augmented state vector, and the augmented state vector includes target motion state parameters and pitch angle error oscillation characteristic parameters; A model establishment module is configured to establish a state transition model and a nonlinear measurement model of the augmented state vector, wherein the nonlinear measurement model represents the pitch angle observation as the sum of an error term composed of the true pitch angle of the target and the pitch angle error oscillation characteristic parameters; A state prediction module is configured to predict a state prediction vector and a prediction covariance matrix at the current time according to the state transition model and the state estimation value at the last time; A filtering update module is configured to perform filtering update on the state prediction vector and the prediction covariance matrix based on the nonlinear measurement model and the actual observation of the radar at the current time, and output the target state estimation value after multipath error compensation and the multipath oscillation characteristic parameter estimation value.

[0012] The technical solution of the third aspect of the present application provides an electronic device, which comprises a processor and a memory connected with the processor; wherein the memory stores instructions executable by the processor, and the instructions are executed by the processor to enable the processor to perform the steps of the low-grazing angle multipath effect real-time compensation method of the technical solution of the first aspect of the present application.

[0013] The technical solution of the fourth aspect of the present application provides a computer readable storage medium, which stores a program for implementing a low-grazing angle multipath effect real-time compensation method, and the program is executed by a processor to implement the steps of the low-grazing angle multipath effect real-time compensation method of the technical solution of the first aspect of the present application.

[0014] The present application has the following advantages: The low-grazing angle multipath effect real-time compensation method provided by the present application establishes an augmented state vector containing target motion state and pitch angle error oscillation characteristic parameters caused by multipath, and constructs a corresponding state transition model and a nonlinear measurement model, so as to separate the pitch angle observation into the sum of the target true pitch angle and the multipath error term, thereby realizing the joint estimation of the target state and the multipath error in a unified filtering framework. Based on the low-grazing angle multipath effect real-time compensation method, the periodic oscillation error of the pitch angle caused by the multipath effect can be estimated and compensated in real time and dynamically online, which significantly improves the pitch angle measurement accuracy. At the same time, the method is realized based on signal processing only, without relying on complex environment modeling or special hardware support, and has the advantages of high calculation efficiency, good real-time performance and easy engineering deployment, which solves the core problem of insufficient radar pitch angle measurement accuracy in the low-grazing angle scene. BRIEF DESCRIPTION OF DRAWINGS

[0015] In order to more clearly illustrate the technical solutions and advantages of the embodiments of the present application or the prior art, the drawings needed in the following embodiment or prior art description will be briefly introduced. Obviously, the drawings in the following description are only some embodiments of the present application, and those skilled in the art can also obtain other drawings according to these drawings without creative labor.

[0016] Figure 1 The method flowchart of the low-grazing angle multipath effect real-time compensation method provided by an embodiment of the present application; Figure 2 The multipath period and amplitude prediction effect schematic diagram provided by an embodiment of the present application; Figure 3 The pitch angle estimation accuracy comparison diagram of different filtering methods provided by an embodiment of the present application; Figure 4 The low grazing angle multipath effect real-time compensation system provided by one embodiment of the present application Figure 3 Comparison chart of pitch angle estimation accuracy of different filtering methods matched with Figure 5 The structure schematic diagram of the low grazing angle multipath effect real-time compensation system provided by one embodiment of the present application. DETAILED DESCRIPTION

[0017] In order to further illustrate the technical means and effects taken by the present application to achieve the predetermined purposes, the following describes in detail the specific implementation, structure, features and effects of the low grazing angle multipath effect real-time compensation method, system, device and medium according to the present application, combined with the accompanying drawings and preferred embodiments. In the following description, different "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. In addition, the specific features, structures or characteristics in one or more embodiments can be combined in any suitable form.

[0018] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which the present application belongs.

[0019] The inventors have found in practical work that the pitch angle error caused by the low grazing angle multipath effect has a key and exploitable feature, i.e. its time-domain form is approximately presented as error, which is expressed as:

[0020] where A is the oscillation amplitude, usually 1°-5°; T is the oscillation period, which varies in the order of 10-100 seconds; is the initial phase; B is the bias, and the formula provides a theoretical basis for suppressing error from the perspective of signal processing.

