Micro-motion target motion detection method based on adaptive unscented Kalman filtering

Abstract

The invention provides a micro-motion target motion detection method based on adaptive unscented Kalman filtering, and the method comprises the steps: receiving an echo signal reflected by a target, carrying out the preprocessing of the echo signal, carrying out the calculation according to the preprocessed echo signal, and obtaining a distance angle unit where the target is located; performing difference frequency processing according to the distance angle units of the two adjacent frames, and performing signal correction through phase unwrapping to obtain initial deformation data; and inputting the initial deformation data into a self-adaptive unscented Kalman filter, outputting to obtain target deformation data, and analyzing according to the target deformation data to obtain a micro-motion target. According to the invention, high-precision detection of tiny deformation can be realized, the detection precision and reliability are greatly improved, and accurate detection data can still be obtained in various complex environments.

Description

Technical Field

[0001] The present invention relates to the field of high-precision detection technology, and in particular to a method for detecting micro-motion target motion based on adaptive unscented Kalman filtering. Background Art

[0002] The application of millimeter wave radar in the field of micro deformation detection is attracting widespread attention. At present, the following methods are mainly used for micro deformation detection:

[0003] The first type is the capacitive micro-deformation sensor, which has a simple structure, fast response speed, is applicable to a variety of materials, and has a low cost. However, it is easily affected by environmental factors, has weak anti-interference ability, and requires additional shielding and compensation measures, which causes inconvenience in use.

[0004] The second category is detection based on deep learning. This method currently needs to be combined with a multimodal data fusion algorithm. It has high accuracy, but is limited by the data set, requires a large amount of data, and is sensitive to data quality. It may perform poorly in unseen samples or under extreme conditions.

[0005] The third type is laser interferometer, which has extremely high measurement accuracy and resolution, and non-contact measurement. However, its performance in harsh environments is poor. For example, under conditions such as rain, snow, or strong wind and sand, the monitoring equipment may not work properly or the monitoring accuracy may drop significantly. Moreover, its cost is extremely high and is not suitable for large-scale industrial applications.

[0006] The fourth category is millimeter wave technology detection, and millimeter waves have strong penetration capabilities and can penetrate obstacles such as rain, fog, and dust, and can still maintain good detection performance in complex environments. However, its accuracy is slightly inferior to the third method. Millimeter wave radar systems usually consume large power, especially in application scenarios with high resolution and long-term monitoring, which may require additional power management and heat dissipation measures. The signal processing of millimeter wave radar is relatively complex, and high-performance computing equipment is required to demodulate and analyze radar echo signals.

[0007] In summary, the first type of method is limited by accuracy and environmental factors, the second type of method has shortcomings in data sets and adaptability, and the third type of method is limited by cost constraints. The fourth type of method consumes a lot of power and has limited accuracy.

[0008] Therefore, there is an urgent need for a micro-motion target motion detection method that can improve detection accuracy and is applicable to a variety of complex environments. Summary of the invention

[0009] Based on this, it is necessary to provide a micro-motion target motion detection method based on adaptive unscented Kalman filtering to address the above technical problems.

[0010] A method for detecting micro-motion target motion based on an adaptive unscented Kalman filter comprises the following steps: receiving an echo signal reflected by a target and performing preprocessing, performing calculations based on the preprocessed echo signal to obtain a distance angle unit where the target is located; performing difference frequency processing based on the distance angle units of two adjacent frames, and performing signal correction through phase unwrapping to obtain preliminary deformation data; inputting the preliminary deformation data into an adaptive unscented Kalman filter, outputting target deformation data, and obtaining a micro-motion target based on analysis of the target deformation data.

[0011] In one of the embodiments, before receiving the echo signal reflected by the target and performing preprocessing, it also includes: arranging a millimeter wave radar facing the target to be measured for transmitting signals, the millimeter wave radar adopts a two-transmit and four-receive transmission mode, and the number of sampling points in the range direction is set to 256.

