Target deformation detection method based on Doppler feature guidance

Abstract

The invention discloses a target deformation detection method based on Doppler feature guidance, and the method comprises the following steps: 1, obtaining radar echo signal data, and calculating distance-slow time two-dimensional data; 2, acquiring signal data at a target position and signal data at a reference position, and calculating a phase at the target position and a phase at the reference position; 3, calculating time-frequency distribution of the signal data at the target, extracting Doppler features of the signal data, and recording an existence time set of the Doppler features; 4, compensating the phase at the target position by using the phase at the reference position to obtain target deformation; and 5, in the Doppler feature existence time set, performing deformation detection on the target deformation to obtain a target deformation displacement set and an occurrence time index set. According to the method, the radar is used for counteracting shaking of the platform perceived by the phase at the reference position, detection is guided by combining Doppler features related to target deformation, and the high deformation detection rate can be achieved with the low false alarm rate.

Description

Technical Field

[0001] This invention belongs to the field of signal processing technology, specifically relating to a target deformation detection method based on Doppler feature guidance. Background Technology

[0002] In fields such as structural health monitoring in civil engineering, human vital sign monitoring, and low-altitude safety protection, microscale dynamic feature detection (such as the instantaneous compression of bridges under heavy vehicle loads, micro-movements on the body surface caused by human breathing and heartbeats, and high-frequency vibrations of drone rotors) is a core technology for ensuring safe operation and accurate perception. These deformation signals typically have extremely small amplitudes (millimeter-level) and are easily masked by environmental noise (such as wind interference, electromagnetic noise, and background motion). Accurate identification of these signals is of significant practical importance for scenarios such as structural safety early warning, vital sign monitoring, and low-altitude target control. Existing target deformation detection technologies mainly rely on the analysis of deformation features themselves. This involves collecting spatiotemporal sequence data of the target through sensing methods such as radar, laser, and optical imaging, extracting key indicators such as displacement, rate of change, and trend characteristics of the deformation, and then using methods such as threshold judgment and machine learning models to achieve deformation detection.

[0003] The problem with the above deformation detection is that when relying solely on the characteristics of the deformation itself for detection, it is difficult to effectively distinguish between real target signals and interference signals. The overlap between the two in feature dimensions (frequency, amplitude, time domain mode) makes it difficult for the detection system to achieve both a high detection rate and a low false alarm rate.

[0004] To address this situation, the target deformation detection technology based on Doppler feature guidance proposed in this patent offers a novel solution: target deformation is often strongly correlated with specific sources (e.g., bridge pressure is related to heavy vehicle movement, breathing and heartbeat are related to the periodic movement of the human chest cavity, and micro-motion of a drone is related to rotor rotation). The movement processes of these sources generate perceptible Doppler features (e.g., radar Doppler frequency shift during heavy vehicle movement, Doppler phase change during chest cavity movement, and Doppler spectral features during rotor rotation). By extracting Doppler features directly related to target deformation, potential deformation regions, time windows, or signal frequency bands can be precisely guided and preliminarily screened to eliminate unrelated environmental interference. Based on this, refined detection is then performed by combining deformation features, thereby maintaining a high detection rate while reducing the false alarm rate caused by misjudgment due to interference signals, providing a more reliable technical path for deformation detection in complex scenarios. Summary of the Invention

[0005] The purpose of this invention is to provide a feasible and effective target deformation detection method based on Doppler feature guidance by using the phase of a reference target to compensate for the effect of the shaking of the sensing platform on the phase, and further using the Doppler features related to the target to be detected to guide the detection of target deformation.

[0006] Technical Solution: To solve the above-mentioned technical problems, this invention proposes a target deformation detection method based on Doppler feature guidance, which includes the following steps:

[0007] Step 1: Acquire radar echo signal data and calculate range-slow time two-dimensional data S. RT (m, n);

[0008] Step 2: Acquire signal data x at the target location tar (n) and the signal data x at the reference point ref (n), and calculate the phase p at the target location. tar (n) and the phase p at the reference point ref (n);

[0009] Step 3: Calculate the signal data x at the target location. tar The time-frequency distribution γ(t, f) of (n) is extracted, its Doppler features are extracted, and the time set T in which the Doppler features exist is recorded;

[0010] Step 4: Utilize the phase p at the reference point ref (n) Phase p at the target tar (n) is compensated to obtain the target deformation d(n);

[0011] Step 5: Within the time set T of Doppler feature existence, perform deformation detection on the target deformation d(n) to obtain the target deformation displacement set D and the occurrence time index set T. D .

[0012] Furthermore, in step 1, radar echo signal data is acquired using the following method, and range-slow time two-dimensional data S is calculated. RT (m, n), specifically including the following steps:

[0013] Step 1.1: Acquire the radar echo signal and organize it according to the fast-time-slow-time format to obtain the fast-time-slow-time two-dimensional data S(k, n), where S(k, n) is the radar echo signal value at the k-th fast time point and the n-th slow time point, k=1, 2, …, K is the discrete fast-time index, K is the number of fast-time samples in one radar transmit-receive cycle, and K is a positive integer ≥ 1024, n=1, 2, …, N is the discrete slow-time index, N is the number of slow times acquired by the radar, i.e., the number of radar transmit-receive cycles, and N≥ 60∙f PRF positive integers, f PRF Let f be the pulse repetition frequency, satisfying f PRF Positive integers ≥400Hz;

[0014] Step 1.2: Calculate the distance-slow time two-dimensional data SRT (m, n):

[0015] ;

[0016] Where m = 1, 2, ..., M are discrete distance gate indices, M is the number of distance cells, and M is a positive integer satisfying M ≥ 1024; n = 1, 2, ..., N are discrete slow-time indices. The above formula represents the distance-slow-time two-dimensional data S obtained by performing an FFT on the fast-time dimension of the fast-time two-dimensional data S(k, n). RT (m, n).

