A data processing method for airborne laser doppler velocimeter radar

Abstract

This invention discloses a data processing method for an airborne laser Doppler velocimetry radar, comprising: S1, an AD converter acquiring echo data from the airborne laser Doppler velocimetry radar and converting it into a digital signal; S2, dividing the converted digital signal into range gates; S3, performing spectral calculations on the digital signal in each range gate; S4, employing a peak prediction and extraction strategy based on particle filtering to identify spectral peaks in the digital signal within each range gate and extracting the peak position of the signal; S5, calculating the velocity based on the extracted peak position. This invention solves the problems of existing peak selection methods for laser Doppler velocimetry radar, which, due to the use of subsequent verification to remove erroneous values, cannot determine whether the extracted peak is the signal peak at the extraction time, and suffer from high algorithm complexity and poor real-time performance.

Description

Technical Field

[0001] This invention relates to, but is not limited to, the field of radar technology, and particularly to a data processing method for an airborne laser Doppler velocimetry radar. Background Technology

[0002] Airborne laser Doppler velocimetry radar requires peak extraction from the detected spectrum when detecting airspeed, and the accuracy of peak extraction directly determines the accuracy of airspeed measurement. The measurement results of airborne laser Doppler velocimetry radar are affected by the aircraft's attitude and speed, and are characterized by large variations in the position of the detected peaks and numerous interference factors.

[0003] Existing peak filtering methods for laser Doppler velocity radar mostly use accuracy indicators such as SNR and relative peak height to filter peaks, such as the tracking correlation method (e.g., CN107942342B). These methods mostly use subsequent verification to eliminate erroneous values, and cannot determine whether the extracted peak value is the signal peak value at the extraction time. The algorithms are highly complex, have poor real-time performance, and are not suitable for airborne wind measurement systems. Summary of the Invention

[0004] The purpose of this invention is to provide a data processing method for airborne laser Doppler velocimetry radar to solve the problems of existing peak value screening methods for laser Doppler velocimetry radar. These methods rely on subsequent verification to remove erroneous values, which makes it impossible to determine whether the extracted peak value is the signal peak value at the extraction time. Furthermore, the methods suffer from high algorithm complexity and poor real-time performance.

[0005] The technical solution of the present invention: The embodiments of the present invention provide a data processing method for an airborne laser Doppler velocity measurement radar, comprising:

[0006] S1, the AD converter acquires the echo data of the airborne laser Doppler velocity radar and converts it into a digital signal;

[0007] S2, divides the converted digital signal into a distance gate;

[0008] S3, perform spectral calculations on the digital signals in each of the divided distance gates;

[0009] S4 employs a peak prediction and extraction strategy based on particle filtering to identify spectral peaks in the digital signal within each distance gate and extract the peak position of the signal.

[0010] S5, calculate the speed based on the extracted peak positions.

[0011] Optionally, in the data processing method of the airborne laser Doppler velocity radar described above,

[0012] The distance gate division in S2 can be used to divide the distance gates into different heights in one direction; or, to divide the data into fixed distances.

[0013] Optionally, in the data processing method of the airborne laser Doppler velocity radar described above,

[0014] S3 includes: spectrum calculation, autocorrelation denoising, and spectrum accumulation processes; wherein the spectrum calculation adopts the FFT method or the maximum likelihood method.

[0015] Optionally, in the data processing method of the airborne laser Doppler velocimetry radar described above, step S4 includes:

[0016] S41, Establish the wind speed and wind direction pre-measurement state equations;

[0017] S42, Establish the wind speed and direction prediction observation equation;

[0018] S43 uses particle filtering to predict wind speed and direction.

[0019] S44, calculate the predicted peak location based on the predicted wind speed and direction;

[0020] S45, Measure the peak position and determine the validity of the peak position based on the predicted value of the peak position;

[0021] S46, at the next moment, repeat the wind speed and direction prediction.