[0012] In one embodiment, the receiving and preprocessing of the echo signal reflected by the target, and the calculation based on the preprocessed echo signal to obtain the distance angle unit where the target is located include:

[0013] receiving an echo signal reflected by a target and preprocessing the echo signal;

[0014] The multiple signal classification algorithm is used to calculate the target angle unit based on the preprocessed echo signal. The formula is:

[0015]

[0016] Where a(θ) is the steering vector, is the noise subspace eigenvector matrix, a T (θ) is the transpose of the steering vector matrix, is the transpose of the noise subspace eigenvector matrix;

[0017] Obtaining the corresponding target distance according to the angle unit, and generating an initial distance angle unit of the target;

[0018] The obtained echo signal is subjected to ordered statistic constant false alarm detection from the distance dimension, and unit average constant false alarm detection is performed from the angle dimension. The noise threshold of point (x, y) is:

[0019]

[0020]

[0021] T os-ca (x,y)=max(T ca (x),T os (y));

[0022] Where i and j represent half of the number of reference units for ordered statistic constant false alarm detection and unit average constant false alarm detection, respectively. f(k,m) represents the value of the received signal amplitude spectrum at (k,m). The maximum value of f(k,m) is taken as the target location. The criterion for judging whether a point (x,y) has a target is:

[0023]

[0024] In the formula, when d(x,y) is not 0, it means there is a target, and when d(x,y) is 0, it means there is no target;

[0025] By judging whether there is a target or not, the initial distance angle unit corresponding to the false alarm target is eliminated to obtain the distance angle unit where the target is located.

[0026] In one embodiment, performing difference frequency processing according to the distance angle unit of two adjacent frames and performing signal correction by phase unwrapping to obtain preliminary deformation data includes:

[0027] The difference frequency processing is performed according to the distance angle unit of two adjacent frames, and the formula is:

[0028]

[0029] In the formula, t represents the frame number of the radar. Indicates the difference frequency between two adjacent frames;

[0030] The echo signal phase is unwrapped according to the difference frequency, and the echo signal exceeding the phase main value is corrected to the echo signal within the phase main value. The formula is:

[0031]

[0032] In the formula, Represents the true value of the phase difference in the t-th frame after correction;

[0033] The moving distance of the target between two adjacent frames is calculated based on the corrected echo signal to obtain the preliminary deformation data. The formula is:

[0034]

[0035] Where Δr is the moving distance of the target between two adjacent frames, and λ is the signal wavelength.

[0036] In one embodiment, the preliminary deformation data is input into an adaptive unscented Kalman filter for filtering, and the target deformation data is outputted, and the micro-motion target is obtained according to the target deformation data analysis, including:

[0037] Construct the sigma point set, and the initialization weight formula of the adaptive Kalman filter is:

[0038]

[0039] In the formula, represents the initial weight of the state matrix, represents the u-th matrix of the state matrix, represents the initial weight of the covariance matrix, represents the u-th matrix of the covariance matrix, λ′ is the scaling parameter, n is the number of sigma point sets, α and β represent the state distribution parameters;

[0040] Among them, the initialization covariance matrix is:

[0041]

[0042] In the formula, det represents the frame interval time of the radar;

[0043] Inputting the preliminary deformation data into an initialized adaptive unscented Kalman filter, filtering through the adaptive unscented Kalman filter, and outputting target deformation data;

[0044] A micro-motion target is obtained according to the target deformation data analysis.

[0045] In one embodiment, it also includes: introducing outlier detection in the adaptive unscented Kalman filter, the formula is:

[0046]

[0047] In the formula, e t Represents the difference between the predicted result and the updated result, Represents the covariance matrix calculated based on the weights at the t-th moment;

[0048] If the calculated outlier value δ t Greater than the preset threshold δ t,L , then the covariance matrix is ​​modified, the formula is:

[0049]

[0050]

[0051]

[0052] ω=max(ω0,(δ t -b·δ t,L ) / δ t );

[0053] In the formula, Q′ trepresents the updated prediction covariance matrix, Q t represents the prediction covariance matrix at time t, K t represents the Kalman coefficient matrix, The transpose of the matrix representing the difference between the predicted result and the updated result, represents the transpose of the Kalman coefficient matrix, R′ t represents the updated observation covariance matrix, R t represents the observation covariance matrix at time t, and ω represent weight coefficients, and ω0 represent the initial values ​​of the weight coefficients, represents the difference between the weighted prediction result and the observed result, represents the weight of the sigma point set, H represents the state transfer matrix, Represents the predicted mean after combining the weights, Z t represents the sensor observation value at time t, and a and b represent the tuning factors.