[0017] Furthermore, in step 2, the signal data x at the target location is obtained using the following method. tar (n) and the signal data x at the reference point ref (n), and calculate the phase p at the target location. tar (n) and the phase p at the reference point ref (n), specifically including the following steps:

[0018] Step 2.1: Calculate the distance-slow time two-dimensional data S RT The sum of signal amplitudes E(m) under different discrete distance gate indices m:

[0019] ;

[0020] Where |∙| represents taking the absolute value, and m=1, 2,…, M are discrete distance gate indices;

[0021] Step 2.2: Calculate the discrete distance gate index set M containing the peak value of the signal amplitude accumulation sum E(m). p :

[0022] ;

[0023] Among them, M p Let f(m) represent the set of all discrete distance gate indices that satisfy E(m) as peak value when m∈{1, 2, …, M}, where max(∙) represents the maximum value and min(∙) represents the minimum value.

[0024] If M p If the set is empty, proceed to step 2.4; otherwise, proceed to step 2.3.

[0025] Step 2.3: In the peak distance gate index set M p Find the gate index m, which represents the distance between the target location and the reference location. tar and m ref :

[0026] ;

[0027] ;

[0028] Where, m p For the peak distance gate index set M p The element in d tar d ref Let M be the actual distance between the target location, the reference location, and the radar, and d0 be the actual distance corresponding to a range gate. The above formula represents the calculation of M... p Estimated distance gate d between the center and the target tar The distance gate index with the smallest difference / d0 is used as the distance gate index m at the target location. tar ; Calculate M p The estimated distance gate d between the center and the reference point ref The distance gate index with the smallest difference / d0 is used as the reference distance gate index m. ref ;

[0029] After completing this step, proceed to step 2.5;

[0030] Step 2.4: Directly calculate the gate index m, the distance between the target and the reference point. tar and m ref :

[0031] ;

[0032] ;

[0033] Where ceil(∙) is the floor function, d tar d ref d0 represents the actual distance between the target location, the reference location, and the radar, where d0 is the actual distance corresponding to a range gate.

[0034] Step 2.5: Acquire signal data x at the target location and the reference location. tar (n) and x ref (n):

[0035] ;

[0036] ;

[0037] Where n=1, 2, …, N are discrete slow time indices;

[0038] Step 2.6: Obtain the phase p between the target location and the reference location. tar (n) and p ref (n):

[0039] ;

[0040] ;

[0041] Among them, angle(∙) represents taking the phase of a complex signal, that is, angle(x)=arctan(imag(x) / real(x)), arctan(∙) is the arctangent function, imag(∙) is taking the imaginary part of a complex signal, real(∙) is taking the real part of a complex signal, and n = 1, 2, …, N is the discrete slow-time index.

[0042] Furthermore, in step 3, the following method is used to calculate the signal data x tar (n) of the time-frequency distribution γ(t, f), extract its Doppler characteristics, and record the set T of the existence time of the Doppler characteristics, which specifically includes the following steps:

[0043] Step 3.1: Initialize the parameters for calculating the time-frequency distribution γ(t, f) of the signal data x tar (n) of the target, extract its Doppler characteristics, and record the set T of the existence time of the Doppler characteristics. The specific initialization includes the following parameters:

[0044] The window length T of the short-time Fourier transform window function is initialized to: T W is initialized as: T W = the positive even number of ceil(c1∙f PRF ), where c1 is the window length coefficient of the short-time Fourier transform window function and is a positive real number satisfying 1 ≤ c1 ≤ 5;

[0045] The step ∆T of the short-time Fourier transform window function is initialized to: ∆T = ceil(c2∙T W ), where c2 is the step coefficient of the short-time Fourier transform window function and is a positive real number satisfying 0 < c2 ≤ 0.5;

[0046] The time length T0 in the short-time Fourier transform is initialized to: T0 = ceil((N - T W ) / ∆T) as a positive integer;

[0047] The Fourier transform length F0 in the short-time Fourier transform is initialized to: F0 = T W as a positive integer;

[0048] The current discrete time index t of the time-frequency distribution for extracting the Doppler characteristics is initialized to: t cur is initialized as: t cur = 1;

[0049] The time sliding window W of the time-frequency distribution for extracting the Doppler characteristics is initialized to: TL is initialized as: as a positive integer, where c3 is the time sliding window coefficient of the time-frequency distribution for extracting the Doppler characteristics and is a positive integer satisfying 2 ≤ c3 ≤ 8;

[0050] Time step ∆W of the time-frequency distribution for extracting Doppler features T Initialize as: ∆W T = ceil(c4 ∙ W TL ) as a positive integer, where c4 is the time step coefficient of the time-frequency distribution for extracting Doppler features, and is a positive real number satisfying 0 < c4 ≤ 0.5;

[0051] Frequency window length W of the time-frequency distribution for extracting Doppler features FL Initialize: W FL = ceil(c5 ∙ F0) as a positive integer, where c5 is the frequency window length coefficient of the time-frequency distribution for extracting Doppler features, and is a positive real number satisfying 0 < c5 < 0.5;

[0052] Doppler feature threshold ρ D Initialize as: 0 < ρ D < 1 as a positive real number;

[0053] The set T of the existence time of Doppler features is initialized as: T is an empty set;

[0054] Step 3.2, Calculate the short-time Fourier transform γ(t, f) of the signal data x tar (n):

[0055] ;

[0056] ;

[0057] Where is the imaginary unit, w(n - t) is the window function, here a rectangular window is used, n = 1, 2, …, N is the discrete slow-time dimension index, t = 1, 2, …, T0 is the discrete time index of the obtained time-frequency distribution, and f = 1, 2, …, F0 is the discrete frequency index of the time-frequency distribution;