[0021] However, the fixed parameter filter in the existing solution cannot adapt to the dynamically changing multipath environment; and the complex adaptive algorithm has the defects of heavy calculation burden (such as O(N³) complexity) or slow convergence speed.

[0022] Based on this, the present application uses the periodic oscillation feature of the multipath error which can be parameterized described, takes the key parameters describing the periodic oscillation as augmented state variables, and introduces them into the Kalman filter framework together with the motion state of the target, to construct a new filter model for joint estimation. This method abandons the traditional complex physical modeling or special hardware support idea, and changes into a pure signal processing error compensation strategy.

[0023] The specific scheme of the low grazing angle multipath effect real-time compensation method, system, device and medium provided by the present application is described in detail below with reference to the accompanying drawings.

[0024] Please refer toFigure 1 FIG. 1 shows a flow chart of a method for real-time compensation of low grazing angle multipath effects according to an embodiment of the present application, the method comprising: Step S100: establishing an augmented state vector, the augmented state vector comprising target motion state parameters and pitch angle error oscillation characteristic parameters; In some embodiments, the target motion state parameters comprise position components and velocity components of the target in horizontal and vertical directions; In some embodiments, the pitch angle error oscillation characteristic parameters comprise amplitude, angular frequency and phase angle of periodic oscillation of the pitch angle caused by multipath effects; The augmented state vector is represented as:

[0025] wherein k represents a time index for distinguishing state quantities at different time points; represents a position component of the target in the horizontal direction, in units of meters (m); represents a height component of the target in the vertical direction, in units of meters (m); represents a motion velocity of the target in the horizontal direction, in units of meters per second (m / s); represents a motion velocity of the target in the vertical direction, in units of meters per second (m / s); represents an oscillation amplitude of the pitch angle caused by multipath effects, in units of degrees (°); represents an angular frequency of the multipath oscillation, in units of radians per second (rad / s); represents a phase angle of the multipath oscillation, in units of radians (rad); This formula integrates the dynamic information of the target motion and the oscillation characteristic information of the multipath error into the same vector space. By constructing the augmented state vector integrating the target motion state and the pitch angle error oscillation characteristic of the multipath effects, the limitation of traditional filtering focusing only on the estimation of the target motion state is broken, the key characteristic parameters of the multipath effects are included in the unified framework of subsequent joint estimation, a basic state space description is provided for real-time tracking of the multipath effects and accurate compensation of the pitch angle error, and the subsequent filtering process can dynamically update the target motion state and the multipath oscillation characteristic, thereby improving the pitch angle measurement accuracy in the low grazing angle scenario.

[0026] Step S200: establishing a state transition model and a nonlinear measurement model of the augmented state vector, wherein the nonlinear measurement model represents the pitch angle observation as the sum of an error term composed of the true pitch angle of the target and the pitch angle error oscillation characteristic parameters; In some embodiments, the state transition model is used to recursively predict the state value at the current time based on the augmented state at the previous time, and reflects the uncertainty in the actual environment through a process noise model. The establishment of the state transition model specifically comprises: Step S210: perform time recursive prediction of the target's position and velocity components based on the target kinematic relationship; specifically, for the target motion state parameters in the augmented state vector, a uniform kinematic model is used for time recursion; the radar system sampling time interval is , in seconds, represents the time difference between two adjacent state updates, so the horizontal position of k at the current moment is From the previous horizontal position With horizontal speed Joint decision, i.e. ; Similarly, the vertical height at the current moment can be obtained; for the velocity component, since the motion state of the low-altitude target changes slowly in a short period of time, the recursive rule of maintaining the velocity at the previous moment is adopted. Through this process, the noiseless recursion of the target motion state from moment k-1 to moment k is completed.

[0027] Step S211: linearly accumulating and updating the phase angle based on the angular frequency; this step is for the phase angle in the augmented state vector , that is, the phase angle of the multipath oscillation at time k-1, using the angular frequency Implementing linear cumulative updates; the physical meaning of the phase angle is the starting position of a sinusoidal oscillation, and its change over time is equal to the product of the angular frequency and the time interval. Through linear operations, the dynamic evolution of the multipath oscillation phase can be accurately characterized. Furthermore, because the angular frequency changes very little over a short period of time, the linear cumulative model can also balance accuracy and computational complexity.