[0054] Compared with the prior art, the advantages and beneficial effects of the present invention are: by receiving the echo signal reflected by the target and preprocessing it, the preprocessed echo signal is used for calculation to obtain the distance angle unit where the target is located; difference frequency processing is performed according to the distance angle units of two adjacent frames, and signal correction is performed through phase unwrapping to obtain preliminary deformation data; the preliminary deformation data is input into an adaptive unscented Kalman filter, and the target deformation data is output, and the micro-motion target is obtained according to the target deformation data analysis. Through advanced signal processing technology and the improvement of the unscented Kalman filter, high-precision detection of tiny deformations is achieved. Compared with traditional detection methods, more subtle deformation changes can be detected, which greatly improves the detection accuracy and reliability, and ensures that accurate monitoring data can still be obtained in various complex environments. BRIEF DESCRIPTION OF THE DRAWINGS

[0055] Figure 1 1 is a flow chart of a method for detecting micro-motion target motion based on adaptive unscented Kalman filtering in one embodiment;

[0056] Figure 2 It is an overall flow chart of a method for detecting micro-motion target motion based on adaptive unscented Kalman filtering in one embodiment;

[0057] Figure 3 A schematic diagram of radar installation in a usage scenario in an embodiment;

[0058] Figure 4 1 is an error diagram of experimental results of the KF algorithm, the UKF algorithm and the AUKF algorithm in one embodiment;

[0059] Figure 5 Graphs of Kalman coefficients for the KF algorithm, the UKF algorithm, and the AUKF algorithm in one embodiment. DETAILED DESCRIPTION

[0060] Before describing the specific embodiments of the present invention, the overall concept of the present invention is described as follows:

[0061] The present invention is mainly developed based on the process of micro-deformation detection. The existing technology has many shortcomings in the field of micro-deformation detection, such as large errors, high costs, poor adaptability to external environments, etc.

[0062] Therefore, the present invention proposes a method for detecting micro-motion target motion based on adaptive unscented Kalman filtering, which receives the echo signal reflected by the target and performs preprocessing, uses the preprocessed echo signal for calculation, and obtains the distance angle unit where the target is located; performs difference frequency processing according to the distance angle units of two adjacent frames, and performs signal correction through phase unwrapping to obtain preliminary deformation data; inputs the preliminary deformation data into an adaptive unscented Kalman filter, outputs the target deformation data, and obtains the micro-motion target based on the target deformation data analysis. This method not only overcomes many shortcomings of the prior art in the field of micro-deformation detection, such as large errors, high costs, and poor adaptability to the external environment, but also can achieve high resolution and non-contact measurement, thereby significantly improving the accuracy and practicality of the micro-deformation system.

[0063] After introducing the overall concept of the present invention, in order to make the purpose, technical solution and advantages of the present invention more clear, the present invention is further described in detail by specific implementation methods in combination with the accompanying drawings. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0064] In one embodiment, Figure 1 and Figure 2 As shown, a method for detecting micro-motion target motion based on adaptive unscented Kalman filtering is provided, comprising the following steps:

[0065] Step S110, receiving the echo signal reflected by the target and performing preprocessing, performing calculations based on the preprocessed echo signal, and obtaining the distance angle unit where the target is located.

[0066] Specifically, a millimeter-wave radar is arranged facing the target to be measured and transmits a signal, an echo signal reflected by the target is received, and the echo signal is preprocessed. The preprocessing can be performed by using a mixer and a low-pass filter to process the echo signal, filtering out noise factors in the echo signal, and calculating the distance angle unit of the target based on the preprocessed echo signal, thereby obtaining the position and angle information of the target.