[0058] Step 3.3, Calculate the amplitude average value A(f) of the time-frequency distribution γ(t, f) at different discrete frequency indices f within the time period {t cur , t cur +1, …, t cur +W TL}:

[0059] ;

[0060] Where represents calculating for t ∈ {t cur , t cur +1, …, t cur +W TLThe mean amplitude of the time-frequency distribution γ(t, f) within the range of different discrete frequency indices f, where f = 1, 2, ..., F0 is the discrete frequency index of the time-frequency distribution;

[0061] Step 3.4: Calculate the frequency index f corresponding to the maximum value of A(f). max :

[0062] ;

[0063] in, This means finding the frequency index f corresponding to the maximum value in {A(1), A(2), …, A(F0)}. max ;

[0064] Step 3.5: Calculate the frequency index f max The corresponding time-frequency distribution energy feature ε of the detection region max and background region time-frequency distribution energy characteristics ε else :

[0065] ;

[0066] ;

[0067] Where sum(∙) is the summation function, F and F max These are the total set of frequency indices and f, respectively. max The corresponding detection region frequency index set, F\F max This means taking the elements from set F and set F max Different frequency indices, i.e., taking F max The complement of the total frequency index set F, the two sets F and F' mentioned above. max for:

[0068] ;

[0069] ;

[0070] If f max -W FL <1, then F max for:

[0071] ;

[0072] If f max +W FL >F0, then F max for:

[0073] ;

[0074] Step 3.6: Calculate the frequency index fmax The corresponding Doppler feature prominence ρ:

[0075] ;

[0076] Step 3.7: Determine the current discrete-time index t cur Does the time-frequency distribution below exhibit Doppler characteristics?

[0077] Judge the following conditions:

[0078] ;

[0079] If the above conditions are met, it means that the current time point does not have prominent Doppler characteristics, so slow time indexing is not recorded, and proceed to step 3.8;

[0080] Conversely, this indicates that the current discrete-time index possesses Doppler characteristics. Calculate the current time-frequency distribution γ(t, f) discrete-time index t. cur Radar slow time index n D :

[0081] ;

[0082] And record the current slow time index n D :

[0083] ;

[0084] The above formula represents the slow time index n D If a feature already exists in the set T of Doppler feature existence times, then set T is not updated; otherwise, n is updated. D Add to set T;

[0085] Step 3.8: Determine whether all discrete-time indices of the time-frequency distribution γ(t, f) have been traversed.

[0086] Judge the following conditions:

[0087] ;

[0088] If the above conditions are met, then let t cur =t cur +∆W T If yes, proceed to step 3.3; otherwise, proceed to step 4.

[0089] Furthermore, in step 4, the phase p at the reference point is used in the following way. ref (n) Phase p at the target tar (n) is compensated to obtain the target deformation d(n), which specifically includes the following steps:

[0090] Step 4.1. Compensate the phase p ref (n) at the target location for the phase p tar (n) at the reference location, and initialize the parameters of the target deformation d(n), specifically including the initialization of the following parameters:

[0091] The current slow-time index n of the target deformation cur Initialize it as: n cur = 1;

[0092] Initialize the detrending window length L of the target deformation as: L = ceil(c6⋅f PRF ) as a positive integer, where c6 is the detrending window length coefficient and is a positive real number satisfying 1 ≤ c6 ≤ 5;

[0093] Initialize the detrending window step ∆S of the target deformation as: ∆S = ceil(c7⋅L) as a positive integer, where c7 is the detrending window step coefficient and is a positive real number satisfying 0 < c7 < 0.5;

[0094] Initialize the target deformation d(n) after detrending as: d(n) = 0, n = 1, 2, …, N;

[0095] Initialize the radar signal wavelength λ as: λ = c / f c , where c = 3×10 8 is the speed of light in nature, with the unit m / s, and f c is the carrier frequency of the device used, with the unit Hz;

[0096] Initialize the angle ψ between the line connecting the target and the radar and the vertical direction as: a positive real number with 0° < ψ < 90°, and the specific angle is set according to the actual perception scenario;

[0097] Step 4.2. Calculate the rough target deformation d rough (n):

[0098] ;

[0099] where p tar (n) and p ref (n) are the target location phase and reference location phase calculated in Step 2.6 respectively, n = 1, 2, …, N is the discrete slow-time index, and the unit of the rough target deformation d rough (n) to be detected is mm;

[0100] Step 4.3. Calculate the trend term d cur (n) of d cur (n) within {n cur , n rough + 1, …, n trend + L}:

[0101] Step 4.3.1, calculate {n} cur , n cur +1, …, n cur +L}inner d rough The target deformation mean of (n) and time mean :

[0102] ;

[0103] ;

[0104] Step 4.3.2: Calculate the fitting trend term parameters a and b using the least squares method:

[0105] ;

[0106] ;

[0107] Step 4.3.3: Calculate the trend term d trend (n):

[0108] ;

[0109] Where, n∈{n cur , n cur +1, …, n cur +L} represents the discrete slow-time index;

[0110] Step 4.4, calculate {n} cur , n cur +1, …, n cur The target deformation data d(n) after the detrending term within +L}:

[0111] ;

[0112] Among them, w S (x) is the smoothing window function, defined as:

[0113] ;

[0114] Step 4.5: Determine if the detrending term's sliding window has reached its endpoint.

[0115] Judge the following conditions:

[0116] ;

[0117] If the above conditions are met, then let n cur = n curIf the value is +∆S, proceed to step 4.3; otherwise, proceed to step 5.