[0028] Step S212: construct a process noise model for each state variable in the augmented state vector. The state transition model is expressed as:

[0029] Where, express The augmented state vector at time t; express The augmented state vector at time t; represents the state transition function; Indicates the radar system sampling time interval in seconds (s); Represents the process noise vector at time k, which is a 7-dimensional column vector. Each component corresponds to horizontal position noise, vertical height noise, horizontal velocity noise, vertical velocity noise, oscillation amplitude noise, angular frequency noise, and phase angle noise, and each component is modeled as having a mean of 0 and a covariance of Gaussian white noise; It is a 7×7 diagonal matrix, and the elements on the diagonal are the variances of the noise components. This example uses but is not limited to the following parameter settings: Position process noise standard deviation , speed process noise standard deviation , amplitude process noise standard deviation , frequency process noise standard deviation , phase process noise standard deviation ; In some embodiments, the role of the nonlinear measurement model is to establish a mapping relationship between the augmented state vector and the actual radar observation value , and to explicitly reflect the influence of multipath error on the elevation angle observation. Establishing a nonlinear measurement model specifically includes: Step S220: calculating the target real elevation angle term according to the geometric projection relationship between the target and the radar; specifically, using the position component of the target in the horizontal direction and the height component in the vertical direction in the augmented state vector, the target real elevation angle is calculated based on the geometric projection relationship; In some embodiments, the slant range between the radar and the target is calculated according to the Pythagorean theorem ; then, through the arctangent function , the ratio of the height to the horizontal distance is converted into an angle to obtain the target real elevation angle, which is the elevation angle that the radar should observe without multipath interference; Step S221: constructing a periodic oscillation error term representing the multipath effect according to the elevation angle error oscillation characteristic parameters; specifically, extracting the elevation angle error oscillation characteristic parameters from the augmented state vector, including the amplitude , phase angle of the periodic oscillation of the elevation angle caused by the multipath effect, and then constructing a periodic oscillation error term representing the multipath effect according to these parameters, which is used to simulate the periodic error generated by the interference of the multipath reflected wave and the direct wave, and the error term reflects the degree of interference of the multipath effect on the elevation angle measurement.

[0030] Step S222: superimposing the target real elevation angle term and the periodic oscillation error term and integrating the radar measurement noise to form a nonlinear measurement model, and the nonlinear measurement transfer model is represented as:

[0031] wherein, represents the radar measurement vector at time k; is the measurement noise, represents the measurement transfer function, represents the square root operation, represents the arctangent operation; the present example selects but is not limited to using the following parameter settings: radar slant range measurement noise standard deviation , elevation angle measurement noise standard deviation , and the present example selects but is not limited to an oscillation amplitude The nonlinear measurement model accurately correlates the physical nature of the target motion state and the radar slant range, true pitch angle through geometric projection relationship, explicitly describes the pitch angle periodic oscillation error caused by low grazing angle multipath interference with a sinusoidal model, and also incorporates measurement noise to reflect the measurement uncertainty of the radar hardware. The nonlinear measurement model provides an observation layer mapping that conforms to the actual detection scene for subsequent extended Kalman filtering, ensures the physical rationality of joint estimation of multipath oscillation parameters and target motion state, and supports effective separation of real signal, multipath error and measurement noise in the filtering process, and finally improves the radar pitch angle measurement accuracy in the low grazing angle scene.

[0032] The state transition model and the nonlinear measurement model established by step S200 accurately depict the dynamic characteristics of the augmented state vector evolution over time and the inherent nonlinear relationship between the radar observation and the state transition model. The state transition model realizes joint prediction of target motion and multipath error characteristics through kinematic recursion and phase linear accumulation, while the nonlinear measurement model effectively separates and characterizes the influence mechanism of multipath error on pitch angle observation through geometric relationship and oscillation term superposition. The establishment of the two models provides a rigorous mathematical basis for subsequent joint optimal estimation of multipath oscillation parameters and target motion state based on the Kalman filtering framework.