[0067] Among them, signal preprocessing includes: passing the transmit signal and the receive signal through a mixer, and then through a low-pass filter. Each antenna signal of each frame of data finally retains only one intermediate frequency signal. The final signal can be expressed as:

[0068] s IF (t) = σexp(j2πτ(αt+f c ));

[0069] Among them, σ represents the amplitude of the received signal, τ represents the round-trip delay of the signal, α represents the modulation frequency of the FMCW signal, and f c Indicates the carrier frequency of FMCW.

[0070] The signal is f s The sampling rate is N s Afterwards it can be expressed as:

[0071]

[0072] The signal is f s The sampling rate is N s Afterwards, a discrete Fourier transform is performed on the signal, and the final signal can be expressed as:

[0073]

[0074] in,

[0075]

[0076] Its distance resolution is:

[0077]

[0078] By changing the radar transmission mode to two transmit and four receive, the received signals of different antennas can be expressed as:

[0079]

[0080] Among them, θ represents the horizontal angle between the target and the radar board plane, l represents the channel number of the antenna, and d represents the distance from the radar to the target. The radar signal can be combined into a matrix according to the distance dimension and the angle dimension.

[0081] Before step S110 , the method further includes: arranging a millimeter wave radar facing the target to be measured for transmitting signals, wherein the millimeter wave radar adopts a two-transmit and four-receive transmission mode, and the number of sampling points in the range direction is 256.

[0082] Specifically, when performing micro-motion target operation detection, Figure 3As shown, the millimeter-wave radar IWR1642 is arranged facing the target to be measured, and the frequency-modulated continuous wave signal is transmitted through the millimeter-wave radar. When transmitting the radar signal, a two-transmit and four-receive transmission mode is adopted, and the number of sampling points in the range direction is set to 256, which is convenient for subsequent echo data calculation.

[0083] Wherein, step S110 includes: receiving the echo signal reflected by the target and preprocessing the echo signal; using a multiple signal classification algorithm to calculate the target angle unit according to the preprocessed echo signal, the formula is:

[0084]

[0085] Where a(θ) is the steering vector, is the noise subspace eigenvector matrix, a T (θ) is the transpose of the steering vector matrix, is the transpose of the noise subspace feature vector matrix; the corresponding target distance is obtained according to the angle unit, and the initial distance angle unit of the target is generated; the obtained echo signal is subjected to ordered statistical constant false alarm detection from the distance dimension, and the unit average constant false alarm detection is performed from the angle dimension, then the noise threshold of the point (x, y) is:

[0086]

[0087]

[0088] T os-ca (x,y)=max(T ca (x),T os (y));

[0089] Where i and j represent half of the number of reference units for ordered statistic constant false alarm detection and unit average constant false alarm detection, respectively. f(k,m) represents the value of the signal amplitude spectrum (k,m) received by the sensor. The maximum value of f(k,m) is taken as the location of the target. The criterion for judging whether a point (x,y) has a target is:

[0090]

[0091] Wherein, when d(x, y) is not 0, it indicates that there is a target, and when d(x, y) is 0, it indicates that there is no target. By judging whether there is a target or not, the initial distance angle unit corresponding to the false alarm target is eliminated to obtain the distance angle unit where the target is located.

[0092] Specifically, the echo signal reflected by the target is received and preprocessed, and the Music algorithm is used

[0093] The Multiple Signal Classification (MSSC) algorithm calculates the target angle unit based on the preprocessed echo signal, separates the signal space and the noise space according to the covariance matrix of the echo signal, and then estimates the signal direction using the orthogonal property, achieving a resolution far exceeding that of traditional beamforming. It can distinguish very close signal sources, thereby achieving accurate identification of the target direction and obtaining the target angle unit.

[0094] The minimum optimization search formula of the above music algorithm is:

[0095]

[0096] The corresponding target distance is calculated according to the obtained angle unit, and the initial distance angle unit of the target is formed according to the angle unit of the target and the target distance.