[0118] Furthermore, in step 5, the following method is used to detect the target deformation d(n) within the Doppler feature existence time set T, thereby obtaining the target deformation displacement set D and the occurrence time index set T. D Specifically, it includes the following steps:

[0119] Step 5.1: Within the time set T of Doppler feature existence, perform deformation detection on the target deformation d(n) to obtain the target deformation displacement set D and the occurrence time index set T. D The parameters are initialized, specifically including the initialization of the following parameters:

[0120] The index i of the time set T in which the Doppler features exist is initialized to: i=1;

[0121] Window length L when relocating the target to the minimum deformation min Initialized as: L min =c8⋅f PRF The positive even number, where c8 is the window length coefficient when the deformation of the repositioning target is minimized, and is a positive integer that satisfies 2≤c8≤8;

[0122] Window length L when finding the maximum value of the target deformation max Initialized as: L max =c9⋅f PRF The positive even number, where c9 is the window length coefficient when searching for the maximum value of the target deformation, and is a positive integer that satisfies 2≤c9≤8;

[0123] Deformation amplitude characteristic threshold ρ A Initialize to: 0 < ρ A Positive real numbers less than 10, in mm;

[0124] Deformation time width feature threshold ρ T Initialize as: 2≤ρ T Positive real numbers ≤ 6, in seconds;

[0125] The target deformation displacement set D is initialized as follows: D is an empty set;

[0126] Target Deformation Occurrence Time Index Set T D Initialized to: T D It is an empty set;

[0127] Step 5.2: Extract the i-th time T from the set of times T in which Doppler features exist. i In T i Centered on the window, with a length of L minIn the target deformation, the discrete slow-time index n corresponding to the minimum value of the repositioning target deformation. min :

[0128] ;

[0129] The above formula represents the expression from max(1, T) i -L min / 2)≤n≤min(N, T i +L min Find the discrete slow time index corresponding to the minimum value of d(n) within the range of / 2);

[0130] Step 5.3, in the case of n min The center window has a length of L max In the target deformation, find the discrete slow-time index n corresponding to the maximum value of the target deformation. max :

[0131] ;

[0132] The above formula represents the expression from max(1, n) min -L max / 2)≤n≤min(N, n min +L max Find the discrete slow time index corresponding to the maximum value of d(n) within the range of / 2);

[0133] Step 5.4, determine n min The target deformation at the location is determined to possess effective deformation characteristics, and the displacement and slow time index corresponding to the minimum effective deformation are recorded in the target deformation displacement set D and the occurrence time index set T, respectively. D :

[0134] Step 5.4.1: Calculate n min The deformation amplitude characteristic evaluation value η of the target deformation A :

[0135] ;

[0136] Step 5.4.2, Calculate n min The deformation time width characteristic evaluation value η of the target deformation T :

[0137] ;

[0138] Step 5.4.3, determine n min Is the target deformation a valid deformation?

[0139] Judge the following conditions:

[0140]

[0141] If the above conditions are met, proceed to step 5.4.4; otherwise, proceed to step 5.5.

[0142] Step 5.4.4: Record the target deformation displacement set D and the occurrence time index set T. D

[0143]

[0144]

[0145] The above formula represents the expression for slow time index n min It already exists in the time index set T D During this period, the target deformation displacement set D and the occurrence time index set T are not updated. D Conversely, when it does not exist in the time index set T D When the minimum effective deformation is reached, the displacement d(n) corresponding to the minimum effective deformation will be calculated. min Add the target deformation displacement set D, and set the slow time index n min Add to the occurrence time index set T D middle;

[0146] Step 5.5: Determine if the traversal of the Doppler feature existence time set T has ended.

[0147] Judge the following conditions:

[0148]

[0149] Where I is the size of the time set T in which the Doppler features exist;

[0150] If the above conditions are met, let i = i + 1 and proceed to step 5.2; otherwise, proceed to step 5.6.

[0151] Step 5.6: Output the target deformation displacement set D and the occurrence time index set T. D .

[0152] Beneficial effects: Compared with the prior art, the technical solution of the present invention has the following beneficial technical effects:

[0153] (1) Unlike fixed sensing platform sensing schemes, this invention allows for more flexible deployment of sensing devices. This avoids the following problems: fixed sensing devices, such as ground-based radar, require the radar equipment to be deployed in advance before sensing, which requires finding a suitable deployment location in advance, debugging the radar equipment in advance, and occupying a certain space and requiring regular maintenance after the radar can work stably. Flexible deployment of sensing devices, such as deploying sensing devices on communication base stations or UAVs, can make related sensing applications more extensive. In order to solve the phase problem caused by the shaking platform, step two proposes the idea of ​​using the phase data of the reference target for compensation, which can effectively obtain the deformation signal contaminated by shaking;

[0154] (2) This invention uses a stationary reference target to compensate for the influence of platform sway on the phase of the received signal, which can better restore deformation with millimeter-level accuracy. Furthermore, due to the influence of the external environment, the platform sway is large, making it impossible to completely remove the influence of platform sway. In step four, this invention adopts a segmented detrending term method to suppress the residual platform sway component and better restore the deformation characteristics of the target to be detected. Furthermore, this method of using reference target data to compensate for platform sway can also be applied to target monitoring of UAV-borne radar and breathing and heartbeat monitoring under sway conditions;

[0155] (3) The present invention utilizes Doppler features related to the target to be detected to guide detection. When Doppler features appear, the related target to be detected is likely to have undergone significant deformation. In step three, a short-time Fourier transform is performed on the complex signal of the signal to be detected to obtain Doppler features, and in step five, these features are used to guide the detection of deformation. This can reduce false alarms while ensuring the deformation detection rate of the target to be detected. Attached Figure Description

[0156] Figure 1 This is a schematic diagram of the target deformation detection method based on Doppler features according to the present invention;

[0157] Figure 2 This is a schematic diagram illustrating a scenario for bridge deformation detection based on radar signals, as an example.