[0033] Step S300: predicting a state prediction vector and a prediction covariance matrix at the current time according to the state transition model and the state estimation value at the previous time; In some embodiments, the one-step prediction of the state vector is represented as:

[0034] wherein, represents the optimal state estimation value based on all observations before and including the k-1 time, which includes target motion state parameters and multipath oscillation characteristic parameters; represents the one-step prediction value of the state based on the k-1 time observation; is a state transition function; In some embodiments, the prediction covariance matrix calculation is represented as:

[0035] wherein, represents the state estimation covariance matrix at the k-1 time, which quantifies the uncertainty of the state estimation at the k-1 time; represents the one-step prediction state covariance matrix; represents a transpose operation; represents a process noise covariance matrix; is the Jacobian matrix of the state transition function ; and is the transpose matrix of ; and The embodiment is based on the time-varying characteristics of target kinematics and multipath oscillation, that is, the physical rationality of target motion and multipath parameter state prediction is ensured; and the uncertainty of the prediction process is accurately quantified by combining the covariance matrix with the Jacobian matrix and the process noise. Prior state information with uncertainty measurement is provided for subsequent filtering update, so that the augmented Kalman filter can adaptively balance between the prior constraint of the prediction value and the posterior correction of the observation value.

[0036] Step S400: based on the nonlinear measurement model and the actual observation of the radar at the current time, filtering update is performed on the state prediction vector and the prediction covariance matrix, and the target state estimation value after multipath error compensation and the multipath oscillation characteristic parameter estimation value are output; Step S400 specifically includes: Step S410: the inverse matrix of the covariance matrix of the measurement prediction residual is solved and the optimal weighting matrix is calculated by combining the prediction covariance matrix and the measurement noise covariance matrix, which can be represented as:

[0037] wherein, K represents the Kalman gain matrix, that is, the optimal weighting matrix, which determines how much the new observation data should be trusted in the state update process relative to the previous prediction value; P k represents the prediction state covariance matrix at the k time, which quantifies the uncertainty of the state prediction; R represents the measurement noise covariance matrix, which is represented as: ; The sum of the state prediction uncertainty propagated after the measurement model and the measurement noise uncertainty is calculated first, and then the inverse is solved to reflect the confidence of the measurement residual; the overall operation logic is to solve the inverse of the sum of the state prediction uncertainty and the measurement noise uncertainty, and then multiply the result by the state prediction value to obtain the Kalman gain matrix The linear approximation relationship between the state prediction uncertainty and the measurement model is associated, and then multiplied by the above inverse matrix to obtain the Kalman gain matrix , which controls the correction weight of the measurement residual to the state estimation.

[0038] Step S420: based on the current state prediction value, the measurement prediction result is calculated through the nonlinear measurement model, and the measurement prediction result is compared with the actual collected radar observation data to obtain the deviation; based on the current state prediction value , the measurement prediction result is calculated through the nonlinear measurement model, and then compared with the actual radar observation data to obtain the deviation , Y represents the actual observation vector of the radar at the current time; Y represents the measurement prediction value obtained based on the state prediction value ; the deviation contains new information in the current observation that is not explained by the prediction model, and is the basis for subsequent state correction; Step S430: Linearly weight the deviation quantity with the optimal weighting matrix, apply the weighted adjusted compensation value to the state prediction vector, complete the estimation of the augmented state vector, which can be expressed as:

[0039] In the formula, The updated state estimation value is represented; the formula weights the deviation between the observation value and the prediction value with the Kalman gain, and applies the weighted correction quantity to the state prediction vector to obtain the updated state estimation value at time k; Step S440: Update the state estimation covariance matrix based on the optimal weighting matrix, the partial derivative matrix of the measurement function, and the prediction covariance matrix, which can be expressed as:

[0040] In the formula, The updated state covariance matrix is represented; is a unit matrix with the same dimension as The formula reflects that the uncertainty of the state estimation is reduced after the new observation information is integrated; It can be understood that the part subtracted from the prediction covariance represents the reduced uncertainty due to the current observation; the updated covariance matrix The prediction step for the next time is performed, thereby starting a new round of filtering recursion.