[0097] In order to adapt to different conditions and background noise, it is also necessary to perform os-ca-cfar (cell averaging / ordered statistic-constant false alarm rate) constant false alarm detection on the obtained echo signal. Among them, os-cfar (ordered statistic-constant false alarm rate) is performed on the echo signal from the distance dimension, and ca-cfar (cell averaging-constant false alarm rate) is performed on the echo signal from the angle dimension to obtain the corresponding noise threshold. The noise threshold is used to implement constant false alarm detection of the angle and distance of the echo signal, thereby improving the detection accuracy. In actual settings, the number of protection units of the os-ca-cfar algorithm is four, and the number of training units is 6.

[0098] When judging whether there is a target, the relationship between the point to be judged and the noise threshold is used to determine whether there is a target at the point to be judged. False alarm targets are eliminated based on the judgment result, and the initial distance angle unit corresponding to the false alarm target is eliminated to obtain the distance angle unit where the real target is located, thereby improving the accuracy of target detection.

[0099] Step S120 , performing difference frequency processing according to the distance angle units of two adjacent frames, and performing signal correction by phase unwrapping to obtain preliminary deformation data.

[0100] Specifically, after obtaining all the distance angle units of the echo signal, difference frequency processing is performed according to the distance angle units of two adjacent frames. By using different frequencies for signal transmission and reception, the performance and anti-interference ability of the system are improved; and phase unwrapping is used to correct the signal that exceeds the main value, so as to obtain preliminary deformation data.

[0101] Wherein, step S120 includes: performing difference frequency processing according to the distance angle unit of two adjacent frames, and the formula is:

[0102]

[0103] In the formula, t represents the frame number of the radar. Represents the difference frequency between two adjacent frames; the echo signal phase is unwrapped according to the difference frequency, and the echo signal exceeding the phase main value is corrected to the echo signal within the phase main value. The formula is:

[0104]

[0105] In the formula, Represents the true value of the phase difference in the tth frame after correction; the moving distance of the target between two adjacent frames is calculated according to the corrected echo signal to obtain the preliminary deformation data, and the formula is:

[0106]

[0107] Where Δr is the moving distance of the target between two adjacent frames, and λ is the signal wavelength.

[0108] Specifically, in order to eliminate the influence of the Doppler effect, the difference frequency processing is performed on the two adjacent frames of the unit where the target is located to obtain the difference frequency between the two adjacent frames, so as to measure the target distance more accurately; since the phase data is periodic and cannot reflect the real physical changes, it is necessary to perform phase unwrapping of the echo signal according to the difference frequency, and correct the echo signal exceeding the phase main value to the echo signal within the phase main value, so as to reflect the continuous real phase change, so as to accurately obtain the deformation information; according to the corrected echo signal, the moving distance of the target between the two adjacent frames is calculated to obtain the preliminary deformation data.

[0109] Step S130, inputting the preliminary deformation data into an adaptive unscented Kalman filter, outputting target deformation data, and obtaining the micro-motion target based on the target deformation data analysis.

[0110] Specifically, in order to obtain more accurate measurement results, the preliminary deformation data is input into an adaptive unscented Kalman filter, which can provide more efficient and high-precision state estimation and is suitable for various nonlinear systems (radar echo data has highly nonlinear characteristics). It further processes the noise and interference in the signal and outputs the target deformation data. The micro-motion target is obtained based on the target deformation data analysis, thereby improving the accuracy of target detection and tracking.

[0111] Wherein, step S130 includes: constructing a sigma point set, and the initialization weight formula of the adaptive unscented Kalman filter is:

[0112]

[0113] In the formula, represents the initial weight of the state matrix, represents the u-th matrix of the state matrix, represents the initial weight of the covariance matrix, represents the u-th matrix of the covariance matrix, λ′ is the scaling parameter, n is the number of sigma point sets, α and β represent the state distribution parameters, which are integers;

[0114] Among them, the initialization covariance matrix is:

[0115]

[0116] Wherein, det represents the frame interval time of the radar; the preliminary deformation data is input into the initialized adaptive unscented Kalman filter, filtered by the adaptive unscented Kalman filter, and the target deformation data is output; and the micro-motion target is obtained according to the target deformation data analysis.

