A method, system, device and medium for real-time velocity measurement of a moving target in a tubular structure

Abstract

The application discloses a kind of tubular structure inside moving target intermediate frequency signal real-time velocity measurement method, system, equipment and medium, it is related to millimeter wave interferometric velocimetry field, this method is executed by FPGA on-chip velocity measurement pipeline, comprising: intermediate frequency echo signal is synchronously cached to generate digital intermediate frequency echo sequence;In initial search interval, the sparse window length transformation of prediction constraint is performed to obtain sparse time-frequency representation;Covariance matrix is constructed based on sparse time-frequency representation, and information geometry fusion is carried out on the symmetric positive definite matrix manifold to obtain fused covariance matrix;Eigenvalue decomposition is carried out to fused covariance matrix, determine main characteristic frequency point and its neighborhood energy, and the target Doppler frequency is calculated;After obtaining the instantaneous velocity of moving target based on this frequency, update initial search interval, carry out the detection of next round.The application can reduce the calculation complexity, improve the stability and precision of velocity measurement under complex multipath and non-stationary noise conditions.

Description

Technical Field

[0001] This application relates to the field of millimeter-wave interferometric velocimetry, and in particular to a method, system, device, and medium for real-time velocimetry of intermediate frequency signals of moving targets inside a tubular structure. Background Technology

[0002] Millimeter-wave interferometric velocimetry devices offer advantages such as short wavelength, high velocity sensitivity, compact structure, and suitability for non-contact measurement. They typically transmit millimeter-wave signals into the tubular structure via a millimeter-wave radio frequency front-end and receive echo signals reflected back from moving targets. Target motion causes Doppler frequency or phase changes in the echoes. After mixing, filtering, and amplification, the millimeter-wave interferometric velocimetry device outputs an intermediate-frequency (IF) echo signal for subsequent digital processing. Since the IF echo signal retains the Doppler frequency and phase information caused by target motion, real-time processing of the IF echo signal is the core component for velocimetry of moving targets inside the tubular structure.

[0003] Existing velocity measurement methods commonly use Fourier transform, short-time Fourier transform, or conventional peak detection to extract Doppler frequencies from intermediate frequency echo signals. While Fourier transform can obtain the overall spectrum of the signal, it struggles to describe the instantaneous frequency characteristics of a moving target's velocity changing rapidly over time. Although short-time Fourier transform offers some time-frequency analysis capabilities, its time and frequency resolutions are limited by a fixed window length, making it difficult to simultaneously meet the real-time and accuracy requirements for high-speed or rapidly changing targets inside tubular structures.

[0004] The S-transform is also a time-frequency transform analysis method with variable window length. It has good time-frequency localization capabilities and can simultaneously preserve amplitude and phase information, making it suitable for analyzing non-stationary Doppler signals. However, traditional global S-transform requires time-frequency calculations on a large number of candidate Doppler frequencies over a wide frequency range, involving numerous complex multiplications and additions, window function calculations, and memory access operations. Performing the global S-transform directly on the Field-Programmable Gate Array (FPGA) chip would result in high resource consumption, significant processing latency, and high memory bandwidth requirements, which is not conducive to real-time processing of intermediate frequency signals.

[0005] On the other hand, reflections from the inner wall of the tubular structure, multipath propagation, sidelobe interference, and non-stationary noise can cause shifts or anomalies in the maximum spectral peak or local time-frequency peak within a single sliding window. Directly calculating velocity based on the single-window time-frequency peak can easily lead to velocity results that fluctuate, exhibit abnormal jumps, or become unstable. Therefore, relying solely on single-window peak detection is insufficient to meet the requirements for reliable velocity measurement in complex tubular structure environments.

[0006] Therefore, there is an urgent need in this field for a real-time speed measurement method for intermediate frequency signals that can perform intermediate frequency echo synchronization buffering, local sparse time-frequency analysis, multi-window statistical fusion and speed calculation on an FPGA chip, so as to improve the stability and accuracy of speed measurement under complex multipath and non-stationary noise conditions while reducing computational complexity. Summary of the Invention

[0007] To address the aforementioned problems in the existing technology, this application provides a method, system, device, and medium for real-time velocity measurement of intermediate frequency signals of moving targets inside a tubular structure.

[0008] To achieve the above objectives, this application provides the following solution: In a first aspect, this application provides a real-time velocity measurement method for intermediate frequency signals of moving targets inside a tubular structure, executed by an on-chip velocity measurement pipeline of an FPGA, the method comprising: Acquire the intermediate frequency echo signal of the moving target and synchronously buffer it to generate a digital intermediate frequency echo sequence; Set an initial search interval, and based on the digital intermediate frequency echo sequence, perform a sparse variable window length transform of the prediction constraint within the initial search interval to obtain a sparse time-frequency representation; Based on the sparse time-frequency representation, multiple sliding window covariance matrices are constructed, and information geometric fusion is performed on the covariance matrices on a symmetric positive definite matrix manifold to obtain a fused covariance matrix. The fusion covariance matrix is ​​decomposed into eigenvalues ​​to determine the main characteristic frequency points and their neighborhood energies, and the target Doppler frequency is obtained by weighted centroid calculation. The instantaneous velocity of the moving target inside the tubular structure is obtained based on the target Doppler frequency; Based on the instantaneous velocity, determine the current Doppler frequency search interval, and use the current Doppler frequency search interval as the initial search interval. Then, return to the step of performing a sparse variable window length transform with prediction constraints within the initial search interval to obtain a sparse time-frequency representation.

[0009] Secondly, this application provides a real-time velocity measurement system for intermediate frequency signals of moving targets inside a tubular structure, embedded in an FPGA processing platform; the FPGA processing platform is used to implement an on-chip velocity measurement pipeline; the real-time velocity measurement system for intermediate frequency signals of moving targets inside a tubular structure includes: The intermediate frequency echo acquisition and buffer module is used to acquire the intermediate frequency echo signal of the moving target and synchronously buffer it to generate a digital intermediate frequency echo sequence. The predictive constraint sparse variable window length time-frequency transform module is used to set an initial search interval and, based on the digital intermediate frequency echo sequence, perform a predictive constraint sparse variable window length transform within the initial search interval to obtain a sparse time-frequency representation. The information geometry fusion module is used to construct multiple sliding window covariance matrices based on the sparse time-frequency representation, and perform information geometry fusion on the covariance matrices on a symmetric positive definite matrix manifold to obtain a fused covariance matrix. The fusion covariance matrix is ​​decomposed into eigenvalues ​​to determine the main characteristic frequency points and their neighborhood energies, and the target Doppler frequency is obtained by weighted centroid calculation. A velocity calculation module is used to obtain the instantaneous velocity of the moving target inside the tubular structure based on the target Doppler frequency; The feedback execution module is used to determine the current Doppler frequency search interval based on the instantaneous velocity, and use the current Doppler frequency search interval as the initial search interval, and return to the step of executing the prediction constraint sparse variable window long time-frequency transformation module.