[0041] Through the filtering update process of step S400, the state prediction value is optimally corrected using the actual observation quantity of the radar at the current time; the confidence between the state prediction and the radar observation is dynamically balanced by calculating the Kalman gain matrix, and the new information in the observation data is fused into the augmented state estimation in the innovation vector (deviation quantity); this process not only outputs a more accurate target motion state estimation value after the multipath error compensation, but also synchronously updates the latest estimation of the multipath oscillation characteristic parameters; at the same time, by updating the state estimation covariance matrix, an accurate uncertainty measure is provided for the filtering recursion at the next time, thereby ensuring the continuous stability and convergence of the entire augmented Kalman filtering process, and finally realizing high-precision, real-time compensation of the target pitch angle in a strong multipath interference environment.

[0042] In some embodiments, the low-grazing-angle multipath effect real-time compensation method further includes noise covariance matrix initialization, specifically including: In some embodiments, the process noise covariance matrix in the state transition process is configured to represent the uncertainty of the state prediction model; the augmented state vector contains target motion state parameters and multipath oscillation characteristic parameters, and the process noise variance needs to be configured for each component, that is, , represents the process noise covariance matrix, and represents , represents a diagonal matrix, is the process noise variance of the horizontal position, is the process noise variance of the vertical height, is the process noise variance of the horizontal velocity, is the process noise variance of the vertical velocity, is the process noise variance of the multipath oscillation amplitude, is the process noise variance of the angular frequency, is the process noise variance of the phase angle, used to represent the random disturbance that is not accurately described by the state transition model; In some embodiments, the measurement noise covariance matrix in the observation process is configured according to the measurement accuracy of the radar sensor, used to represent the uncertainty of the radar observation data; the radar observation vector contains the slant range and the pitch angle, and the slant range component can be configured according to the ranging accuracy index of the radar, and the pitch angle component can be configured according to the performance of the radar pitch angle measurement unit, such as the beam pointing accuracy of the phased array antenna, the influence of the receiver noise on the angle measurement, etc., to determine the pitch angle measurement noise variance through the parameter document or the actual measurement.

[0043] In some embodiments, based on the process noise covariance matrix and the measurement noise covariance matrix, the confidence between the predicted value and the observed value is dynamically weighted and balanced in the filtering process; in the prediction stage, the prediction covariance matrix is obtained; if increases, the uncertainty of state transition increases, and the confidence of the predicted value depending only on the state transition model will be indirectly reduced.

[0044] In the update stage, the Kalman gain matrix is obtained; if is large, the prediction uncertainty is high, and the radar observation value will tend to increase, so that the radar observation value accounts for a higher proportion in the state update; if is large, the observation uncertainty is high, and the state prediction value will tend to decrease, so that the state prediction value accounts for a higher proportion. In this way, the relative uncertainty of the model and the observation is used to adaptively adjust the contribution weight of the two to the final state estimation.

[0045] In summary, the noise covariance matrix initialization provides the basis for the prediction and update process of the augmented Kalman filter by precisely configuring the process noise covariance matrix (matching the random disturbance characteristics of target motion and multipath environment) and the measurement noise covariance matrix (matching the measurement accuracy limit of radar hardware); on this basis, the filter process can adaptively balance the confidence of the state prediction value obtained from the prior model relying on the target motion and multipath characteristics and the radar observation value reflecting the real-time detection information, avoiding the estimation drift caused by the deviation of the state transition model assumption, and weakening the interference of the radar observation noise on the estimation result, finally ensuring the precision and stability of the joint estimation of target motion state and multipath oscillation characteristic parameters in the low grazing angle scene, and realizing the real-time and high-precision compensation of the elevation angle error.