[0117] Specifically, a sigma point set (i.e., a set of sampling points) is constructed. After being mapped by a nonlinear function, the probability density function of the original distribution can be approximated. The adaptive unscented Kalman filter can directly process the probability density distribution of nonlinear functions with higher accuracy and robustness.

[0118] By inputting preliminary deformation data, the adaptive unscented Kalman filter filters the data between two adjacent frames, eliminates interference such as noise and error, and outputs the filtered target deformation data. The micro-motion target is obtained according to the target deformation data analysis, that is, the micro-motion target with deformation between two adjacent frames is within the radar measurement range, thereby achieving high-precision detection of tiny deformations, being able to detect more subtle deformation changes, greatly improving the detection accuracy and reliability, and ensuring that accurate detection data can still be obtained in various complex environments.

[0119] In one embodiment, it further includes: introducing outlier detection in the adaptive unscented Kalman filter, the formula is:

[0120]

[0121] In the formula, e t Represents the difference between the predicted result and the updated result, Represents the covariance matrix calculated based on the weights at the t-th moment;

[0122] If the calculated outlier value δ t Greater than the preset threshold δ t,L , then the covariance matrix is ​​modified, the formula is:

[0123]

[0124]

[0125]

[0126] ω=max(ω0,(δ t -b·δ t,L ) / δ t );

[0127] In the formula, Q′ t represents the updated prediction covariance matrix, Q t represents the prediction covariance matrix at time t, K t represents the Kalman coefficient matrix, The transpose of the matrix representing the difference between the predicted result and the updated result, represents the transpose of the Kalman coefficient matrix, R′ t represents the updated observation covariance matrix, R t represents the observation covariance matrix at time t, and ω represent weight coefficients, and ω0 represent the initial values ​​of the weight coefficients, represents the difference between the weighted prediction result and the observed result, represents the weight of the sigma point set, H represents the state transfer matrix, Represents the predicted mean after combining the weights, Z t represents the sensor observation value at time t, and a and b represent the tuning factors.

[0128] Specifically, in order to further improve the filtering accuracy of the adaptive Kalman filter, outlier detection may be introduced therein. When the calculated outlier is greater than a preset threshold, the covariance matrix is ​​corrected to ensure the filtering effect.

[0129] In this embodiment, the echo signal reflected by the target is received and preprocessed, and the preprocessed echo signal is used for calculation to obtain the distance angle unit where the target is located; difference frequency processing is performed according to the distance angle units of two adjacent frames, and signal correction is performed through phase unwrapping to obtain preliminary deformation data; the preliminary deformation data is input into an adaptive unscented Kalman filter, and the target deformation data is output. The micro-motion target is obtained based on the target deformation data analysis. Through advanced signal processing technology and the improvement of the unscented Kalman filter, high-precision detection of tiny deformations is achieved, which greatly improves the detection accuracy and reliability, ensuring that accurate monitoring data can still be obtained in various complex environments.

[0130] The present invention also has powerful all-weather monitoring capabilities and is not affected by adverse weather conditions. Whether it is heavy rain, heavy fog, strong wind or night environment, the millimeter wave radar can work stably to ensure the continuity and accuracy of the monitoring data. It can be calculated in real time and displayed through a GUI interface. It is particularly important for application scenarios that require long-term and continuous detection, such as building safety monitoring and landslide warning.

[0131] The present invention adopts a non-contact monitoring method, avoiding the physical influence and interference of traditional contact sensors on the monitored object, and is particularly suitable for areas where direct contact is not suitable, such as mountain slopes, thereby improving the safety and applicability of monitoring.

[0132] In addition, the present invention can be applied to surface deformation monitoring of dams, bridges or other buildings, and is particularly suitable for building safety monitoring, landslide warning and other industrial application fields that require high-sensitivity and high-precision deformation monitoring. It improves the accuracy of millimeter-wave radar in monitoring target deformation and has wide application value.