[0022] Optionally, in the data processing method of the airborne laser Doppler velocimetry radar described above, step S41 includes:

[0023] Based on the second-order uniform velocity model, the wind speed and direction state vector at time n is defined as follows:

[0024]

[0025] The corresponding wind speed and direction prediction state equation is:

[0026] W(n+1)=F CV W(n)+G Cv v(n); (1)

[0027] Where Vw(n) and Let Vwθ(n) be the wind speed and the rate of change of wind speed at time n, respectively. These represent the wind direction and the rate of change of wind direction at time n, respectively.

[0028]

[0029]

[0030] v(n) = [v w (n), v w (n), v θ (n), v θ (n)] T v is the state noise vector; w (n) is the wind speed acceleration, v θ (n) is the wind direction acceleration, and ΔT represents the time interval between two consecutive wind speed and direction predictions.

[0031] Optionally, in the data processing method of the airborne laser Doppler velocity radar described above, the radar is selected as a three-beam airborne Doppler laser velocity radar, and S42 includes:

[0032] S42-1, the wind speed values ​​measured by the three laser beams are v1, v2, and v3, respectively:

[0033] v1 = V*cos(θ-n);

[0034] v2 = V * cos(θ);

[0035] v3 = V*cos(θ+σ);

[0036] The time delay for the signal from the Mth distance gate to the rth measurement position (three measurement positions for three laser beams) is:

[0037] τ1=0,

[0038]

[0039] Where σ is the angle between the three laser beams and the axis of rotation of the cone, V is the wind speed to be measured, θ is the wind direction angle to be measured; r = 2, 3; D is the range gate length of the laser Doppler wind radar; c is the speed of laser propagation in air;

[0040] The loss of airborne laser wind radar speed measurement data for S42-2 is:

[0041]

[0042] Where f = 1 / λ, λ is the laser wavelength; j is the imaginary unit;

[0043] S42-3, the received signal vector of the airborne laser wind measuring radar is the wind speed and direction prediction equation:

[0044] Z(n)=ξ(θ,V)S(n)|W(n); (2)

[0045] Where n = 1, 2, ..., L, L is the number of sampling points in one period, S(n) is the source signal, and W(n) is the noise vector at three locations;

[0046] The wind speed and direction prediction state equation and the wind speed and direction prediction observation equation are combined to form the state space model for wind speed and direction prediction.

[0047] Optionally, in the data processing method of the airborne laser Doppler velocity radar described above, step S43 includes: recursively estimating wind speed and direction values ​​using a particle filter based on the wind speed and direction pre-side state equation and the wind speed and direction pre-side observation equation to achieve wind speed and direction prediction. The entire prediction process is divided into two stages: prediction and update.

[0048] (1) Prediction stage: Construct the likelihood function p(x) based on the state-space model of wind speed and direction prediction. o ), the particle swarm that generates spectral peak positions from p(xo) Each sample obtains prior samples at time k through the state equation, and the wind speed and direction prediction state of the i-th particle is obtained as follows:

[0049]

[0050] in, This refers to the iterative method of the state equation, i.e., the prior probability density function. This represents the wind speed and direction state vector of the i-th particle at time k; This represents the wind speed and direction state vector obtained from the iteration at the first k-1 time steps of the i-th particle;

[0051] (2) Update phase: The measurement result Z is received at time k. k Then, the probability of each prior sample is evaluated, and the particle weights are updated as follows:

[0052]

[0053] And normalized to:

[0054]

[0055] A new set of particles is obtained after resampling. By synthesizing particles, the posterior probability density function is obtained:

[0056]

[0057] Where x k Let k be the wind speed and direction state vector at time k. Let be the observation vector obtained by the i-th particle through the first k time steps of the iterative process. For x k The Dirac function.