[0046] In some embodiments, the filtering update of the state prediction vector and the prediction covariance matrix further includes model linearization, specifically including: calculating the partial derivative matrix of the state transition function with respect to the augmented state vector to obtain a linear approximation of the state transition function; calculating the partial derivative matrix of the measurement function with respect to the augmented state vector to obtain a linear approximation of the measurement function; and performing covariance matrix prediction and update calculation of the augmented state Kalman filter using the linear approximations of the state transition function and the measurement function. Specifically, first, the state transition Jacobian matrix is calculated to describe the linear propagation relationship of a small change in state in the transition process, which is expressed as:

[0047] wherein denotes the derivative operation on the augmented state vector ; the matrix realizes the first-order Taylor linear approximation of the nonlinear state transition process near the current state, providing a linear mapping relationship for subsequent covariance propagation; Then, the measurement Jacobian matrix is calculated to describe the linear influence relationship of a small change in state on the observation, which is expressed as:

[0048] wherein denotes the instantaneous slant range between the radar and the target; is the measurement function; Further, the and are directly used in the covariance prediction and update formulas of the standard Kalman filter to replace the state transition matrix and the measurement matrix in the linear system, i.e., the prediction covariance matrix: , the Kalman gain matrix: and the updated state estimation covariance matrix: The model linearization process successfully linearizes the nonlinear system describing the joint evolution of target state and multipath error around the current state estimate or prediction at each filtering cycle by computing the Jacobian matrix of the state transition function and the measurement function in real time; the linearization makes the standard Kalman filter covariance prediction and update rules applicable to the highly nonlinear augmented state system; finally, this step ensures that the extended Kalman filter can effectively handle the nonlinear problems caused by complex geometric relationships and multipath periodic oscillations, thereby ensuring the stable convergence and high-precision output of the joint estimation process of target state and multipath parameters.

[0049] Please refer to Figure 2 , as shown in Figure 2 , in order to predict the multipath period and amplitude of the low grazing angle multipath effect real-time compensation method and system based on the augmented state Kalman filter in the embodiments of the present application, the figure contains three curves: Real oscillation component: represents the real trend of the periodic error of the pitch angle caused by the multipath effect in an ideal case; Estimated oscillation component: represents the multipath error estimated by the augmented state Kalman filter method proposed in the present application; Measured oscillation: represents the multipath oscillation component contained in the actual observation data of the radar, which usually also contains measurement noise and other interference; Where the abscissa is time (unit: seconds) and the ordinate is the size of the oscillation component (unit: degrees); according to Figure 2 The three curves in the figure can be understood as follows: after a period of initial iterative convergence process, the estimated oscillation component is highly consistent with the real oscillation component, indicating that the method of the present application can accurately and stably track the amplitude and period changes of the multipath error, and the method has dynamic adaptability to the amplitude and frequency of the multipath oscillation, and is suitable for time-varying multipath environment.

[0050] Please refer to Figure 3 and Figure 4 , which show the performance of the method of the present application and the classical filtering method in pitch angle estimation by comparison, Figure 3 and Figure 4 include: Real pitch angle: the real pitch angle change trajectory of the target; Noisy measurement: the actual observed pitch angle data of the radar, containing multipath error and sensor noise; Standard KF estimation: the pitch angle estimation result obtained by using the classical Kalman filter method; Augmented state KF estimation: the pitch angle estimation result obtained by using the augmented state Kalman filter method proposed in the present application.

[0051] By comparing the performance indicators of the standard KF and the augmented state KF, the root mean square error (RMSE) of the standard KF is 0.9266 degrees, and the maximum error is 4.0887 degrees; the RMSE of the augmented state KF is 0.2772 degrees, and the maximum error is 1.2360 degrees, so it can be known that the augmented state Kalman filtering method has a significant improvement in the estimation accuracy of the pitch angle, and the RMSE is reduced by 70.08%, and the maximum error is reduced by 69.77%; According to Figure 3 and Figure 4 , respectively showing the pitch angle error of the standard KF and the augmented state KF changing with time, it can be understood that the curve with noise measurement shows obvious periodic oscillation, verifying the error characteristics caused by the multipath effect; the standard KF estimation result still contains significant multipath error, indicating that the traditional method is difficult to effectively suppress this type of structured error; The augmented state KF estimation result is closer to the true pitch angle, and the error is significantly reduced, indicating that the method of the application can significantly improve the pitch angle measurement accuracy; the error curve shows that the method of the application maintains a low error level throughout the time range, and has good stability and consistency. Based on this, the application successfully realizes real-time and high-precision modeling and compensation of the low sweep angle multipath error through the augmented state Kalman filtering method; the method has good convergence and tracking ability, and can adapt to the change of the multipath characteristics; compared with the classical filtering method, the application has a significant improvement in the estimation accuracy of the pitch angle, and the root mean square error and the maximum error are greatly reduced.