[0133] like Figure 4 and Figure 5 As shown, the actual error and Kalman gain effect diagram of the KF (Kalman filtering, Kalman filtering) algorithm, the UKF (unscented Kalman filter, unscented Kalman filtering) algorithm and the AUKF (Adaptive Unscented Kalman Filter, Adaptive Unscented Kalman Filter) algorithm of the present invention. Figure 4 It can be seen that the error of the AUKF algorithm of the present invention is the smallest, the error of the UKF algorithm is second, and the error of the KF algorithm is the highest, and it will increase rapidly with the increase of the number of iterations. The Kalman gain represents the degree of change of uncertainty after each fusion of data. The smaller the Kalman gain, the higher the prediction accuracy. Figure 5 It can be seen that with the increase of the number of iterations, the Kalman gain of the AUKF algorithm of the present invention is the smallest, and a prediction result with higher accuracy can be obtained compared with the UKF algorithm and the KF algorithm.

[0134] In summary, the adaptive unscented Kalman filter algorithm of the present invention has higher accuracy than the existing algorithms.

[0135] A person skilled in the art can understand that all or part of the processes in the above-mentioned embodiments can be implemented by instructing the relevant hardware through a computer program, and the program can be stored in a computer-readable storage medium, and when the program is executed, it can include the processes of the embodiments of the above-mentioned methods. The storage medium can be a disk, an optical disk, a read-only memory (ROM) or a random access memory (RAM), etc.

[0136] Obviously, those skilled in the art should understand that the modules or steps of the present invention described above can be implemented by a general-purpose computing device, they can be concentrated on a single computing device, or distributed on a network composed of multiple computing devices, and optionally, they can be implemented by a program code executable by a computing device, so that they can be stored in a computer storage medium (ROM / RAM, magnetic disk, optical disk) and executed by the computing device, and in some cases, the steps shown or described can be executed in a different order than that here, or they can be made into individual integrated circuit modules, or multiple modules or steps therein can be made into a single integrated circuit module for implementation. Therefore, the present invention is not limited to any specific combination of hardware and software.

[0137] The above contents are further detailed descriptions of the present invention in combination with specific implementation methods, and it cannot be determined that the specific implementation of the present invention is limited to these descriptions. For ordinary technicians in the technical field to which the present invention belongs, several simple deductions or substitutions can be made without departing from the concept of the present invention, which should be regarded as falling within the scope of protection of the present invention.

Claims

1. A method for detecting micro-motion target motion based on adaptive unscented Kalman filtering, characterized in that: The following steps are involved: Receive the echo signal reflected by the target and pre-process it, and perform calculations based on the pre-processed echo signal to obtain the distance angle unit where the target is located; Perform difference frequency processing according to the distance angle units of two adjacent frames, and perform signal correction through phase unwrapping to obtain preliminary deformation data; The preliminary deformation data is input into an adaptive unscented Kalman filter, and the target deformation data is outputted, and the micro-motion target is obtained according to the target deformation data analysis.

2. The method for detecting micro-motion target motion based on adaptive unscented Kalman filtering according to claim 1, characterized in that: Before receiving the echo signal reflected by the target and performing preprocessing, the method further includes: A millimeter-wave radar is arranged facing the target to be measured to transmit signals. The millimeter-wave radar adopts a two-transmit and four-receive transmission mode, and the number of sampling points in the range direction is set to 256.