[0158] Figure 3 A schematic diagram of the amplitude accumulation and E(m) under partial distance units;

[0159] Figure 4 The phase time-domain waveforms at the target location and the reference location are shown.

[0160] Figure 5 A schematic diagram of the Doppler characteristics of a heavy vehicle passing by;

[0161] Figure 6 A schematic diagram illustrating the effect of platform sway compensation using the phase at a reference point;

[0162] Figure 7 This is a schematic diagram illustrating the effect of using Doppler features for bridge deformation detection. Detailed Implementation

[0163] To better understand the purpose, structure, and function of this invention, the invention will be further described below with reference to the accompanying drawings.

[0164] like Figure 1 As shown, this invention proposes a target deformation detection method based on Doppler feature guidance, comprising the following steps:

[0165] Step 1: Acquire radar echo signal data and calculate range-slow time two-dimensional data S. RT (m, n);

[0166] Step 2: Acquire signal data x at the target location tar (n) and the signal data x at the reference point ref (n), and calculate the phase p at the target location. tar (n) and the phase p at the reference point ref (n);

[0167] Step 3: Calculate the signal data x at the target location. tar The time-frequency distribution γ(t, f) of (n) is extracted, its Doppler features are extracted, and the time set T in which the Doppler features exist is recorded;

[0168] Step 4: Utilize the phase p at the reference point ref (n) Phase p at the target tar (n) is compensated to obtain the target deformation d(n);

[0169] Step 5: Within the time set T of Doppler feature existence, perform deformation detection on the target deformation d(n) to obtain the target deformation displacement set D and the occurrence time index set T. D .

[0170] Example:

[0171] The embodiment test was conducted using measured data. The radar signal parameters are: carrier frequency f c =4.9 GHz, wavelength λ≈0.06 m, pulse repetition frequency f PRF =400 Hz, the angle ψ between the integrated sensing base station antenna and the observed target in the vertical direction is approximately 78°, and the schematic diagram of the sensing scene is as follows. Figure 2 As shown;

[0172] Based on step 1, acquire the fast-time and slow-time two-dimensional data S(k, n) of the radar echo signal, and obtain the range-slow-time two-dimensional data S by calculating the FFT on the fast-time dimension. RT(m, n). Initialize the fast-time sampling number K=1944, the number of range gates M=1944, and the number of slow-time sampling points N=30×60×400=720000, corresponding to the number of radar sampling points in 30 minutes;

[0173] Based on step 2, obtain the signal data x at the target location and the reference location. tar (n) and x ref (n), and calculate their respective phases p tar (n) and p ref (n). Initialize the actual distance d0≈5.14 m corresponding to a range cell, and the approximate actual distance d between the target and the radar. tar =280 m, the approximate actual distance d between the reference point and the radar. ref =320 m. Calculate the set of cell indices M containing the peak value of the cumulative amplitude sum E(m). p =[43, 47, 50, 53, 57, 59, 65, 67, 70, 72, 74, 77, 80] (Only some peak values ​​between 40 and 80 are listed here), such as Figure 3 As shown, the corresponding m can be estimated. tar =56,m ref =64, further extract the corresponding distance gate data, x tar (n) represents the signal data at the target location, x ref (n) represents the signal data at the reference point, and the phase p at the target point is calculated. tar (n) is in phase p with the reference point ref (n), such as Figure 4 As shown;

[0174] Based on step 3, calculate the signal data x at the target location. tar The time-frequency distribution γ(t, f) of (n) is used to extract Doppler features, and the time set T in which the Doppler features exist is recorded. The window length T of the short-time Fourier transform window function is initialized. W =400, Short-Time Fourier Transform window function step ∆T=100, Time length in Short-Time Fourier Transform T0=7196, Fourier transform length in Short-Time Fourier Transform F0=400, Time index t for extracting Doppler features from the time-frequency distribution. cur =1, time-frequency distribution time sliding window W for extracting Doppler features TL =16, time step ∆W for extracting Doppler features and time-frequency distribution T =8, the frequency window length W for extracting Doppler features in the time-frequency distribution. FL =50, Doppler feature threshold ρ D=0.5, the Doppler feature exists in the time set T which is empty. By calculating the short-time Fourier transform, it can be observed that the time points where obvious deformation occurs at the target are Doppler features in the time-frequency distribution under the corresponding time index, such as... Figure 5 As shown, the Doppler feature existence time set T can be calculated based on this time-frequency distribution Doppler feature;

[0175] Based on step 4, using the phase data p at the reference point ref (n) Phase p at the target tar Compensation is performed on n to obtain the target deformation d(n). The initial slow time index of the target deformation start is initialized n. cur =1, the target deformation detrending term sliding window length L=4000, the target deformation detrending term sliding window step ∆S=800, the target deformation after detrending term d(n)=0, n=1, 2,…, N, the radar signal wavelength λ, the angle ψ between the line connecting the observed target and the radar and the vertical direction have been initialized in the embodiment description. Based on this, the target deformation after detrending can be obtained, such as... Figure 6 As shown, the compensated target shape changes are closer to the true value data of the sensor.