[0010] Thirdly, this application provides a computer device, including: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of the above-described method for real-time velocity measurement of intermediate frequency signals of moving targets inside a tubular structure.

[0011] Fourthly, this application provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the above-described method for real-time velocity measurement of intermediate frequency signals of moving targets inside a tubular structure.

[0012] According to the specific embodiments provided in this application, this application has the following technical effects: This application provides a method, system, device, and medium for real-time velocity measurement of intermediate frequency (IF) signals of moving targets inside a tubular structure. By performing a sparse variable window length transform with prediction constraints within an initial search interval based on a digital IF echo sequence, a sparse time-frequency representation is obtained. This transforms the global time-frequency transformation into a local sparse time-frequency analysis within the predicted Doppler interval, reducing the number of candidate Doppler frequency points, complex multiplication and addition operations, and storage access operations, thus lowering the computational complexity of real-time IF echo processing. Furthermore, by constructing covariance matrices with multiple sliding windows based on the sparse time-frequency representation and performing information geometric fusion on the covariance matrices on a symmetric positive definite matrix manifold, the effects of reflection from the inner wall of the tubular structure, multipath propagation, and side effects can be reduced. This method addresses the impact of lobe and non-stationary noise on single-window velocity measurement results, improving the stability and reliability of velocity measurement. By fusing the covariance matrix for eigenvalue decomposition and calculating the target Doppler frequency through weighted centroid calculation, a Doppler frequency estimate of the sub-frequency grid can be obtained, improving the accuracy of instantaneous velocity calculation. By executing the real-time velocity measurement method of the intermediate frequency signal of the moving target inside the tubular structure provided in this application on the FPGA on-chip velocity measurement pipeline, it is possible to achieve low-latency, high-precision real-time velocity output. This solves the problems in the existing technology where relying solely on single-window peak detection is prone to offset or anomalies, easily causing velocity result jitter, abnormal jumps or instability, and making it difficult to meet the reliable velocity measurement requirements in complex tubular structure environments. Attached Figure Description

[0013] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0014] Figure 1 A flowchart illustrating a real-time velocity measurement method for a moving target inside a tubular structure using intermediate frequency signals, provided in an embodiment of this application. Figure 2 A flowchart of sparse variable window long time-frequency transform and orthogonal matching pursuit processing for prediction constraints provided in an embodiment of this application; Figure 3 A flowchart of information geometry fusion based on a symmetric positive definite matrix manifold provided in one embodiment of this application; Figure 4 This is a flowchart of eigenvalue decomposition and weighted centroid velocity calculation provided in an embodiment of this application; Figure 5 A schematic diagram of the system structure of the millimeter-wave interferometric velocimetry device, dual-channel analog-to-digital converter, and FPGA processing platform provided in an embodiment of this application; Figure 6This is a schematic diagram of the structure of an on-chip speed measurement pipeline for an FPGA provided in an embodiment of this application; Figure 7 This is a schematic diagram of the structure of a computer device provided in an embodiment of this application. Detailed Implementation

[0015] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0016] To make the above-mentioned objectives, features and advantages of this application more apparent and understandable, the application will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0017] In one exemplary embodiment, this application provides a real-time velocity measurement method for intermediate frequency signals of moving targets inside a tubular structure, executed by an on-chip velocity measurement pipeline of an FPGA. Figure 1 As shown, the method includes: Step 100: Acquire the intermediate frequency echo signal of the moving target and synchronously buffer it to generate a digital intermediate frequency echo sequence.

[0018] Step 101: Set the initial search interval, and perform a sparse variable window length transform of the prediction constraint within the initial search interval based on the digital intermediate frequency echo sequence to obtain the sparse time-frequency representation.

[0019] Step 102: Construct covariance matrices for multiple sliding windows based on sparse time-frequency representation, and perform information geometric fusion on the covariance matrices on a symmetric positive definite matrix manifold to obtain a fused covariance matrix.

[0020] Step 103: Perform eigenvalue decomposition on the fused covariance matrix to determine the main characteristic frequency points and their neighborhood energies, and obtain the target Doppler frequency by calculating the weighted centroid.

[0021] Step 104: Obtain the instantaneous velocity of the moving target inside the tubular structure based on the target Doppler frequency.

[0022] Step 105: Determine the current Doppler frequency search interval based on the instantaneous velocity, and use the current Doppler frequency search interval as the initial search interval. Return to step 101 to perform the next round of velocity measurement.

[0023] By implementing steps 100-105 above, this application can reduce the computational complexity of global time-frequency analysis, improve the velocity measurement stability under conditions of reflection from the inner wall of the tubular structure, multipath propagation, and non-stationary noise, and achieve low-latency, high-precision real-time velocity output through an FPGA on-chip pipeline.

[0024] In an exemplary embodiment of this application, the intermediate frequency (IF) echo signal includes an analog reference IF echo signal and an analog measurement IF signal containing the target Doppler frequency component. Both the analog reference IF echo signal and the analog measurement IF signal are analog IF signals. The analog reference IF echo signal is used to provide an IF carrier and a phase reference. The analog measurement IF signal is used to carry the Doppler frequency and phase changes caused by the target motion. Based on this, the implementation process of step 100 provided above in this application can be described as follows: Under the control of the same sampling clock or synchronous trigger signal, the analog reference intermediate frequency (IF) echo signal and the analog measurement IF signal are synchronously sampled respectively. After synchronous sampling, the signals enter the FPGA interface logic, where data alignment, bit width conversion, clock domain synchronization, and windowed buffering are completed to obtain the digital IF echo sequence. The digital IF echo sequence includes a reference digital IF echo sequence and a measurement digital IF echo sequence. The reference digital IF echo sequence is obtained by sampling the analog reference IF echo signal through the first sampling channel of the FPGA interface and performing analog conversion, while the measurement digital IF echo sequence is obtained by sampling the analog measurement IF signal through the second sampling channel of the FPGA interface and performing analog conversion.

