Millimeter wave non-line-of-sight detection method and system

Abstract

The present invention discloses a millimeter-wave non-line-of-sight detection method and system, belonging to the field of target detection technology. For complex and rough environments, a rough simulated surface model composed of multiple micro-surfaces with random slopes and lengths is constructed. The echo signal scattered back by a hidden target through the rough simulated surface is collected and decomposed into sub-echo signals on multiple scattering paths. The positional relationship between each scattering point and the hidden target is obtained based on the information carried by each sub-echo signal, thereby obtaining the position of the hidden target. The present invention expands the detection target signal path by utilizing the positional relationship between each scattering point and the hidden target. Even under a randomly rough surface without a specular reflection component, the position of the hidden target can be accurately sensed, and non-line-of-sight hidden target detection can be accurately and reliably performed in a complex and rough environment.

Description

Technical Field

[0001] The present invention belongs to the technical field of target detection, and more specifically, relates to a millimeter-wave non-line-of-sight detection method and system. Background Art

[0002] Millimeter-wave non-line-of-sight detection aims to recover information about obscured objects through indirect reflections of electromagnetic waves off visible surfaces. Decoding these echo signals can significantly expand a sensor's detection range and has a wide range of applications, including autonomous driving, disaster relief, and remote sensing. For example, at an intersection, if a sensor can detect the location of pedestrians around a corner, it can avoid them when making turns or planning the next step, reducing collisions and accidents.

[0003] However, existing non-line-of-sight detection methods suffer from the following problems: 1) Current NLOS detection systems assume that the visible relay reflective surface is an ideal smooth plane, ignoring the scattered signals generated by non-ideal rough surfaces in outdoor scenes. Specifically, for a single hidden target, the scattering effect of the rough surface will generate new virtual ghost targets, causing ambiguity in target recovery and making it impossible to accurately recover information such as the true target location. 2) The reflective micro-surfaces on random rough surfaces are random and depend on the geometric relationship between the NLOS scene and the rough surface. This results in the blocking of reliable mirror paths, making traditional NLOS sensing methods based on the third-order bounce principle difficult to operate effectively, resulting in poor robustness and significant detection errors.

[0004] Therefore, the existing NLOS millimeter-wave detection schemes are not suitable for concealed target reconstruction in complex and rough environments (such as field environments). It is of great value to study millimeter-wave non-line-of-sight detection based on rough surface scattering. Summary of the Invention

[0005] In response to the above defects or improvement needs of the existing technology, the present invention provides a millimeter-wave non-line-of-sight detection method and system to solve the technical problem that the existing technology cannot accurately and reliably detect non-line-of-sight hidden targets in complex and rough environments.

[0006] In order to achieve the above objectives, in a first aspect, the present invention provides a millimeter wave non-line-of-sight detection method, comprising the following steps:

[0007] S1. Using a sensor at a vertical distance h from the rough simulated surface, transmit a modulated signal to the rough simulated surface and collect echo signals scattered back by the covert target through the rough simulated surface; wherein the rough simulated surface is composed of multiple micro-surfaces with random slopes and lengths;

[0008] S2. Estimate the path distance and arrival angle information of the sub-echo signals on multiple scattering paths of the echo signal; the i-th scattering path where the i-th sub-echo signal is located is from the hidden target through the scattering point A on the i-th micro surface. i The signal path to the sensor has a path distance of r i , the arrival angle is θ i ;

[0009] S3, based on the path distance and arrival angle information of the sub-echo signal, calculate the actual distance between each scattering point on the rough simulation surface and the hidden target T; where the scattering point A i The actual distance dis from the hidden target T i =r i -b i ; b i =h / sinθ i Scattering point A i The distance from the sensor S;

[0010] S4. The position of the hidden target is obtained by minimizing the difference between the distance between each scattering point and the hidden target and its actual distance.

[0011] Preferably, in an optional embodiment, step S4 includes: establishing a cost function with the goal of minimizing the difference between the distance between each scattering point and the hidden target and its actual distance, thereby obtaining the position of the hidden target; wherein the cost function is:

[0012]

[0013] Where (x, y) is the position coordinate of the hidden target; I is the number of scattering paths; Scattering point A i The location coordinates of .

[0014] Preferably, in an optional embodiment, step S4 includes: performing a first-order Taylor expansion on the square root term in the cost function and minimizing the result to obtain the position of the hidden target.

[0015] in,

[0016] Preferably, in an optional implementation manner, step S2 includes:

[0017] S21, performing tensor decomposition on the echo signal after performing tensor representation to obtain a two-dimensional matrix (2-mode matrix) corresponding to each dimension of information carried by the echo signal;

[0018] S22. Perform singular value decomposition on the two-dimensional matrix corresponding to each dimension information to obtain the eigenvalue matrix and right singular value matrix corresponding to each dimension information;

[0019] S23. Based on the Bayesian information theory criterion, determine the eigenvalues ​​in the eigenvalue matrix corresponding to the q-th dimensional information to obtain the number of scattering paths corresponding to the q-th dimensional information; and take the maximum value of the number of scattering paths corresponding to all dimensional information as the number of scattering paths I of the echo signal.