[0058] Optionally, in the data processing method for airborne laser Doppler velocimetry radar described above, the calculation steps of the particle filter algorithm in S43 include:

[0059] (1) Initialization: Sample N s There are particles, each with a weight of . The initial wind speed and direction of each particle were estimated using the MUSIC algorithm, and the rate of change of wind speed was calculated. For accurate wind speed measurement, the rate of change of wind direction is required. For the accuracy of wind direction measurement;

[0060] (2) Update weights: At time k, the eigenvalues ​​are decomposed into the noise subspace according to the covariance matrix in the MUSIC algorithm. The particle state is obtained by extracting the particles according to equation (3), and the signal flow vector ξ(θ) is obtained. (i) V (i) Then, the likelihood function is calculated using the MUSIC algorithm, and the particle weights are calculated according to equation (4).

[0061] (3) Resampling: Calculate the normalized weights according to equation (5), and resample N. s Each particle is assigned a weight.

[0062] (4) State estimation: Calculate the estimated value of the state according to equation (6);

[0063] (5) Prediction: Based on the wind speed and direction prediction state equation (1), predict the wind speed and direction state at the next moment.

[0064] (6) Enter S44.

[0065] Optionally, in the data processing method of the airborne laser Doppler velocimetry radar described above, step S44 includes:

[0066] Based on the predicted wind speed and direction at the next moment Obtain the wind speed prediction value Vw that meets a certain set confidence level. k+1 Wind direction forecast value θw k+1 Based on the current aircraft ground speed Vp k and roll angle α k Pitch angle β k and yaw angle γ k Calculate the predicted peak position P k+1 ;

[0067] The S45 includes:

[0068] The set of measured peak points is obtained, and the predicted peak position P is determined accordingly. k+1 Select the optimal peak point as the measured peak point for the current location;

[0069] S46: At time k = k + 1, go to S43 and continue until the measurement ends.

[0070] Optionally, in the data processing method of the airborne laser Doppler velocity radar described above, step S5 includes:

[0071] The final airspeed measurement result is synthesized based on the measurement speed in each laser direction and the laser pointing direction.

[0072] The beneficial effects of this invention are as follows: This invention provides a data processing method for airborne laser Doppler velocimetry radar. It utilizes particle filtering to predict peak position changes over a short period, providing a more reliable standard for peak extraction and increasing its accuracy. This effectively suppresses data drift caused by noise in the measuring equipment. Furthermore, the method reduces the computational load of subsequent peak verification and processing, enabling rapid and accurate extraction of signal peaks using a simpler method with minimal computational complexity. Attached Figure Description

[0073] The accompanying drawings are provided to further understand the technical solutions of the present invention and constitute a part of the specification. They are used together with the embodiments of this application to explain the technical solutions of the present invention and do not constitute a limitation on the technical solutions of the present invention.

[0074] Figure 1 A flowchart illustrating a data processing method for an airborne laser Doppler velocimetry radar, provided as an embodiment of the present invention;

[0075] Figure 2 A flowchart illustrating the peak prediction and extraction process based on particle filtering in the data processing method for airborne laser Doppler velocimetry radar provided in this embodiment of the invention;

[0076] Figure 3 A schematic diagram of the three-beam laser Doppler radar wind measurement structure in the data processing method of the airborne laser Doppler velocity radar provided in the embodiments of the present invention. Detailed Implementation

[0077] To make the objectives, technical solutions, and advantages of the present invention clearer, the embodiments of the present invention will be described in detail below with reference to the accompanying drawings. It should be noted that, unless otherwise specified, the embodiments and features described in this application can be arbitrarily combined with each other.

[0078] As explained in the background section, the existing screening method for peak values ​​of laser Doppler velocity radar cannot determine whether the extracted peak value is the signal peak value at the extraction time because it uses subsequent verification to remove erroneous values. Furthermore, the algorithm is highly complex and has poor real-time performance, making it unsuitable for airborne wind measurement systems.