[0052] In summary, the low sweep angle multipath effect real-time compensation method, by constructing an augmented state vector containing target motion state and multipath oscillation characteristic parameters, combining a state transition model and a nonlinear measurement model, and using augmented Kalman filtering to realize joint estimation of multipath oscillation parameters and target motion state, can accurately strip the multipath oscillation error in the original pitch angle measurement, significantly improving the pitch angle measurement accuracy; at the same time, the algorithm is based on pure signal processing, without the need to modify the radar hardware, reducing the system modification cost; it has low calculation delay, can meet the real-time demand of high-speed target tracking, and through the design of parameter self-initialization, it is suitable for complex scenes such as cities, deserts and forests, and can also output multipath characteristic parameters to assist environmental perception, providing high-precision protection for low-altitude unmanned aerial vehicle monitoring, low-altitude logistics navigation and other scenes.

[0053] Please refer to Figure 5 , which shows the structure schematic diagram of the low sweep angle multipath effect real-time compensation system provided by an embodiment of the application, the system comprises: a state vector construction module configured to establish an augmented state vector, the augmented state vector comprising target motion state parameters and pitch angle error oscillation characteristic parameters; The model establishing module is configured to establish a state transition model of the augmented state vector and a nonlinear measurement model, wherein the nonlinear measurement model represents the pitch angle observation as a sum of an error term composed of a target real pitch angle and a pitch angle error oscillation characteristic parameter. The state prediction module is configured to predict a state prediction vector and a prediction covariance matrix at a current time according to the state transition model and a state estimation value at a previous time. The filter updating module is configured to perform filter updating on the state prediction vector and the prediction covariance matrix based on the nonlinear measurement model and actual observation of the radar at the current time, and output a target state estimation value after multipath error compensation and a multipath oscillation characteristic parameter estimation value.

[0054] The third aspect of the present application provides an electronic device, which comprises a processor and a memory connected with the processor; wherein the memory stores instructions executable by the processor, and the instructions are executed by the processor to enable the processor to perform the steps of the low-grazing angle multipath effect real-time compensation method of the first aspect of the present application.

[0055] The fourth aspect of the present application provides a computer readable storage medium, which stores a program for implementing a low-grazing angle multipath effect real-time compensation method, and the program is executed by a processor to perform the steps of the low-grazing angle multipath effect real-time compensation method of the first aspect of the present application.

[0056] It should be noted that the above-mentioned embodiments of the present application are only for description, and do not represent the advantages and disadvantages of the embodiments. The processes depicted in the drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are possible or may be advantageous.

[0057] Each embodiment in the specification is described in a progressive manner, and the same or similar parts between each embodiment can be referred to each other. Each embodiment mainly describes the difference from other embodiments.

Claims

1. A method for real-time compensation of low-grazing-angle multipath effects, characterized in that, The method comprises: establishing an augmented state vector comprising target motion state parameters and pitch angle error oscillation characteristic parameters; establishing a state transition model and a nonlinear measurement model of the augmented state vector, wherein the nonlinear measurement model represents a pitch angle observation as a sum of a target true pitch angle and an error term composed of the pitch angle error oscillation characteristic parameters; predicting a state prediction vector and a prediction covariance matrix at a current time according to the state transition model and a state estimation value at a previous time; filtering and updating the state prediction vector and the prediction covariance matrix based on the nonlinear measurement model and actual observation of the radar at the current time, and outputting a target state estimation value after multipath error compensation and a multipath oscillation characteristic parameter estimation value.

2. The low-grazing-angle multipath effect real-time compensation method of claim 1, wherein, The target motion state parameters comprise position components and velocity components of the target in horizontal and vertical directions; The pitch angle error oscillation characteristic parameters comprise an amplitude, an angular frequency and a phase angle of periodic oscillation of the pitch angle caused by the multipath effect.