[0020] S24, obtain the smallest I eigenvalues ​​in the eigenvalue matrix corresponding to the qth dimension information, form the first singular value matrix with the right singular value vectors corresponding to the I eigenvalues, and divide the first singular value matrix into two sub-matrices V1 and V2 with the same number of elements based on the rotation invariance technology, and calculate the matrix bundle The generalized eigenvalue decomposition of is performed to obtain the I generalized eigenvalues ​​under the qth dimension information, and then the path distance, arrival angle information and speed information of each sub-echo signal and their expressions are obtained;

[0021] The dimensional information includes: path distance, arrival angle information and speed information; q = 1, 2, 3.

[0022] Further preferably, the tensor of the echo signal is expressed as:

[0023]

[0024] Where I is the number of scattering paths; ρ i is the attenuation of the i-th scattering path; represents the outer product; r i is the path distance of the i-th sub-echo signal; q i is the arrival angle information of the i-th sub-echo signal; v i is the velocity vector information carried by the i-th sub-echo signal; K is the sampling number of the modulation signal; B is the bandwidth of the sensor transmission signal; c is the speed of light; T m is the duration of the modulation signal; f s is the sampling rate; L is the number of antennas; f c is the carrier frequency; d is the antenna spacing; is Gaussian white noise;

[0025] The expression of the i-th sub-echo signal is:

[0026]

[0027] k is the path distance index; l is the arrival angle index;

[0028] The path distance r of the i-th sub-echo signal i and arrival angle information q i They are:

[0029]

[0030]

[0031] Among them, λ r,i is the i-th generalized eigenvalue under the path distance dimension information; q,i is the i-th generalized eigenvalue under the arrival angle dimension information; λ v,i is the i-th generalized eigenvalue under the velocity dimension information; ∠(·) is the function representing the extracted phase angle.

[0032] Preferably, in an optional embodiment, the millimeter wave non-line-of-sight detection method further includes step S5 performed after step S4 to simultaneously detect the position and velocity of the concealed target; in this case, step S2 further includes: estimating velocity information carried by sub-echo signals on multiple scattering paths of the echo signal;

[0033] The above step S5 includes:

[0034] S51, based on the speed information carried by the sub-echo signal, obtain the true speed of the hidden target on each scattering path; wherein, for the i-th scattering path, based on the position of the hidden target and the scattering point A i The actual distance between the hidden target T and the hidden target is obtained by the angle α between the velocity vector carried by the i-th sub-echo signal and the moving direction of the hidden target. i , and then based on the cosine theorem, the true speed of the hidden target on the i-th scattering path is obtained; the moving direction of the hidden target includes horizontal and vertical directions;

[0035] S52. Predict the speed of the concealed target by minimizing the difference between the speed of the concealed target and the true speed of the concealed target on each scattering path.

[0036] Further preferably, when it is necessary to simultaneously estimate the path distance, arrival angle information, and velocity information carried by the sub-echo signals on multiple scattering paths of the echo signal; the above-mentioned step S2 includes:

[0037] S21, performing tensor decomposition on the echo signal after performing tensor representation to obtain a two-dimensional matrix (3-mode matrix) corresponding to each dimension of information carried by the echo signal;

[0038] S22. Perform singular value decomposition on the two-dimensional matrix corresponding to each dimension information to obtain the eigenvalue matrix and right singular value matrix corresponding to each dimension information;

[0039] S23. Based on the Bayesian information theory criterion, determine the eigenvalues ​​in the eigenvalue matrix corresponding to the q-th dimensional information to obtain the number of scattering paths corresponding to the q-th dimensional information; and take the maximum value of the number of scattering paths corresponding to all dimensional information as the number of scattering paths I of the echo signal.

[0040] S24, obtain the smallest I eigenvalues ​​in the eigenvalue matrix corresponding to the qth dimension information, form the first singular value matrix with the right singular value vectors corresponding to the I eigenvalues, and divide the first singular value matrix into two sub-matrices V1 and V2 with the same number of elements based on the rotation invariance technology, and calculate the matrix bundle The generalized eigenvalue decomposition of is performed to obtain the I generalized eigenvalues ​​under the qth dimension information, thereby obtaining the path distance, arrival angle information and the speed information carried by each sub-echo signal;

[0041] The dimensional information includes: path distance, arrival angle information and speed information; q = 1, 2, 3.