[0079] Particle filtering, also known as sequential Monte Carlo filtering, is a recursive Bayesian estimation algorithm for nonlinear, non-Gaussian systems based on the Monte Carlo method. Its sampling is concentrated in high-probability regions, and the sampling calculation process is simple. For linear Gaussian systems, the minimum variance estimate of the state can be given by the Kalman filter equation. However, for nonlinear, non-Gaussian models, it is difficult to obtain the optimal solution through Kalman filtering, requiring approximate algorithms to obtain the Bayesian estimate of the state. The particle filtering method employs the Monte Carlo method, making it suitable for optimal signal processing of nonlinear, non-Gaussian systems; and it uses sequential inference, updating the inference results for some or all elements upon receiving a measurement value, making it suitable for real-time online data processing.

[0080] For airborne wind measurement systems, and based on the aforementioned characteristics of particle filtering, this invention provides a data processing method for airborne laser Doppler velocity radar, specifically a data processing method for laser Doppler wind measurement radar based on particle filtering.

[0081] The present invention provides the following specific embodiments, which can be combined with each other. For the same or similar concepts or processes, they may not be described again in some embodiments.

[0082] Figure 1 This is a flowchart illustrating a data processing method for an airborne laser Doppler velocimetry radar provided in an embodiment of the present invention. The data processing method for an airborne laser Doppler velocimetry radar provided in this embodiment of the present invention includes the following steps:

[0083] S1, the AD converter acquires the echo data of the airborne laser Doppler velocity radar and converts it into a digital signal;

[0084] S2, divides the digital signal obtained from the conversion in S1 into a distance gate;

[0085] S3, perform spectrum calculation on the digital signal in each of the divided distance gates;

[0086] This step is to remove noise and highlight the echo signal.

[0087] S4 employs a peak prediction and extraction strategy based on particle filtering to identify spectral peaks in the digital signal within each distance gate and extract the peak position of the signal.

[0088] S5, calculate the speed based on the extracted peak positions.

[0089] In the data processing method provided in the embodiments of the present invention, the distance gate division in S2 can be performed by dividing distance gates at different heights in one direction, or by using data at a fixed distance. It should be noted that in the following data processing methods provided in the embodiments of the present invention,

[0090] In the spectral operation method of S3 in the embodiment of the present invention, the process usually includes spectrum calculation, autocorrelation denoising and spectrum accumulation. The spectrum calculation can use the FFT method or the maximum likelihood method, mainly to remove noise and highlight the peak value of the echo signal.

[0091] In a specific implementation of this invention, the peak position of the signal is extracted through the peak prediction and extraction strategy based on particle filtering employed in S4, such as... Figure 2 The diagram shows a flowchart of the peak prediction and extraction process based on particle filtering in the data processing method for airborne laser Doppler velocimetry radar provided in an embodiment of the present invention. The specific process includes the following implementation steps:

[0092] S41, Establish the state equation for wind speed and direction prediction;

[0093] There are many ways to establish the state equations for wind speed and direction. This embodiment of the invention takes a second-order uniform velocity model as an example:

[0094] Let the wind speed and direction state vector at time n be:

[0095]

[0096] Where Vw(n) and Wind speed and rate of change of wind speed at time n, and wind acceleration, respectively. Where v w (n) is the wind speed acceleration, which follows a mean of 0 and a variance of . Gaussian distribution;

[0097] Vw θ (n) and Wind direction and rate of change of wind direction at time n, respectively, and wind acceleration. Where v θ (n) is the wind acceleration, which has a mean of 0 and a variance of δ. vθ The Gaussian distribution. Under the second-order uniform (CV) model:

[0098]

[0099]

[0100]

[0101]

[0102] Where ΔT represents the time interval between two consecutive wind speed and direction predictions, in seconds.

[0103] The wind speed and direction prediction state equation established based on the above second-order uniform velocity model is:

[0104] W(n+1)=F CV W(n)+G Cv v(n); (1)

[0105] In the above formula (1),

[0106]

[0107]

[0108] v(n) = [v w (n), v w (n), v θ (n), v θ (n)] T This is the state noise vector.