3. The method for detecting micro-motion target motion based on adaptive unscented Kalman filtering according to claim 1, characterized in that: The receiving and preprocessing of the echo signal reflected by the target, and the calculation based on the preprocessed echo signal to obtain the distance angle unit where the target is located, include: receiving an echo signal reflected by a target and preprocessing the echo signal; The multiple signal classification algorithm is used to calculate the target angle unit based on the preprocessed echo signal. The formula is: Where a(θ) is the steering vector matrix, is the noise subspace eigenvector matrix, a T (θ) is the transpose of the steering vector matrix, is the transpose of the noise subspace eigenvector matrix; Obtaining the corresponding target distance according to the angle unit, and generating an initial distance angle unit of the target; The obtained echo signal is subjected to ordered statistic constant false alarm detection from the distance dimension, and unit average constant false alarm detection is performed from the angle dimension. The noise threshold of point (x, y) is: T os-ca (x,y)=max(T ca (x),T os (y)); Where i and j represent half of the number of reference units for ordered statistic constant false alarm detection and unit average constant false alarm detection, respectively. f(k,m) represents the value of the received signal amplitude spectrum at (k,m). The maximum value of f(k,m) is taken as the target location. The criterion for judging whether a point (x,y) has a target is: In the formula, when d(x,y) is not 0, it means there is a target, and when d(x,y) is 0, it means there is no target; By judging whether there is a target or not, the initial distance angle unit corresponding to the false alarm target is eliminated to obtain the distance angle unit where the target is located.

4. The method for detecting micro-motion target motion based on adaptive unscented Kalman filtering according to claim 1, characterized in that: The difference frequency processing is performed according to the distance angle unit of two adjacent frames, and the signal is corrected by phase unwrapping to obtain preliminary deformation data, including: The difference frequency processing is performed according to the distance angle unit of two adjacent frames, and the formula is: In the formula, t represents the frame number of the radar. Indicates the difference frequency between two adjacent frames; The echo signal phase is unwrapped according to the difference frequency, and the echo signal exceeding the phase main value is corrected to the echo signal within the phase main value. The formula is: In the formula, Represents the true value of the phase difference in the t-th frame after correction; The moving distance of the target between two adjacent frames is calculated based on the corrected echo signal to obtain the preliminary deformation data. The formula is: Where Δr is the moving distance of the target between two adjacent frames, and λ is the signal wavelength.

5. The method for detecting micro-motion target motion based on adaptive unscented Kalman filtering according to claim 1, characterized in that: The step of inputting the preliminary deformation data into an adaptive unscented Kalman filter for filtering, outputting target deformation data, and obtaining a micro-motion target according to the target deformation data analysis includes: Construct the sigma point set, and the initialization weight formula of the adaptive Kalman filter is: In the formula, represents the initial weight of the state matrix, represents the u-th matrix of the state matrix, represents the initial weight of the covariance matrix, represents the u-th matrix of the covariance matrix, λ ′ is the scaling parameter, n is the number of sigma point sets, α and β represent the state distribution parameters; Among them, the initialization covariance matrix is: In the formula, det represents the frame interval time of the radar; Inputting the preliminary deformation data into an initialized adaptive unscented Kalman filter, filtering through the adaptive unscented Kalman filter, and outputting target deformation data; A micro-motion target is obtained according to the target deformation data analysis.

6. The method for detecting micro-motion target motion based on adaptive unscented Kalman filtering according to claim 5, characterized in that: Also includes: Introducing outlier detection in the adaptive unscented Kalman filter, the formula is: In the formula, e t Represents the difference between the predicted result and the updated result, Represents the covariance matrix calculated based on the weights at the t-th moment; If the calculated outlier value δ t Greater than the preset threshold δ t,L , then the covariance matrix is ​​modified, the formula is: θ=max(θ0,(δ t -a·d t,L ) / d t ); ω=max(ω0,(δ t -b·d t,L ) / d t ); In the formula, Q′ t represents the updated prediction covariance matrix, Q t represents the prediction covariance matrix at time t, K t represents the Kalman coefficient matrix, e t A matrix representing the difference between the predicted result and the updated result, The transpose of the matrix representing the difference between the predicted result and the updated result, represents the transpose of the Kalman coefficient matrix, R′ t represents the updated observation covariance matrix, R t represents the observed covariance matrix at time t, θ and ω represent weight coefficients, θ0 and ω0 represent the initial values ​​of the weight coefficients, represents the difference between the weighted prediction result and the observed result, represents the weight of the sigma point set, H represents the state transfer matrix, Represents the predicted mean after combining the weights, Z t represents the sensor observation value at time t, and a and b represent the tuning factors.

Images