[0025] The specific implementation of step 100 above is not a simple caching operation, but rather a process that sequentially executes two-channel sampling point pairing, internal fixed-point bit width unification, cross-clock domain transmission, and segmented caching according to a preset window length through the FPGA interface logic. After this process, each sliding window contains a time-aligned reference digital intermediate frequency echo sequence and a measured digital intermediate frequency echo sequence, and has a unique window number and timestamp, which can be used for subsequent Doppler feature sequence generation, frequency search interval prediction, and velocity calculation. Each sliding window corresponds to a window number and timestamp, used for subsequent Doppler frequency search interval prediction, velocity result output, and data feedback.

[0026] In an exemplary embodiment of this application, in order to mitigate the impact of fixed intermediate frequency carrier, channel common disturbance, and amplitude fluctuation on Doppler frequency estimation, the FPGA performs difference frequency processing, phase correction, carrier removal, or amplitude normalization processing on the measured digital intermediate frequency echo sequence based on the reference digital intermediate frequency echo sequence to obtain a Doppler characteristic sequence for sparse variable window long time-frequency conversion. The simulated reference intermediate frequency echo signal of the sliding window is denoted as... Then we have: .

[0027] In the formula, This indicates the amplitude of the reference intermediate frequency echo signal. Represents the imaginary unit. The intermediate frequency carrier signal representing the reference intermediate frequency echo signal. This indicates the initial phase of the reference intermediate frequency echo signal.

[0028] The first The analog measurement intermediate frequency signal of the sliding window is denoted as Then we have: .

[0029] In the formula, This indicates the amplitude of the measured intermediate frequency echo signal. Indicates the first The target Doppler frequency of a sliding window This indicates the initial phase of the measured intermediate frequency echo signal.

[0030] The conjugate multiplication of the analog measured intermediate frequency signal and the analog reference intermediate frequency echo signal can be expressed as: .

[0031] In the formula, This represents the result of conjugate multiplication, primarily including the Doppler phase and frequency changes caused by target motion. After low-pass filtering, phase correction, and normalization, It can be used as a Doppler feature sequence for subsequent sparse variable window long time-frequency transformation to reduce the impact of fixed intermediate frequency carrier, channel common disturbance and amplitude fluctuation on Doppler frequency estimation.

[0032] In an exemplary embodiment of this application, in step 100 above, for the initial sliding window that has just started or where the target has entered the measurement area, since the target velocity at the previous moment (i.e., the measured instantaneous velocity) has not yet been obtained, the initial search interval can be set using a preset initial Doppler frequency search interval, a full-band search interval, or a priori velocity range. After obtaining an instantaneous velocity estimate, the FPGA writes the instantaneous velocity into the prediction register and updates the Doppler frequency search interval of the subsequent sliding window based on the instantaneous velocity at the previous moment (i.e., the newly determined current Doppler frequency search interval is used as the new initial search interval).

[0033] In practical applications, moving targets inside tubular structures exhibit motion continuity between adjacent sliding windows, and the Doppler frequency of the current sliding window typically does not jump irregularly across the entire frequency domain. Therefore, to narrow down the candidate Doppler frequency range for subsequent time-frequency analysis, the process of determining the current Doppler frequency search interval based on instantaneous velocity in step 105 of this application can be described as... Figure 2 The first part, as shown, is the prediction constraint. This interval determination process includes: Step (1): Determine the Doppler center prediction frequency based on the instantaneous velocity and millimeter wave wavelength.

[0034] Let the target velocity at the previous moment be... The current instantaneous velocity is In one practical application, we have: .

[0035] Based on the single Doppler velocity relationship, the Doppler center frequency is predicted by calculating the current instantaneous velocity and millimeter-wave wavelength: .

[0036] In the formula, It is the wavelength of millimeter waves.

[0037] Step (2): Using the Doppler center prediction frequency as the center, determine the current Doppler frequency search interval according to the preset frequency margin. The preset frequency margin is set based on the target acceleration change, sampling error, channel error, and environmental disturbance, that is, considering the target acceleration change, sampling error, channel error, and environmental disturbance, the preset frequency margin is set, denoted as . .

[0038] The current Doppler frequency search range is represented as follows: : .

[0039] In the formula, Indicates the first Doppler center prediction frequency of a sliding window.

[0040] In an exemplary embodiment of this application, the variable-window long-time-frequency transform can be understood as extracting a local signal using a frequency-dependent Gaussian window function at each time position and candidate Doppler frequency, and then performing a matching calculation with the complex exponential function corresponding to that candidate Doppler frequency. For example, the discrete form can be expressed as: .

[0041] In the formula, This refers to the digital intermediate frequency echo sequence or Doppler characteristic sequence of the k-th sliding window. The sampling point number, The candidate Doppler frequency (i.e., the frequency of the candidate Doppler frequency point). The Gaussian window function associated with the candidate Doppler frequencies. Sampling rate, For the k-th sliding window at the candidate Doppler frequency The amplitude of the variable window long-time frequency complex coefficients at the location. This indicates the strength and phase of the corresponding frequency component within the sliding window. The phase information of this frequency component can be represented by the energy, which can be expressed as: .

[0042] Traditional global S-transform requires calculation of a large number of candidate Doppler frequencies over a wide frequency range. This application utilizes the sparse characteristic of the Doppler energy of a moving target concentrating in the main frequency and its neighborhood within a short time window. It performs high-density time-frequency calculations using a first frequency step size within the current Doppler frequency search interval, and low-density time-frequency calculations or skips calculations using a second frequency step size outside the search interval. In the FPGA implementation, the current sliding window data is read from the cache, and the corresponding Gaussian window function and complex exponential coefficients are read or generated based on the candidate Doppler frequency index. The variable-window-length time-frequency transform at the candidate Doppler frequency point is calculated using the FPGA's internal multiply-accumulate operation resources. Based on this, the process of performing a sparse variable-window-length transform with predictive constraints within the initial search interval based on the digital intermediate frequency echo sequence in step 101 of this application to obtain the sparse time-frequency representation can be corresponding to... Figure 2 The second part (sparse variable window long time-frequency transform) and the third part (OMP (Orthogonal Matching Pursuit) sparse reconstruction) include: Step 1): Within the initial search interval, the first density time-frequency (i.e., high density time-frequency) calculation is performed on the digital intermediate frequency echo sequence using the first frequency step size to obtain the first time-frequency result.