[0176] Based on step 5, within the time period near the Doppler feature existence time set T, deformation detection is performed on the target deformation d(n) to obtain the target deformation displacement set D and the occurrence time index set T. D Initialize the index i=1 of the time set T where Doppler features exist, and set the window length L for relocating the target deformation to its minimum. min =1600, the window length L when searching for the maximum target deformation. max =1200, the characteristic threshold ρ of the downward pressure amplitude of the target deformation. A =1, the characteristic threshold ρ of the compression time width of the target deformation. T =2, the target deformation displacement set D is initialized to an empty set, and the target deformation occurrence time index set T is initialized to an empty set. D Initialize to an empty set. Figure 7 This demonstrates a comparison between direct pressure detection and pressure detection combined with Doppler guidance, such as... Figure 7 As shown in (a), directly using a threshold to determine whether undervoltage exists can introduce false undervoltage judgments. For example, in the 400-500 s time period shown in the figure, multiple sensors show no obvious undervoltage, but after processing, the phase shows undervoltage, which is misjudged as true undervoltage. Figure 7 As shown in (b), the pressure detection combined with Doppler features can eliminate false alarms of 400-500 s and maintain a high detection rate.

[0177] It is understood that the present invention has been described through some embodiments, and those skilled in the art will recognize that various changes or equivalent substitutions can be made to these features and embodiments without departing from the spirit and scope of the invention. Furthermore, under the teachings of the present invention, these features and embodiments can be modified to adapt to specific situations and materials without departing from the spirit and scope of the invention. Therefore, the present invention is not limited to the specific embodiments disclosed herein, and all embodiments falling within the scope of the claims of this application are within the protection scope of the present invention.

Claims

1. A target deformation detection method based on Doppler feature guidance, characterized in that, The method includes the following steps: Step 1: Acquire radar echo signal data and calculate range-slow time two-dimensional data S. RT (m, n); Step 2: Acquire signal data x at the target location tar (n) and the reference signal data x ref (n), and calculate the phase p at the target location. tar (n) and the phase p at the reference point ref (n); Step 3: Calculate the signal data x at the target location. tar The time-frequency distribution γ(t, f) of (n) is extracted, its Doppler features are extracted, and the time set T in which the Doppler features exist is recorded; Step 4: Utilize the phase p at the reference point ref (n) Phase p at the target tar (n) is compensated to obtain the target deformation d(n); Step 5: Within the time set T of Doppler feature existence, perform deformation detection on the target deformation d(n) to obtain the target deformation displacement set D and the occurrence time index set T. D .

2. The target deformation detection method based on Doppler feature guidance according to claim 1, characterized in that, In step 1, radar echo signal data is acquired using the following method, and range-slow time two-dimensional data S is calculated. RT (m,n), specifically including the following steps: Step 1.1: Acquire the radar echo signal and organize it according to the fast-time-slow-time format to obtain the fast-time-slow-time two-dimensional data S(k, n), where S(k, n) is the radar echo signal value at the k-th fast time point and the n-th slow time point, k=1, 2, …, K is the discrete fast-time index, K is the number of fast-time samples in one radar transmit-receive cycle, and it is a positive integer satisfying K≥1024, n=1, 2, …, N is the discrete slow-time index, N is the number of slow times acquired by the radar, i.e., the number of radar transmit-receive cycles, and satisfies N≥60∙f PRF positive integers, f PRF Let f be the pulse repetition frequency, which satisfies f PRF Positive integers ≥400 Hz; Step 1.2: Calculate the distance-slow time two-dimensional data S RT (m, n): ； Where m = 1, 2, ..., M are discrete distance gate indices, M is the number of distance cells, and M is a positive integer satisfying M ≥ 1024; n = 1, 2, ..., N are discrete slow-time indices. The above formula represents the distance-slow-time two-dimensional data S obtained by performing an FFT on the fast-time dimension of the fast-time two-dimensional data S(k,n). RT (m,n).

3. The target deformation detection method based on Doppler feature guidance according to claim 2, characterized in that, In step 2, the signal data x at the target location is obtained using the following method. tar (n) and the reference signal data x ref (n), and calculate the phase p at the target location. tar (n) and the phase p at the reference point ref (n), specifically including the following steps: Step 2.1: Calculate the distance-slow time two-dimensional data S RT The sum of signal amplitudes E(m) under different discrete distance gate indices m: ； where, |∙| represents taking the absolute value, and m = 1, 2, …, M is the discrete distance gate index; Step 2.2: Calculate the discrete distance gate index set M containing the peak value of the signal amplitude accumulation sum E(m). p : ； Among them, M p Let f(m) represent the set of all discrete distance gate indices that satisfy E(m) as peak value when m∈{1, 2, …, M}, where max(∙) represents the maximum value and min(∙) represents the minimum value. If M p If the set is empty, proceed to step 2.4; otherwise, proceed to step 2.

3. Step 2.3: In the peak distance gate index set M p Find the gate index m, which represents the distance between the target location and the reference location. tar and m ref : ； ； Where, m p For the peak distance gate index set M p The element in d tar d ref Let M be the actual distance between the target location, the reference location, and the radar, and d0 be the actual distance corresponding to a range gate. The above formula represents the calculation of M... p Estimated distance gate d between the center and the target tar The distance gate index with the smallest difference at d0 is used as the distance gate index m at the target location. tar ; Calculate M p The estimated distance gate d between the center and the reference point ref The distance gate index with the smallest difference at d0 is used as the reference distance gate index m. ref ; After completing this step, proceed to step 2.5; Step 2.4: Directly calculate the gate index m, the distance between the target and the reference point. tar and m ref : ； ； Where ceil(∙) is the floor function, d tar d ref d0 represents the actual distance between the target location, the reference location, and the radar, where d0 is the actual distance corresponding to a range gate. Step 2.5: Acquire signal data x at the target location and the reference location. tar (n) and x ref (n): ； ； where, n = 1, 2, …, N is the discrete slow time index; Step 2.6: Obtain the phase p between the target location and the reference location. tar (n) and p ref (n): ； ； where, angle(∙) represents taking the phase of the complex signal, that is, angle(x)=arctan(imag(x) / real(x)), arctan(∙) is the arctangent function, imag(∙) is taking the imaginary part of the complex signal, real(∙) is taking the real part of the complex signal, and n = 1, 2, …, N is the discrete slow time index.