[0109] S42, Establish the wind speed and direction prediction observation equation;

[0110] Based on the wind measurement structure of the Doppler velocity radar, a guiding vector is constructed. Figure 3 This is a schematic diagram of the three-beam laser Doppler radar wind measurement structure in the data processing method of the airborne laser Doppler velocity radar provided in this embodiment of the invention. Taking the three-beam airborne Doppler laser velocity radar as an example, the three lasers are emitted from the same laser emission point and are evenly distributed in the same cone angle. The angles between the three lasers and the axis of rotation of the cone are all σ. V is the magnitude of the wind speed to be measured, θ is the wind direction angle to be measured, and the wind direction is north-northeast θ. Then the wind speed values ​​measured by the three lasers are v1, v2, and v3, respectively:

[0111] v1 = V*cos(θ-σ);

[0112] v2 = V * cos(θ);

[0113] v3 = V*cos(θ+π);

[0114] The time delay for the signal from the Mth distance gate to the rth measurement position (three measurement positions for three laser beams) is:

[0115] τ1=0,

[0116]

[0117] Where r = 2, 3; D is the range gate length of the laser Doppler wind radar, in meters; and c is the speed of laser propagation in air. The velocity loss of the airborne laser wind radar is then:

[0118]

[0119] Where f = 1 / λ, λ is the laser wavelength, and j is the imaginary unit.

[0120] The radar received signal vector can be obtained as follows:

[0121] Z(n)=ξ(θ,V)S(n)+W(n); (2)

[0122] The formula for the radar received signal vector is the wind speed and wind direction prediction equation.

[0123] Where n = 1, 2, ..., L, L is the number of sampling points in one period. The number of sampling points for FFT is generally a multiple of 2, such as 512, 1024, 2048, etc. Z(n) is the received signal vector, ξ(θ, V) is the radar velocity loss, S(n) is the transmitted source signal, and W(n) is the noise vector at three locations, W(n) = [w1(n), w2(n), w3(n)]. T w1(n), w2(n), and w3(n) are Gaussian white noise with a mean of 0 and a variance of δ2 received from the three laser beams.

[0124] Equation (1) represents the variation law of wind speed and direction to be estimated, and can be used as the state equation;

[0125] Equation (2) represents the relationship between the quantity to be estimated and the observed quantity, and serves as the observation equation;

[0126] The combination of equations (1) and (2) constitutes the state-space model for wind speed and direction prediction.

[0127] S43 uses particle filtering to predict wind speed and direction.

[0128] Based on the received radar signal vector Z(n), the observation data of L sampling points at time k are obtained as follows:

[0129] Z k ={Z(1+kL), Z(2+kL),…, Z(L+kL)};

[0130] Where Z(1+kL) represents the first sample point data received at time k, Z(2+kL) represents the second sample point data received at time k, and so on, Z(L+kL) is the Lth sample point data received at time k.

[0131] Use Z 1：k ={Z1, Z2, ..., Z k [] represents the discrete-time measurement sequence, and K is the sampling period. Based on the wind speed and direction prediction state equation and the wind speed and direction prediction observation equation, a particle filter method is used to recursively estimate the wind speed and direction values, thus achieving wind speed and direction prediction. The entire process is divided into two stages: prediction and update.

[0132] (1) Prediction stage: Construct the likelihood function p(x) based on the state-space model of wind speed and direction prediction.o ), by p(x o ) Particle swarm generating spectral peak positions Each sample obtains prior samples at time k through the state equation, and the wind speed and direction prediction state of the i-th particle is obtained as follows:

[0133]

[0134] It should be noted that the above samples represent the wind speed and direction within the door at a specified time.

[0135] in, This refers to the iterative method of the state equation, i.e., the prior probability density function. This represents the wind speed and direction state vector of the i-th particle at time k; This represents the wind speed and direction state vector obtained from the iteration at the first k-1 time steps of the i-th particle.