[0043] Step 2): Outside the initial search interval, use a second frequency step size to calculate the second density time-frequency (i.e., low density time-frequency) of the digital intermediate frequency echo sequence, or skip the calculation, to obtain the second time-frequency result. The second frequency step size is larger than the first frequency step size.

[0044] Step 3): Map the first and second time-frequency results to the frequency index set according to the candidate Doppler frequency index. The candidate Doppler frequency index is generated based on the initial search interval.

[0045] Step 4): To further extract the main Doppler components, the example identifies the variable-window long-time-frequency transform of the prediction constraint as a sparse representation problem. For the current sliding window, a candidate dictionary is constructed, consisting of variable-window long-time-frequency atoms corresponding to the candidate Doppler frequencies. The variable-window long-time-frequency atoms are windowed complex exponential basis functions composed of complex exponential functions corresponding to the candidate Doppler frequencies and frequency-related Gaussian window functions. The candidate dictionary can be represented as... : .

[0046] In the formula, Candidate Doppler frequencies The corresponding variable window length time-frequency transform atom.

[0047] Step 5): Based on the frequency index set, extract the Doppler sparse components from the candidate dictionary using a finite-iteration orthogonal matching pursuit algorithm, including: (1): The residual signal is determined based on the frequency index set and the candidate dictionary. The current window signal is represented as: .

[0048] In the formula, The digital intermediate frequency echo sequence or Doppler characteristic sequence of the current sliding window. This is a sparse coefficient vector. Since moving targets inside a tubular structure are typically dominated by the principal Doppler component within a single short time window... Only a few coefficients are significantly nonzero.

[0049] The orthogonal matching pursuit algorithm with finite iterations is used to extract the main Doppler sparse components from the candidate dictionary. During initialization, the residual signal is denoted as... : .

[0050] (2) Determine the correlation between the residual signal and the variable window long time frequency atom.

[0051] Among them, the selected atom set is empty, and in the th... m In this iteration, the correlation between the current residual signal and the variable-window long-time-frequency atom can be expressed as: .

[0052] In the formula, The correlation between the current residual signal and the variable-window long-time-frequency atoms, Candidate Doppler frequencies The corresponding conjugate transpose of the variable window long time-frequency transform atom For the first m The residual signal of -1 iteration.

[0053] (3) Select the variable window long time frequency atom with the highest correlation and add it to the selected atom set. Then, use the least squares method to update the coefficients of each variable window long time frequency atom in the selected atom set and update the residual signal according to the updated coefficients.

[0054] Among them, the variable-window long-time-frequency atom with the highest correlation is denoted as: .

[0055] In the formula, The variable-window long-time-frequency atom with the highest correlation. This refers to the argmax function.

[0056] Add the variable-window long-time-frequency atom with the highest correlation to the selected atom set. Let the dictionary formed by the selected atom set be... The coefficients of the variable-window long-time-frequency atoms in the selected atom set are updated using the least squares method, and the coefficients of the variable-window long-time-frequency atoms are denoted as... : .

[0057] The residual signal is then updated based on the updated coefficients, and the updated residual signal is denoted as: .

[0058] (4) When the number of iterations reaches the preset number, the selected set of atoms and their coefficients are taken as the Doppler sparse components. The Doppler sparse components include the selected candidate Doppler frequency points, the corresponding variable window length time-frequency atoms, and their corresponding coefficients. In one embodiment, the preset number of iterations can be set to 3-5 times to ensure that the on-chip processing delay of the FPGA is determined.

[0059] Step 6): Form a sparse time-frequency representation based on the Doppler sparse components. The sparse time-frequency representation includes candidate Doppler frequency indexes, sliding window indexes, variable window length time-frequency coefficients, and at least one of amplitude, phase, or energy obtained from the variable window length time-frequency coefficients.

[0060] In one exemplary embodiment of this application, in practical applications, sparse variable window long time-frequency transform and orthogonal matched pursuit mainly extract candidate Doppler time-frequency features within a single sliding window. To improve the stability of velocity measurement in complex tubular structure environments, such as... Figure 3 As shown, in the implementation of step 102 above, the covariance matrix is ​​constructed by utilizing the statistical consistency among multiple consecutive sliding windows.

[0061] For the k-th sliding window, a time-frequency feature vector can be constructed based on the sparse time-frequency representation, and denoted as . : .

[0062] In the formula, Candidate Doppler frequencies Energy at the Road It can include features such as amplitude, phase, inter-channel phase difference, inter-channel amplitude ratio, and OMP residual information. The time-frequency feature vectors within multiple consecutive sliding windows constitute a time-frequency feature vector sequence.

[0063] To perform multi-window sparse time-frequency feature extraction, for example, by taking a sliding window of length L, the sliding window of length L is denoted as... : .

[0064] In the formula, This represents the time-frequency feature vector within a sliding window of length L.

[0065] Calculate the mean of the feature vectors within the sliding window, and denote the mean as... : .

[0066] In the formula, .

[0067] Calculate the corresponding covariance matrix, and denote the covariance matrix as... : .

[0068] In the formula, H This represents the conjugate transpose. Covariance matrix. This is used to describe the statistical correlation between multiple candidate Doppler frequencies, phase features, and channel features. The true target Doppler features typically have a relatively stable energy and phase structure over multiple consecutive windows, while noise peaks, multipath peaks, or sidelobe peaks usually exhibit local anomalies, thus leading to deviations in the covariance structure.

[0069] In cases with finite samples, high feature dimensionality, or correlations among features, the covariance matrix may be positive semi-definite or near-singular. To ensure subsequent distance and mean calculations are performed on a symmetric positive definite (SPD) manifold, this application diagonally loads the covariance matrix, resulting in: .

[0070] In the formula, The covariance matrix after diagonal loading. A positive diagonal loading coefficient. It is the identity matrix. The diagonally loaded covariance matrix satisfies the symmetric positive definite condition and can be used for subsequent information geometric fusion.

[0071] The diagonally loaded covariance matrix can be viewed as a point on the stream of symmetric positive definite matrices. Unlike directly comparing matrices element-by-element in Euclidean space, distance measurement and fusion on the stream of symmetric positive definite matrices can more reasonably describe the differences between different sliding window statistical structures.

[0072] For two symmetric positive definite matrices and Its affine invariant Riemann distance is denoted as : .

[0073] In the formula, Represents the matrix logarithm. This represents the Frobenius norm. This distance is used to measure the geodesic distance between two covariance matrices on a symmetric positive definite matrix manifold, thus characterizing the difference between the statistical structures of the time-frequency features of two sliding windows.