[0042] Further preferably, the tensor of the echo signal is expressed as:

[0043]

[0044] Where I is the number of scattering paths; ρ i is the attenuation of the i-th scattering path; represents the outer product; r i is the path distance of the i-th sub-echo signal; q i is the arrival angle information of the i-th sub-echo signal; v i is the velocity vector information carried by the i-th sub-echo signal; K is the sampling number of the modulation signal; B is the bandwidth of the sensor transmission signal; c is the speed of light; T m is the duration of the modulation signal; f s is the sampling rate; L is the number of antennas; f c is the carrier frequency; d is the antenna spacing; M is the total number of modulated signals; T c is the pulse period of the modulation signal; is Gaussian white noise;

[0045] The expression of the i-th sub-echo signal is:

[0046]

[0047] Where k is the path distance index; n is the speed index; l is the arrival angle index;

[0048] The path distance r of the i-th sub-echo signal i, arrival angle information q i and the velocity vector information v i They are:

[0049]

[0050]

[0051]

[0052] Among them, λ r,i is the i-th generalized eigenvalue under the path distance dimension information; q,i is the i-th generalized eigenvalue under the arrival angle dimension information; λ v,i is the i-th generalized eigenvalue under the velocity dimension information; ∠(·) is the function representing the extracted phase angle.

[0053] Preferably, in an optional implementation manner, in step S52, the concealed target velocity is gradually updated in an iterative manner until convergence is reached, so as to minimize the difference between the concealed target velocity and the true velocity of the concealed target on each scattering path; wherein the concealed target velocity at the t+1th iteration is:

[0054]

[0055] I is the number of scattering paths; v i is the velocity vector carried by the i-th sub-echo signal; v t is the hidden target velocity at the tth iteration.

[0056] In a second aspect, the present invention provides a millimeter wave non-line-of-sight detection system, comprising: a memory and a processor, wherein the memory stores a computer program, and the processor executes the millimeter wave non-line-of-sight detection method provided by the first aspect of the present invention when executing the computer program.

[0057] In a third aspect, the present invention further provides a computer-readable storage medium, which includes a stored computer program, wherein when the computer program is executed by a processor, the device where the storage medium is located is controlled to execute the millimeter wave non-line-of-sight detection method provided in the first aspect of the present invention.

[0058] In general, the above technical solutions conceived by the present invention can achieve the following beneficial effects:

[0059] 1. The present invention provides a millimeter-wave non-line-of-sight detection method. For complex and rough environments, a rough simulated surface model composed of multiple micro-surfaces with random slopes and lengths is constructed. The echo signal scattered back by a hidden target through the rough simulated surface is collected and decomposed into sub-echo signals on multiple scattering paths. The positional relationship between each scattering point and the hidden target is obtained based on the information carried by each sub-echo signal, thereby obtaining the position of the hidden target. By utilizing the positional relationship between each scattering point and the hidden target, the present invention expands the detection target signal path. Even under a randomly rough surface without a specular reflection component, the position of the hidden target can be accurately perceived, and non-line-of-sight hidden target detection can be accurately and reliably performed in complex and rough environments.

[0060] 2. Furthermore, the millimeter-wave non-line-of-sight detection method provided by the present invention, when solving the position of a hidden target, takes into account the presence of a square root expression in the corresponding cost function, and the amount of direct calculation is large. Therefore, the above-mentioned cost function is transformed to resolve the square root expression therein, and the square root expression is subjected to a low-order Taylor expansion, and a polynomial function that is easy to calculate is used to approximate the smooth cost function, and the simplified approximate cost function is obtained through the least squares algorithm, thereby reducing the complexity and greatly improving the computational efficiency.

[0061] 3. Furthermore, the millimeter-wave non-line-of-sight detection method provided by the present invention, when estimating the relevant information of the sub-echo signal, first performs tensor representation on the echo signal and then performs tensor decomposition to obtain a two-dimensional matrix corresponding to each dimensional information carried by the echo signal, thereby reducing the complexity of three-dimensional signal processing. Then, the two-dimensional matrix corresponding to each dimensional information is singular value decomposition is performed respectively to obtain the eigenvalue matrix and right singular value matrix corresponding to each dimensional information; and based on the Bayesian information theory criterion, the number of scattering paths of the echo signal is accurately obtained to avoid the problem of low target recovery accuracy caused by a large number of invalid paths; finally, the relevant information of each sub-echo signal is obtained based on the rotational invariance technology; the path distance and arrival angle information of each sub-echo signal can be estimated, and in addition, the speed information carried can also be estimated. The present invention uses the correlation between signals to improve the accuracy of parameter estimation, realizes the accurate acquisition of information carried by each scattering path, and is conducive to recovering the true position, speed and other information of the hidden target.