[0136] (2) Update phase: The measurement result Z is received at time k. k Then, the probability of each prior sample is evaluated, and the particle weights are updated as follows:

[0137]

[0138] And normalized to:

[0139]

[0140] A new set of particles is obtained after resampling. By synthesizing particles, the posterior probability density function is obtained:

[0141]

[0142] Where x k Let k be the wind speed and direction state vector at time k. Let be the observation vector obtained by the i-th particle through the first k time steps of the iterative process. For x k The Dirac function. Where the likelihood function p(x) O It can be constructed using the multi-classification MUSIC algorithm.

[0143] The specific steps of implementing the particle filter algorithm in this process include:

[0144] (1) Initialization: Sample N s There are particles, each with a weight of . The initial wind speed and direction of each particle were estimated using the MUSIC algorithm, and the rate of change of wind speed was calculated. For accurate wind speed measurement, the rate of change of wind direction is required. For the accuracy of wind direction measurement;

[0145] (2) Update weights: At time k, the eigenvalues ​​are decomposed into the noise subspace according to the covariance matrix in the MUSIC algorithm. The particle state is obtained by extracting the particles according to equation (3), and the signal flow vector ξ(θ) is obtained. (1) V (1) Then, the likelihood function is calculated using the MUSIC algorithm, and the particle weights are calculated according to equation (4).

[0146] (3) Resampling: Calculate the normalized weights according to equation (5), resample Ns particles and assign weights to them.

[0147] (4) State estimation: Calculate the estimated value of the state according to equation (6);

[0148] (5) Prediction: Based on the wind speed and direction prediction state equation (1), predict the wind speed and direction state at the next moment.

[0149] (6) Enter S44.

[0150] S44: Calculate the predicted peak location P based on the predicted wind speed and direction. k+1 .

[0151] Based on the predicted wind speed and direction at the next moment Obtain the wind speed prediction value Vw that meets a certain set confidence level. k+1 Wind direction forecast θW k+1 Based on the current aircraft ground speed Vp k and roll angle α k Pitch angle β k and yaw angle γ k Calculate the predicted peak position P k+1 .

[0152] (1) Convert the predicted wind speed in the direction the laser is pointing to the aircraft ground speed Vp. k and pitch angle α k Roll angle β k and yaw angle γ k In the local geodetic coordinate system, the coordinate transformation method can be expressed as:

[0153]

[0154] Where L LOS The direction of the laser beam is indicated in the geodetic coordinate system; x k y k x k These represent the projection magnitudes of the unit vector pointing to the laser at this moment in the X, Y, and Z directions in the Earth coordinate system; This is the direction of the laser beam in the machine's coordinate system at this moment:

[0155]

[0156] Where θ k At this moment, the laser is pointing in the azimuth direction. This is the elevation angle of the laser's pointing direction at this moment. H1, H2, and H3 are the coordinate transformation matrices about the three axes, as shown in the following formulas:

[0157]

[0158]

[0159]

[0160] Then, at this moment, the laser line-of-sight velocity v caused by the aircraft's motion... LOS_pk :

[0161]

[0162] In the formula: V px_k Vp y_k Vp z_k These represent the velocity components of the aircraft in the three coordinate directions in the Earth coordinate system.

[0163] (2) Calculate the laser line-of-sight wind speed at the current predicted location.

[0164]

[0165] in For a size of Vw k+1 The direction is θW k+1 The predicted wind speed vector.

[0166] (3) Calculate the predicted peak position P k+1 .

[0167] Predicted measurement velocity along the laser line of sight It is the sum of the components of the wind speed and the aircraft speed at that location relative to the laser line of sight speed:

[0168]

[0169] The predicted peak position is:

[0170]

[0171] Where λ is the laser wavelength and P0 is the initial frequency shift relative to the 0-frequency position.