[0074] Calculate the Fréchet mean of multiple sliding window covariance matrices over a symmetric positive definite matrix manifold, and use the Fréchet mean as the manifold center of the multiple covariance matrices. The manifold center is denoted as: .

[0075] In the formula, The population center is then identified. Subsequently, the fusion weights are determined based on the affine-invariant Riemann distance between each covariance matrix and the population center. Sliding windows with smaller distances have higher weights, while those with larger distances have lower weights. Larger distances typically indicate that the statistical structure of the time-frequency characteristics deviates significantly from the overall trend, potentially influenced by multipath propagation, noise, or anomalous sidelobes.

[0076] Based on the obtained fusion weights, multiple covariance matrices are weighted and fused to obtain a fused covariance matrix (i.e., the fusion result). The fused covariance matrix integrates the stable time-frequency statistical characteristics of multiple continuous sliding windows, which can reduce the impact of single-window anomalous peaks on the final velocity result.

[0077] Furthermore, the fusion weights can be modified based on the KL divergence or symmetric KL divergence between each covariance matrix and the epidemic center to reduce the weight of the outlier window.

[0078] In one exemplary embodiment of this application, such as Figure 4 As shown, when performing step 103 above, in practical applications, after obtaining the fusion covariance matrix, the fusion covariance matrix is... Eigenvalue decomposition is performed, and the decomposition formula is as follows: .

[0079] In the formula, To integrate the covariance matrix, To the candidate Doppler frequency point The components (i.e., eigenvectors) on the vector. These are the eigenvalues.

[0080] Alternatively, the fusion covariance matrix can be written in matrix form as follows: .

[0081] In the formula, It is an eigenvalue diagonal matrix. Let be the eigenvector matrix. Select the eigenvector corresponding to the largest eigenvalue as the principal eigenvector. Since there is a correspondence between the eigendimensionality of the fused covariance matrix and the candidate Doppler frequency points or their corresponding time-frequency eigendimensionality, the principal eigenfrequency points can be determined based on the component energy of the principal eigenvector at each candidate Doppler frequency point.

[0082] The characteristic energy formula is denoted as: .

[0083] In the formula, The main characteristic component is energy.

[0084] The principal characteristic component energy of candidate Doppler frequencies can also be combined with fused time-frequency energy or sliding window confidence weights for confirmation. A local frequency neighborhood is formed by selecting the principal characteristic frequency and its adjacent frequencies, and a weighted centroid is calculated based on this local frequency neighborhood. The formula for calculating the weighted centroid is denoted as: .

[0085] In the formula, The target Doppler frequency is defined as follows. Through weighted centroid calculation, the output target Doppler frequency can be located between discrete candidate Doppler frequency points, thus achieving frequency estimation of the subfrequency grid.

[0086] Furthermore, when using a simulated reference intermediate frequency echo signal and a simulated measured intermediate frequency signal, the target Doppler frequency can also be obtained from the difference between the measured intermediate frequency estimate and the reference intermediate frequency estimate.

[0087] In one exemplary embodiment of this application, based on the above description, as follows Figure 4 As shown, the implementation process of step 104 can be described as follows: Based on the target Doppler frequency millimeter wave wavelength or system calibration coefficients Calculate the instantaneous velocity of the moving target inside the tubular structure. This instantaneous velocity is expressed as: .

[0088] In the formula, The instantaneous velocity of the moving target inside the tubular structure.

[0089] This application reduces the number of candidate Doppler frequencies in the global time-frequency transformation by predictive constraints, extracts the main Doppler sparse components through a finite-iteration orthogonal matching pursuit algorithm, suppresses the abnormal window effects caused by multipath and non-stationary noise through information geometric fusion, and realizes Doppler frequency estimation of the sub-frequency grid through weighted centroid. It can realize low-latency, high-precision and reliable mid-frequency real-time speed measurement on FPGA.

[0090] Based on the same inventive concept, this application also provides a real-time velocity measurement system for intermediate frequency signals of moving targets inside a tubular structure, used to implement the aforementioned method for real-time velocity measurement of intermediate frequency signals of moving targets inside a tubular structure. The solution provided by this system is similar to the solution described in the above method. Therefore, the specific limitations in one or more embodiments of the real-time velocity measurement system for intermediate frequency signals of moving targets inside a tubular structure provided below can be found in the limitations of the real-time velocity measurement method for intermediate frequency signals of moving targets inside a tubular structure described above, and will not be repeated here.

[0091] In one exemplary embodiment, this application provides a real-time velocity measurement system for intermediate frequency signals of moving targets inside a tubular structure, which is embedded in an FPGA processing platform. This FPGA processing platform is used to implement an on-chip FPGA velocity measurement pipeline. The system includes: an intermediate frequency echo acquisition and buffering module, a predictive constraint sparse variable window long time-frequency transform module, an information geometric fusion module, a velocity calculation module, and a feedback execution module.

[0092] The intermediate frequency echo acquisition and buffer module is used to acquire the intermediate frequency echo signal of the moving target and synchronously buffer it to generate a digital intermediate frequency echo sequence.

[0093] The predictive constraint sparse variable window length time-frequency transform module is used to set the initial search interval and, based on the digital intermediate frequency echo sequence, to perform the predictive constraint sparse variable window length transform within the initial search interval to obtain the sparse time-frequency representation.

[0094] The information geometry fusion module is used to construct covariance matrices with multiple sliding windows based on sparse time-frequency representations, and to perform information geometry fusion on the covariance matrices on a symmetric positive definite matrix manifold to obtain a fused covariance matrix.

[0095] The velocity calculation module performs eigenvalue decomposition on the fused covariance matrix to determine the principal characteristic frequency points and their neighborhood energies, and obtains the target Doppler frequency through weighted centroid calculation. Based on the target Doppler frequency, the instantaneous velocity of the moving target inside the tubular structure is obtained.

[0096] The feedback execution module is used to determine the current Doppler frequency search interval based on the instantaneous velocity, and use the current Doppler frequency search interval as the initial search interval, and return to the steps of executing the predictive constraint sparse variable window long time-frequency transformation module.

[0097] As an optional implementation, the intermediate frequency echo sampling buffer module acquires the intermediate frequency echo signal output by the millimeter-wave interferometric velocimetry device, and synchronously samples it through the dual-channel analog-to-digital conversion unit in the FPGA processing platform to obtain the reference digital intermediate frequency echo sequence and the measurement digital intermediate frequency echo sequence. The reference digital intermediate frequency echo sequence and the measurement digital intermediate frequency echo sequence are then buffered in the FPGA-controlled buffer unit.