[0062] 4. Furthermore, the millimeter-wave non-line-of-sight detection method provided by the present invention can also be used to detect the speed information of concealed targets based on the information carried by each sub-echo signal, with high accuracy and more complete detected information. BRIEF DESCRIPTION OF THE DRAWINGS

[0063] Figure 1 A flowchart of a millimeter-wave non-line-of-sight detection method provided by an embodiment of the present invention;

[0064] Figure 2 Schematic diagram of scattered signal detection in a rough NLOS scenario provided by an embodiment of the present invention;

[0065] Figure 3 A diagram of the NLOS detection scenario under a real rough reflecting surface provided by an embodiment of the present invention;

[0066] Figure 4 A schematic diagram of the hidden target position estimation result provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0067] In order to make the objectives, technical solutions and advantages of the present invention more clearly understood, the present invention is further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely for the purpose of explaining the present invention and are not intended to limit the present invention. In addition, the technical features involved in the various embodiments of the present invention described below may be combined with each other as long as they do not conflict with each other.

[0068] In order to further illustrate the millimeter wave non-line-of-sight detection method provided by the present invention, a specific embodiment is described below:

[0069] In order to solve the problems of low detection accuracy and low detection efficiency in complex environments in existing technologies, this embodiment provides a millimeter wave non-line-of-sight detection method based on rough surface scattering, which can simultaneously detect the position and speed of hidden targets. Figure 1 As shown in the figure, it includes rough NLOS scene establishment, recording of rough scene geometry information, signal acquisition and preprocessing, target state estimation and positioning detection modules.

[0070] Detecting moving targets in a rough and complex NLOS environment. The simulation scenario is as follows Figure 2 As shown, in this embodiment, considering that the millimeter wave radar is small, highly portable, and easy to implement, the radar is used as the sensor (it should be noted that the sensor type in the present invention is not limited to the millimeter wave radar, and can also be millimeter wave WIFI, etc.), and the radar is placed At this point, it can be seen that the length of the relay surface is L ω , the distance between the sensor and the surface is h, and the position of the hidden moving target in the experiment is where x p >0,y p <h, and it moves parallel to the surface of the relay wall at a speed v. The radar's transmitted signal frequency is 77 GHz, with a bandwidth of B = 1 GHz, and the interval between two adjacent receiving antennas is half a wavelength. Figure 3 This is the NLOS detection scene image under a real rough reflecting surface.

[0071] More specifically, the processing steps according to the present invention are:

[0072] (1) Record the rough geometric information of the environment in which the MIMO millimeter-wave radar is located and estimate the vertical distance h = 5m from the radar antenna to the rough surface.

[0073] (2) A random rough surface model was established to obtain a rough simulated surface. Specifically, based on the principles of ray tracing and electromagnetic scattering, the observable rough surface was described as a set of micro-surfaces with random slopes and lengths, and the scattering points on the rough surface were assumed to be w. i .

[0074] According to the electromagnetic scattering characteristics of millimeter waves, a geometric model of observable rough surfaces was established based on ray tracing and random geometry theory. That is the above-mentioned rough simulated surface; wherein, is the topology and operation symbol; M is the number of microsurfaces; is the topological structure of the mth microsurface; are the length, slope, and starting point coordinates of the mth microsurface respectively.

[0075] (3) Build Figure 2 In the simulation environment shown, a radar (whose model, settings, and transmitted signals are exactly the same as those in a real environment) transmits a frequency-modulated continuous wave (FMCW) toward an observable rough surface to detect hidden targets. The radar then collects the echo signals scattered back by the rough micro-scattering surface and performs preliminary preprocessing to obtain a three-dimensional echo tensor signal of range, velocity, and angle.

[0076] Specifically, the radar sensor is a MIMO FMCW radar system having a uniform linear array consisting of L units, and the FMCW chirp sequence is a chirp duration T m and the pulse repetition interval T c Then the carrier frequency f c The FM continuous wave signal on band B starts: When t = t c +nT c ,t c ∈[0,T c ) decomposes the periodic emission signal into the fast time domain t c and slow time domain Where n = 0, 1, 2, ..., N-1, and N is the total number of chirps. The transmitted signal g0(t) and the signal scattered back through the rough micro-scattering surface are Perform cross-correlation and other operations to obtain the intermediate frequency signal Among them, r i represents the path distance of the i-th scattering path, represents the velocity component of the hidden target on the i-th scattering path; q i is the arrival angle of the i-th sub-echo signal, ρ i is the attenuation of the i-th scattering path; then, I independent scattering paths are superimposed, and the three-dimensional echo tensor of range-velocity-angle is expressed as: in, represents the outer product, Represents fast-time sinusoidal vectors, velocities, angles, and so on.