4. The target deformation detection method based on Doppler feature guidance according to claim 3, characterized in that, In step 3, the signal data x at the target location is calculated using the following method. tar The time-frequency distribution γ(t, f) of (n) is used to extract its Doppler features. And record the set T of Doppler feature existence time, which specifically includes the following steps: Step 3.1: Calculate the signal data x at the target location. tar The time-frequency distribution γ(t, f) of (n) is analyzed, its Doppler features are extracted, and the parameters of the time set T in which the Doppler features exist are initialized. Specifically, the initialization of the following parameters is performed: Short-time Fourier transform window function window length T W Initialized to: T W =ceil(c1∙f PRF ) is a positive even number, where c1 is the window length coefficient of the short-time Fourier transform window function, and is a positive real number that satisfies 1≤c1≤5; The step ∆T of the short-time Fourier transform window function is initialized as: ∆T = ceil(c2 ∙ T W ) as a positive integer, where c2 is the step coefficient of the short-time Fourier transform window function and is a positive real number satisfying 0 < c2 ≤ 0.5; In the short-time Fourier transform, the time length T0 is initialized as: T0 = ceil((NT W A positive integer of () / ∆T); In the short-time Fourier transform, the Fourier transform length F0 is initialized as: F0 = T W Positive integers; Extracting the time-frequency distribution of Doppler features: Current discrete-time index t cur Initialized to: t cur =1; Time-frequency distribution time sliding window W for extracting Doppler features TL Initialize to: A positive integer, where c3 is the time-sliding window coefficient of the time-frequency distribution for extracting Doppler features, and is a positive integer satisfying 2≤c3≤8; Time step ∆W of the time-frequency distribution for extracting Doppler features T Initialize it as: ∆W T = ceil(c4 ∙ W TL ) as a positive integer, where c4 is the time step coefficient of the time-frequency distribution for extracting Doppler features, and is a positive real number satisfying 0 < c4 ≤ 0.5; Frequency window length W of the time-frequency distribution for extracting Doppler features FL Initialization: W FL = positive integer of ceil(c5 ∙ F0), where c5 is the frequency window length coefficient of the time-frequency distribution for extracting Doppler features, and is a positive real number satisfying 0 < c5 < 0.5; Doppler feature threshold ρ D Initialize to: 0 < ρ D Positive real numbers less than 1; The set T of Doppler feature existence time is initialized as: T is an empty set; Step 3.2: Calculate the signal data x at the target location. tar The short-time Fourier transform γ(t, f) of (n): ； ； in, The imaginary unit is w(nt), which is the window function. A rectangular window is used here. n=1, 2, …, N are the discrete slow time dimension indices. t=1, 2, …, T0 are the discrete time indices of the obtained time-frequency distribution. f=1, 2, …, F0 are the discrete frequency indices of the time-frequency distribution. Step 3.3, Calculate {t} cur , t cur +1, …, t cur +W TL The average amplitude A(f) of the time-frequency distribution γ(t, f) over a time period at different discrete frequency indices f: ； in, This indicates the calculation of t∈{t cur , t cur +1, …, t cur +W TL The mean amplitude of the time-frequency distribution γ(t, f) within the range of different discrete frequency indices f, where f = 1, 2, ..., F0 is the discrete frequency index of the time-frequency distribution; Step 3.4: Calculate the frequency index f corresponding to the maximum value of A(f). max : ； in, This means finding the frequency index f corresponding to the maximum value in {A(1), A(2), …, A(F0)}. max ; Step 3.5: Calculate the frequency index f max The corresponding time-frequency distribution energy feature ε of the detection region max and background region time-frequency distribution energy characteristics ε else : ； ； Where sum(∙) is the summation function, F and F max These are the total set of frequency indices and f, respectively. max The corresponding detection region frequency index set, F\F max This means taking the elements from set F and set F max Different frequency indices, i.e., taking F max The complement of the total frequency index set F, the two sets F and F' mentioned above. max for: ； ； If f max -W FL <1, then F max for: ； If f max +W FL >F0, then F max for: ； Step 3.6: Calculate the frequency index f max The corresponding Doppler feature prominence ρ: ； Step 3.7: Determine the current discrete-time index t cur Does the time-frequency distribution below exhibit Doppler characteristics? Judge the following conditions: ； If the above conditions are satisfied, it means that there is no prominent Doppler feature at the current time point, and the slow time index is not recorded, and proceed to step 3.8; Conversely, this indicates that the current discrete-time index possesses Doppler characteristics. Calculate the current time-frequency distribution γ(t, f) discrete-time index t. cur Radar slow time index n D : ； And record the current slow time index n D : ； The above formula represents the slow time index n D If a feature already exists in the set T of Doppler feature existence times, then set T is not updated; otherwise, n is updated. D Add to set T; Step 3.8, judge whether all discrete time indexes of the time-frequency distribution γ(t, f) have been traversed: Judge the following conditions: ； If the above conditions are met, then let t cur =t cur +∆W T If yes, proceed to step 3.3; otherwise, proceed to step 4.