[0098] As an optional implementation, the predictive constraint sparse variable window length time-frequency transform module is used to predict the current Doppler frequency search interval based on the target velocity of the previous time step (i.e., the target velocity of the previous time step after reading the velocity calculation completed in the previous sliding window and writing it into the FPGA prediction register), and to perform a predictive constraint sparse variable window length time-frequency transform within the current Doppler frequency search interval to obtain a sparse time-frequency representation. In a specific implementation, the predictive constraint sparse variable window length time-frequency transform module is also used to construct a candidate dictionary and extract the main Doppler sparse components from the candidate dictionary through a finite-iteration orthogonal matching pursuit algorithm.

[0099] As an optional implementation, the FPGA processing platform uses the KC705 development board or a customized processing board based on the Kintex-7XC7K325T FPGA.

[0100] When the FPGA processing platform uses the KC705 development board, the KC705 development board can be equipped with a Kintex-7 XC7K325T FPGA. The dual-channel high-speed analog-to-digital converter (ADC) is connected to the FPGA processing platform through the FPGA Mezzanine Card (FMC) interface.

[0101] The first sampling channel of the dual-channel high-speed ADC is used to acquire the analog reference intermediate frequency echo signal, and the second sampling channel is used to acquire the analog measurement intermediate frequency signal containing the target Doppler frequency component.

[0102] As an optional implementation, the FPGA processing platform includes DDR3 memory and on-chip FPGA memory resources.

[0103] The KC705 development board's DDR3 memory can serve as the main cache for long-frame data or sliding window data. The FPGA's internal on-chip storage resources are used for interface buffering and clock domain conversion before ADC data is written to the DDR3 memory, and for temporarily storing window data, sparse time-frequency representations, residual signals, covariance matrices, and intermediate quantities for speed calculations during algorithm pipeline processing. The FPGA's internal on-chip storage resources include one or more of the following: Block Random Access Memory (BRAM), First-In-First-Out (FIFO) data buffers, or register arrays.

[0104] As an alternative implementation, the FPGA processing platform can also employ a customized processing board based on a Kintex-7 XC7K325T FPGA. This customized processing board retains the FPGA's configuration interface and Ethernet communication interface, and integrates or connects an external dual-channel high-speed ADC for receiving two intermediate frequency signals output from the millimeter-wave interferometric velocimetry device.

[0105] In the Kintex-7 FPGA, the FPGA processing platform includes the DSP48E1 computing resource. The DSP48E1 computing resource is used for hardware acceleration of at least some of the multiplication-addition operations in sparse S-transform, correlation calculation in orthogonal matching pursuit, covariance matrix calculation, matrix multiplication-addition operations in eigenvalue decomposition, and weighted centroid calculation. Matrix logarithms, matrix square roots, reciprocals, sine and cosine, or other complex functions can be implemented using lookup tables, a Coordinate Rotation Digital Computer (CORDIC) unit, or vertex approximation. The multiplication-addition operations include one or more of the following: complex multiplication-addition operations in sparse variable-window long-time-frequency transform, correlation calculation in orthogonal matching pursuit, covariance matrix calculation, matrix multiplication-addition operations in eigenvalue decomposition, and weighted centroid calculation.

[0106] To meet the real-time processing requirements of intermediate frequency signals, a window-level pipelined processing approach can be adopted. Data acquisition and buffering in sliding window n+3, sparse S-transformation in sliding window n+2, information geometric fusion in sliding window n+1, and speed calculation in sliding window n can be performed in parallel at different processing stages on the FPGA chip. After the pipeline is filled, the system can continuously output speed results according to the sliding window cycle.

[0107] As an optional implementation, the FPGA can encapsulate the instantaneous velocity, timestamp, sliding window, and system status of the moving target inside the tubular structure output by the velocity calculation module into a velocity data frame, and send it to the host computer via Gigabit Ethernet. This data feedback method is used to monitor and record the velocity measurement results and does not constitute a limitation on the core on-chip velocity measurement algorithm of this invention.

[0108] As an optional implementation, after the dual-channel analog-to-digital converter outputs the reference digital intermediate frequency (IF) echo sequence and the measurement IF echo sequence, the FPGA interface logic first determines the starting sampling point of the two sequences based on the synchronous sampling trigger signal, the sampling validity flag, or the sampling count value. For the fixed transmission delay between the two channels, a configurable delay register, shift register, or FIFO is used to compensate for the delay of one of the data channels, ensuring that the nth sampling point in the reference IF echo sequence and the nth sampling point in the measurement IF echo sequence are output as a pair within the same processing cycle, thus completing data alignment.

[0109] Subsequently, the FPGA interface logic performs bit-width conversion on the sampled data according to the ADC output format. For example, it converts the N-bit parallel sampled data output by the ADC into W-bit signed data required for fixed-point operations within the FPGA. The bit-width conversion includes one or more of the following: offset binary to two's complement format conversion, sign extension, truncation, rounding, or saturation processing, so that the two digital intermediate frequency sequences meet the bit-width requirements of fixed-point operations for subsequent difference frequency processing, phase correction, and sparse variable window long time-frequency transformation.

[0110] Subsequently, the FPGA interface logic synchronizes the data written in the ADC sampling clock domain to the FPGA algorithm processing clock domain or the DDR3 cache access clock domain through asynchronous FIFO, dual-port RAM, or cross-clock domain handshake mechanism. During cross-clock domain synchronization, the reference digital intermediate frequency echo sequence and the measurement digital intermediate frequency echo sequence are written and read out in the form of paired sampled data to maintain the sampling correspondence between the two data streams.

[0111] After the above processing, the FPGA performs windowed buffering of the paired sampled data according to a preset sliding window length and window sliding step size. Specifically, the FPGA counts consecutive sampling points using a window counter. When the number of sampling points reaches the preset window length, the reference digital intermediate frequency echo sequence and the measured digital intermediate frequency echo sequence are written as a sliding window into the BRAM, FIFO, or DDR3 buffer, and a window number and timestamp are assigned to this sliding window. Adjacent sliding windows can slide with or without overlap at preset steps. The window number and timestamp are used for subsequent Doppler frequency search interval prediction, velocity result output, and data feedback.

[0112] Based on the above description, the data processing structure formed by the millimeter-wave interferometric velocimetry device, the dual-channel analog-to-digital converter, and the FPGA processing platform is as follows: Figure 5 As shown.