[0077] (4) Decompose the echo signal into sub-echo signals on multiple scattering paths; the i-th scattering path where the i-th sub-echo signal is located is the path from the hidden target through the scattering point A on the i-th micro-surface. i The signal path to the sensor has a path distance of r i , the arrival angle is θ i ;

[0078] Specifically, the echo signal is subjected to tensor decomposition and denoising. The path distance, angle of arrival, and velocity information carried by each echo signal are accurately estimated using a rotationally invariant algorithm to obtain each sub-echo signal. Possible tensor decomposition methods include Tucker decomposition and CP decomposition. Algorithms such as MUSIC, ESPRIT, and FFT can be used to accurately estimate the path distance, angle of arrival, and velocity information carried by each echo signal.

[0079] In an optional embodiment, step (4) includes the following sub-steps:

[0080] (4-1) The echo signal is represented as a tensor and then decomposed into tensors to obtain a two-dimensional matrix corresponding to each dimension of information carried by the echo signal, thereby reducing the complexity of three-dimensional signal processing.

[0081] Similarly, the tensor decomposition methods that can be used include Tucker decomposition method, CP decomposition method, etc. In this embodiment, the radar echo tensor Perform Tucker tensor decomposition: Where × j is the j-mode product, resulting in the core tensor represents the unit tensor of H in terms of path distance r, arrival angle θ, and velocity v, and F r 、F v and F θ is the corresponding two-dimensional mode vector.

[0082] (4-2) Perform singular value decomposition on the two-dimensional matrix corresponding to each dimension information to obtain the eigenvalue matrix and right singular value matrix corresponding to each dimension information;

[0083] (4-3) Based on the Bayesian information theory criterion, the eigenvalues ​​in the eigenvalue matrix corresponding to the qth dimension information are determined to obtain the number of scattering paths corresponding to the qth dimension information; the maximum number of scattering paths corresponding to all dimensional information is taken as the number of scattering paths I of the echo signal; wherein the dimensional information includes: path distance, arrival angle information, and velocity information; q = 1, 2, 3;

[0084] F r For example, for F r Perform singular value decomposition (such as SVD, HOSVD, etc.) to obtain the corresponding eigenvalue matrix Λ r =diag(λ r,1 ',λ r,2 ',...,λ r,K ') and the right singular value matrix. Based on the Bayesian information theory criterion The number of scattering paths corresponding to the path distance dimension information can be estimated Using the same method, we can also get the number of scattering paths corresponding to the arrival angle dimension information. The number of scattering paths corresponding to the velocity dimension information Then we can get the number of scattering paths of the echo signal The number of scattering paths of the echo signal is accurately calculated based on the Bayesian information theory criterion, which avoids the problem of low target recovery accuracy caused by a large number of invalid paths.

[0085] (44) Obtain the smallest I eigenvalues ​​in the eigenvalue matrix corresponding to the qth dimension information, form the first singular value matrix with the right singular value vectors corresponding to the I eigenvalues, and divide the first singular value matrix into two sub-matrices V1 and V2 with the same number of elements based on the rotation invariance technology, and calculate the matrix bundle The generalized eigenvalue decomposition of is performed to obtain the I generalized eigenvalues ​​under the qth dimension information, and then the path distance, arrival angle information and speed information of each sub-echo signal and their expressions are obtained;

[0086] Specifically, taking the dimension information of path distance as an example, the smallest I eigenvalues ​​in the eigenvalue matrix corresponding to the dimension information of path distance are obtained, and the right singular value vectors corresponding to the I eigenvalues ​​form the first singular value matrix. Based on the rotation invariance technology, the first singular value matrix is ​​divided into two sub-matrices V1 and V2 with the same number of elements, and the matrix bundle is calculated. The generalized eigenvalue decomposition of is used to obtain the I generalized eigenvalues ​​under the path distance dimension information, and then the path distance of each sub-echo signal is obtained; the arrival angle and velocity information of each sub-echo signal can also be obtained by the same method. Specifically, the path distance r of the i-th sub-echo signal is i , arrival angle information q iand the velocity vector information v i They are:

[0087]

[0088]

[0089]

[0090] Among them, λ r,i is the i-th generalized eigenvalue under the path distance dimension information; q,i is the i-th generalized eigenvalue under the arrival angle dimension information; λ v,i is the i-th generalized eigenvalue under the velocity dimension information; ∠(·) is the function representing the extracted phase angle.

[0091] (5) Based on the path distance and arrival angle information of the sub-echo signal, the actual distance between each scattering point on the rough simulation surface and the hidden target T is calculated; where the scattering point A i The actual distance dis from the hidden target T i =r i -b i ; b i =h / sinθ i Scattering point A i The distance from the sensor S;

[0092] (6) Calculate the coordinates of the scattering points formed by each scattering path on the rough simulation surface, and calculate the distance from each scattering point to the hidden target.