[0172] S45: Obtain the set of measured peak points, and select the optimal peak point as the measured peak point at the current position based on the predicted peak position.

[0173] In practice, this step involves measuring the peak position. With P k+1 By comparing the results, the difference Δ is obtained. k+1 One method for measuring peak values ​​is to select the peak point with the highest SNR from a stored set of peak values. If Δ k+1 If the value is less than the first threshold, the peak value is considered valid. If Δ k+1 If the value exceeds the first threshold, the peak value is considered invalid and requires further processing. This processing could involve discarding the peak value or finding the second-largest SNR peak from the stored peak value set for comparison. This effectively suppresses data drift caused by noise from the measurement equipment.

[0174] S46: At time k = k + 1, go to S43, and repeat S43 to S46 until the measurement ends.

[0175] In S5 of this embodiment, the final airspeed measurement result is synthesized based on the measurement speed in each laser direction and the laser pointing direction.

[0176] In its implementation, this step includes: spectral correction; if an initial frequency shift exists, S5 also includes a frequency shift extraction step to obtain the final frequency difference. Finally, the final velocity result is calculated by combining the Doppler principle and the velocity measurement results of the three laser beams.

[0177] The data processing method for airborne laser Doppler velocimetry radar provided in this invention utilizes particle filtering to predict peak position changes over a short period, providing a more reliable standard for peak extraction and increasing its accuracy. This effectively suppresses data drift caused by noise in the measuring equipment. Furthermore, the method reduces the computational load of subsequent peak verification and processing, enabling rapid and accurate extraction of signal peaks using a simpler method with minimal computational complexity.

[0178] While the embodiments disclosed in this invention are as described above, they are merely illustrative of the embodiments to facilitate understanding of the invention and are not intended to limit the invention. Any person skilled in the art to which this invention pertains may make any modifications and variations in the form and details of the implementation without departing from the spirit and scope disclosed herein; however, the scope of patent protection for this invention shall still be determined by the scope defined in the appended claims.