[0113] As an optional implementation, the real-time velocity measurement system for intermediate frequency signals of moving targets inside a tubular structure provided in this application is configured to encapsulate the instantaneous velocity, timestamp, sliding window number, and system status flag of the moving target inside the tubular structure output by the velocity calculation module into a velocity data frame, and send it to the host computer via gigabit Ethernet.

[0114] Based on the above description, the specific implementation of the real-time velocity measurement system for intermediate frequency signals of moving targets inside a tubular structure provided in this application is described in steps 100-105 above, and the structure of its FPGA on-chip velocity measurement pipeline can be as follows: Figure 6 As shown.

[0115] In summary, the real-time velocity measurement method and system for intermediate frequency (IF) signals of moving targets inside a tubular structure provided in this application transforms the global time-frequency transformation into a local sparse time-frequency analysis within the predicted Doppler interval by using prediction constraints based on the target velocity at the previous moment. This reduces the number of candidate Doppler frequency points, the number of complex multiplications and additions, and the number of memory accesses, thereby lowering the computational complexity of real-time IF echo processing. Utilizing the sparse characteristics of the moving target echoes inside the tubular structure in the time-frequency domain, a finite-iteration orthogonal matched pursuit algorithm is used to extract the main Doppler sparse components, making the on-chip processing delay deterministic and suitable for real-time IF signal velocity measurement. By constructing the sparse time-frequency features of multiple sliding windows into a covariance matrix and performing information geometric fusion on a symmetric positive definite matrix manifold, the influence of reflections from the inner wall of the tubular structure, multipath propagation, sidelobes, and non-stationary noise on the single-window velocity measurement results can be reduced, improving the stability and reliability of the velocity measurement. By performing eigenvalue decomposition on the fused covariance matrix and calculating the weighted centroid of the principal characteristic frequency points and their neighborhood energies, Doppler frequency estimation of the sub-frequency grid is obtained, improving the accuracy of instantaneous velocity calculation. Through on-chip FPGA velocity measurement pipeline execution of synchronous buffering, predictive constraint frequency search, sparse variable window long time-frequency transformation, information geometric fusion, eigenvalue decomposition, and weighted centroid calculation, low-latency, high-precision real-time velocity output can be achieved.

[0116] In one exemplary embodiment, a computer device is provided, which may be a server or a terminal, and its internal structure diagram may be as follows. Figure 7 As shown, the computer device includes a processor, memory, input / output (I / O) interfaces, and a communication interface. The processor, memory, and I / O interfaces are connected via a system bus, and the communication interface is also connected to the system bus via the I / O interfaces. The processor provides computational and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system, computer programs, and a database. The internal memory provides the environment for the operating system and computer programs in the non-volatile storage media to run. The database stores real-time velocity measurement data of intermediate frequency signals from moving targets inside a tubular structure. The I / O interfaces are used for exchanging information between the processor and external devices. The communication interface is used for communication with external terminals via a network connection. When the computer program is executed by the processor, it implements a method for real-time velocity measurement of intermediate frequency signals from moving targets inside a tubular structure.

[0117] Those skilled in the art will understand that Figure 7 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer equipment to which the present application is applied. Specific computer equipment may include, for example, [the following is a list of possible additional structures]. Figure 7The diagram shows more or fewer components, or combinations of certain components, or different component arrangements.

[0118] In one exemplary embodiment, a computer device is provided, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps in the above-described method embodiments.

[0119] In one exemplary embodiment, a computer-readable storage medium is provided storing a computer program that, when executed by a processor, implements the steps in the above-described method embodiments.

[0120] In one exemplary embodiment, a computer program product is provided, including a computer program that, when executed by a processor, implements the steps in the above-described method embodiments.

[0121] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, data stored, data displayed, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of the relevant data must comply with relevant regulations.

[0122] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments described above. Any references to memory, databases, or other media used in the embodiments provided in this application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (RRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM).

[0123] The databases involved in the embodiments provided in this application may include at least one type of relational database and non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the embodiments provided in this application may be general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, quantum computing-based data processing logic devices, etc., and are not limited to these.

[0124] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0125] This document uses specific examples to illustrate the principles and implementation methods of this application. The descriptions of the above embodiments are only for the purpose of helping to understand the methods and core ideas of this application. Furthermore, those skilled in the art will recognize that, based on the ideas of this application, there will be changes in the specific implementation methods and application scope. Therefore, the content of this specification should not be construed as a limitation of this application.

Claims

1. A method for real-time velocity measurement of intermediate frequency signals of moving targets inside a tubular structure, characterized in that, The real-time velocity measurement method for the intermediate frequency signal of the moving target inside the tubular structure is executed by the on-chip FPGA velocity measurement pipeline. Acquire the intermediate frequency echo signal of the moving target and synchronously buffer it to generate a digital intermediate frequency echo sequence; Set an initial search interval, and based on the digital intermediate frequency echo sequence, perform a sparse variable window length transform of the prediction constraint within the initial search interval to obtain a sparse time-frequency representation; Based on the sparse time-frequency representation, multiple sliding window covariance matrices are constructed, and information geometric fusion is performed on the covariance matrices on a symmetric positive definite matrix manifold to obtain a fused covariance matrix. The fusion covariance matrix is ​​decomposed into eigenvalues ​​to determine the main characteristic frequency points and their neighborhood energies, and the target Doppler frequency is obtained by weighted centroid calculation. The instantaneous velocity of the moving target inside the tubular structure is obtained based on the target Doppler frequency; Based on the instantaneous velocity, determine the current Doppler frequency search interval, and use the current Doppler frequency search interval as the initial search interval. Then, return to the step of performing a sparse variable window length transform with prediction constraints within the initial search interval to obtain a sparse time-frequency representation.

2. The method for real-time velocity measurement of intermediate frequency signals of moving targets inside a tubular structure according to claim 1, characterized in that, The intermediate frequency echo signal includes an analog reference intermediate frequency echo signal and an analog measurement intermediate frequency signal containing the target Doppler frequency component; Acquire the intermediate frequency (IF) echo signal of the moving target and synchronously buffer it to generate a digital IF echo sequence, including: The analog reference intermediate frequency echo signal and the analog measurement intermediate frequency signal are synchronously sampled and converted from analog to digital to obtain a reference digital intermediate frequency echo sequence and a measurement digital intermediate frequency echo sequence; the digital intermediate frequency echo sequence includes the reference digital intermediate frequency echo sequence and the measurement digital intermediate frequency echo sequence.