[0093] In this embodiment, a Cartesian coordinate system is established with the sensor as the origin, the direction parallel to the rough simulated surface as the x-axis, and the direction perpendicular to the rough simulated surface as the y-axis; according to the geometric relationship, the coordinates of the i-th scattering point are obtained. is (h / tanq i ,h). It should be noted that this embodiment only takes the Cartesian coordinate system as an example, and other coordinate systems, such as the polar coordinate system, can also be used for solving.

[0094] (7) The position of the hidden target is obtained by minimizing the difference between the distance between each scattering point and the hidden target and its actual distance.

[0095] Specifically, the hidden target coordinates are calculated by minimizing the difference between the distance between each scattering point and the hidden target and its actual distance. Taking the Cartesian coordinate system as an example, the hidden target coordinates satisfy the following cost function: Where (x, y) is the position coordinate of the hidden target; I is the number of scattering paths; Scattering point A iThe location coordinates of .

[0096] The cost function can be solved by gradient descent method, Newton iteration method and other methods. Considering that there is a square root expression in the cost function, the amount of calculation required for direct calculation is large. Preferably, in an optional implementation, the cost function is transformed to reduce the complexity and improve the calculation efficiency by solving the square root. Specifically, the first-order Taylor expansion is introduced to reduce the calculation complexity. At this point, the cost function can be simplified to: in, In order to minimize the cost function, we have in,

[0097] According to the above cost function, the least squares solution, that is, the true hidden position coordinates, is calculated as: like Figure 4 shown.

[0098] When solving the position of the hidden target, this embodiment takes into account the presence of a square root expression in the corresponding cost function, and the amount of computation required for direct calculation is large. Therefore, the above-mentioned cost function is transformed to resolve the square root expression, and the square root expression is subjected to a low-order Taylor expansion. A polynomial function that is easy to calculate is used to approximate the smooth cost function, and the simplified approximate cost function is obtained through the least squares algorithm, thereby reducing complexity and greatly improving computational efficiency.

[0099] (8) Further detecting the speed of the concealed target includes the following steps:

[0100] (8-1) Based on the velocity information carried by the sub-echo signal, the true velocity of the hidden target on each scattering path is obtained; wherein, for the i-th scattering path, based on the position of the hidden target and the scattering point A i The actual distance between the hidden target T and the hidden target is obtained by the angle α between the velocity vector carried by the i-th sub-echo signal and the moving direction of the hidden target. i , and then based on the cosine theorem, the true speed of the hidden target on the i-th scattering path is obtained; the moving direction of the hidden target includes horizontal and vertical directions;

[0101] (8-2) The speed of the hidden target is predicted by minimizing the difference between the speed of the hidden target and the true speed of the hidden target on each scattering path.

[0102] Similarly, the above cost function can be solved using methods such as gradient descent and Newton's iteration. Preferably, in an optional embodiment, the hidden target velocity is updated incrementally through iteration until convergence is reached, so as to minimize the difference between the hidden target velocity and the true velocity of the hidden target on each scattering path; wherein, the hidden target velocity at the t+1th iteration is:

[0103]

[0104] I is the number of scattering paths; v i is the velocity vector carried by the i-th sub-echo signal, specifically is the vector from the hidden target to the i-th scattering point, is the coordinate of the i-th scattering point. t is the hidden target velocity at the tth iteration. As long as the intermediate iteration v t Not with any point v i Consistently, the method converges to the geometric median of all velocity values.

[0105] It should be noted that the velocity estimation problem is transformed into a problem of solving the geometric center point, that is, finding the shortest sum of the distances from a point to all estimated points; the fixed-point iteration method is used here, without considering the iteration direction and iteration distance, and the current point is directly substituted into the formula to obtain the next more approximate point. The iteration is stopped by customizing the limit of the iterative optimization, avoiding the non-convergence problem caused by the fact that the gradient of the objective function does not exist everywhere.

[0106] It should be noted that the specific method for detecting only the position of the concealed target is the same as steps (1)-(7) in the above embodiment of simultaneously detecting the position and speed of the concealed target, with only adaptive modifications being made, and will not be elaborated here.

[0107] In summary, the present invention utilizes the angular resolution of millimeter-wave MIMO signals to detect the positions of scattering points at different locations on a rough reflective surface, and then calculates the position of the hidden target based on the positional relationship between each scattering point and the hidden target. This avoids detection failures caused by random rough surfaces, is not limited by the assumption that the relay reflective surface is smooth, does not depend on the presence of a mirror reflection path, and will not fail due to ambient light or outdoor conditions. It is suitable for detecting and locating moving targets such as hidden humans and vehicles in environments with random, irregular surfaces (such as the wild, at intersections, and on curved streets), reducing the risk of target collisions in blind spots. It is of great significance in the fields of intelligent transportation, security, etc. At the same time, it does not need to suppress multipath information in complex environments. Instead, target detection is achieved through the position information of multiple scattering points caused by multipath. The present invention is not complex in engineering implementation and does not require a large amount of computation. These results verify the practicality and effectiveness of the present invention for non-line-of-sight detection based on millimeter-wave radar on rough surfaces in the wild.