Claims

1. A data processing method for an airborne laser Doppler velocity measuring radar, characterized in that, include: S1, the AD converter acquires the echo data of the airborne laser Doppler velocity radar and converts it into a digital signal; S2, divides the converted digital signal into a distance gate; S3, perform spectral calculations on the digital signals in each of the divided distance gates; S4 employs a peak prediction and extraction strategy based on particle filtering to identify spectral peaks in the digital signal within each distance gate and extract the peak position of the signal. S5, calculate the speed based on the extracted peak positions; S4 includes: S41, Establish the wind speed and wind direction pre-measurement state equations; S42, Establish the wind speed and direction prediction observation equation; S43 uses particle filtering to predict wind speed and direction. S44, calculate the predicted peak location based on the predicted wind speed and direction; S45, Measure the peak position and determine the validity of the peak position based on the predicted value of the peak position; S46, at the next moment, repeat the wind speed and direction prediction; S41 includes: Based on the second-order uniform velocity model, the wind speed and direction state vector at time n is defined as follows: ； The corresponding wind speed and direction prediction state equation is: ；（1） in, and Let n be the wind speed and the rate of change of wind speed at time n, respectively. and These represent the wind direction and the rate of change of wind direction at time n, respectively. ； ； This is the state noise vector; It is wind speed acceleration. It is wind acceleration. T represents the time interval between two consecutive wind speed and direction predictions; The radar is an airborne Doppler laser velocimetry radar using three laser beams. S42 includes: S42-1, the wind speed values ​​measured by the three laser beams are as follows: : ； ； ； The time delay for the signal from the Mth distance gate to the rth measurement position is: ， ； Where σ is the angle between the three laser beams and the axis of rotation of the cone, V is the wind speed to be measured, θ is the wind direction angle to be measured; r = 2, 3; D is the range gate length of the laser Doppler wind radar; c is the speed of laser propagation in air; The loss of airborne laser wind radar speed measurement data for S42-2 is: ； in, λ is the laser wavelength; j The imaginary unit; S42-3, the received signal vector of the airborne laser wind measuring radar is the wind speed and direction prediction equation: ；（2） Where n = 1,2,…,L, L is the number of sampling points in one period, and S(n) is the source signal; W (n) represents the noise vector at the three locations; The wind speed and direction prediction state equation and the wind speed and direction prediction observation equation are combined as the state space model for wind speed and direction prediction. S43 includes: recursively estimating wind speed and direction values ​​using a particle filter method based on the wind speed and direction pre-side state equation and the wind speed and direction pre-side observation equation to achieve wind speed and direction prediction. The entire prediction process is divided into two stages: prediction and update. (1) Prediction stage: Construct the likelihood function based on the state-space model of wind speed and direction prediction. ,Depend on Particle swarm generating spectral peak positions Each sample obtains prior samples at time k through the state equation, and the wind speed and direction prediction state of the i-th particle is obtained as follows: ；（3） in, This refers to the iterative method of the state equation, i.e., the prior probability density function. This represents the wind speed and direction state vector of the i-th particle at time k; This represents the wind speed and direction state vector obtained from the iteration at the first k-1 time steps of the i-th particle; (2) Update phase: In Receive measurement results in real time Then, the probability of each prior sample is evaluated, and the particle weights are updated as follows: ； （4） And normalized to: ； （5） A new set of particles is obtained after resampling. Synthesize particles to obtain the posterior probability density function: ；（6） in Let k be the wind speed and direction state vector at time k. Let be the observation vector obtained by the i-th particle through the first k time steps of the iterative process. for The Dirac function; S44 includes: Based on the predicted wind speed and direction at the next moment Obtain wind speed prediction values ​​that meet a certain set confidence level. Wind direction forecast Based on the current aircraft ground speed and roll angle Pitch angle and yaw angle Calculate the predicted peak position ; The S45 includes: The set of measured peak points is obtained, and the peak positions are predicted accordingly. Select the optimal peak point as the measured peak point for the current location; S46: Time k = k +1, go to S43, until the measurement is complete.

2. The data processing method for airborne laser Doppler velocity measurement radar according to claim 1, characterized in that, The distance gate division in S2 can be used to divide the distance gates into different heights in one direction; or, to divide the data into fixed distances.

3. The data processing method for airborne laser Doppler velocity measurement radar according to claim 1, characterized in that, S3 includes: spectrum calculation, autocorrelation denoising, and spectrum accumulation processes; wherein the spectrum calculation adopts the FFT method or the maximum likelihood method.

4. The data processing method for airborne laser Doppler velocity measurement radar according to claim 1, characterized in that, The calculation steps of the particle filter algorithm in S43 include: (1) Initialization: Sampling There are particles, each with a weight of . The initial wind speed and direction of each particle were estimated using the MUSIC algorithm, and the rate of change of wind speed was calculated. For accurate wind speed measurement, the rate of change of wind direction is required. For the accuracy of wind direction measurement; (2) Update weights: At time k, the eigenvalues ​​are decomposed into the noise subspace according to the covariance matrix in the MUSIC algorithm. The particle state is obtained by extracting the particles according to equation (3), and the signal flow vector is obtained. Then, the likelihood function is calculated using the MUSIC algorithm, and the particle weights are calculated according to equation (4). (3) Resampling: Calculate the normalized weights according to equation (5) and resample. Each particle is assigned a weight. ; (4) State estimation: Calculate the estimated state value according to equation (6); (5) Prediction: Based on the wind speed and direction prediction state equation (1), predict the wind speed and direction state at the next moment. ; (6) Enter S44.

5. The data processing method for airborne laser Doppler velocity measurement radar according to claim 1, characterized in that, S5 includes: The final airspeed measurement result is synthesized based on the measurement speed in each laser direction and the laser pointing direction.

Images