3. The low-grazing-angle multipath effect real-time compensation method of claim 2, wherein, The state transition model is established specifically by: time recursively predicting the position components and the velocity components of the target according to a target kinematics relationship; linearly accumulating and updating the phase angle based on the angular frequency; and constructing a process noise model for each state variable in the augmented state vector.

4. The low-grazing-angle multipath effect real-time compensation method of claim 1, wherein, The nonlinear measurement model is established specifically by: calculating a target true pitch angle term according to a geometric projection relationship between the target and the radar; constructing a periodic oscillation error term representing the multipath effect according to the pitch angle error oscillation characteristic parameters; and superimposing and integrating the target true pitch angle term and the periodic oscillation error term into radar measurement noise to form the nonlinear measurement model.

5. The low-grazing-angle multipath effect real-time compensation method of claim 1, wherein, The filtering and updating of the state prediction vector and the prediction covariance matrix specifically comprises: solving an inverse matrix of a covariance matrix of a measurement prediction residual by combining the prediction covariance matrix and a measurement noise covariance matrix, and calculating an optimal weighting matrix; calculating a measurement prediction result through the nonlinear measurement model based on a current state prediction value, and comparing the measurement prediction result with actual collected radar observation data to obtain a deviation; linearly weighting the deviation by using the optimal weighting matrix, and applying a weighted adjusted compensation value to the state prediction vector to complete estimation of the augmented state vector; and updating a state estimation covariance matrix based on the optimal weighting matrix, a partial derivative matrix of a measurement function and the prediction covariance matrix.

6. The low-grazing-angle multipath effect real-time compensation method of claim 1, wherein, The method further comprises noise covariance matrix initialization, specifically comprising: configuring a process noise covariance matrix in a state transition process to represent uncertainty of a state prediction model; configuring a measurement noise covariance matrix in an observation process according to measurement accuracy of a radar sensor to represent uncertainty of radar observation data; and dynamically weighting and balancing confidence between a prediction value and an observation value in a filtering process based on the process noise covariance matrix and the measurement noise covariance matrix.

7. The low-grazing-angle multipath effect real-time compensation method of claim 1, wherein, The filtering and updating of the state prediction vector and the prediction covariance matrix further comprises model linearization, specifically comprising: calculating a partial derivative matrix of a state transition function with respect to the augmented state vector to obtain a linear approximation expression of the state transition function; and The partial derivative matrix of the measurement function with respect to the augmented state vector is calculated to obtain a linear approximation of the measurement function; The covariance matrix prediction and update calculation of the augmented state Kalman filter is performed using the linear approximation of the state transition function and the measurement function.

8. A low angle multi-path effect real-time compensation system, characterized in that, The system comprises the low-grazing-angle multipath effect real-time compensation method according to any one of claims 1 to 7. The state vector construction module is configured to establish an augmented state vector, and the augmented state vector comprises target motion state parameters and pitch angle error oscillation characteristic parameters. The model establishment module is configured to establish a state transition model and a nonlinear measurement model of the augmented state vector, wherein the nonlinear measurement model represents the pitch angle observation as a sum of an error term composed of a target real pitch angle and the pitch angle error oscillation characteristic parameters. The state prediction module is configured to predict a state prediction vector and a prediction covariance matrix at a current time according to the state transition model and a state estimation value at a previous time. The filter update module is configured to perform filter update on the state prediction vector and the prediction covariance matrix based on the nonlinear measurement model and actual observation of the radar at the current time, and output a target state estimation value and a multipath oscillation characteristic parameter estimation value after multipath error compensation.

9. An electronic device, comprising: The electronic device comprises a processor and a memory connected with the processor in communication; wherein the memory stores instructions executable by the processor, and the instructions are executed by the processor to enable the processor to perform the steps of the low-grazing-angle multipath effect real-time compensation method according to any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a program for implementing the low-grazing-angle multipath effect real-time compensation method, and the program is executed by the processor to implement the steps of the low-grazing-angle multipath effect real-time compensation method according to any one of claims 1 to 7.

Images