[0108] In a second aspect, the present invention provides a millimeter wave non-line-of-sight detection system, comprising: a memory and a processor, wherein the memory stores a computer program, and the processor executes the millimeter wave non-line-of-sight detection method provided by the first aspect of the present invention when executing the computer program.

[0109] The relevant technical solution is the same as the millimeter wave non-line-of-sight detection method provided in the first aspect, and will not be described in detail here.

[0110] In a third aspect, the present invention further provides a computer-readable storage medium, which includes a stored computer program, wherein when the computer program is executed by a processor, the device where the storage medium is located is controlled to execute the millimeter wave non-line-of-sight detection method provided in the first aspect of the present invention.

[0111] The relevant technical solution is the same as the millimeter wave non-line-of-sight detection method provided in the first aspect, and will not be described in detail here.

[0112] It will be easily understood by those skilled in the art that the above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A millimeter wave non-line-of-sight detection method, characterized in that: The following steps are involved: S1. Using a sensor at a vertical distance h from the rough simulated surface to transmit a modulated signal to the rough simulated surface, and collecting the echo signal scattered back by the hidden target through the rough simulated surface; the rough simulated surface is composed of multiple micro-surfaces with random slopes and lengths; S2. Estimate the path distance and arrival angle information of the sub-echo signals on multiple scattering paths of the echo signal; the i-th scattering path where the i-th sub-echo signal is located is from the hidden target through the scattering point A on the i-th micro surface. i The signal path to the sensor has a path distance of r i , the arrival angle is θ i ; S3, based on the path distance and arrival angle information of the sub-echo signal, calculate the actual distance between each scattering point on the rough simulation surface and the hidden target; where scattering point A i The actual distance to the hidden target dis i =r i -b i ; b i =h / sinθ i Scattering point A i The distance from the sensor S; S4. The position of the hidden target is obtained by minimizing the difference between the distance between each scattering point and the hidden target and its actual distance.

2. The millimeter wave non-line-of-sight detection method according to claim 1, characterized in that: The step S4 includes: establishing a cost function with the goal of minimizing the difference between the distance between each scattering point and the hidden target and its actual distance, thereby obtaining the position of the hidden target; wherein the cost function is: Where (x, y) is the position coordinate of the hidden target; I is the number of scattering paths; Scattering point A i The location coordinates of .

3. The millimeter wave non-line-of-sight detection method according to claim 2, characterized in that: The step S4 includes: performing a first-order Taylor expansion on the square root term in the cost function and minimizing the result to obtain the position of the hidden target. in, 4. The millimeter wave non-line-of-sight detection method according to claim 1, characterized in that: The step S2 comprises: S21, performing tensor decomposition on the echo signal after performing tensor representation to obtain a two-dimensional matrix corresponding to each dimension of information carried by the echo signal; S22. Perform singular value decomposition on the two-dimensional matrix corresponding to each dimension information to obtain the eigenvalue matrix and right singular value matrix corresponding to each dimension information; S23. Based on the Bayesian information theory criterion, determine the eigenvalues ​​in the eigenvalue matrix corresponding to the q-th dimensional information to obtain the number of scattering paths corresponding to the q-th dimensional information; and take the maximum value of the number of scattering paths corresponding to all dimensional information as the number of scattering paths I of the echo signal. S24, obtain the smallest I eigenvalues ​​in the eigenvalue matrix corresponding to the qth dimension information, form the first singular value matrix with the right singular value vectors corresponding to the I eigenvalues, and divide the first singular value matrix into two sub-matrices V1 and V2 with the same number of elements based on the rotation invariance technology, and calculate the matrix bundle The generalized eigenvalue decomposition of is performed to obtain the I generalized eigenvalues ​​under the qth dimension information, thereby obtaining the path distance and arrival angle information of each sub-echo signal; The dimensional information includes: path distance and arrival angle information; q = 1, 2.