3. The method for real-time velocity measurement of intermediate frequency signals of moving targets inside a tubular structure according to claim 1, characterized in that, Based on the digital intermediate frequency echo sequence, a sparse variable window length transform with prediction constraints is performed within the initial search interval to obtain a sparse time-frequency representation, including: Within the initial search interval, a first density time-frequency calculation is performed on the digital intermediate frequency echo sequence using a first frequency step size to obtain a first time-frequency result; Outside the initial search interval, a second frequency step size is used to perform a second density time-frequency calculation or skip the calculation on the digital intermediate frequency echo sequence to obtain a second time-frequency result; the second frequency step size is greater than the first frequency step size. The first time-frequency result and the second time-frequency result are mapped to a frequency index set according to the candidate Doppler frequency point index; the candidate Doppler frequency point index is generated based on the initial search interval; Construct a candidate dictionary composed of variable-window long-time-frequency atoms; the variable-window long-time-frequency atoms are windowed complex exponential basis functions composed of complex exponential functions corresponding to candidate Doppler frequency points and frequency-related Gaussian window functions; Based on the frequency index set, Doppler sparse components are extracted from the candidate dictionary using a finite-iteration orthogonal matching pursuit algorithm; A sparse time-frequency representation is formed based on the Doppler sparse components; the sparse time-frequency representation includes candidate Doppler frequency index, sliding window index, variable window length time-frequency coefficients, and at least one of amplitude, phase, or energy obtained from the variable window length time-frequency coefficients.

4. The method for real-time velocity measurement of intermediate frequency signals of moving targets inside a tubular structure according to claim 3, characterized in that, Based on the frequency index set, Doppler sparse components are extracted from the candidate dictionary using a finite-iteration orthogonal matching pursuit algorithm, including: The residual signal is determined based on the frequency index set and the candidate dictionary; Determine the correlation between the residual signal and the variable-window long-time-frequency atom; The variable-window long-time frequency atom with the highest correlation is selected and added to the selected atom set. The coefficients of each variable-window long-time frequency atom in the selected atom set are updated using the least squares method. The residual signal is then updated based on the updated coefficients. When the number of iterations reaches a preset number, the selected set of atoms and its coefficients are taken as the Doppler sparse components.

5. The method for real-time velocity measurement of intermediate frequency signals of moving targets inside a tubular structure according to claim 1, characterized in that, Information geometry fusion is performed on the covariance matrix on a symmetric positive definite matrix manifold to obtain a fused covariance matrix, including: The covariance matrices of multiple sliding windows are used as points on the symmetric positive definite matrix manifold, and the Fréchet mean of the multiple covariance matrices on the symmetric positive definite matrix manifold is determined. Using the Fréchet mean as the manifold center of the multiple covariance matrices, the fusion weight of each sliding window is determined based on the affine invariant Riemann distance between the covariance matrix and the manifold center. The multiple covariance matrices are weighted and fused based on the fusion weights to obtain the fused covariance matrix.

6. The method for real-time velocity measurement of intermediate frequency signals of moving targets inside a tubular structure according to claim 3, characterized in that, The fused covariance matrix is ​​subjected to eigenvalue decomposition to determine the principal characteristic frequency points and their neighborhood energies, and the target Doppler frequency is obtained by weighted centroid calculation, including: The fusion covariance matrix is ​​decomposed into eigenvalues ​​to obtain multiple eigenvalues ​​and eigenvectors corresponding to the eigenvalues. The eigenvector corresponding to the largest eigenvalue is selected as the principal eigenvector, and the component energy of the principal eigenvector at each candidate Doppler frequency point is determined. The energy of the main feature frequency point and its neighborhood is determined based on the component energy of the main feature vector at each candidate Doppler frequency point. The main characteristic frequency point and its adjacent frequency points are selected to form a local frequency neighborhood; The target Doppler frequency is obtained by calculating the weighted centroid of each candidate Doppler frequency point and its corresponding weight in the local frequency neighborhood.

7. The method for real-time velocity measurement of intermediate frequency signals of moving targets inside a tubular structure according to claim 1, characterized in that, Determining the current Doppler frequency search interval based on the instantaneous velocity includes: The Doppler center prediction frequency is determined based on the instantaneous velocity and millimeter-wave wavelength; Centered on the predicted Doppler center frequency, the current Doppler frequency search interval is determined according to a preset frequency margin; the preset frequency margin is set based on target acceleration changes, sampling errors, channel errors, and environmental disturbances.

8. A real-time velocity measurement system for intermediate frequency signals of moving targets inside a tubular structure, characterized in that, Embedded in the FPGA processing platform; The FPGA processing platform is used to implement the on-chip FPGA speed measurement pipeline; the real-time speed measurement system for intermediate frequency signals of moving targets inside the tubular structure includes: The intermediate frequency echo acquisition and buffer module is used to acquire the intermediate frequency echo signal of the moving target and synchronously buffer it to generate a digital intermediate frequency echo sequence. The predictive constraint sparse variable window length time-frequency transform module is used to set an initial search interval and, based on the digital intermediate frequency echo sequence, perform a predictive constraint sparse variable window length transform within the initial search interval to obtain a sparse time-frequency representation. The information geometry fusion module is used to construct multiple sliding window covariance matrices based on the sparse time-frequency representation, and perform information geometry fusion on the covariance matrices on a symmetric positive definite matrix manifold to obtain a fused covariance matrix. The fusion covariance matrix is ​​decomposed into eigenvalues ​​to determine the main characteristic frequency points and their neighborhood energies, and the target Doppler frequency is obtained by weighted centroid calculation. A velocity calculation module is used to obtain the instantaneous velocity of the moving target inside the tubular structure based on the target Doppler frequency; The feedback execution module is used to determine the current Doppler frequency search interval based on the instantaneous velocity, and use the current Doppler frequency search interval as the initial search interval, and return to the step of executing the prediction constraint sparse variable window long time-frequency transformation module.

9. A computer device, comprising: A memory, a processor, and a computer program stored in the memory and capable of running on the processor, characterized in that the processor executes the computer program to implement the method for real-time velocity measurement of intermediate frequency signals of moving targets inside a tubular structure as described in any one of claims 1-7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When executed by a processor, the computer program implements the method for real-time velocity measurement of intermediate frequency signals of moving targets inside a tubular structure as described in any one of claims 1-7.

Images