5. The target deformation detection method based on Doppler feature guidance according to claim 4, characterized in that, In step 4, the following method is used to utilize the phase p at the reference point. ref (n) Phase p at the target tar (n) is compensated to obtain the target deformation d(n), which specifically includes the following steps: Step 4.1: Using the phase p at the reference point ref (n) Phase p at the target tar (n) is compensated, and the parameters of the target deformation d(n) are initialized, specifically including the initialization of the following parameters: Target deformation current slow time index n cur Initialized to: n cur =1; The target deformation detrending term sliding window length L is initialized as: L = ceil(c6⋅f PRF () is a positive integer, where c6 is the sliding window length coefficient of the detrending term, and is a positive real number that satisfies 1≤c6≤5; The sliding window step ∆S of the target deformation detrending term is initialized as: ∆S is a positive integer of ceil(c7⋅L), where c7 is the detrending term sliding window step coefficient, which is a positive real number satisfying 0 < c7 < 0.5; The target deformation d(n) after detrending is initialized as: d(n)=0, n = 1, 2, …, N; The radar signal wavelength λ is initialized as follows: λ = c / f c Where, c = 3 × 10 8 The speed of light in nature, measured in m / s, f c The carrier frequency of the equipment used, in Hz; The angle ψ between the line connecting the target and the radar and the vertical direction is initialized as: a positive real number with 0° < ψ < 90°, and the specific angle is set according to the actual perception scenario; Step 4.2: Calculate the rough target deformation d rough (n): ； Where, p tar (n) and p ref (n) represents the target phase and reference phase calculated in step 2.6, respectively, where n = 1, 2, ..., N is the discrete slow-time index, and the approximate deformation d of the target to be detected. rough The unit of (n) is mm; Step 4.3, calculate {n} cur , n cur +1, …, n cur +L}inner d rough The trend term d of (n) trend (n): Step 4.3.1, calculate {n} cur , n cur +1, …, n cur +L}inner d rough The target deformation mean of (n) and time mean : ； ； Step 4.3.2, calculate the fitting trend term parameters a and b using the least squares method: ； ； Step 4.3.3: Calculate the trend term d trend (n): ； Where, n∈{n cur , n cur +1, …, n cur +L} represents the discrete slow-time index; Step 4.4, calculate {n} cur , n cur +1, …, n cur The target deformation data d(n) after the detrending term within +L}: ； Among them, w S (x) is the smoothing window function, defined as: ； Step 4.5, judge whether the sliding window of the detrending term has reached the end: Judge the following conditions: ； If the above conditions are met, then let n cur = n cur If the value is +∆S, proceed to step 4.3; otherwise, proceed to step 5.

6. The target deformation detection method based on Doppler feature guidance according to claim 5, characterized in that, In step 5, the following method is used to detect the target deformation d(n) within the Doppler feature existence time set T, thereby obtaining the target deformation displacement set D and the occurrence time index set T. D Specifically, it includes the following steps: Step 5.1: Within the time set T of Doppler feature existence, perform deformation detection on the target deformation d(n) to obtain the target deformation displacement set D and the occurrence time index set T. D The parameters are initialized, specifically including the initialization of the following parameters: The index i of the set T of Doppler feature existence time is initialized as: i = 1; Window length L when relocating the target to the minimum deformation min Initialized as: L min =c8⋅f PRF The positive even number, where c8 is the window length coefficient when the deformation of the repositioning target is minimized, and is a positive integer that satisfies 2≤c8≤8; Window length L when finding the maximum value of the target deformation max Initialized as: L max =c9⋅f PRF The positive even number, where c9 is the window length coefficient when searching for the maximum value of the target deformation, and is a positive integer that satisfies 2≤c9≤8; Deformation amplitude characteristic threshold ρ A Initialize to: 0 < ρ A Positive real numbers less than 10, in mm; Deformation time width feature threshold ρ T Initialize as: 2≤ρ T Positive real numbers ≤ 6, in seconds; The set D of target deformation displacements is initialized as: D is an empty set; Target Deformation Occurrence Time Index Set T D Initialized to: T D It is an empty set; Step 5.2: Extract the i-th time T from the set of times T in which Doppler features exist. i In T i Centered on the window, with a length of L min In the target deformation, the discrete slow-time index n corresponding to the minimum value of the repositioning target deformation. min : ； The above formula represents the expression from max(1, T) i -L min / 2)≤n≤min(N, T i +L min Find the discrete slow time index corresponding to the minimum value of d(n) within the range of / 2); Step 5.3, in the case of n min The center window has a length of L max In the target deformation, find the discrete slow-time index n corresponding to the maximum value of the target deformation. max : ； The above formula represents the expression from max(1, n) min -L max / 2)≤n≤min(N, n min +L max Find the discrete slow time index corresponding to the maximum value of d(n) within the range of / 2); Step 5.4, determine n min The target deformation at the location possesses effective deformation characteristics, and the displacement and slow time index corresponding to the minimum effective deformation are recorded in the target deformation displacement set D and the occurrence time index set T, respectively. D : Step 5.4.1: Calculate n min The deformation amplitude characteristic evaluation value η of the target deformation A : ； Step 5.4.2, Calculate n min The deformation time width characteristic evaluation value η of the target deformation T : ； Step 5.4.3, determine n min Is the target deformation a valid deformation? Judge the following conditions: ； If the above conditions are satisfied, proceed to step 5.4.4; otherwise, proceed to step 5.5; Step 5.4.4: Record the target deformation displacement set D and the occurrence time index set T. D ； ； The above formula represents the expression when the slow time index n min Already exists in the time index set T D During this period, the target deformation displacement set D and the occurrence time index set T are not updated. D Conversely, when it does not exist in the time index set T D When the minimum effective deformation is reached, the displacement d(n) corresponding to the minimum effective deformation will be calculated. min Add the target deformation displacement set D, and set the slow time index n min Add to the occurrence time index set T D middle; Step 5.5, judge whether the traversal of the set T of Doppler feature existence time is over: ； Judge the following conditions: where, I is the size of the set T of Doppler feature existence time; If the above conditions are satisfied, let i = i + 1 and proceed to step 5.2; otherwise, proceed to step 5.6; Step 5.6: Output the target deformation displacement set D and the occurrence time index set T. D .

Images