5. The millimeter wave non-line-of-sight detection method according to claim 4, characterized in that: The tensor representation of the echo signal is: Where I is the number of scattering paths; ρ i is the attenuation of the i-th scattering path; represents the outer product; r i is the path distance of the i-th sub-echo signal; θ i is the arrival angle information of the i-th sub-echo signal; v i is the velocity vector information carried by the i-th sub-echo signal; K is the sampling number of the modulation signal; B is the bandwidth of the sensor transmission signal; c is the speed of light; T m is the duration of the modulation signal; f s is the sampling rate; L is the number of antennas; f c is the carrier frequency; d is the antenna spacing; is Gaussian white noise; The expression of the i-th sub-echo signal is: k is the path distance index; l is the arrival angle index; The path distance r of the i-th sub-echo signal i and arrival angle information θ i They are: Among them, λ r,i is the i-th generalized eigenvalue under the path distance dimension information; θ,i is the i-th generalized eigenvalue under the arrival angle dimension information; λ v,i is the i-th generalized eigenvalue under the velocity dimension information; ∠(·) is the function representing the extracted phase angle.

6. The millimeter wave non-line-of-sight detection method according to any one of claims 1 to 5, characterized in that: The method further includes step S5 performed after step S4, wherein step S2 further includes: estimating velocity information carried by sub-echo signals on multiple scattered paths of the echo signal; The step S5 comprises: S51, based on the speed information carried by the sub-echo signal, obtain the true speed of the hidden target on each scattering path; wherein, for the i-th scattering path, based on the position of the hidden target and the scattering point A i The actual distance between the hidden target T and the hidden target is obtained by the angle α between the velocity vector carried by the i-th sub-echo signal and the moving direction of the hidden target. i , and then based on the cosine theorem, the true speed of the hidden target on the i-th scattering path is obtained; the moving direction of the hidden target includes horizontal and vertical directions; S52. Predict the speed of the concealed target by minimizing the difference between the speed of the concealed target and the true speed of the concealed target on each scattering path.

7. The millimeter wave non-line-of-sight detection method according to claim 6, characterized in that: The step S2 comprises: S21, performing tensor decomposition on the echo signal after performing tensor representation to obtain a two-dimensional matrix corresponding to each dimension of information carried by the echo signal; S22. Perform singular value decomposition on the two-dimensional matrix corresponding to each dimension information to obtain the eigenvalue matrix and right singular value matrix corresponding to each dimension information; S23. Based on the Bayesian information theory criterion, determine the eigenvalues ​​in the eigenvalue matrix corresponding to the q-th dimensional information to obtain the number of scattering paths corresponding to the q-th dimensional information; and take the maximum value of the number of scattering paths corresponding to all dimensional information as the number of scattering paths I of the echo signal. S24, obtain the smallest I eigenvalues ​​in the eigenvalue matrix corresponding to the qth dimension information, form the first singular value matrix with the right singular value vectors corresponding to the I eigenvalues, and divide the first singular value matrix into two sub-matrices V1 and V2 with the same number of elements based on the rotation invariance technology, and calculate the matrix bundle The generalized eigenvalue decomposition of is performed to obtain the I generalized eigenvalues ​​under the qth dimension information, thereby obtaining the path distance, arrival angle information and the speed information carried by each sub-echo signal; The dimensional information includes: path distance, arrival angle information and speed information; q = 1, 2, 3.

8. The millimeter wave non-line-of-sight detection method according to claim 7, characterized in that: The tensor representation of the echo signal is: Where I is the number of scattering paths; ρ i is the attenuation of the i-th scattering path; represents the outer product; r i is the path distance of the i-th sub-echo signal; θ i is the arrival angle information of the i-th sub-echo signal; v i is the velocity vector information carried by the i-th sub-echo signal; K is the sampling number of the modulation signal; B is the bandwidth of the sensor transmission signal; c is the speed of light; T m is the duration of the modulation signal; f s is the sampling rate; L is the number of antennas; f c is the carrier frequency; d is the antenna spacing; M is the total number of modulated signals; T c is the pulse period of the modulation signal; is Gaussian white noise; The expression of the i-th sub-echo signal is: k is the path distance index; n is the speed index; l is the arrival angle index; The path distance r of the i-th sub-echo signal i , arrival angle information θ i and the velocity vector information v i They are: λ r,i is the i-th generalized eigenvalue under the path distance dimension information; θ,i is the i-th generalized eigenvalue under the arrival angle dimension information; λ v,i is the i-th generalized eigenvalue under the velocity dimension information; ∠(·) is the function representing the extracted phase angle.

9. The millimeter wave non-line-of-sight detection method according to claim 6, characterized in that: In step S52, the hidden target velocity is updated step by step in an iterative manner until convergence is reached, so as to minimize the difference between the hidden target velocity and the true velocity of the hidden target on each scattering path; wherein, the hidden target velocity at the t+1th iteration is: I is the number of scattering paths; v i is the velocity vector carried by the i-th sub-echo signal; v t is the hidden target velocity at the tth iteration.

10. A millimeter wave non-line-of-sight detection system, characterized in that: include: A memory and a processor, wherein the memory stores a computer program, and when the processor executes the computer program, the millimeter wave non-line-of-sight detection method according to any one of claims 1 to 9 is executed.

Images